Cosmology Part I: The Homogeneous Universe

Hiranya V. Peiris1, ∗
1
Department of Physics and Astronomy, University College London, Gower Street, London, WC1E 6BT, U.K.

Classes

Thursdays, Spring Term

13/1 Cruciform B.3.01 10.00 − 13.00
20/1 Physics E1 10.00 − 13.00
27/1 Physics E7 10.00 − 13.00
03/2 Physics E1 10.00 − 13.00
10/2 Cruciform B.3.01 09.00 − 12.00 (note different time)
17/2 NO LECTURE
24/2 Physics E1 10.00 − 13.00
03/3 Physics E1 10.00 − 13.00
10/3 Physics E7 10.00 − 13.00
17/3 Physics E1 10.00 − 13.00
24/3 Physics E1 10.00 − 13.00

Office

G04, Kathleen Lonsdale Building.

Course Website

http://zuserver2.star.ucl.ac.uk/∼hiranya/PHASM336/

Introductory Reading:

1. Liddle, A. An Introduction to Modern Cosmology. Wiley (2003)

Complementary Reading

1. Dodelson, S. Modern Cosmology. Academic Press (2003) ∗

2. Carroll, S.M. Spacetime and Geometry. Addison-Wesley (2004) ∗
3. Liddle, A.R. and Lyth, D.H. Cosmological Inflation and Large-Scale Structure. Cambridge (2000)
4. Kolb, E.W. and Turner, M.S. The Early Universe. Addison-Wesley (1990)
5. Weinberg, S. Gravitation and Cosmology. Wiley (1972)
6. Peacock, J.A. Cosmological Physics. Cambridge (2000)
7. Mukhanov, V. Physical Foundations of Cosmology. Cambridge (2005)

Books denoted with a ∗ are particularly recommended for this course.

Acknowledgements

These notes are distilled from a variety of excellent presentations of the subject in courses and
textbooks I have encountered over the years. It is, of course, a topic that can be treated with
a wide range of sophistication and difficulty. I have elected to aim it at an accessible level with
more emphasis on developing physical intuition rather than on mathematical principles. In terms
of textbooks, it most closely parallels the treatments found in Dodelson and Carroll, which are
my current personal favourites, and borrows with gratitude from course notes by (in no particular
order) Paul Steinhardt, Paul Shellard, Daniel Baumann, and Anthony Lasenby. I thank Anthony
Challinor for reading through a huge sheaf of rough notes and providing helpful comments. I am
supported by STFC, the European Commission, and the Leverhulme Trust.

Errata

Any errata contained in the following notes are solely my own. Reports of any typos or unclear
explanations in the notes will be gratefully received at the email address below. The notes are
evolving, and the most up-to-date version at any given time will be found on the website above.

∗ Electronic address: [email protected]

 2

Contents

I. INTRODUCTION 4
A. Brief history of the universe 4
B. The universe observed 5

II. THE METRIC 7

A. The cosmological principle 7
B. Example metrics 7
C. The metric and the Einstein summation convention 9
D. Special relativity metric 9
E. General relativity metric 10
F. Metric of a spatially-flat expanding universe 11

III. THE GEODESIC EQUATION 12

A. Transforming a Cartesian basis 12
B. Christoffel symbol 13
C. Particles in an expanding universe 14
D. Redshift 15

IV. THE EINSTEIN EQUATION 16

A. Energy-momentum tensor 16
1. Dust 17
2. Perfect fluid 17
3. Evolution of energy 17
B. Friedmann equations in a flat universe 19

V. GENERAL FRIEDMANN-ROBERTSON-WALKER METRIC 20

A. The cosmological principle revisited 20
1. Maximally symmetric spaces 20
2. Spatial curvature 21
3. General FRW metric 22
4. General Friedmann equations 23
B. Dynamics of the FRW universe 24
1. Terminology 24
2. Evolution of the scale factor 25
C. Matter-radiation equality 26
D. Cosmological constant 26
E. Spatial curvature and destiny 27

VI. TIME AND DISTANCE 30

A. Conformal time 31
B. Lookback distance and lookback time 31
C. Instantaneous physical distance 32
D. Luminosity distance 32
E. Angular diameter distance 34

VII. PARTICLES AND FIELDS IN COSMOLOGY 35

A. General particle motion 35
B. Classical field theory 35
C. Energy-momentum tensor from the action 37

VIII. BASICS OF INFLATION 41

A. Shortcomings of the Standard Big Bang Theory 41
1. Homogeneity Problem 41
2. Flatness Problem 41
3. Horizon Problem 42
4. On the Problem of Initial Conditions 43
5. What got the Big Bang going? 43
B. Possible resolution of the Big Bang puzzles 43
1. A Crucial Idea 43
2. Comoving Horizon during Inflation 44
3. Flatness Problem Revisited 45
4. Horizon Problem Revisited 46
5. Origin of primordial inhomogeneities 46
6. Conformal Diagram of Inflationary Cosmology 46

IX. INFLATION: IMPLEMENTATION WITH A SCALAR FIELD 48

A. Negative pressure 48
B. Implementation with a scalar field 48
C. Slow-Roll Inflation 50
 3

D. What is the Physics of Inflation? 51

E. Slow-Roll Inflation in the Hamilton-Jacobi Approach (NON–EXAMINABLE) 52
1. Hamilton-Jacobi Formalism 52
2. Hubble Slow-Roll Parameters 52
3. Slow-Roll Inflation 53
 4

I. INTRODUCTION

A. Brief history of the universe

The discovery of the expansion of the universe by Edwin Hubble in 1929 heralded the dawn of observational
cosmology. If we mentally rewind the expansion, we find that the universe was hotter and denser in its past. At very
early times the temperature was high enough to ionize the material that filled the universe. The universe therefore
consisted of a plasma of nuclei, electrons and photons, and the number density of free electrons was so high that the
mean free path for the Thomson scattering of photons was extremely short. As the universe expanded, it cooled,
and the mean photon energy diminished. The universe transitioned from being dominated by radiation to being
matter-dominated. Eventually, at a temperature of about 3000◦ K, the photon energies became too low to keep the
universe ionized. At this time, known as recombination, the primordial plasma coalesced into neutral atoms, and the
mean free path of the photons increased to roughly the size of the observable universe. Initial inhomogeneities present
in the primordial plasma grew under the action of gravitational instability during the matter-dominated era into all
the bound structures we observe in the universe today. Now, 13.7 billion years later, it appears that the universe has
entered an epoch of accelerated expansion, with its energy density dominated by the mysterious “dark energy”.

Composition of and Key Events During the Evolution of the Universe

present energy
Ωtot= 1 (k = 0) density
fraction of energy density

dark energy
radiation

dark
dark matter energy (73%)

dark
baryons matter (23.6%)
white = well understood, darkness proportional to poor understanding
baryons (4.4%)

scale factor
-42 -33 -22 -16 -12
Time t=10 s t=10 s t=10 s t=10 s t=10 s t=1 sec t = 380kyr
acoustic oscillations * t=14Gyr
electroweak symmetry

structure formation*
baryon asymmetry?

reionization *
recombination*
GUT symmetry

neutrino freezeout
nucleosynthesis
generation of primordial

(limit of accelerators?
physical theories)
Timeline of
Key Events

(breakdown of
planck energy

perturbations *

ILC x 100)

* generate observable
signatures in the CMB
T = 100TeV
in the CMB
Signatures

generation of gravity waves initial density perturbations CMB Emitted non-linear structure imparts
and initial density perturbations grow imparting fluctuations caries signature of acoustic signature on CMB through
which seed structure formation to CMB oscillations and potentially primordial gravitational lensing
gravity waves

FIG. 1 Composition of, and key events during, the history of the universe. Figure credit: Jeff McMahon.

The history of the universe is summarized in Fig. 1, emphasizing the “known unknowns” in our understanding of
its composition and evolution. Bear in mind that there might be “unknown unknowns” as well!
 5

The following table [reproduced from Liddle and Lyth] summarizes key events in the history of the universe and
the corresponding time– and energy–scales:

t ρ1/4 Event

10−42 s 1018 GeV Inflation begins?

10−32±6 s 1013±3 GeV Inflation ends, Cold Big Bang begins?
10−18±6 s 106±3 GeV Hot Big Bang begins?
10−10 s 100 GeV Electroweak phase transition?
10−4 s 100 MeV Quark-hadron phase transition?
10−2 s 10 MeV γ, ν, e, ē, n, and p in thermal equilibrium
1s 1 MeV ν decoupling, eē annihilation.
100 s 0.1 MeV Nucleosynthesis (BBN)
104 yr 1 eV Matter-radiation equality
105 yr 0.1 eV Atom formation, photon decoupling (CMB)
∼ 109 yr 10−3 eV First bound structures form
Now 10−4 eV (2.73 K) The present.

During most of its history, the universe is very well described by the hot Big Bang theory – i.e. the idea that the
universe was hot and dense in the past and has since cooled by expansion. The observational pillars underlying the
Big Bang Theory are:
• the Hubble diagram,
• Big Bang Nucleosynthesis (BBN),
• the Cosmic Microwave Background (CMB).
In this course, amongst (many) other things, we will explore the theoretical underpinnings of our understanding
of these observations. Cosmological observations have indicated several properties of the universe that cannot be
explained within the “standard model” (both the hot Big Bang theory and the Standard Model of particle physics):
• dark matter,
• dark energy,
• anisotropies of the CMB.
We will treat the first two phenomenologically in this course, and develop a basic understanding of the third.

B. The universe observed

The observed universe has the following properties:

1. homogeneous and isotropic when averaged over the largest scales (homogeneous means isotropic from every
vantage point).
2. topologically trivial (not periodic within 3000 Mpc).
3. expanding (Hubble flow).
4. hotter in the past, cooling (CMB).
5. predominantly matter (vs. antimatter).
6. chemical composition (75% H, 25% He, trace “metals”: hot Big Bang plus processing in stars).
7. highly inhomogeneous today and locally ( 100 Mpc).
 6

8. highly homogeneous when it was 103 × smaller.

9. (maybe) negligible curvature.
10. (maybe) dark energy/acceleration.

In these lectures, we will use natural units, ~ = c = kB = 1 , unless explicitly knowing the dependence on these
quantities is necessary to develop understanding.

Most of cosmology can be learnt with only a passing knowledge of general relativity (GR). We will need the concepts
of the metric and the geodesic, and apply Einstein’s equations to the Friedmann-Robertson-Walker metric, relating
the metric parameters to the (energy) density of the universe. In this section of the course, we will apply Einstein’s
equations to the homogeneous universe. In the second section of the course, we will apply them to the perturbed
universe. With the experience we gain here, there will be nothing difficult later. The principles are identical; only the
algebra will be a bit harder.
 7

II. THE METRIC

A. The cosmological principle

On the largest scales, the universe is assumed to be uniform. This idea is called the cosmological principle. There
are two aspects of the cosmological principle:
• the universe is homogeneous. There is no preferred observing position in the universe.
• the universe is isotropic. The universe looks the same in every direction.
Fig. 2 illustrates these concepts. We will make them more precise in due course.

FIG. 2 Departures from homogeneity and isotropy illustrated. Figure credit: This image was produced by Nick Strobel, and
obtained from Nick Strobel’s Astronomy Notes at http://www.astronomynotes.com.

There is an overwhelming amount of observational evidence that the universe is expanding. This means that early
in the history of the universe, the distant galaxies were closer to us than they are today. It is convenient to describe
the scaling of the coordinate grid in an expanding universe by the scale factor. In a smooth, expanding universe, the
scale factor connects the coordinate distance with the physical distance. More generally,

coordinate distance =⇒ metric =⇒ physical distance . (1.2.1)

The metric is an essential tool to make quantitative predictions in an expanding universe.

B. Example metrics

Cartesian coordinates:

ds2 = dx2 + dy 2 . (1.2.2)

Polar coordinates:

ds2 = dr2 + r2 dθ2 6= dr2 + dθ2 . (1.2.3)

 8

ds
dy

dx
FIG. 3 2D Cartesian Coordinates.

A metric turns observer-dependent coordinates into invariants. In 2D (Fig. 3),

X
ds2 = gij dxi dxj , (1.2.4)
i,j=1,2

where the metric gij is a 2 × 2 symmetric matrix. In Cartesian coordinates,

!
1 2 1 0
x = x, x = y, gij = , (1.2.5)
0 1
while in polar coordinates,
!
1 2 1 0
x = r, x = θ, gij = . (1.2.6)
0 r2

Another way to think about a metric is to take a pair of vectors on a topographical map (Fig. 4), of the same
length (same coordinate distance). But the actual physical distance depends on the topography ↔ metric.

FIG. 4 Contour map of a mountain, with closely spaced contours near the centre corresponding to rapid elevation gain.

The great advantage of a metric is that it incorporates gravity. In classical Newtonian mechanics, gravity is an
external force, and particles move in a gravitational field. In GR, gravity is encoded in the metric, and the particles
move in a distorted/curved spacetime. In 4 (3+1) dimensions, the invariant includes time intervals as well:
3
X
2
ds = gµν dxµ dxν , (1.2.7)
µ,ν=0

where µ, ν −→ {0, 1, 2, 3} with dx0 = dt reserved for the timelike coordinate, and dxi for the spacelike coordinates.
From now, we will use the summation convention where repeated indices are summed over.
 9

C. The metric and the Einstein summation convention

~ has three components: Ai ,

In 3D, a vector A i = {1, 2, 3}.
3
~·B
~ =
X
⇒A Ai B i ≡ Ai B i , (1.2.8)
i=1

where we sum over repeated indices. For example, the matrix product: M · N = (M · N)ij = Mik Nkj , where
summation over k is implied.
In relativity, there are two generalizations. First, there is a zeroth component, time. Spatial indices run from 1 → 3,
and it is conventional to use 0 for the time component: Aµ = (A0 , Ai ). Secondly, there is a distinction between the
upper (vectors) and lower (1–forms) indices. One goes back and forth between the two using the metric:
Aµ = gµν Aν ; Aµ = g µν Aν . (1.2.9)
A vector and a 1–form can be contracted to produce an invariant; a scalar. For example, the statement “The squared
4–momentum of a massless particle must vanish.” is equivalent to:
p2 ≡ pµ pµ = gµν pµ pν = 0. (1.2.10)
Contraction can be thought of as counting the contours of the topographic map crossed by a vector, in our previous
analogy. The metric can be used to raise and lower indices on tensors with an arbitrary number of indices. For
example,
g µν = g µα g νβ gαβ . (1.2.11)
Taking α = ν, we see that

g νβ gαβ = δ να , (1.2.12)
where δ να is the Kronecker delta:
(
1 (ν = α)
δ να = . (1.2.13)
0 (ν 6= α)
This is the definition of g µν as the inverse of gµν . The metric gµν is
• necessarily symmetric,
• in principle, has 4 diagonal and 6 off-diagonal components,
• provides the connection between the values of the coordinates and the more physical measure of the interval
ds2 .
In this course we will use the following convention for the signature of the metric: (+.−, −, −). Beware, while
this convention is commonly used by particle physicists, the convention used in relativity and cosmology is often
(−, +, +, +).

D. Special relativity metric

The Minkowski metric ηµν is the metric of special relativity, and it describes flat space. Its line element is given by,
ds2 = ηµν dxµ dxν , (1.2.14)
2 2 2 2


= dt − dx + dy + dz , (1.2.15)
and the metric is
ηµν = diag(1, −1, −1, −1) . (1.2.16)
SR applies in inertial frames, or locally, in those falling freely in a gravitational field. Because it is locally equivalent
only, we can’t say in general that there is a frame where
∆s2 = ∆t2 − ∆x2 − ∆y 2 − ∆z 2 = 0 . (1.2.17)
We want to be able to transform to non-free-fall coordinate systems and relate variables over distances. This will
require a more general set of transformations than the Lorentz transformations (translations, rotations, and boosts).
There is thus no global SR frame, and gµν has to reflect the curvature of spacetime.
 10

E. General relativity metric

Let’s consider two forms of a key principle.

Strong Equivalence Principle (SEP): At any point in a gravitational field, in a frame moving with the free fall
acceleration at that point, all the laws of physics have their usual Special Relativity (SR) form, except gravity, which
disappears locally.

Weak Equivalence Principle (WEP): At any point in a gravitational field, in a frame moving with the free fall
acceleration at that point, the laws of motion of free test particles have their usual Special Relativity (SR) form, i.e.
particles move in straight lines with uniform velocity locally.

Instead of thinking of particles moving under a force, the WEP allows us to think of them in a frame without
gravity, but moving with the free fall acceleration at that point. This is a very powerful idea:

EQUIVALENCE : gravity ⇐⇒ acceleration , (1.2.18)

which has two important consequences. First, it explains the equivalence of gravitational mass and inertial mass.
Second, it says that the motion of a test body in a gravitational field only depends on its position and instantaneous
velocity in spacetime. In other words, gravity determines a geometry . The equivalence principle works locally where
the gravitational field can be taken to be uniform.

lightlike
timelike

spacelike

FIG. 5 A lightcone on a spacetime diagram. Points that are spacelike, lightlike, and timelike separated from the origin are
indicated.

There is a component of “true” gravity, not transformable into acceleration, that has the distance dependence of
“tidal forces” (∝ r−3 for spherical polar coordinates). True gravity manifests itself via the separation or coming
together of test particles initially on parallel tracks. There is no global Lorentz frame in the presence of a non-uniform
gravitational field.
A freely-falling particle follows a geodesic in spacetime. The metric links the concepts of “geodesic” and “spacetime”:

ds2 = gµν dxµ dxν (1.2.19)

 11

where ds2 is the proper interval, gµν is the metric tensor, and xµ is a four-vector. There are three possible kinds of
intervals (see Fig. 5):

ds2 < 0 : spacelike (1.2.20)

ds2 = 0 : null / lightlike (1.2.21)
ds2 > 0 : timelike . (1.2.22)

F. Metric of a spatially-flat expanding universe

What is the metric of an expanding universe?

a(t)y0 t t0 y0

a(t)x0

x0
time

FIG. 6 If the comoving distance today at time t0 is x0 , the physical distance between the two points at some earlier time t < t0
was a(t)x0 .

If the comoving distance today is x0 , the physical distance between two points at some earlier time t was a(t)x0
(see Fig. 6). At least in a flat (as opposed to open or closed) universe, the metric must be ∼ Minkowski, except
that the distance must be multiplied by the scale factor a(t). Thus, the metric of a flat, expanding universe is the
Friedmann-Robertson-Walker metric:

gµν = diag 1, −a2 (t), −a2 (t), −a2 (t) .

(1.2.23)

The evolution of the scale factor depends on the density of the universe. When perturbations are introduced,
the metric will become more complicated, and the perturbed part of the metric will become determined by the
inhomogeneities in the matter and the radiation.
 12

III. THE GEODESIC EQUATION

A. Transforming a Cartesian basis

In Minkowski space, particles travel in straight lines unless they are acted upon by an external force. In more
general spacetimes, the concept of a “straight line” gets replaced by the “geodesic”, which is the path followed by a
2
particle in the absence of any external forces. Let us generalize Newton’s law with no forces, ddt2~x = 0, to the expanding
universe. We will start with particle motion in a Euclidean 2D plane. Equations of motion in Cartesian coordinates
xi = (x, y) for a free particle are

d2 xi
=0. (1.3.1)
dt2
i
What are the equations of motion in polar coordinates, x0 = (r, θ)? The basis vectors for polar coordinates, r̂, θ̂ vary
in the plane! Therefore, in polar coordinates,
i
d2 ~x d2 x0
=0 ; =0 (1.3.2)
dt2 dt2
i
for x0 = (r, θ). Let us start from the Cartesian equation and transform:
 i  0j
dxi dx dx
= 0j
, (1.3.3)
dt dx dt

where the term in the brackets on the RHS is a transformation matrix going from one basis to another, i.e. it is the
determinant of the Jacobean. To transform from Cartesian to polar coordinates,

x1 = x01 cos x02 , x2 = x01 sin x02 , (1.3.4)

with transformation matrix,

!
dxi cos x02 −x01 sin x02
= . (1.3.5)
dx0j sin x02 x01 cos x02

Therefore, the geodesic equation is,

d dxi d dxi dx0j

   
= =0. (1.3.6)
dt dt dt dx0j dt

If the transformation was linear, the derivative acting on the transformation matrix would vanish, and the geodesic
2 0i
equation in the new basis would still be ddtx2 = 0. In polar coordinates, the transformation is not linear, and using
the chain rule, we have

d dxi dx0k ∂ 2 xi
 
= . (1.3.7)
dt dx0j dt ∂x0k ∂x0j

The geodesic equation therefore becomes,

d dxi dx0j ∂xi d2 x0j ∂ 2 xi dx0k dx0j

 
= + . (1.3.8)
dt dx0j dt ∂x0j dt2 ∂x0j ∂x0k dt dt

Multiplying by the inverse of the transformation matrix, we obtain the geodesic equation in a non-Cartesian basis:
 
−1 !`
d2 x0`  ∂ 2 xi  dx0k dx0j

∂x
+ =0. (1.3.9)
dt2 ∂x0 ∂x0j ∂x0k dt dt
i
 13

EXERCISE: Check that it works for polar coordinates!

The term in the brackets is the Christoffel symbol,

 
 −1 !` 2 i
∂x ∂ x
Γ`jk =   , (1.3.10)
∂x0 ∂x0j ∂x0k
i

2 i
which is symmetric in j, k. In Cartesian coordinates, Γ`jk = 0, and the geodesic equation is simply ddtx2 = 0. In
general, Γ`jk 6= 0 describes geodesics in non-trivial coordinate systems. The geodesic equation is a very powerful
concept, because in a non-trivial spacetime (e.g. the expanding universe) it is not possible to find a fixed Cartesian
coordinate system. So we need to know how particles travel in the more general case.
To import this concept into relativity, we need
• to allow indices to range from 0 → 3 to include time and space.
• time is now one of our coordinates! We can’t use it to describe an evolution parameter.

xμ(λ2)

xμ(λ1)
λ

FIG. 7 A particle’s path parametrized by λ, which monotonically increases from its initial value λ1 to its final value λ2 .

Take a parameter λ (Fig. 7) which monotonically increases along the particle’s path. The geodesic equation
becomes:

d2 xµ α
µ dx dx
β
+ Γ αβ =0 . (1.3.11)
dλ2 dλ dλ

B. Christoffel symbol

Rather than the previous definition of the Christoffel symbol obtained by transforming a Cartesian basis, it is
almost always more convenient to obtain it from the metric:

g µν
 
∂gαν ∂gβν ∂gαβ
Γµαβ = + − . (1.3.12)
2 ∂xβ ∂xα ∂xν

EXERCISE: Verify that Eq. (1.3.10) is consistent with the definition (1.3.12) in the case of a flat spacetime.

Be careful: raised indices are important. g µν is the inverse of gµν . So g µν in the flat FRW metric is identical to
gµν , except that the spatial elements are − a12 instead of −a2 .
 14

The components of the Christoffel symbol in the flat FRW universe (with overdots denoting d/dt) are:

ȧ i
Γ000 = 0, Γ00i = Γ0i0 = 0, Γ0ij = δij ȧa Γi0j = Γij0 = δ , Γijk = 0, Γi00 = 0 . (1.3.13)
a j

EXERCISE: Use the flat FRW metric to derive the components of the Christoffel symbol.

C. Particles in an expanding universe

Let us apply the geodesic equation to a single particle. How does a particle’s energy change as the universe expands?
Most measurements we make in cosmology has to do with intercepting photons which have arrived on the earth after
being emitted at various epochs during the evolution of the universe. Therefore we will consider a massless particle,
which has energy-momentum 4–vector, pα = (E, p~), and use this to implicitly define the parameter λ:

dxα
pα = . (1.3.14)
dλ
d dx0 d d
Eliminate λ by noting that dλ = dλ dx0 = E dt , and the 0–component of the geodesic equation,

dE
E = −Γ0ij pi pj ,
dt
= −δij ȧa pi pj . (1.3.15)

For a massless particle, the energy-momentum vector has zero magnitude:

gµν pµ pν = E 2 − δij a2 pi pj = 0 . (1.3.16)

Since p~ measures motion on the comoving grid, the physical momentum which measures changes in the physical
distance is related to p~ by a factor of a, hence the factor of a2 here. This leads to
dE ȧ
+ E=0. (1.3.17)
dt a
We see that the energy of a massless particle decreases as the universe expands:

1
E∝ . (1.3.18)
a

This accords with the intuition from a handwaving argument: E ∝ λ−1 (where λ is the wavelength) and λ ∝ a is
stretched along with the expansion. The frequency of a photon emitted with frequency νem will therefore be observed
with a lower frequency νobs as the universe expands:
νobs aem
= . (1.3.19)
νem aobs
 15

D. Redshift

Cosmologists like to speak of this in terms of the redshift z between two events, defined by the fractional change in
wavelength:
λobs − λem
zem = . (1.3.20)
λem

So if the observation takes place today (aobs = a0 = 1), this implies

1
aem = . (1.3.21)
1 + zem

So the redshift of an object tells us the scale factor when the photon was emitted.
Notice that this redshift is not the same as the conventional Doppler effect. It is the expansion of space, not the
relative velocities of the observer and emitter, that leads to the redshift.
Measuring the redshifts of distant objects is one of the most basic tools in the observer’s toolkit. It is a rather
amazing notion that every time one measures a redshift, one is directly detecting the curvature of spacetime.
 16

IV. THE EINSTEIN EQUATION

So far, we have not used General Relativity. The concept of the metric and the realization that non-trivial metrics
affect geodesics exist independently of GR. The part of GR that is hidden here is that gravitation can be described by
a metric. There is another aspect of GR which we will need now, which connects the metric to the matter/energy.
This is described by the Einstein equation, which relates geometry to energy:
1
Gµν ≡ Rµν − gµν R = 8πGTµν , (1.4.1)
2
where Gµν is the Einstein tensor, Rµν is the Ricci tensor, the Ricci scalar R = g µν Rµν is the contraction of the Ricci
tensor, and the energy-momentum (or stress-energy) tensor Tµν is a symmetric tensor describing the constituents of
the universe. The Ricci tensor is defined as,

Rµν = Γα α α β α β
µν,α − Γµα,ν + Γβα Γµν − Γβν Γµα , (1.4.2)

where commas denote derivatives with respect to x,

∂Γα
µν
Γα
µν,α ≡ . (1.4.3)
∂xα
It looks like hard work but we have already done the hard bit by computing Γα
µν in a flat FRW universe, and it has
only two sets of non-vanishing components: µ, ν = 0 and µ, ν = i. Using them (and noting that δ ii = 3) we can show
that,
 
ä
, Rij = δij 2ȧ2 + äa .
 
R00 = −3 (1.4.4)
a
Contracting the Ricci tensor, we then obtain the Ricci scalar for the flat, homogeneous FRW universe as:
R ≡ g µν Rµν
1
= R00 − 2 δ ij Rij
" a
 2 #
ä ȧ
= −6 + . (1.4.5)
a a

EXERCISE: Verify the components of the Ricci tensor and Ricci scalar given above for the flat, homogeneous FRW
universe.

A. Energy-momentum tensor

Consider the 4-momentum (should be familiar from SR) of a particle: pµ = mU α , where m is the “rest mass” of the
particle, independent of inertial frame. The energy of the particle is E = p0 (the timelike component of the momentum
4-vector). Note that E is not invariant under Lorentz transformations.pIn the particle’s rest frame, p0 = m, which
is just the famous result E = mc2 . We also have pµ pµ = m2 , or E = m2 + |~ p|2 , where p~2 = δij pi pj . pµ provides
a complete description of the energy-momentum of a particle. In cosmology, we need to describe extended systems
comprised of large numbers of particles: fluids. A fluid is a continuum characterised by macroscopic properties such
as density, pressure, entropy, and viscosity.
A single momentum 4-vector field is insufficient to describe the energy-momentum of a fluid. We need to replace
it by the energy-momentum tensor (also called the stress-energy tensor) T µν , which is a symmetric tensor describing
the flux of 4-momentum pµ across a surface of constant xν .
This definition is not that useful. Later, in the section on inflation, we will define T µν in terms of a functional
derivative of the action with respect to the metric. This leads to a more algorithmic procedure for finding an explicit
expression for T µν . For now, we will use the above definition to gain some physical intuition.
Consider an infinitesimal element of the fluid in its rest frame, where there are no bulk motions. Then,
 17

• T 00 : the flux of p0 (energy) in the x0 (time) direction (i.e. the rest frame energy density ρ),
• T 0i = T i0 : momentum density,
• T ij = T ji : momentum flux (or stress) (i.e. forces between neighbouring infinitesimal elements of the fluid).
Off-diagonal elements in T ij are shearing terms, e.g. due to viscosity. Diagonal terms T ii give the i–th component
of the force/unit area exerted by a fluid element in the i–direction. The pressure has three components given in the
fluid rest frame. Let us now consider some example fluids.

1. Dust

Start with “dust” (cosmologists tend to use “matter” as a synonym for dust): in a flat spacetime, a collection of
particles at rest with eachother. The number-flux 4-vector is N µ = nU µ , where n is the number density measured
in the rest frame. N 0 is the number density of the particles in any other frame. N i is the particle flux in the xi
direction. Imagine that each particle has mass m. The rest frame energy density is ρ = mn. ρ completely specifies
dust, since pressure is by definition zero, as dust has no random motions within the fluid. Since N µ = (n, 0, 0, 0),
pµ = (m, 0, 0, 0), we have
µν
Tdust = pµ N ν = mn U µ U ν = ρ U µ U ν . (1.4.6)
This is not general enough to describe all interesting cosmological fluids, so we need to make a slight generalization.

2. Perfect fluid

A perfect fluid can be completely defined by a rest frame energy density ρ and an isotropic rest frame pressure P
(which serves to specify pressure in every direction). Thus, T µν is diagonal in its rest frame, with no net flux of any
component of momentum in an orthogonal direction. The non-zero spacelike components are all equal. For dust, we
had T µν = ρ U µ U ν . For a perfect fluid, the general form in any frame is:

T µν = (ρ + P )U µ U ν − P g µν . (1.4.7)

EXERCISE: Check this by writing out the terms in e.g. flat space, ηµν .

We might have seemed to arrived at this arbitrarily, but given that (1.4.7) reduces to T µν = diag(ρ, −g ii P ) in the
rest frame, and is a tensor equation (and therefore coordinate–independent), it must be valid in any frame. A perfect
fluid is general enough to describe a wide variety of cosmological fluids, given their equation of state,

P
w= . (1.4.8)
ρ
ρ 1
Dust has P = 0, w = 0. Radiation has P = 3, w = 3. Vacuum energy is proportional to the metric: T µν =
−ρvac g µν , Pvac = −ρvac , w = −1.

3. Evolution of energy

Consider a perfect isotropic fluid. Then, the energy-momentum tensor can be written with one index raised in the
following metric-independent form:
 
ρ 0 0 0
 0 −P 0 0 
 
T µν =   , (1.4.9)
 0 0 −P 0 
0 0 0 −P
 18

where ρ is the energy density, and P is the pressure of the fluid. How do the components of T µν evolve with time?
Consider the case where there is no gravity and velocities are negligible. Then, the pressure and energy evolve as:
∂ρ
Continuity Equation : =0, (1.4.10)
∂t
∂P
Euler Equation : =0. (1.4.11)
∂xi
We need to promote this to a 4–component conservation equation for the energy-momentum tensor:
∂T µν
= 0. (1.4.12)
∂xµ
However, in an expanding universe, the conservation criterion must be modified. In this context, conservation implies
the vanishing of the covariant derivative:

∂T µν
T µν;µ ≡ + Γµαµ T αν − Γα µ
νµ T α . (1.4.13)
∂xµ

The importance of the covariant derivative is that, to paraphrase Misner, Thorne, & Wheeler, Gravitation (W.H.
Freeman, 1973) p. 387,
A consistent replacement of regular partial derivatives by covariant derivatives carries the laws of physics (in
component form) from flat (Lorentzian) spacetime into the curved (non-Lorentzian) spacetime of general
relativity. Indeed, this substitution may be taken as a mathematical statement of Einstein’s principle of
equivalence.
We will call this the “comma goes to semi-colon” rule. Thus, the conservation criterion becomes

T µν;µ = 0 . (1.4.14)

There are four separate equations to be considered here. Consider first the ν = 0 component. By isotropy, T i0
vanishes, yielding the continuity equation:

∂ρ ȧ
+ 3 (ρ + P ) = 0 . (1.4.15)
∂t a

EXERCISE: Verify the above assertion.

Rearranging this equation,

 
∂ ȧ
a−3 ρa3 = −3


P . (1.4.16)
∂t a
Immediately this yields information about the scaling of both matter and radiation. Pressureless matter has P = 0
by definition, so
∂
ρm a3 = 0 ⇒ ρm ∝ a−3 .


MATTER : (1.4.17)
∂t
This is expected if you consider that mass remains constant while number density scales as inverse volume.
Radiation has P = ρ3 , giving
 
∂ρr ȧ ∂
ρr = a−4 ρr a4 = 0 .


+4 (1.4.18)
∂t a ∂t
Thus, radiation scales as

RADIATION : ρr ∝ a−4 . (1.4.19)

 19

ρ
EXERCISE: Show that P = 3 for radiation.

B. Friedmann equations in a flat universe

To understand the evolution of the scale factor in a homogeneous expanding universe, we only need to consider the
time-time component of the Einstein equation:
1
R00 − g00 R = 8πGT00 , (1.4.20)
2
leading to the first Friedmann equation for a flat universe:
 2
ȧ 8πG
= ρ (FLAT) . (1.4.21)
a 3

EXERCISE: Verify the above assertion.

There is a second Friedmann equation. Consider the space-space component of Einstein’s equation:
1
Rij − gij R = 8πGTij . (1.4.22)
2
Using the flat FRW terms we worked out previously in Eqs. (1.3.13, 1.4.4, 1.4.5), with gij = −δij a2 we find:
"  2 #
 2  a2 ä ȧ
LHS : δij 2ȧ + äa − δij 6 + . (1.4.23)
2 a a

Noting the mixed form for the perfect fluid energy-momentum tensor, (1.4.9), we see that,

RHS : 8πGTij = 8πGgik T kj = 8πGa2 δij P . (1.4.24)

Equating these terms we obtain,

 2
ä 1 ȧ
+ = −4πGP (FLAT) . (1.4.25)
a 2 a

Combining with the first Friedmann equation (1.4.21), this leads us to the second Friedmann equation:

ä 4πG
=− (ρ + 3P ) (FLAT) . (1.4.26)
a 3

EXERCISE: Derive Eq. (1.4.26) by another route, by differentiating the first Friedmann equation (1.4.21) with respect
to time t, and combining with the continuity equation (1.4.15).
 20

V. GENERAL FRIEDMANN-ROBERTSON-WALKER METRIC

A. The cosmological principle revisited

1. Maximally symmetric spaces

The mathematics in this subsection are NON-EXAMINABLE, but you are expected to fully grasp the physical
concepts therein.

We leave the details to the GR course in the Maths Department and/or any GR text, and note that the Riemann
tensor:
Rρσµν = ∂µ Γρνσ − ∂ν Γρµσ + Γρµλ Γλνσ − Γρνλ Γλµσ , (1.5.1)
Rρσµν = gρλ Rλσµν , (1.5.2)
quantifies curvature and is non-zero when the metric departs from flatness. The Ricci tensor is formed by contracting
the Riemann tensor:
Rµν = g ασ Rαµσν = Rαµαν . (1.5.3)
It has some elegant symmetry properties, satisfying the following index symmetries:
Rαµνσ = −Rµανσ = −Rαµσν , Rαµνσ = Rνσαµ , Rαµνσ + Rασµν + Rανσµ = 0 , (1.5.4)
and the Bianchi identities,
Rλαµν;σ + Rλασµ;ν + Rλανσ;µ = 0 . (1.5.5)
In particular, in a maximally symmetric space (details left to GR course) of n dimensions,
R
Rρσµν = (gρµ gσν − gρν gσµ ) , (1.5.6)
n(n − 1)
where the Ricci scalar R is constant over the manifold. Conversely, if the Riemann tensor satisfies (1.5.6), the space
is maximally symmetric. Setting
R
K= , (1.5.7)
n(n − 1)
since at any given point the metric can be put into its canonical form gµν = ηµν , the kinds of maximally symmetric
manifolds are characterized locally by the metric signature and the sign of K. For the metric signature (+, −, −, −),
K=0: Minkowski space (1.5.8)
K<0: de Sitter space (1.5.9)
K>0: anti-de Sitter space . (1.5.10)
We said “locally” to account for possible global differences, such as between the plane and the torus.
Do any of these describe the real universe? Let’s consider its properties. Contemporary cosmological models are
based on the idea that, at “sufficiently large scales”, the Copernican principle applies: the universe is pretty much
the same everywhere. This is encoded more rigorously in the ideas of,
• isotropy: at some specified point in the manifold, space looks the same in whatever direction you look.
• homogeneity: the metric is the same throughout the manifold.
A manifold can be homogeneous but nowhere isotropic, or isotropic around a point but nowhere homogeneous. If
a space is isotropic everywhere, then it is also homogeneous. If a space is isotropic around one point and also
homogeneous, it will be isotropic everywhere.
The CMB shows that the universe is isotropic on the order of 10−5 , and since by the Copernican principle, we don’t
believe that we are the centre of the universe, we assume both homogeneity and isotropy.
However, observations tell us that the universe is expanding, so the Copernican principle only applies in space, not
in time. So the maximally symmetric spacetimes itemized above don’t describe our universe (or any universe with a
dynamically interesting amount of matter and/or radiation).
 21

2. Spatial curvature

So let’s give up the “perfect” Copernican principle and posit that the universe is spatially homogeneous and isotropic:

ds2 = dt2 − R2 (t)dσ 2 , (1.5.11)

where t is a timelike coordinate, R(t) is the scale factor, and dσ 2 is the metric on a maximally symmetric 3–manifold
Σ:

dσ 2 = γij (u)dui duj , (1.5.12)

where (u1 , u2 , u3 ) are coordinates on Σ, and γij is a maximally symmetric 3D metric. R(t) tells us how big the
spacelike slice is at time t. These are comoving coordinates.
We want to know the possible maximally symmetric 3-metrics γij . They obey
(3)
Rijkl = k (γik γjl − γil γjk ) , (1.5.13)

where
(3)
R
k= , (1.5.14)
6
and (3) reminds us that we are dealing with 3-metric γij , not the entire spacetime metric. The Ricci tensor is then
(EXERCISE: CHECK!)
(3)
Rjl = 2kγjl . (1.5.15)

Maximally symmetric ⇒ spherically symmetric, so the metric can be put in the form (cf. Schwarzschild metric):

dσ 2 = γij dui duj = e2β(r̃) dr̃2 + r̃2 dΩ2 , (1.5.16)

where r̃ is the radial coordinate, and dΩ2 = dθ2 +sin2 θdφ2 is the metric on the 2-sphere. Working out the components
leads to:

(3) 2
R11 = ∂1 β , (1.5.17)
r̃
(3)
R22 = e−2β (r̃∂1 β − 1) + 1 , (1.5.18)
(3)
= e−2β (r̃∂1 β − 1) + 1 sin θ .
 
R33 (1.5.19)

Setting ∝ the metric via (1.5.15) , we get

1
β = − ln 1 − kr̃2 .


(1.5.20)
2
Thus the metric on Σ is:

dr̃2
dσ 2 = + r̃2 dΩ2 . (1.5.21)
1 − kr̃2

The value of k sets the curvature, and therefore the size, of the spatial surfaces. It is common to normalize such that
k ∈ {−1, 0, +1}, and absorb the physical size of the manifold into R(t). The geometry is then classified as:

k = −1 : constant negative curvature on Σ (OPEN) (1.5.22)

k=0: no curvature on Σ (FLAT) (1.5.23)
k = +1 : constant positive curvature on Σ (CLOSED) . (1.5.24)

The physical meaning of these cases becomes more apparent by redefining the radial coordinate:
dr̃
dχ = √ . (1.5.25)
1 − kr̃2
 22

Integrating,

r̃ = Sk (χ) , (1.5.26)

where

 sin χ
 k = +1
Sk (χ) ≡ χ k=0 , (1.5.27)

 sinh χ k = −1

such that

dσ 2 = dχ2 + Sk2 (χ)dΩ2 . (1.5.28)

EXERCISE: Verify Eq. (1.5.26)–(1.5.28).

To summarize, we have:

k=0: dσ 2 = dχ2 + χ2 dΩ2 = dx2 + dy 2 + dz 2 (flat Euclidean space) (1.5.29)

k = +1 : dσ 2 = dχ2 + sin2 χ dΩ2 (metric of a 3-sphere) (1.5.30)
k = −1 : dσ 2 = dχ2 + sinh2 χ dΩ2 (a hyperboloid space) . (1.5.31)

A note on the hyperboloid case: globally, such a space could be infinite – the origin of “open” – but could also describe
a non-simply-connected compact space, so it is not really a good description.

3. General FRW metric

The metric on spacetime describes one of these maximally symmetric hypersurfaces evolving in size:
dr̃2
 
ds2 = dt2 − R2 (t) + r̃ 2
dΩ 2
. (1.5.32)
1 − kr̃2
Normalizing the coordinates to the present epoch, subscript “0”,
R(t)
a(t) = , r = R0 r̃ , (1.5.33)
R0
we can define a curvature parameter of dimensions [length]−2 :
k
κ= . (1.5.34)
R0
Note that κ can take any value, not just {+1, 0, −1}. We obtain the general FRW metric:

dr2
 
ds2 = dt2 − a2 (t) + r2 dΩ2 . (1.5.35)
1 − κr2

da
Setting ȧ = dt , the Christoffel symbols are:

Γ011 = aȧ/(1 − κr2 ) Γ111 = κr/(1 − κr2 )

Γ022 = aȧr2 Γ033 = aȧr2 sin2 θ
Γ101 = Γ202 = Γ303 = ȧ/a
(1.5.36)
Γ122 = −r(1 − κr2 ) Γ133 = −r(1 − κr2 ) sin2 θ
Γ212 = Γ313 = 1/r
Γ233 = − sin θ cos θ Γ323 = cot θ ,
 23

or related to these by symmetry. Non-zero components of the Ricci tensor are:

ä
R00 = −3 (1.5.37)
a
aä + 2ȧ2 + 2κ
R11 = (1.5.38)
1 − κr2
R22 = r (aä + 2ȧ2 + 2κ)
2
(1.5.39)
R33 = r2 (aä + 2ȧ2 + 2κ) sin2 θ , (1.5.40)

and the Ricci scalar is

"  2 #
ä ȧ κ
R = −6 + + 2 . (1.5.41)
a a a

EXERCISE: Verify the components of the Christoffel symbols, the Ricci tensor, and the Ricci scalar given above.

4. General Friedmann equations

The first Friedmann equation (1.4.21) becomes:

 2
ȧ 8πG κ
= ρ− 2 , (1.5.42)
a 3 a

and the second Friedmann equation does not change due to κ (EXERCISE: CHECK), and we repeat it here for com-
pleteness:

ä 4πG
=− (ρ + 3P ) . (1.5.43)
a 3

Notice that, in an expanding universe (i.e. ȧ > 0 at all times) filled with ordinary matter (i.e. matter satisfying
the strong energy condition: ρ + 3P ≥ 0), Eq. (1.5.43) implies ä < 0 at all times. This indicates the existence of a
singularity in the finite past: a(t = 0) = 0. Of course, this conclusion relies on the assumption that general relativity
and the Friedmann equations are applicable up to arbitrarily high energies. This assumption is almost certainly not
true and it is expected that a quantum theory of gravity will resolve the initial big bang singularity.
 24

B. Dynamics of the FRW universe

1. Terminology

The expansion rate of the FRW universe is characterized by the Hubble parameter,

ȧ
H(t) = . (1.5.44)
a
The expansion rate at the present epoch, H(t0 ), is called the Hubble constant, H0 1 . Often you will see the dimensionless
number h, where
H0 = 100h km s−1 Mpc−1 . (1.5.45)
The astronomical length scale of a megaparsec (Mpc) is equal to 3.0856 × 1024 cm, and h should not be confused with
Planck’s constant. Observationally, h ∼ 0.7. Typical cosmological scales are set by the Hubble length,
dH = H0−1 c = 9.25 × 1027 h−1 cm = 3.00 × 103 h−1 Mpc . (1.5.46)
The Hubble time is,
tH = H0−1 = 3.09 × 1017 h−1 sec = 9.78 × 109 h−1 yr. (1.5.47)
Since we usually set c = 1, H0−1 is referred to as both the Hubble length and the Hubble time. The deceleration
parameter,
aä
q=− , (1.5.48)
ȧ2
measures the rate of change of the expansion.
The density parameter, which counts the energy density from all forms of constituents of the universe, is defined as

8πG ρ
Ω= 2
ρ= , (1.5.49)
3H ρcrit

where the critical density

3H 2
ρcrit = , (1.5.50)
8πG
changes with time, and is so called because the Friedmann equation (1.5.42) can be written:
κ
Ω(a) − 1 = 2 2 . (1.5.51)
H a
Thus, sign(κ) is defined by sign(Ω − 1):

ρ < ρcrit ↔ Ω < 1 ↔ κ < 0 ↔ open

ρ = ρcrit ↔ Ω = 1 ↔ κ = 0 ↔ flat
ρ > ρcrit ↔ Ω > 1 ↔ κ > 0 ↔ closed

The density parameter thus tells us which of the three FRW geometries describes our universe. Our universe is obser-
vationally indistinguishable from the flat case. We can further streamline our expressions by treating the contribution
of the spatial curvature as a fictitious energy density,
3κ
ρκ = − , (1.5.52)
8πGa2
with a corresponding density parameter,
κ
Ωκ = − . (1.5.53)
H 2 a2

1 The “0” subscript is used to denote the present epoch: t = t0 , a(t0 ) = a0 = 1.

 25

2. Evolution of the scale factor

An immediate consequence of the two Friedmann equations is the continuity equation, which we previously derived
in Eq. (1.4.15) from considering the conservation of the energy-momentum tensor:
dρ
+ 3H (ρ + P ) = 0 . (1.5.54)
dt
More heuristically this also follows from the first law of thermodynamics

dU = −P dV
d ln ρ
d(ρa3 ) = −P d(a3 ) ⇒ = −3(1 + w) , (1.5.55)
d ln a
where, w is the equation of state, reminding ourselves of Eq. (1.4.8). The continuity equation (1.4.15) can be
integrated to give

ρ ∝ a−3(1+w) . (1.5.56)

Together with the Friedmann equation (1.5.42) this leads to the time evolution of the scale factor,

a ∝ t2/3(1+w) ∀ w 6= −1 . (1.5.57)

In particular, constituents of our universe follow the following scalings:

component wi ρ(a) a(t)

non-relativistic matter 0 ∝ a−3 ∝ t2/3
radiation/relativistic matter 13 ∝ a−4 ∝ t1/2
curvature − 13 ∝ a−2 ∝t
cosmological constant −1 ∝ a0 ∝ exp(Ht)

For each species i we define the present ratio of the energy density relative to the critical density,
ρi0
Ωi,0 ≡ , (1.5.58)
ρcrit,0
and the corresponding equations of state
Pi
wi ≡ . (1.5.59)
ρi
This allows one to rewrite the first Friedmann equation (1.5.42) as
 2 X
H
= Ωi,0 a−3(1+wi ) + Ωκ,0 a−2 , (1.5.60)
H0 i

which implies the following consistency relation

X
Ωi + Ω κ = 1 . (1.5.61)
i

The second Friedmann equation (1.5.43) evaluated at t = t0 becomes

H2 X
 
ä
=− 0 Ωi (1 + 3wi ) . (1.5.62)
a t=t0 2 i

Observations of the cosmic microwave background (CMB) and the large-scale structure (LSS) find that the universe
is flat (Ωκ ∼ 0) and composed of 4% atoms, 23%, cold dark matter and 73% dark energy: Ωb,0 = 0.04, Ωcdm,0 = 0.23,
ΩΛ,0 = 0.73, with wΛ ≈ −1. In the following, we will sometimes drop the suffix “0” that denotes present-day values
of the cosmological parameters unless this is not clear from the context.
 26

C. Matter-radiation equality

The epoch of matter-radiation equality, when ρr = ρm , has special significance for the generation of large scale
structure and the development of CMB anisotropies because perturbations grow at different rates in the two different
eras. It is given by,
ρr 4.15 × 10−5 Ωr Ωm
= 2 4
≡ 4 = 3 , (1.5.63)
ρcrit h a a a
where Ωr and Ωm are specified at the present epoch, yielding
4.15 × 10−5
aEQ = . (1.5.64)
Ωm h2
In terms of redshift,

1 + zEQ = 2.4 × 104 Ωm h2 . (1.5.65)

As Ωm h2 increases, equality is pushed back to higher redshifts and earlier times. It is very important that zEQ is at
least a factor of a few larger than the redshift where photons decouple from matter, z? ' 1100, and that the photons
decouple when the universe is well into the MD era.

D. Cosmological constant

A characteristic feature of GR is that the source for the gravitational field is the entire energy-momentum tensor.
In the absence of gravity, only changes in energy from one state to another are measurable; the normalization of the
energy is arbitrary. However, in gravitation, the normalization of the energy matters. This opens up the possibility of
vacuum energy: the energy density of empty space. We want the vacuum energy not to pick out a preferred direction.
This implies that the associated energy-momentum tensor is Lorentz-invariant in locally inertial coordinates:
(vac)
Tµν = ρvac ηµν . (1.5.66)

Generalizing to an arbitrary frame,

(vac)
Tµν = ρvac gµν . (1.5.67)

Comparing to the perfect fluid energy-momentum tensor, Tµν = (ρ + P )Uµ Uν − P gµν , the vacuum looks like a perfect
fluid with an isotropic pressure opposite in sign to the density:

Pvac = −ρvac . (1.5.68)

(M)
If we decompose the energy-momentum tensor into a matter piece Tµν plus a vacuum piece (1.5.67), the Einstein
equation is:
1 h
(M)
i
Rµν − gµν R = 8πG Tµν + ρvac gµν . (1.5.69)
2
Einstein tried to get a static universe by adding a cosmological constant,
1 (M)
Rµν − gµν R − Λgµν = 8πGTµν , (1.5.70)
2
and this concept is interchangeable with vacuum energy:
Λ
ρΛ = ρvac = . (1.5.71)
8πG
−2
Λ has dimensions of [length] while ρvac has units of [energy/volume]. So Λ defines a scale (whereas GR is otherwise
scale-free). What should be the value of ρvac ? There is no known way to precisely calculate this at present, but since
the reduced Planck mass is
1
MP = √ ∼ 1018 GeV , (1.5.72)
8πG
 27

one might guess that

ρ(guess)
vac ∼ MP4 ∼ (1018 GeV)4 . (1.5.73)

However, this is a dramatically bad guess compared to the observational measurement:

−3
ρ(obs)
vac ∼ (10 eV)4 , (1.5.74)

which is 30 orders of magnitude smaller.

EXERCISE: Taking the universe to be at the critical density, ρcrit (i.e., Ω = 1), and ΩΛ ∼ 0.75, compute ρvac .

There are three conundrums associated with Λ:

• Why is Λ so small?
• Why is Λ 6= 0?
• Why is ΩΛ ∼ Ωm ?
The last is the so-called coincidence problem. Note that
ΩΛ ρΛ
= ∝ a3 . (1.5.75)
Ωm ρm
The vacuum density rapidly overwhelms the matter density! It is a big coincidence to find ourselves at the epoch
where we can observe the transition.

E. Spatial curvature and destiny

In the current universe, the radiation density is significantly lower than the matter density, but both the vacuum
and matter are dynamically important. Parameterized as

Ωκ = 1 − Ωm − Ω Λ , (1.5.76)

these densities evolve differently as a function of time:

ΩΛ ∝ Ωκ a2 ∝ Ωm a3 . (1.5.77)

As a → 0 in the past, curvature and vacuum will be negligible and the universe will behave as Einstein-de Sitter till
the radiation becomes important. As a → ∞ in the future, curvature and matter will be negligible and the universe
will asymptote to de Sitter, unless the scale factor never reaches ∞ because the universe begins to recollapse at a
finite time. Some possibilities the evolution of such a universe are:
• ΩΛ < 0: always decelerates and recollapses (as vacuum energy is always going to dominate).
• ΩΛ ≥ 0: recollapse is possible if Ωm is sufficiently large that it halts the universal expansion before ΩΛ dominates.
To determine the dividing line between perpetual expansion and eventual recollapse, note that collapse requires H
to pass through 0 as it changes from positive to negative:
8πG
H2 = 0 = ρm,0 a−3 −2


? + ρΛ,0 + ρκ,0 a? , (1.5.78)
3
where a? is the scale-factor at turnaround. Dividing by H0 , using Ωκ,0 = 1 − Ωm,0 − ΩΛ,0 , and rearranging, we obtain

ΩΛ,0 a3? + (1 − Ωm,0 − ΩΛ,0 )a? + Ωm,0 = 0 . (1.5.79)

 28

positive
spatial
curvature expands
forever

negative recollapses
spatial
curvature

FIG. 8 Properties of universes dominated by matter and vacuum energy, as a function of the density parameters Ωm and ΩΛ .
The ellipse in the upper-left corner corresponds roughly to the observationally favoured region from a current data compilation
as of 2008.

But what we really care about is not really a? but the range of ΩΛ,0 given Ωm,0 for which there is a real solution to
(1.5.79). The range of ΩΛ,0 for which the universe will expand forever is given by:
(
0 0 ≤ Ωm,0 ≤ 1
ΩΛ,0 ≥ h 
1−Ωm,0
 i . (1.5.80)
4Ωm,0 cos3 31 cos−1 Ωm,0 + 4π
3 Ωm,0 > 1

EXERCISE: Verify Eq. (1.5.80).

When ΩΛ,0 = 0,
• open and flat universes (Ω0 = Ωm,0 ≤ 1) will expand forever.
• closed universes (Ω0 = Ωm,0 > 1) will recollapse.
There is a “folk wisdom” that this correspondence is always true, but it is only true in the absence of vacuum
energy. The current cosmological data favours Ωm,0 ∼ 0.25, ΩΛ,0 ∼ 0.75, Ωκ ∼ 0, which is well into the regime of
perpetual expansion, under the assumption that vacuum energy remains truly constant (which it might not). These
considerations are illustrated in Fig. 8.
 29

We will end this discussion by noting the difficulty of finding static solutions to the Friedmann equations. To be
static, we must have not only ȧ = 0, but also ä = 0.

EXERCISE: Verify that one can only get a static solution, ä = ȧ = 0, if

ρ
P =− , (1.5.81)
3
and the spatial curvature is non-vanishing:
κ 8πG
2
= ρ. (1.5.82)
a 3

Because the energy density and pressure must be of the opposite sign, these conditions can’t be fulfilled in a universe
containing only radiation and matter. Einstein looked for a static solution because at the time, the expansion of the
universe had not yet been discovered. He added the cosmological term, whereby one can satisfy the static conditions
with
1
ρΛ = ρm along with the appropriate positive spatial curvature. (1.5.83)
2
This Einstein-static universe is empirically of little interest today, but extremely useful to theorists, providing the
basis for the construction of conformal diagrams.
 30

VI. TIME AND DISTANCE

Measuring distances in an expanding universe is a tricky business! Which distance should one consider? Some
obvious definitions immediately come to mind:
• comoving distance (remains fixed as the universe expands).
• physical distance (grows simply because of the expansion).
Frequently, neither of these is what we want; e.g. a photon leaves a quasar at z ∼ 6 when the scale factor was 17 of its
present value, and arrives on the earth today, when the universe has expanded by a factor of 7 (see Fig. 9). How can
we relate the luminosity of the quasar to the flux we see?

current metric
distance
to quasar

photon
worldine

Milky Way
worldine quasar
worldine

FIG. 9 Euclidean embedding of a part of the ΛCDM spacetime geometry, showing the Milky Way (brown), a quasar at redshift
z = 6.4 (yellow), light from the quasar reaching the Earth after approximately 12 billion years (red), and the present-era
metric distance to the quasar of approximately 28 billion light years (orange). Lines of latitude (purple) are lines of constant
cosmological time, spaced by 1 billion years; lines of longitude (cyan) are worldlines of objects moving with the Hubble flow,
spaced by 1 billion light years in the present era (less in the past and more in the future). Figure credit: Ben Rudiak-Gould /
Wikimedia Commons.
 31

A. Conformal time

The fundamental measure from which all others may be calculated is the distance on the comoving grid. If the
universe is flat, as we will assume throughout most of these lectures, computing distances on the comoving grid is easy.
One very important comoving distance is the distance travelled by light since t = 0 (in the absence of interactions).
Recalling that we are working in units with c = 1, in time dt, light travels a distance dx = dt
a ; thus, the total comoving
distance light travels is:
Z t
dt0
η≡ 0
. (1.6.1)
0 a(t )

No information could have propagated faster than η on the comoving grid since the beginning of time; thus η is called
the causal horizon or comoving horizon. A related concept is the particle horizon dH , the proper radius travelled by
light since t = 0:
Z t
dt0
dH ≡ a(t) 0
= a(η)η . (1.6.2)
0 a(t )

Regions separated by distances > dH are not causally connected; if they appear similar, we should be suspicious! (cf.
the cosmic microwave background – see later!). We can think of η (which increases monotonically) as a time variable
and call it the conformal time. In terms of η, the FRW metric becomes

dr2
 
ds = a (η) dη 2 −
2 2
− r2 dΩ2 . (1.6.3)
1 − κr2

Just like {t, T, z, a}, η can be used to discuss the evolution of the universe. η is the most useful time variable for
most purposes! In the analysis of the evolution of perturbations, we will use it instead of t. Conformal spacetime
diagrams are easy to construct in terms of η. Consider radial null geodesics in a flat FRW spacetime:
dt
ds2 = a2 (dη 2 − dr2 ) ≡ 0 =⇒ dη = ± ≡ ±dr . (1.6.4)
a(t)
In conformal coordinates, null geodesics (photon worldlines) are always at 45◦ angles, and light cones are Minkowskian
since the metric is conformally flat: gµν = a2 ηµν .
In simple cases, η can be expressed analytically in terms of a. In particular, during radiation domination (RD) and
matter domination (MD),

RD : ρ ∝ a−4 , η ∝ a (1.6.5)
√
MD : ρ ∝ a−3 , η ∝ a . (1.6.6)

EXERCISE: Show that the conformal time as a function of scale-factor in a flat universe containing only matter and
radiation is
η p √
= a + aEQ − aEQ . (1.6.7)
η0
where aEQ denotes the epoch of matter-radiation equality.

B. Lookback distance and lookback time

Another important comoving distance is the distance between us and a distant emitter, the lookback distance. The
comoving distance to an object at scale factor a (or redshift z = a1 − 1) is:
t0 1
dt0 da0
Z Z
dlookback (a) = = , (1.6.8)
t(a) a(t0 ) a a2 (t0 )H(a0 )
 32

where we have used da

dt = aH. Typically, we can see objects out to z . 6. At these late times, radiation can be
ignored. During matter domination, H ∝ a−3/2 , so
2  √ 
FLAT MD : dlookback (a) = 1− a (1.6.9)
H0
 
2 1
dlookback (z) = 1− √ . (1.6.10)
H0 1+z

For small z, dlookback → Hz0 , which is the Hubble Law. At very early times, z  1, we find the limit dlookback → H20 .
Similarly, one can define the lookback time, elapsed between now and when light from redshift z was emitted:
Z t0 Z 1
0 da0
tlookback (a) = dt = 0 0
. (1.6.11)
t(a) a a(t )H(a )

For a flat, matter-dominated universe, the lookback time to redshift z is:

2 h i
FLAT MD : tlookback (z) = 1 − (1 + z)−3/2 . (1.6.12)
3H0
The total age of a matter-dominated universe is obtained by letting z → ∞:
2
t0 (FLAT MD) = . (1.6.13)
3H0
2
For universes that are not totally matter-dominated, the factor of 3 will not be quite right, but for reasonable values
of the cosmological parameters, we usually get t0 ∼ H0−1 .

C. Instantaneous physical distance

Another distance we might want to know is the distance between us and the location of a distant object along our
current spatial hypersurface. Let us write the FRW metric in the form

ds2 = dt2 − a2 (t)R02 dχ2 + Sk2 (χ)dΩ2 ,

 
(1.6.14)

where Sk (χ) is defined by (1.5.27) and k ∈ {+1, 0, −1}. In this form, the instantaneous physical distance dP as
measured at time t between us at χ = 0 and an object at comoving radial coordinate χ is,

dP (t) = a(t)R0 χ , (1.6.15)

where χ remains constant because we assume that both we and the observed object are perfectly comoving (they
might not be, but it is trivial to include the corrections due to so-called “peculiar velocities”). As expected, this
definition of distance also leads to the Hubble Law when the redshift is small,
ȧ
v = d˙P = ȧR0 χ = dP −→ v = H0 dP (1.6.16)
a
when evaluated today.
The instantaneous physical distance, while a convenient construct, is not that useful, because observations always
refer to events on our past lightcone, not on our current spatial hypersurface. For various kinds of observations, we
can define a kind of distance that is what we would infer if space were Euclidean and the universe were not expanding,
and relate it to observables in the FRW universe.

D. Luminosity distance

A classic way of measuring distances in astronomy is to measure the flux from an object of known luminosity, a
standard candle. Let us neglect expansion for a moment, and consider the observed flux F at a distance dL from a
source of known luminosity L:
L
F = . (1.6.17)
4πd2L
 33

This definition comes from the fact that in flat space, for a source at distance d, the flux over the luminosity is just
the inverse of the area of a sphere centred around the source, 1/A(d) = 1/4πd2 . In an FRW universe, however, the
flux will be diluted. Conservation of photons tells us that all of the photons emitted by the source will eventually
pass through a sphere at a comoving distance χ from the emitter. But the flux is diluted by two additional effects:
the individual photons redshift by a factor (1 + z), and the photons hit the sphere less frequently, since two photons
emitted a time δt apart will be measured at a time (1 + z)δt apart. Therefore we will have

F 1
= . (1.6.18)
L (1 + z)2 A

The area A of a sphere centred at a comoving distance χ can be derived from the coefficient of dΩ2 in (1.6.14), yielding

A = 4πR02 Sk2 (χ) , (1.6.19)

where we have set a(t) = 1 because the photons are being observed today. Comparing with (1.6.17), we obtain the
luminosity distance:

dL = (1 + z)R0 Sk (χ) . (1.6.20)

Here, we must point out a caveat: the observed luminosity is related to emitted luminosity at a different wavelength.
Here, we have assumed that the detector counts all photons.
The luminosity distance dL is something we might hope to measure, since there are some astrophysical sources
which are standard candles. But χ is not an observable, so we should rephrase it in terms of something we can
measure. On a radial null geodesic, we have

0 = ds2 = dt2 − a2 R02 dχ2 , (1.6.21)

or
1 dt 1 da
Z Z
χ= = . (1.6.22)
R0 a R0 a2 H(a)

It’s conventional to convert the scale factor to redshift using a = 1/(1 + z), so we have
Z z
1 dz 0
χ(z) = , (1.6.23)
R0 0 H(z 0 )

leading to the luminosity distance,

z
dz 0
 
1
Z
dL = (1 + z)R0 Sk . (1.6.24)
R0 0 H(z 0 )

Note that R0 drops out when k = 0, which is good because in that case it is a completely arbitrary parameter. Even
when it is not arbitrary, it is still more common to speak in terms of Ωκ,0 = −k/R02 H02 , which can either be determined
though measurements of the spatial curvature, or by measuring the matter density and using Ωκ,0 = 1 − Ωm,0 . In
terms of this parameter, we have

H −1
q
R0 = H0−1 −k/Ωκ,0 = p 0 . (1.6.25)
|Ωκ,0 |

Thus we can write the luminosity distance in terms of measurable cosmological parameters as

H −1
Z z
dz 0
 q 
dL = (1 + z) p 0 Sk H0 |Ωκ,0 | 0
, (1.6.26)
|Ωκ,0 | 0 H(z )

where the integral can be evaluated by making use of the Friedmann equation. Though it appears unwieldy, this
equation is of fundamental importance in cosmology. Given the observables H0 and Ωi,0 , we can calculate dL to an
object any redshift z; conversely, we can measure dL (z) for objects at a range of redshifts, and from that extract H0
or the Ωi,0 .
 34

E. Angular diameter distance

Another classic distance measurement in astronomy is to measure the angle δθ subtended by an object of known
physical size `, known as a standard ruler. The angular diameter distance is then defined as,
`
dA = , (1.6.27)
δθ
where δθ is small. At the time when the light was emitted, when the universe had scale factor a, the object was at
redshift z at comoving coordinate χ (assuming again that we are at χ = 0). Hence, from the angular part of the
metric, ` = aR0 Sk (χ)δθ, and comparing with (1.6.27) we have the angular diameter distance

R0 Sk (χ)
dA = . (1.6.28)
1+z
Fortunately, the unwieldy dependence on cosmological parameters is common to all distance measures, and we are
left with a simple dependence on redshift:

dL = (1 + z)2 dA . (1.6.29)

Note that dA is equal to the comoving distance at low redshift! But it actually decreases at very large redshift. In
a flat universe, objects at large redshift appear larger than they would at intermediate redshift.
Both distances: dA , dL are larger in a universe with a cosmological constant than in one without. This follows since
the energy density, and hence the expansion rate, is smaller in a Λ universe. The universe was therefore expanding
more slowly early on, and light had more time to travel from distant objects to us. Distant objects will therefore appear
fainter in a Λ-dominated universe than if the universe was MD today. This observation (using Type Ia supernovae as
standardizable candles) is exactly what lead to the discovery of dark energy in the 1990’s.

EXERCISE: Simplify (1.6.26) and (1.6.28) to obtain expressions for the luminosity distance and the angular diameter
distance in a flat universe.
 35

VII. PARTICLES AND FIELDS IN COSMOLOGY

A. General particle motion

We previously considered the redshifting of a massless particle. Let us consider the case of a general particle for
completeness. This is, of course, governed by the geodesic equation
duµ
+ Γµνσ uν uσ = 0 , (1.7.1)
ds
dxµ
where uµ = ds . In the metric (1.5.35), we have
(
µ 0 2 i j 0 2 2 0 photon
u uµ = g00 (u ) + gij u u ≡ (u ) − |ũ| = (1.7.2)
1 massive particle

where |ũ|2 = −gij ui uj is the physical velocity2 . The latter result comes from the fact that for the 4-momentum
pµ = (E, p~), pµ pµ = 0 for photons, and the on-shell condition pµ pµ = m2 holds for massive particles, with pµ = muµ .
For both cases,

du0 d|ũ|
u0 = |ũ| . (1.7.3)
ds ds
For the metric (1.5.35), the relevant Christoffel symbol can be expressed as Γ0ij = − ȧa gij , so the 0th component of
(1.7.1) becomes

du0 ȧ du0 ȧ
− gij ui uj = + |ũ|2 = 0 , (1.7.4)
ds a ds a
and thus by (1.7.3),

|ũ| d|ũ| ȧ 2
+ |ũ| = 0 . (1.7.5)
u0 ds a
But u0 = dt ˙ + ȧ |ũ| = 0, or
ds , so |ũ| a

˙
|ũ| ȧ
=− . (1.7.6)
|ũ| a

Hence, for both massless and massive particles, momentum always redshifts as

1
|ũ| ∝ . (1.7.7)
a

The particle therefore slows down with respect to the comoving coordinates as the universe expands. In fact this is
an actual slowing down, in the sense that a gas of particles with initially high relative velocities will cool down as the
universe expands.

B. Classical field theory

When we make the transition from SR to GR, the Minkowski metric ηµν is promoted to a dynamical tensor field,
gµν (x). GR is an example of a classical field theory. Let’s get a feel for how such theories work by considering classical
fields in a flat spacetime. We will not discuss quantum fields here, though this will become relevant in the second part
of the course when the origin of primordial perturbations is discussed.

2 Recall p again that p ~ is the comoving momentum, and the physical momentum p̃ measuring changes in physical distance is given by
|p̃| = |gij pi pj | = a|~p|. Similarly, for 4-velocity uµ = (1, ~ u to the physical velocity as |ũ| =
u), we relate the comoving velocity ~
p
i j
|gij u u | = a|~ u|.
 36

Begin with the familiar example of the classical mechanics of a single particle in 1D with coordinate q(t). Equations
of motion for such a particle comes from using the principle of least action: search for critical points (as a function of
the trajectory) of an action S,
Z
S = dtL(q, q̇) , (1.7.8)

where L(q, q̇) is the Lagrangian. The Lagrangian in point-particle interactions is typically of the form L = K − V ,
where K is the kinetic energy and V is the potential energy. Using the calculation of variations procedure (cf. any
advanced classical mechanics textbook), the critical points of the action, i.e. trajectories q(t) for which the action S
remains stationary under small variations, are those that satisfy the Euler-Lagrange equations:
 
∂L d ∂L
− =0 . (1.7.9)
∂q dt ∂ q̇

For example, L = 12 q̇ 2 − V (q) leads to the equation of motion q̈ = − dV dq . For field theory, we replace the single
coordinate q(t) by a set of spacetime-dependent fields Φi (xµ ), and the action becomes a functional of these fields. i
labels individual fields. A functional is a function of an infinite number of variables, e.g. the values of a field in some
region of spacetime.
In field theory, the Lagrangian can be expressed as an integral over the space of a Lagrangian density, L, which is
a function of the fields Φi and their spacetime derivatives ∂µ Φi :
Z
L = d3 x L(Φi , ∂µ Φi ) . (1.7.10)

Then the action becomes,

Z Z
S= dtL = d4 x L(Φi , ∂µ Φi ) . (1.7.11)

L is a Lorentz scalar. It is most convenient to define a field theory by specifying the Lagrange density, from which
all equations of motion can be derived. The Euler-Lagrange equations again come from requiring that S be invariant
under small variations of the field,

Φi → Φi + δΦi , ∂µ Φi → ∂µ Φi + δ(∂µ Φi ) = ∂µ Φi + ∂µ (δΦi ) . (1.7.12)

The expression for variation in ∂µ Φi is simply the derivative of the variation of Φi . Since δΦi is assumed to be small,
we can Taylor-expand the Lagrangian under this variation,

L(Φi , ∂µ Φi ) → L(Φi + δΦi , ∂µ Φi + ∂µ δΦi ) ,

∂L ∂L
= L(Φi , ∂µ Φi ) + i
δΦi + ∂µ (δΦi ) . (1.7.13)
∂Φ ∂(∂µ Φi )

Correspondingly, the action goes to S → S + δS, with

 
∂L ∂L
Z
4 i i
δS = d x δΦ + ∂µ (δΦ ) . (1.7.14)
∂Φi ∂(∂µ Φi )
 37

EXERCISE: (NON-EXAMINABLE) Factor out the δΦi term from the integrand, by integrating the second term
by parts. You will obtain one term which is a total derivative – the integral of something of the form ∂µ V µ – that
can be converted to a surface term by the four-dimensional version of Stokes’ Theorem. Since we are considering
variational problems, we can choose to consider variations that vanish at the boundary, along with their derivatives.
It is therefore traditional in such contexts to integrate by parts with complete impunity, always ignoring the boundary
conditions. Sometimes this is not okay, but fortunately we will not encounter such situations in this course. Assuming
that the variations and their derivatives vanish at the boundaries, show that
  
∂L ∂L
Z
4
δS = d x − ∂µ δΦi . (1.7.15)
∂Φi ∂(∂µ Φi )

The functional derivative δS/δΦi of a functional S with respect to a function Φi is defined to satisfy

δS
Z
δS = d4 x i δΦi (1.7.16)
δΦ
when such an expression is valid. We can therefore express the notion that S is at a critical point by saying that the
functional derivative vanishes. Finally we arrive at the Euler-Lagrange equations of motion for a field theory in flat
spacetime:
 
δS ∂L ∂L
i
= − ∂µ =0 . (1.7.17)
δΦ ∂Φi ∂(∂µ Φi )

You will not be asked to derive (1.7.17) in the exam for this course, but it is one of the most important equations
in theoretical physics, and you are therefore requested to make sure that you have gone through this derivation by
yourself, even if it is with the aid of a textbook such as Carroll.

There are several benefits of introducing the Lagrangian formulation:

• The simplicity of positing a single scalar-valued function of spacetime, the Langrange density, rather than a
number of (perhaps tensor-valued) equations of motion.
• Demanding that the action be invariant under a symmetry assures that the dynamics respect the symmetry too.
• The action leads via a direct procedure to a unique energy-momentum tensor.

C. Energy-momentum tensor from the action

As promised earlier, we will now consider the last point. First, we need to generalize the previous discussion to
curved space. Recalling our experience with the geodesic equation, first we will replace the partial derivative by the
covariant derivative, defined e.g.

∂T αβ
∇µ T αβ ≡ + Γαγµ T
γβ
+ Γβγµ T αγ ≡ T αβ;µ , (1.7.18)
∂xµ
∂Tαβ
∇µ Tαβ ≡ − Γγαµ Tγβ − Γγβµ Tαγ ≡ Tαβ;µ , (1.7.19)
∂xµ
∂T α
∇µ T α ≡ + Γα γ α
γµ T ≡ T ;µ , (1.7.20)
∂xµ
∂Tα
∇µ Tα ≡ − Γγαµ Tγ ≡ Tα;µ , (1.7.21)
∂xµ
∂T
∇µ T ≡ ≡ T ;µ , (1.7.22)
∂xµ
gαβ;µ = 0. (1.7.23)
 38

Now in n dimensions,
Z
S= dn x L(Φi , ∇µ Φi ) . (1.7.24)

Note that, since dn x is a density rather than a tensor, L is also a density. We typically write,
√
L = −g L̂ , (1.7.25)

where L̂ is a scalar, and g = det gµν . The associated Euler-Lagrange equations make use of the scalar L̂, and are like
those in flat space but with covariant instead of partial derivatives:
!
∂ L̂ ∂ L̂
− ∇µ =0. (1.7.26)
∂Φ ∂(∇µ Φ)

e.g., the curved space generalization of the action for a single scalar field is:

n √
 
1 µν
Z
Sφ = d x −g g (∇µ φ)(∇ν φ) − V (φ) , (1.7.27)
2
which would lead to an equation of motion,
dV
φ + =0, (1.7.28)
dφ
where
φ = ∇ν ∇ν φ = g µν ∇µ ∇ν φ . (1.7.29)
With that warm-up, let’s think about the construction of an action for GR. The dynamical variable is now gµν , the
metric. What scalars can we make out of the metric to serve as a Lagrangian? Since we know that the metric can be
set equal to its canonical form (i.e., gµν = ηµν ) and its derivatives set to zero at any one point, any non-trivial scalar
must involve at least second derivatives of the metric.
We have already encountered the Ricci scalar. It turns out to be the only independent scalar constructed from the
metric, which is no higher than second order in its derivatives. Hilbert figured that this was the simplest possible
choice for a Lagrangian, and proposed:

√
Z
SH = − dn x −gR (Einstein–Hilbert Action) . (1.7.30)

He was right! Beware the sign convention of the metric signature here. If you see this expression with no minus-sign
in the literature, remember that Rµν is invariant under sign-change of the metric, and hence R = g µν Rµν flips sign
under sign change of the metric. Cutting a very long story short, varying the action,

n √
 
1
Z
δSH = − d x −g Rµν − R gµν δg µν , (1.7.31)
2
we find that at stationary points,
1 δSH 1
−√ = Rµν − R gµν = 0 . (1.7.32)
−g δg µν 2
Voila – we have recovered Einstein’s equation in a vacuum! We got the result in a vacuum because we only included
the gravitational part of the action, not the matter fields. To include them, consider
1
S= SH + SM , (1.7.33)
16πG
where SM is the matter action, and we have presciently normalized the gravitational action to get the right answer.
Again, cutting out the details of the variational procedure, which you are welcome to work through on your own, one
obtains,
 
1 δS 1 1 1 δSM
√ =− Rµν − R gµν + √ =0. (1.7.34)
−g δg µν 16πG 2 −g δg µν
 39

Let’s boldly define the energy-momentum tensor as:

2 δSM
Tµν = √ . (1.7.35)
−g δg µν

We immediately see that this definition leads to Einstein’s equation:

1
Rµν − R gµν = 8πG Tµν . (1.7.36)
2
Consider again the action (1.7.27) for a scalar field. Now vary this action with respect to gµν , not φ:

√ gµν g ρσ
  
1
Z
δSφ = dn x −g δg µν ∇µ φ∇ν φ − ∇ρ φ∇σ φ − V (φ) , (1.7.37)
2 2 2

we obtain from (1.7.35) the energy-momentum tensor for a scalar field,

 
(φ) 1 ρσ
Tµν = ∇µ φ∇ν φ − gµν g ∇ρ φ∇σ φ − V (φ) . (1.7.38)
2

Here, it is worth pointing out that you will find different sign conventions to (1.7.38) in the cosmology literature. This
can be traced to the fact that most of the cosmology literature uses the opposite metric signature to ours: (−, +, +, +).
For a homogeneous field (∂i φ = 0), we don’t need the covariant derivative (“semi-colon goes to a comma rule”), and
this reduces to
 
(φ) 1 ρσ
Tµν = ∂µ φ∂ν φ − gµν g ∂ρ φ∂σ φ − V (φ) . (1.7.39)
2

EXERCISE: If you want to double check this derivation (NON-EXAMINABLE) via the variational procedure,
you might find it useful to note that the Kronecker delta δ µν is unchanged under any variation, and first derive the
following results:

δgµν = −gµρ gνσ δg ρσ , (1.7.40)

δg = −g(gµν δg µν ) , (1.7.41)

where the latter expression requires the variation of the identity ln(det M) = Tr(ln M), and M is a square matrix
with non-vanishing determinant, with exp(ln M) = M. It is helpful to remember the cyclic property of the trace.
You will need to show also that the term ∝ δRµν leads to an integral with respect to the natural volume element of
the covariant divergence of a vector. By Stokes’ Theorem, this is equal to a boundary condition at infinity, which
contributes nothing to the total variation. Consider subtleties which may affect this conclusion.

In Minkowski space there is an alternative definition for the “canonical energy-momentum tensor” (often discussed
in books on electromagnetism or field theory). In this context, energy-momentum conservation arises as a consequence
of symmetry of the Lagrangian under spacetime translations, via Noether’s theorem. Applying this procedure to a
Lagrangian that depends on some fields Φi and their first derivatives ∂µ Φi (in flat spacetime), one obtains the canonical
energy momentum tensor:

µν δL
T(canonical) = ∂ ν Φi − η µν L = ∂ µ Φi ∂ ν Φi − η µν L , (1.7.42)
δ(∂µ Φi )

where a sum over i is implied. We will not discuss this procedure further here, firstly because it does not always
generalize to a curved spacetime, and also because (1.7.35) is what appears on the RHS of Einstein’s equation when
it is derived from an action.
 40

Before we move on, let us remember that we arrived at the Hilbert Lagrangian L̂H = R by looking for the simplest
possible scalar constructed from the metric. Of course, an even simpler one is a constant, so

4 √
 
1
Z
(M)
S = d x −g − (R + 2Λ) + L̂ , (1.7.43)
16πG

giving

L̂(vac) = −ρvac . (1.7.44)

So it is easy to introduce vacuum energy; however, we have no insight into its expected value, since it enters as an
arbitrary constant.
 41

VIII. BASICS OF INFLATION

A. Shortcomings of the Standard Big Bang Theory

A number of fundamental questions about the universe are raised by the Big Bang theory:
1. Why is the universe spatially flat on large scales?
2. Why is the universe so homogeneous on large scales?
3. What started the “Big Bang”?
4. What was the origin of the primordial fluctuations which lead to the complex structures observed in the universe
today?
Let us discuss these puzzles in more detail.

1. Homogeneity Problem

We previously derived the FRW metric assuming the homogeneity and isotropy of the universe. Why is this a good
assumption? This is particularly surprising given that inhomogeneities are gravitationally unstable and therefore grow
with time (cf. the second part of the course). Observations of the CMB show that the inhomogeneities were much
smaller in the past (at decoupling) than today. One thus expects that these inhomogeneities were even smaller at
earlier times. How do we explain the homogeneity of the early universe?

2. Flatness Problem

Why is the universe so closely approximated by flat Euclidean space? To understand the severity of the problem
in more detail consider the Friedmann equation in the form of (1.5.51), reproduced here for completeness:
κ
Ω(a) − 1 = . (1.8.1)
(aH)2

In the standard Big Bang cosmology containing just matter and radiation, the comoving Hubble radius, (aH)−1 ,
grows with time and from (1.5.51), the quantity |Ω − 1| must thus diverge with time. Ω = 1 is an unstable fixed
point. Therefore, in standard Big Bang cosmology, the near-flatness observed today (Ω0 ∼ 1) requires an extreme
fine-tuning of Ω close to 1 in the early universe. More specifically, one finds that the deviation from flatness at Big
Bang nucleosynthesis (BBN, ∼ 0.1 MeV ), during the GUT era (∼ 1015 GeV) and at the Planck scale (∼ 1019 GeV),
respectively has to satisfy the following conditions

|Ω(aBBN ) − 1| ≤ O(10−16 ) (1.8.2)

|Ω(aGUT ) − 1| ≤ O(10−55 ) (1.8.3)
|Ω(aPl ) − 1| ≤ O(10−61 ) . (1.8.4)

Another way of understanding the flatness problem is by differentiating (1.5.51) and using the continuity equation
in the form of (1.5.55) to obtain

dΩ
= (1 + 3w)Ω(Ω − 1) . (1.8.5)
d[ln a]

This makes it apparent that Ω = 1 is an unstable fixed point if the strong energy condition is satisfied

d|Ω − 1|
>0 ⇐⇒ 1 + 3w > 0 . (1.8.6)
d[ln a]

Again, why is Ω0 ∼ O(1) and not much smaller or much larger?

 42

EXERCISE: Ω(t) = 8πGρ(t)

3H 2 (t) . Assume Ω0 ∼ 0.3 today, and ΩΛ = 0. Plot Ω(t) − 1 as a function of a(t) assuming that
the universe was first radiation-dominated and then matter-dominated. How close is Ω(t) − 1 to zero at the Planck
epoch, a ∼ 10−32 ? This fine-tuning of the initial conditions is known as the flatness problem. If not for fine-tuning,
an open universe would be obviously open, and Ω0 would be almost exactly zero today. Instead, we observe that
|Ω0 − 1| . O(0.01) today.

3. Horizon Problem

The comoving Hubble radius, (aH)−1 , characterizes the fraction of comoving space in causal contact. During the
Big Bang expansion (aH)−1 grows monotonically and the fraction of the universe in causal contact increases with
time.
Consider the CMB today. It is very close to isotropic and its temperature is the same everywhere to one part in
10−5 . The largest observed scales have just entered the horizon, long after the decoupling of photons from baryons;
i.e. causal physics have just begun to operate on them. Before decoupling, the wavelengths of these modes were so
large that no causal physics could force deviations from smoothness to go away. After decoupling, the photons (to
a good approximation) simply free stream. But the near-homogeneity of the CMB tells us that the universe was
extremely homogeneous at the time of decoupling on a scale encompassing many regions that a priori are causally
independent. How is this possible? This is a profound problem that we’ve glossed over by simply assuming that the
temperature is uniform and perturbations about the zero order temperature are small.
To be more specific, remember the definition (1.6.1) of the conformal time η, which is also the causal (or comoving)
horizon, characterizing the distance travelled by light since t = 0. We can rewrite this in the following useful form:
a a
da0
 
1
Z Z
0
η≡ = d[ln a ] . (1.8.7)
0 Ha02 0
0
aH

1
During the standard cosmological expansion the increasing comoving Hubble radius, aH , is therefore associated with
3
an increasing comoving horizon , η. As we saw previously, for RD and MD universes we find
Z a
1 da0
η =
H0 Ω1/2 (a0 )a02
( 0
a RD
∝ . (1.8.8)
a1/2 MD

This means that the comoving horizon grows monotonically with time (at least in an expanding universe) which
implies that comoving scales entering the horizon today have been far outside the horizon at CMB decoupling. Why
is the CMB uniform on large scales (of order the present horizon)?

EXERCISE: Calculate the angular size of the comoving horizon at the last scattering surface at z ' 1100 as projected
on to our present CMB sky, assuming a flat FRW cosmology, and that the universe contains only radiation and matter.
You will need to calculate the angular diameter distance to last scattering and the particle horizon at last scattering
(since proper distances, not comoving distances, appear in the derivation of the angular diameter distance).

3 This justifies the common practice of often using the terms ‘comoving Hubble radius’ and ‘comoving horizon’ interchangeably. Although
these terms should conceptually be clearly distinguished, this inaccurate use of language has become standard.
 43

4. On the Problem of Initial Conditions

The flatness and horizon problems are not strict inconsistencies in the standard cosmological model. If one assumes
that the initial value of Ω was extremely close to unity and that the universe began homogeneously (but with just the
right level of inhomogeneity to explain structure formation) then the universe will continue to evolve homogeneously
in agreement with observations. The flatness and horizon problems are therefore really just severe shortcomings in
the predictive power of the Big Bang model. The dramatic flatness of the early universe cannot be predicted by the
standard model, but must instead be assumed in the initial conditions. Likewise, the striking large-scale homogeneity
of the universe is not explained or predicted by the model, but instead must simply be assumed.

5. What got the Big Bang going?

Recall the second Friedmann equation,

4πGa
ä = − (ρ + 3P ) . (1.8.9)
3
Since ordinary matter (ρ + 3P > 0) can only cause deceleration (ä < 0), what got the Big Bang going?

B. Possible resolution of the Big Bang puzzles

1. A Crucial Idea

All Big Bang puzzles are solved by a beautifully simple idea: invert the behavior of the comoving Hubble radius i.e.
make it decrease sufficiently in the very early universe (Fig. 10). The corresponding condition is that

H −1 d2 a
 
d
<0 ⇒ >0 ⇒ ρ + 3P < 0 . (1.8.10)
dt a dt2

start
COMOVING
Hubble length

now end

smooth patch

FIG. 10 Evolution of the comoving Hubble radius, (aH)−1 , in the inflationary universe. Figure credit: Adapted from Liddle
and Lyth by Daniel Baumann. The comoving Hubble sphere shrinks during inflation and expands after inflation.
 44

In a general sense, this is the essential idea about a paradigm for describing the early universe known as “inflation”.
The three equivalent conditions necessary for inflation are:
• Decreasing comoving horizon
 
d 1
< 0. (1.8.11)
dt aH

• Accelerated expansion
d2 a
> 0. (1.8.12)
dt2

• Violation of the strong energy condition – negative pressure

1
P < − ρ. (1.8.13)
3

Inflation presents a logical way out of the previous argument by realizing that an early epoch of rapid expansion
solves the horizon problem. Einstein’s equation tells us what type of energy is needed in order to produce this rapid
expansion, showing that negative pressure is required. We will consider a scalar field theory and show that negative
pressure is easy to accommodate in such a theory.

2. Comoving Horizon during Inflation

Recall the definition of the comoving horizon (= conformal time) as a logarithmic integral of the comoving Hubble
radius
Z a
1
η= d[ln a0 ] 0 . (1.8.14)
0 a H(a0 )
If particles are separated by comoving distances greater than aH, they cannot currently communicate. Note the
subtle distinction with η: if particles are separated by comoving distances greater than η, they could never have
communicated. If they are separated by comoving distances greater than (aH)−1 , they can’t talk to eachother now.
Therefore it is possible that η  (aH)−1 now, so particles can’t communicate now, but were in causal contact early
on. For example, this would happen if (aH)−1 early on was much larger than it is now, so the η integral gets most
of its contribution early on. Unfortunately this does not happen during the RD and MD epochs. In those epochs,
(aH)−1 increases with time (EXERCISE: CHECK!), so the largest contributions come from the most recent times.
But so far, we have been assuming that the universe was RD all the way back to a = 0! This suggests a solution:
perhaps early on, the universe was not dominated by matter or radiation. For at least a brief time, (aH)−1 decreased
dramatically. Then η would get most of its contribution from early times before the rapid expansion of the grid.
How must a(t) evolve to solve the horizon problem? To give away some of the punch line, inflationary models
typically operate at energy scales 1015 GeV or larger. We can get a qualitative answer by assuming that the universe
was RD since the end of inflation, and ignoring the relatively recent MD epoch. Remembering that H ∝ a−2 during
RD, the scale factor at the end of inflation is
 2
a0 H0 a0 ae
= = ae . (1.8.15)
ae He ae a0

If Te ∼ 1015 GeV,
T0 T0
ae ∼ ∼ 15 ∼ 10−28 , (1.8.16)
Te 10 GeV
i.e., (aH)−1 at the end of inflation was 1028 times smaller than it is today.
For inflation to solve the horizon problem, (aH)−1 at the start of inflation was smaller than the current comoving
Hubble radius, i.e. the largest scales observable today. Thus, during inflation, (aH)−1 had to decrease by ∼ 1028 . The
most common way to arrange this it to set up H ∼ const during inflation:
da
= Hdt, a(t) = ae eH(t−te ) , t < te , (1.8.17)
a
 45

cosmological scales
1

comoving distance (Mpc)

10− 10 Hubble radius
inflation ( aH ) − 1 radiation
dominated

10− 20
10− 40 10− 20 1
scale factor a
red: scales are smaller than horizon, and subject to microphysical processes
FIG. 11 Solution of the Horizon Problem. Scales of cosmological interest were larger than the Hubble radius until a ∼ 10−5 .
However, very early on, before inflation operated, all scales of interest were smaller than the Hubble radius and therefore
susceptible to microphysical processing. Similarly, at very late time, scales of cosmological interest came back within the
Hubble radius. Adapted from Dodelson.

where te is the time at the end of inflation. Thus, decreasing (aH)−1 during inflation is solely due to the exponential
increase in a(t). For the scale factor to increase by 1028 , H(t − te ) must be ∼ ln(1028 ) ∼ 64.

EXERCISE: Assuming ae corresponded to 1015 GeV, compute the number of “e–folds” of inflation needed to solve the
horizon problem, accounting for the RD/MD transition at aEQ .

Notice the symmetry of the inflationary solution (cf. Fig. 11). Scales just entering the horizon today, ∼ 60 e-folds
after the end of inflation, left the horizon ∼ 60 e-folds before the end of inflation.
So far, we have discussed inflation in terms of comoving coordinates. But it is also profitable to think of exponential
expansion in terms of physical coordinates. The physical size of a causally connected region exponentially increases
during inflation. So regions that we observe to be cosmological today were actually microscopically small before
inflation, and in causal contact.
The total comoving horizon is not an effective time parameter after inflation, because it becomes very large early
on, then changes very little during RD and MD. We can rectify this by subtracting off the primordial part
Z t
dt0
η= 0
, (1.8.18)
te a(t )

so that the total comoving horizon is ηprim + η. So during inflation, η is negative, but monotonically increasing. A
scale leaves the horizon when k|η| < 1, and reenters the horizon when k|η| > 1.

3. Flatness Problem Revisited

Recall the Friedmann equation in the form

κ
|1 − Ω(a)| = . (1.8.19)
(aH)2
 46

Inflation (H ≈ const., a = eHt ) is characterized by a decreasing comoving horizon which drives the universe toward
flatness (rather than away from it),
1
|1 − Ω(a)| ∝ = e−2Ht → 0 as t → ∞. (1.8.20)
a2
This solves the flatness problem! Ω = 1 is an attractor during inflation.

EXERCISE: Assume Ω0 ∼ 0.3 today, and ΩΛ = 0. Extrapolate Ω(t) − 1 back to the end of inflation, then through 60
e-folds of inflation. What is Ω(t) − 1 right before the 60 e-folds of inflation?

4. Horizon Problem Revisited

A decreasing comoving horizon means that large scales entering the present horizon were inside the horizon be-
fore inflation (see Fig. 11). Causal physics before inflation therefore established thermal equilibrium and spatial
homogeneity. The uniformity of the CMB is not a mystery.

5. Origin of primordial inhomogeneities

Besides solving the Big Bang puzzles the decreasing comoving horizon during inflation is the key feature required
for the quantum generation of cosmological perturbations described in the second part of the course. During inflation,
quantum fluctuations are generated on subhorizon scales, but exit the horizon once the Hubble radius becomes smaller
than their comoving wavelength. In physical coordinates this corresponds to the superluminal expansion stretching
perturbations to acausal distances. They become classical superhorizon density perturbations which reenter the
horizon in the subsequent Big Bang evolution and then undergo gravitational collapse to form the large scale structure
in the universe.

6. Conformal Diagram of Inflationary Cosmology

A truly illuminating way of visualizing inflation is with the aid of a conformal spacetime diagram. Reminder: the
flat FRW metric in conformal time η becomes

ds2 = a2 (η) dη 2 − d~x2 ,

 
(1.8.21)

and during matter or radiation domination the scale factor evolves as

(
η RD
a(η) ∝ (1.8.22)
η 2 MD .

If the universe had always been dominated by matter or radiation, this would imply the existence of the big bang
singularity at η = 0,

a(η ≡ 0) = 0 . (1.8.23)

The conformal diagram corresponding to standard Big Bang cosmology is given in Fig. 12. The horizon problem is
apparent. Each spacetime point in the conformal diagram has an associated past light cone which defines its causal
past. Two points on a given η = constant surface are in causal contact if their past light cones intersect at the Big
Bang, η = 0. This means that the surface of last scattering (ηCMB ) consisted of many causally disconnected regions
that won’t be in thermal equilibrium. The uniformity of the CMB on large scales hence becomes a serious puzzle.
In de Sitter space, the scale factor is
1
a(η) = − , (1.8.24)
Hη
 47

Conformal Time Today

Past Light-Cone

CMB
Big Bang

FIG. 12 Conformal diagram of Big Bang cosmology. The CMB at last scattering consists of 105 causally disconnected regions.
Figure credit: Daniel Baumann.

Conformal Time Today

Past Light-Cone

CMB
0 Reheating

Inflation

Big Bang
FIG. 13 Conformal diagram of inflationary cosmology. Inflation extends conformal time to negative values. The end of inflation
creates an “apparent” Big Bang at η = 0. There is, however, no singularity at η = 0 and the light cones intersect at an earlier
time iff inflation lasts for at least 60 e-folds. Figure credit: Daniel Baumann.

and the singularity, a = 0, is pushed to the infinite past, η → −∞. The scale factor (1.8.24) becomes infinite at
η = 0! This is because we have assumed de Sitter space with H = const., which means that inflation will continue
forever with η = 0 corresponding to the infinite future t → +∞. In reality, inflation ends at some finite time, and the
approximation (1.8.24) although valid at early times, breaks down near the end of inflation. So the surface η = 0 is
not the Big Bang, but the end of inflation. The initial singularity has been pushed back arbitrarily far in conformal
time η  0, and light cones can extend through the apparent Big Bang so that apparently disconnected points are in
causal contact. This is summarized in the conformal diagram in Fig. 13.
 48

IX. INFLATION: IMPLEMENTATION WITH A SCALAR FIELD

Inflation, an epoch in which the universe accelerates, solves a number of puzzles associated with the standard
Big Bang cosmology. During a phase of accelerated expansion, H −1 (the physical Hubble radius) remains fixed. So
particles initially in causal contact can no longer communicate. Regions separated by cosmological distances today
were actually in causal contact before/during inflation. At that time, these regions were given the necessary initial
conditions, smoothness, and small perturbations about smoothness we observe today. The second part of the course
will detail how these small perturbations came about.
In this lecture, we will consider how to implement inflation using a scalar field with special dynamics. Although no
fundamental scalar field has yet been detected in experiments, there are fortunately plenty of such fields in theories
beyond the standard model of particle physics. In fact, in string theory, for example, there are numerous scalar fields
(moduli), although it proves very challenging to find just one with the right characteristics to serve as an inflaton
candidate. In the following we will therefore describe the dynamics of a generic scalar field leaving the connection with
fundamental particle theory for a future revolution in theoretical physics. We will leave the physics of the generation
of primordial perturbations during inflation to the discussion of perturbation theory later in the course, only dealing
with them in a heuristic way to describe observables of this theory.

A. Negative pressure

From the second Friedmann equation (1.4.26), an accelerating universe requires

ρ 1
ä > 0 ⇒ (ρ + 3P ) < 0 ⇒ P < − ⇒w<− . (1.9.1)
3 3
Since ρ ≥ 0, this implies negative pressure, P < 0. This can’t be ordinary matter or radiation!

B. Implementation with a scalar field

Can a scalar field have ρ + 3P < 0? We can rewrite the energy momentum tensor for a homogeneous scalar field φ,
(1.7.39) in the mixed form,
 
µ µβ µ 1 λα
T ν = g φ,β φ,ν − δ ν g φ,α φ,λ − V (φ) , (1.9.2)
2

and equate it to (1.4.9). Here, V (φ) is the potential, e.g., for a free field with mass m, V (φ) = 12 m2 φ2 . For a
homogeneous field, only time-derivatives are relevant. Let us assume φ(~x, t) = φ(0) (t) + δφ(~x, t) and derive the
density, pressure, and time-evolution of the homogeneous part. Equating the time-time component, we find:
2
1 ∂φ(0)

ρφ = + V (φ(0) ) , (1.9.3)
2 ∂t | {z }
| {z } potential energy
kinetic energy

which looks like the energy density of a single particle with position φ(0) moving in a potential V (φ). Equating the
space-space component, we find that:
2
∂φ(0)

1
Pφ = − V (φ(0) ) . (1.9.4)
2 ∂t

Thus, for negative pressure,

2
1 ∂φ(0)

(0)
P < 0 =⇒ V (φ )> . (1.9.5)
2 ∂t
| {z }
PE>KE
 49

The resulting equation of state

1 2
2 φ̇ −V
w= 1 2
(1.9.6)
2 φ̇ +V
shows that a scalar field can lead to negative pressure (w < 0) and accelerated expansion (w < −1/3) if the potential
energy V dominates over the kinetic energy 21 φ̇2 .
As an example, consider a field trapped in a false vacuum. The energy density is constant since φ(0) is a constant.
Thus,
r
ȧ 8πGρ
= = const . (1.9.7)
a 3
Thus, we obtain exponential expansion with H ∝ ρ1/2 = const:
1 h H(te −tb ) i
ηprim = e −1 , (1.9.8)
He ae
where tb and te denote the beginning and end of inflation, respectively. So if the field is trapped for 60 e-foldings,
H(te − tb ) > 60, the horizon problem is solved. This is actually “Old Inflation”, the original Guth (1981) formulation.
It doesn’t work because “bubbles” of localized regions tunnelling into the true vacuum never coalesce in order for the
universe as a whole to move to the true vacuum. In “New Inflation”, Linde, Steinhardt and Albrecht postulated that
the scalar field rolls slowly toward the true vacuum in a nearly flat potential, where ρ ∼ const.

reheating

FIG. 14 Example of an inflaton potential. Acceleration occurs when the potential energy of the field, V (φ), dominates over its
kinetic energy, 12 φ̇2 . Inflation ends at φend when the kinetic energy has grown to become comparable to the potential energy,
1 2
2
φ̇ ≈ V . CMB fluctuations are created by quantum fluctuations δφ about 60 e-folds before the end of inflation. At reheating,
the energy density of the inflaton is converted into radiation. Figure credit: Daniel Baumann.

What is the equation of motion of a scalar field in an FRW background? Reminding ourselves of the scalar field
action of Eq. (1.7.27), for a scalar field minimally coupled to gravity, the action is
√
 
R 1
Z
(minimally−coupled)
Sφ = d4 x −g − + g µν (∇µ φ)(∇ν φ) − V (φ) . (1.9.9)
16πG 2
Hence, reminding ourselves of Eq. (1.7.28), the equation of motion for the homogeneous field (∂i φ = 0) is
dV
g µν ∇µ (∂ν φ) + =0. (1.9.10)
dφ
Noting the presence of the covariant derivative and using the Christoffel symbols (1.3.13) for the flat FRW cosmology,
the equation of motion can be written as:

dV (φ(0) )
φ̈(0) + 3H φ̇(0) + =0 , (1.9.11)
dφ
 50

with Hubble parameter

 
8πG 1  (0) 2
H2 = φ̇ + V (φ(0) ) . (1.9.12)
3 2

EXERCISE: Derive Eq. (1.9.11).

By directly varying the action with respect to δφ, (cf. Eq. (1.7.17)) the equation of motion can also be written in
the following useful form:

δS 1 √ dV (φ)
= √ ∂µ ( −g∂ µ φ) + =0 , (1.9.13)
δφ −g dφ
√
where in the flat FRW background, −g = a3 (t).

EXERCISE: Obtain Eq. (1.9.11) from yet another route by appropriately combining the Friedmann equations for a
scalar-field dominated universe. By changing the independent variable to the conformal time η, show that the equation
of motion can be written in the following useful form:

d2 φ(0) dφ(0) 2 dV (φ
(0)
)
2
+ 2aH + a =0 . (1.9.14)
dη dη dφ

C. Slow-Roll Inflation

−2
In the following discussion, for notational convenience we will set 8πG ≡ MPl ≡ 1.

Inflation occurs if the field is evolving slow enough that the potential energy dominates over the kinetic energy,
and the second time derivative of φ is small enough to allow this slow-roll condition to be maintained for a sufficient
period. Thus, inflation requires
φ̇2  V (φ) (1.9.15)
|φ̈|  |3H φ̇|, |V,φ | . (1.9.16)
Satisfying these conditions requires the smallness of two dimensionless quantities known as slow-roll parameters
 2
1 V,φ
V (φ) ≡ (1.9.17)
2 V
V,φφ
ηV (φ) ≡ . (1.9.18)
V
In the slow-roll regime
V , |ηV |  1 (1.9.19)
the background evolution is
1
H2 ≈ V (φ) ≈ const , (1.9.20)
3
V,φ
φ̇ ≈ − , (1.9.21)
3H
 51

and the spacetime is approximately de Sitter

a(t) ∼ eHt , H ≈ const . (1.9.22)

EXERCISE: To understand the relation between the slow-roll condition (1.9.19) and inflation convince yourself that
the following is true
!
ä Ḣ
= H 1 + 2 ≈ H 2 (1 − V ) .
2
(1.9.23)
a H

Consequently, if the slow-roll approximation is valid (V  1), then inflation is guaranteed. However, this condition
is sufficient but not necessary since the validity of the slow-roll approximation was required to establish the second
equality in (1.9.23).
Inflation ends when the slow-roll conditions (1.9.19) are violated4

V (φend ) ≈ 1 . (1.9.24)

The number of e-folds before inflation ends is

aend
N (φ) ≡ ln
a
Z tend
= Hdt
t
φ
V
Z
≈ dφ . (1.9.25)
φend V,φ

D. What is the Physics of Inflation?

The inflationary proposal requires a huge extrapolation of the known laws of physics. It is therefore not surprising
that the physics governing this phase of superluminal expansion is still very uncertain. In the absence of a complete
theory standard practice has been a phenomenological approach, where an effective potential V (φ) is postulated.
Ultimately, V (φ) has to be derived from a fundamental theory.
Understanding the (micro)physics of inflation remains one of the most important open problems in modern cos-
mology and theoretical physics. Explicit particle physics models of inflation remain elusive. A natural microscopic
explanation for inflation has yet to be uncovered. Nevertheless, there have recently been interesting efforts to derive
inflation from string theory. Inflation in string theory is still in its infancy, but it seems clear that our understanding
of inflation will benefit greatly from a better understanding of moduli stabilization and supersymmetry breaking in
string theory. Hopefully, this will give some insights into which models of inflation are microscopically viable, mean-
ing that they can be derived from explicit string compactifications. Given the prospect of explicit and controllable
models of inflation in string theory, one is led to ask whether these theories have specific observational signatures.
In particular, it will be interesting to explore whether there are predictions that while non-generic in effective field
theory, may still have a well-motivated microscopic origin in string theory.

4 This can be made exact by the use of Hubble slow-roll parameters instead of the potential slow-roll parameters we introduced here (see
§ IX.E).
 52

E. Slow-Roll Inflation in the Hamilton-Jacobi Approach (NON–EXAMINABLE)

−2
In the following discussion, for notational convenience we will set 8πG ≡ MPl ≡ 1.

1. Hamilton-Jacobi Formalism

The Hamilton-Jacobi approach treats the Hubble expansion rate H(φ) = H/a as the fundamental quantity, con-
sidered as a function of time. Consider
H0 − a1 (H2 − H0 ) φ0
H,φ = = = − , (1.9.26)
φ0 φ0 2a

where we used H2 − H0 = a2 (ρ + P )/2 = (φ0 )2 /2 and primes are derivatives with respect to conformal time. This
gives the master equation

dφ φ0
= = −2H,φ . (1.9.27)
dt a

This allows us to rewrite the Friedmann equation

"   #
2
2 1 1 dφ
H = + V (φ) (1.9.28)
3 2 dt

in the following way

3 1
[H,φ ]2 − H 2 = V (φ) . (1.9.29)
2 2

Notice the following important consequence of the Hamilton-Jacobi equation (1.9.29): For any specified function H(φ),
it produces a potential V (φ) which admits the given H(φ) as an exact inflationary solution. Integrating (1.9.27)
1 dφ
Z Z
dt = − (1.9.30)
2 H 0 (φ)

relates φ to proper time t. This enables us to obtain H(t), which can be integrated to give a(t). The Hamilton-Jacobi
formalism can be used to generate infinitely many inflationary models with exactly known analytic solutions for the
background expansion. Here we are more concerned with the fact that it allows an elegant and intuitive definition of
the slow-roll parameters.

2. Hubble Slow-Roll Parameters

During slow-roll inflation the background spacetime is approximately de Sitter. Any deviation of the background
equation of state

P (φ0 )2 /2a2 − V
w= = 0 2
ρ (φ ) /2a2 + V
from the perfect de Sitter limit w = −1 may be defined by the parameter

3
H ≡ (1 + w) . (1.9.31)
2

We can express the Friedmann equations

1 2
H2 = a ρ (1.9.32)
3
1
H0 = − a2 (ρ + 3P ) (1.9.33)
6
 53

in terms of H
1 (φ0 )2
H2 = (1.9.34)
3 H
H0 = H2 (1 − H ) . (1.9.35)
Hence,
H0 d(H −1 )
H = 1 − = . (1.9.36)
H2 dt
Note that this can be interpreted as the rate ot change of the Hubble parameter during inflation H with respect to
the number of e-foldings dN = Hdt = − 12 H(φ)
H,φ dφ

 2
d[ln H] H,φ
H =− =2 . (1.9.37)
dN H

Analogously we define the second slow-roll parameter as the rate of change of H,φ

d[ln |H,φ |] H,φφ

ηH = − =2 (1.9.38)
dN H
Using the Hamilton-Jacobi master equation (1.9.27) this is also
d[ln |dφ/dt|]
ηH = . (1.9.39)
dN

3. Slow-Roll Inflation

By definition, slow-roll corresponds to a regime where all dynamical characteristics of the universe, measured in
physical (proper) units, change little over a single e-folding of expansion. This ensures that the primordial perturba-
tions are generated with approximately equal power on all scales, leading to a scale-invariant perturbation spectrum.
Since H and ηH characterize the rate of change of H and H,φ with e-foldings, slow-roll is naturally defined by
H  1 (1.9.40)
|ηH |  1 . (1.9.41)
The first slow-roll condition implies
1 (φ0 )2
H  1 ⇒ H2 =  (φ0 )2 , (1.9.42)
3 H
so that the slow-roll limit of the first Friedmann equation is
1 2
H2 ≈ a V. (1.9.43)
3
The second slow-roll condition implies
d[ln |dφ/dt|] 1 d2 φ
ηH = = 1 ⇒ |d2 φ/dt2 |  H|dφ/dt| (1.9.44)
dN H|dφ/dt| dt2
so that the Klein-Gordon equation reduces to
a2 V 0
φ̇ ≈ −
. (1.9.45)
3H
In section IX.C we defined a second set of common slow-roll parameters in terms of the local shape of the potential
V (φ)
 2
1 V,φ
V ≡ (1.9.46)
2 V
V,φφ
ηV ≡ . (1.9.47)
V
 54

H (φend ) ≡ 1 is an exact definition of the end of inflation, while V (φend ) = 1 is only an approximation. In the
slow-roll regime the following relations hold

H ≈ V (1.9.48)
η H ≈ η V − V . (1.9.49)