Cosmology

V.A. Rubakov
Institute for Nuclear Research of the Russian Academy of Sciences, 117312 Moscow, Russia

Abstract
These lectures aim at a pedagogical introduction to cosmology from the view-
point of a high-energy physicist. After introducing some theoretical back-
ground, a brief overview of cosmological data is presented, and the gross fea-
tures of the late and present Universe are outlined. Then the main properties
of the hot stage of the cosmological evolution are discussed, and inflationary
theory with its predictions for cosmological perturbations is described.

1 Introduction
The evolution of the Universe is determined, to a large extent, by microscopic laws of physics—the same
laws that govern particle interactions at high energies. Hence, discoveries in particle physics are of direct
relevance to the theory of the Universe. Conversely, cosmology provides important insights for high-
energy physics. Amazingly, many fundamental aspects of cosmology require dramatic extensions beyond
known physics; it may even happen that some aspects of cosmology will be possible to understand only
by invoking hints from string theory.
How well do we understand the present and the early Universe? Why are cosmologists so confident
when inventing new physics, even though it has not been discovered yet by the high-energy physics
community? What lessons should high-energy physicists learn from advances in cosmology? These
lectures are an attempt to address these issues, which are at the core of high-energy physicists’ interest
in cosmology.
There are many excellent books and reviews on cosmology and astrophysics. Here we mention a
few [1–12]. When discussing specific subjects in these lectures, we shall mostly refer to review papers;
an interested reader may find references to original literature there.

2 Expanding universe
2.1 Friedmann–Robertson–Walker metric
Two basic facts about the observable part of the Universe are that it is homogeneous and isotropic at
large spatial scales, and that it expands.
There are three types of homogeneous and isotropic three-dimensional spaces, which are conven-
tionally labelled by a parameter κ = 0, +1, −1. These are 1 three-sphere (κ = 1), flat space (κ = 0) and
three-hyperboloid (κ = −1). Were our space a three-sphere, the sum of angles of a triangle would be
greater than π; for a three-hyperboloid it would be smaller than π, while for a flat three-space it would
be equal to π. Accordingly, one speaks about a closed, a flat, and an open Universe (κ = +1, 0 and −1,
respectively); in the latter two cases the spatial size of the Universe is infinite, whereas in the former the
Universe is compact.
The homogeneity and isotropy of the Universe mean that its hypersurfaces of constant time are
either three-spheres or three-planes or three-hyperboloids. The distances between points may (and in
fact, do) depend on time, i.e., the interval has the form
ds2 = dt2 − a2 (t)dx2 (1)
1
Strictly speaking, this statement is valid only locally: in principle, a homogeneous and isotropic Universe may have
complex global properties. As an example, a spatially flat Universe may have the topology of a three-torus. There is some
discussion of such a possibility in the literature, and fairly strong limits have been obtained by the analyses of cosmic microwave
background [13].

197
 V. A. RUBAKOV

where dx2 is the distance on unit three-sphere/plane/hyperboloid. It is straightforward to see that a point
mass (e.g., a galaxy) put at some fixed value of x at zero velocity will stay at the same x. For this
reason the coordinates x are often called comoving coordinates. The physical distance between two such
neighbouring points at a given moment of time is
dl2 = a2 dx2 (2)
and thus it depends on time. Note that time t is the proper time of an observer staying at fixed x; this is
physical ‘cosmic time’. To an excellent approximation it is this time that we measure by our watches.
Metric (1) is usually called the Friedmann–Robertson–Walker (FRW) metric, and a(t) is called
the scale factor. In our Universe
da
ȧ ≡ >0
dt
which means that the distance (2) between points of fixed spatial coordinates x grows. The Universe
expands.
As the Universe expands, non-relativistic objects lose their velocities ẋ, i.e., they get frozen in the
comoving coordinate frame. Today the galaxies, to a good approximation, stay at certain fixed values of
x, so the physical distance between them grows according to (2).

2.2 Redshift and the Hubble law

Let us discuss the propagation of photons in an expanding Universe. Instead of dealing with a vector
field, let us consider a massless scalar field; this study will be used in what follows for other purposes,
while the main results are valid for photons and other massless particles. We are going to consider a
spatially flat Universe (κ = 0); it is clear that the results will be valid for a closed and an open Universe
as well, provided that the radius of spatial curvature is much greater than the wavelengths of the waves
of interest (this is certainly the case in the present Universe for photons of all relevant wavelengths).
To begin with, let us find out the action for the massless scalar field ϕ(t, x) in the expanding
Universe. In Minkowski space–time, the action is
Z  
3 1 2 1 2
SMink = dt d x (ϕ̇) − (∇i ϕ)
2 2
where the dot denotes a time-derivative, and ∇ i is a spatial gradient (hereafter i = 1, 2, 3 is a spatial
index). To generalize this action to the expanding background (1), we note that the physical spatial
gradient is
∂ϕ
∇=
∂l
where l is a vector of the physical distance (2), so that
1 ∂ϕ
∇i ϕ =
a(t) ∂xi
1
≡ ∂i ϕ .
a(t)
Another modification is that the physical volume is
d3 l = a3 (t)d3 x .
Thus, we arrive at the following expression for the action in the expanding Universe 2
Z  
3 3 1 2 1 2
S= dt d x a (t) (ϕ̇) − 2 (∂i ϕ) . (3)
2 2a (t)
2
TheR generalization of the action of massless scalar field to an arbitrary geometry described by metric ds 2 = gµν dxµ dxν
√
is S = d3 x −g 12 g µν ∂µ ϕ∂ν ϕ. The expression (3) specifies the latter formula to the particular metric (1).

2
198
C OSMOLOGY

By varying this action with respect to ϕ we obtain the field equation, generalizing the Klein–Gordon
equation to the expanding background,
ȧ 1
ϕ̈ + 3 ϕ̇ − 2 ∂i ∂i ϕ = 0 . (4)
a a
We note in the first place that solutions to this equation are collections of plane waves
i
ϕ(t, x) = ϕk (t)e−iki x (5)

where the coordinate wave vector (coordinate momentum) k is independent of time. The coordinate
wavelength 2π/|k| stays constant, but the physical wavelength
2π
λ(t) = a(t)
k
increases. The physical momentum
k
p(t) =
a(t)
decreases in time, i.e., the waves get red-shifted. The mode ϕ k (t) obeys the equation

ȧ k2
ϕ̈k + 3 ϕ̇k − 2 ϕk = 0 .
a a
The WKB solution to this equation is
const i R ω(t)dt
ϕk (t) = e (6)
a(t)
where
k
ω(t) =
a(t)
is the physical frequency, which decreases in time in the same way as the physical momentum. All this is
in accord with an intuitive understanding of the expansion of space: as it expands, all lengths, including
wavelengths of photons, increase.
We shall always label the present values of time-dependent quantities by the subscript 0: the
present physical wavelength of a photon is thus denoted by λ 0 , the present time is t0 , the present value
of the scale factor is a0 ≡ a(t0 ), etc. If a photon was emitted at some moment of time t e in the past, and
its physical wavelength at the moment of emission was λ e (λe is fixed by physics of the source; say, it is
the wavelength of a photon emitted by an excited hydrogen atom), then we receive today a photon whose
physical wavelength is longer,
λ0 a0
≡1+z = .
λe a(te )
Here we introduced the redshift z which, on the one hand, is directly measurable 3 , and, on the other hand,
is related to the time of emission, and hence to the distance to the source. To summarize, the relation
between redshift and scale factor is
a0
1 + z(t) = . (7)
a(t)
It is worth noting that in a spatially flat Universe, the scale factor does not have absolute meaning: at a
given moment of time it may be set equal to 1 by rescaling the spatial coordinates. What is physically
meaningful is the ratio of scale factors at different times; only this ratio enters Eq. (7), as well as all
formulas for a flat Universe.
3
One identifies a series of emission or absorption lines, thus obtaining λe , and measures their actual wavelength λ0 . These
spectroscopic measurements give very accurate values of z even for distant sources.

3
199
 V. A. RUBAKOV

Let us consider a ‘nearby’ source, for which

z 1.

This corresponds to relatively small (t 0 − te ). Expanding a(te ), one writes

a(te ) = a0 − ȧ(t0 )(t0 − te ) . (8)

To the leading order in z, the difference between the present time and the emission time is equal to the
distance to the source r (the speed of light is set equal to 1). Let us define the Hubble parameter

ȧ(t)
H(t) =
a(t)

and denote its present value by H0 . Then Eq. (8) takes the form

a(te ) = a0 (1 − H0 r)

and we get for the redshift, again to the leading non-trivial order in z,
1
1+z = = 1 + H0 r .
1 − H0 r
In this way we obtain the Hubble law,

z = H0 r , z1. (9)

Traditionally, one tends to interpret the expansion of the Universe as a runaway of galaxies from each
other, and the redshift as the Doppler effect. Then at small z one writes z = v, where v is the radial
velocity of the source with respect to the Earth, so H 0 is traditionally measured in ‘velocity per distance’
units. Observational data, which we shall discuss briefly in Section 3, give
km
H0 ≈ 70 (10)
s · Mpc
with about five per cent precision. Traditionally, the present value of the Hubble parameter is written as
km
H0 = h · 100 (11)
s · Mpc

(1 Mpc ≈ 3 mln. light yrs. ≈ 3 · 1024 cm). Thus

h ≈ 0.7 .

We shall use this value in further estimates.

Let us point out that the interpretation of the redshift in terms of the Doppler shift is actually not
adequate, at least for large enough z. In fact, there is no need in this interpretation at all: the ‘radial
velocity’ enters neither theory nor observations, so this notion may be safely dropped. A physically
meaningful quantity is the redshift z itself.
A final comment is that H0−1 has dimension of time, or length. Clearly, this quantity sets the
cosmological scales of time and distance at the present epoch,

H0−1 ≈ 14 · 109 yrs ≈ 4300 Mpc .

We shall discuss this point further in Section 2.5.

4
200
C OSMOLOGY

2.3 Hot Universe

Our Universe is filled with a cosmic microwave background (CMB). The cosmic microwave background
as observed today consists of photons with an excellent black-body spectrum of temperature

T0 = 2.725 ± 0.001 K . (12)

The spectrum has been precisely measured by various instruments and does not show any deviation from
the Planck spectrum. Using the Planck distribution, one calculates all properties of the photon gas; as an
example, the present number density of CMB photons is
2ζ(3) 1
nγ,0 = · T03 = 410 . (13)
π 2 cm3
The entropy density in the present Universe is of the same order of magnitude. The present energy
density of the CMB photons is
π2 GeV
ργ,0 = · T04 = 2.7 · 10−10 . (14)
15 cm3
Thus the present Universe is ‘warm’. The earlier Universe was warmer; it cooled down because of the
expansion. While the CMB photons freely propagate today, it was not so at the early stage. When
the Universe was hot, the usual matter (electrons and protons with a rather small admixture of light
nuclei) was in the plasma phase. At that time electromagnetic interactions between electrons, photons
and protons in the plasma were strong, so all these particles were in thermal equilibrium. As the Universe
cooled down, electrons ‘recombined’ with protons into neutral hydrogen atoms, and the Universe became
transparent to photons. The temperature scale of recombination is, very crudely speaking, determined
by the ionization energy of hydrogen, which is of order 10 eV. In fact, recombination occurred at lower
temperature4 ,
Trec ≈ 3000 K .
An important point is that the recombination process lasted quite a bit less than the Hubble time at that
epoch; to a reasonable approximation, recombination occurred instantaneously.
Another point is that even though after recombination photons were no longer in thermal equilib-
rium with anything, the shape of the photon distribution function did not change, except for the overall
redshift. Indeed, the thermal distribution function for ultra-relativistic particles, the Planck distribution,
depends only on the ratio of frequency to temperature,
ω 
p
fPlanck (p, T ) = f , ωp = |p| .
T
As the Universe expands, the frequency gets red-shifted, ω p → ωp /(1 + z), but the shape of the spectrum
remains Planckian, with temperature T /(1 + z). Hence, the Planckian form of the observed spectrum is
no surprise. Generally speaking, this property does not hold for massive particles 5 .
At even earlier times, the temperature of the Universe was even higher. The earliest time which
has been observationally probed to date is the Big Bang Nucleosynthesis epoch (which we shall briefly
discuss below), and corresponds to a temperature of order 1 MeV.
4
The reason is that the number density of electrons and protons was small compared to the number density of photons, and
that the thermal velocities of electrons and protons were small. At temperature T considerably smaller than 10 eV, photons
from the tail of thermal distribution still disintegrated hydrogen atoms into electrons and protons, while an electron had to travel
for a relatively long time to find a proton to recombine with. It was only at T = Trec that the rate of disintegration became small
enough.
5
A similar property holds, however, for particles that decouple being non-relativistic (hydrogen, cold dark matter). At de-
coupling, they have a Maxwell–Boltzmann distribution function, which is a function of the ratio p2 /(2mT ). As the momentum
gets red-shifted, p → p/(1+z), the shape of this distribution function remains Maxwell–Boltzmann, with effective temperature
T /(1 + z)2 . Incidentally, this explains the fact, mentioned above, that non-relativistic free massive particles get frozen into the
comoving frame.

5
201
 V. A. RUBAKOV

To summarize, the effective temperature of photons scales as

1
T (t) ∝ (15)
a(t)

that is
T 1
= .
T0 1+z
This behaviour is characteristic of ultra-relativistic free species (at zero chemical potential) only. The
same formula is valid for ultra-relativistic particles (at zero chemical potential) which are in thermal equi-
librium. Thermal equilibrium means adiabatic expansion; during adiabatic expansion, the temperature
of ultra-relativistic gas scales as the inverse size of the system, hence Eq. (15).
Both for free photons and for photons in thermal equilibrium, the number density behaves as
follows,
1
nγ = const · T 3 ∝ 3
a
and the energy density is
π2 1
ργ = · 2 · T4 ∝ 4 (16)
30 a
where the factor 2 accounts for two photon polarizations.
Let us now turn to non-relativistic particles: baryons, massive neutrinos, possible ‘dark matter’
species, etc. If they are not destroyed during the evolution of the Universe (that is, they are stable and do
not co-annihilate), their number density merely gets diluted,
1
n∝ . (17)
a3
This means, in particular, that the baryon-to-photon ratio stays constant,
nB
η≡ = const . (18)
nγ

The energy density of non-relativistic particles scales as

1
ρ(t) = m · n(t) ∝ (19)
a3 (t)

in contrast with more rapid fall off (16) characteristic to ultra-relativistic species.
As we shall discuss later, there exists strong evidence for dark energy in the Universe, whose
density does not decrease in time as fast as in Eqs. (16) or (19). For the moment suffice it to mention that
this property holds for vacuum, whose energy density stays constant (in a locally Lorentz frame),

ρvac = const (20)

while the vacuum energy-momentum tensor in the arbitrary frame is, by general covariance,
vac
Tµν = ρvac gµν ,

i.e., the vacuum has negative pressure [see Eq. (24)],

pvac = −ρvac . (21)

In this context, the vacuum energy density is the same thing as the cosmological constant, or Λ term.

6
202
C OSMOLOGY

2.4 Friedmann equation

The basic equation governing the expansion rate of the Universe is the Friedmann equation
 2
2 ȧ 8π κ
H ≡ = Gρ − 2 (22)
a 3 a
where the dot again denotes a derivative with respect to time t, ρ is the total energy density in the
Universe, the parameter κ has been introduced in Section 2.1 and distinguishes closed (κ = +1), flat
(κ = 0) and open (κ = −1) Universes, and G is Newton’s gravity constant. In natural units
G = MPl−2
where
MPl = 1.2 · 1019 GeV
is the Planck mass. The first and second terms on the right-hand side of Eq. (22) may be viewed as the
contributions of all kinds of matter/energy and spatial curvature, respectively, to the expansion rate. As
we shall discuss in Section 3, observational data show that the curvature term is very close to zero today,
1/a20
< 0.02 , (23)
H02
i.e., it currently makes up not more than 2% of the right-hand side of the Friedmann equation. In fact,
inflationary theory suggests that the spatial curvature vanishes to an excellent approximation.
The Friedmann equation (22) is nothing but one of the Einstein equations of General Relativity
specialized to homogeneous and isotropic space. It has to be supplemented by the law of evolution for
the energy density ρ. To obtain the latter, we use the first law of thermodynamics, and take into account
the fact that the expansion of the Universe is slow, and hence adiabatic, at interesting temperatures. We
apply the first law to a comoving volume V = a 3 and write, using entropy conservation in a comoving
volume,
d(ρV ) = −pdV
where p is pressure. We obtain
dρ · V = −(p + ρ)dV
and get finally6
dρ da
= −3 . (25)
ρ+p a
To close the system of equations, one needs also an equation of state
p = p(ρ) .
The latter equation is governed by the composition of the Universe at a given moment of time.
Other Einstein equations are satisfied automatically for this simple geometry. The relations (16),
(19) and (20) may in fact be viewed as consequences of Eq. (25), since p = ρ/3, p = 0 and p = −ρ for
relativistic matter, non-relativistic particles and vacuum, respectively.
As we pointed out above, the Universe is spatially flat today to a good approximation; the curvature
term κ/a2 in Eq. (22) today is less than two per cent of the matter term. This can also be phrased in the
following way. One defines the critical density ρ c according to
8π
Gρc = H02 . (26)
3
6
This equation is in fact a consequence of the covariant conservation of energy-momentum,∇µ T µν = 0. For the energy-
momentum tensor of a fluid one has
Tνµ = diag (ρ, −p, −p, −p) . (24)
In the background FRW metric the covariant conservation law reduces precisely to Eq. (25).

7
203
 V. A. RUBAKOV

The meaning of this quantity is as follows. If the actual energy density ρ of all forms of matter in the
Universe (including vacuum, quintessence, etc.) is larger than ρ c , then κ = +1, and the Universe is
closed; if ρ < ρc , the Universe is open, and it is flat for ρ = ρ c . From (23) we see that observationally,

ρ = (1 ± 0.02)ρc . (27)

At earlier times, the curvature term was even less significant, so we shall neglect it in the study of the
evolution of the Universe, and write the Friedmann equation as
 2
ȧ 8π
= Gρ . (28)
a 3

To get an idea of numerics, one plugs the present value of the Hubble parameter (11) into the definition
(26) and converts ρc into
GeV GeV
ρc = h2 · 1 · 10−5 ≈ 5 · 10−6 (29)
cm3 cm3
where we use our fiducial value h = 0.7.

2.5 Sample solutions

Solutions to the Friedmann equation (28) are most easily obtained in cases when matter of a definite type
gives a dominant contribution to the energy density ρ. Let us present some solutions of this sort.

2.5.1 Matter-dominated Universe

If the dominant contribution to the energy density comes from non-relativistic particles, then p = 0, and
from (25) one finds
ρ = const · a−3 .
Making use of this relation, one finds that the Friedmann equation reads
ȧ const
= 3/2 .
a a
The solution is
a = const · t2/3 . (30)
Note that this solution describes a decelerating Universe,

ä < 0 .

As t → 0, the scale factor tends to zero, and the energy density tends to infinity. This is a cosmological
singularity, ‘beginning of the Universe’ (‘Big Bang’), and t is the lifetime of the Universe. Note that the
lifetime is related to the Hubble parameter, since
ȧ 2
H≡ = .
a 3t
Until fairly recently, our Universe was indeed matter dominated, so this relation may be used to obtain a
crude estimate of its present age,
2
t0 ∼ H0−1 ≈ 1 · 1010 yrs
3
(again with h = 0.7). In fact, the lifetime of 10 billion years was a bit of a problem about 10 years ago,
as independent estimates of lifetimes of old objects in our Universe suggested that their lifetimes were
close to, and sometimes even larger than 10 billion years. This was one of the reasons for suggesting,

8
204
C OSMOLOGY

even before the observational evidence for an accelerating Universe, that our Universe is not in the
matter-dominated stage today—rather, it is in a dark-energy-dominated stage.
Let us make use of this solution to introduce another notion, the cosmological horizon. Suppose
that at t = 0, signals were emitted everywhere in space, and then propagated in the Universe with the
speed of light. We ask at what distance today are the sources of signals we receive now. This sphere is
precisely the cosmological horizon for the solution (30): the interior of this sphere is causally connected
to us, while the part of the Universe outside this sphere is causally disconnected from us. The world line
of a signal propagating with the speed of light obeys ds = 0, that is along this world line

dt = a(t)dr .

This implies that the coordinate distance to the horizon is

Z t
dt0
rH = 0
.
0 a(t )

The physical distance to the horizon at the time t is thus

Z t
dt0
lH (t) = a(t)rH = a(t) . (31)
0 a(t0 )
For the solution (30) this distance is finite,

lH (t) = 3t = 2H −1 .

Hence the present size of the visible part of the Universe is estimated as

lH,0 ∼ 2H0−1 ≈ 3 · 1028 cm ≈ 104 Mpc .

Note that in a matter-dominated Universe, the integral in Eq. (31) is saturated at large t 0 , so the relic
photons, which were actually emitted somewhat after the Big Bang, travelled almost the same distance
as lH,0 .
Let us stress that the above estimates for the present age and horizon size are not quite correct,
since during a good part of the evolution, the expansion of the Universe was not dominated by non-
relativistic matter; rather, it was dark-energy dominated. Taking that into account, the present lifetime of
the Universe is
t0 = 14 · 109 yrs .
We do not need the refined estimate of the present size of the horizon.

2.5.2 Radiation-dominated Universe

An elementary calculation shows that the equation of state for ultra-relativistic matter (not necessarily in
thermal equilibrium) is
1
p= ρ.
3
If the dominant contribution to the energy density comes from ultra-relativistic particles, then from (25)
it follows that [see also Eq. (16)]
ρ = const · a−4
and the solution to the Friedmann equation (28) is

a(t) = const · t1/2 . (32)

Qualitative features of this solution are similar to the matter-dominated case: the Universe starts from
the singularity, its expansion decelerates, age and horizon size are finite at given t.

9
205
 V. A. RUBAKOV

2.5.3 Vacuum-dominated Universe

A qualitatively different solution occurs if the Universe is vacuum dominated. The vacuum energy
density ρvac is time-independent, so the solution to the Friedmann equation (28) is

a(t) = const · eHvac t (33)

where the time-independent Hubble parameter is determined by the vacuum energy density,
r
8π
Hvac = Gρvac .
3
One notices that the Universe accelerates, rather than decelerates,

ä > 0 .

We note in passing that a spatially flat FRW metric with scale factor (33) describes (part of) the de Sitter
space–time. Unlike other cosmological solutions, de Sitter geometry does not have a past singularity.

2.5.4 Equation of state p = wρ

Let us now consider the general case of a fluid with the equation of state

p = wρ

where w is a constant. For definiteness, let us restrict ourselves to the case

w > −1 .

Then the solution to Eq. (25) is

const
ρ= .
a3(1+w)
With κ = 0 (spatially flat Universe) one finds from Eq. (22)

a = const · tα (34)

where
2 1
α= .
31+w
The behaviour of solutions is qualitatively different for w > −1/3 and w < −1/3, i.e., for α < 1 and
α > 1:
1
w>− : ä < 0 , decelerated expansion
3
1
w<− : ä > 0 , accelerated expansion . (35)
3
Thus, accelerated expansion of the Universe requires negative pressure.
It is worth noting that the two cases differ in another respect: in the former case there exists a
cosmological horizon, while in the latter the entire Universe is causally connected. Indeed, we have seen
that the cosmological horizon exists, if the following integral converges [see Eq. (31)],
Z t
dt0
0
.
0 a(t )

This integral is convergent for α < 1, i.e., w > −1/3. Otherwise this integral diverges, so the cosmo-
logical horizon is absent. Note that this observation has to do with early times, t → 0; it is of relevance
for inflation rather than for the present epoch.

10
206
C OSMOLOGY

2.6 Changing regimes

As we shall discuss in Sections 3 and 4, the present expansion of the Universe is dark-energy dominated;
to a reasonable approximation, the expansion of the Universe today follows the exponential law (33).
This was not so at earlier times. Indeed, the dark energy density is constant in time (or almost constant),
while the energy density of non-relativistic particles scales like a −3 (t). Hence, even though the present
energy density of non-relativistic matter is smaller than dark energy, non-relativistic matter dominated
at an earlier epoch, when the scale factor was smaller than today. Even earlier, the expansion of the
Universe was dominated by ultra-relativistic matter. Indeed, the energy density of the latter scales as
a−4 (t), so at small enough a(t) it was higher than the energy density of non-relativistic particles. The
curvature term κ/a2 never dominated the expansion of the Universe: at present it contributes much less
than not only the dark energy density, but also the matter energy density. Its contribution was even less
significant at earlier times, as it scales as a −2 , whereas the matter energy density scales as a −3 . This is
why it is legitimate to neglect the curvature term for describing the entire evolution of the Universe.
We shall quantify this discussion later, when we have a better idea of the composition of the
Universe.

3 Overview of cosmological data

In the last 10 to 15 years, cosmology has become a qualitative science. Detailed data on the present and
earlier Universe are now available, and even more precise data are due to come. Before going into further
theoretical discussions, let us briefly consider what kinds of data there are, and what gross features of the
Universe they show.

3.1 Distribution of luminous matter

Distribution of luminous matter (galaxies, quasars) is obtained by deep surveys. The largest surveys (2dF,
SDSS) measure angular positions of, and distances (redshifts) to hundreds of thousands of galaxies, with
a depth of the order of 2000 Mpc (about 6 billion light years). This is a fairly large portion of the present
Universe. An even larger part of the Universe is sampled by quasars. Thus, by now we have a ‘map’ of
our ‘neighbourhood’, and can discuss matter distribution on various length scales.
At large scales, the Universe is homogeneous and isotropic, as illustrated in Fig. 1.

Fig. 1: Spatial distribution of galaxies (left panel) and quasars (right panel), according to SDSS survey [14]. Shown
are samples of usual and brighter galaxies and quasars. The parameter h is defined in the text.

11
207
 V. A. RUBAKOV

At shorter scales, the Universe is of course inhomogeneous. The largest structures visible (su-
perclusters of galaxies, giant voids) extend to several dozens of Mpc. The comparison of the observed
structure to simulations, at scales ranging from a few kpc (size of a galaxy) and smaller, to thousands of
Mpc, tells a lot about the primordial density perturbations in the early Universe, the composition of the
Universe, and the history of its expansion.

3.2 Observation of ‘standard candles’

These are the objects whose absolute luminosity is believed to be known. What is measured is the visible
luminosity F and redshift z. At relatively short distances, the visible luminosity is related to the distance,
L
F = (36)
4πr 2
where L is the absolute luminosity. This relation becomes confusing for z ∼ 1, yet one often defines the
luminosity distance by Eq. (36), and talks about the redshift–distance relation (the notion of velocity is
usually not used for large z).
At z  1, the relation between z and r follows a linear Hubble law (9), as shown in Fig. 2. The
present value of the Hubble parameter consistent with virtually all measurements is
km
H0 = (73 ± 3) . (37)
s · Mpc

40000
Velocity (km/sec)

30000

20000

10000

0
0 100 200 300 400 500 600 700
Distance (Mpc)

Fig. 2: Hubble diagram for Supernovae 1a as standard candles, see Refs. [7, 15]. The straight line corresponds to
the Hubble law.

At large z, the linear relation (9) no longer holds. The relation between z and r tells us about the
expansion rate of the Universe at a relatively late epoch. Currently, the data for large z come from the
observations of type 1a supernovae (SNe 1a), Figs. 3 and 4. Surprisingly, they show that the Universe
undergoes accelerated expansion today (and at relatively small z, i.e., at late times), while at higher
redshift the expansion was decelerating. We shall discuss the significance of this result shortly.

3.3 Cosmic-microwave-background radiation

The cosmic microwave background (CMB) is an extremely important source of information about the
properties of the earlier Universe.

12
208
C OSMOLOGY

44
MLCS
42

m-M (mag)
40

38
ΩM=0.28, ΩΛ=0.72
36 ΩM=0.20, ΩΛ=0.00

34 ΩM=1.00, ΩΛ=0.00

1.0

0.5
∆(m-M) (mag)

0.0

-0.5
Preliminary
-1.0
0.100.01 1.00
z
Fig. 3: Upper panel: stellar magnitude (proportional to − log(visible luminosity)) of SNe 1a as a function of
redshift, according to relatively old data [16]. The larger the stellar magnitude, the dimmer the object. For a
definition of Ω’s see the text. Lower panel: the same figure, with expectation from the model Ω M = 0.2, ΩΛ = 0
subtracted.

1.0

0.5
∆(m-M) (mag)

0.0

-0.5
Ground Discovered
HST Discovered
-1.0

1.0
)
q(z)=q0+z(dq/dz) , dq/dz=0 (j0=0
eleration, q0=-
0.5 Constant Acc
∆(m-M) (mag)

0.0
Constant Deceleration
Coasting, q(z)=0 , q0=+, dq/dz=0 (j =0)
-0.5 0
Acceleration+Deceleration, q0=-, dq/dz=++
Acceleration+Jerk, q0=-, j0=++
-1.0

0.0 0.5 1.0 1.5 2.0

z

Fig. 4: More recent data [17] on SNe 1a; subtracted is the expectation from a model in which the Universe expands
at constant velocity, ȧ = const

13
209
 V. A. RUBAKOV

The CMB photons were last scattered/emitted at the recombination epoch, when the Universe was
only about 3 · 105 years old; for comparison, the present age of the Universe is about 1.4 · 10 10 years. The
observations of CMB give the photographic picture (literally) of the ‘young’ Universe, which had quite
different properties than it has today. One of these properties is a much higher level of homogeneity and
isotropy: photons coming from different directions in the sky have almost (but not exactly!) the same
temperature. Crudely speaking, relative angular anisotropy of CMB temperature, δT /T 0 , is of order
10−5 . This means that the Universe was homogeneous and isotropic at the level of 10 −5 when it was
300 thousand years old.
Yet the angular anisotropy of the CMB temperature exists, and has been measured at various
angular scales. In a wide range of angular scales this anisotropy was accurately measured by the WMAP
satellite, see Figs. 5–7, while at smaller angular scales data are available from measurements made by
ground-based interferometers.

Fig. 5: WMAP data [18]: temperature of photons coming from different directions in the sky. Darker regions
correspond to lower temperatures. The average temperature T 0 and the dipole component are subtracted. The
angular variation of the temperature is at the level of δT ∼ 100 µK, i.e., δT /T 0 ∼ 10−4 –10−5 .

It is convenient to decompose the temperature, as a function of the direction ~n, in spherical har-
monics Ylm (~n), which make a complete set of functions on a sphere. One writes 7
X
δT (~n) ≡ T (~n) − T0 − δTdipole = Clm Ylm (~n) .
l,m

The angular momentum l corresponds to fluctuations with typical angluar scale π/l.
Figures 6 and 7 show the measured anisotropy as a function of l.
To understand what is shown, one notices that the data are grossly consistent with Gaussian fluc-
tuations [19], for which Clm are statistically independent. For an isotropic Universe this means
hClm Cl∗0 m0 i = Cl2 δll0 δmm0 .
The coefficients Cl determine the correlation function
X 2l + 1
hδT (~n1 )δT (~n2 )i = Cl2 Pl (cos θ) ,
4π
l
7
The dipole component is due to the motion of the Earth with respect to the rest frame of the CMB photon gas.

14
210
C OSMOLOGY

Fig. 6: Measured angular anisotropy of cosmic microwave background [18, 20, 21]. Shown are data points from
the first-year WMAP observations and from ground-based observations by CBI and ACBAR.

Fig. 7: Recent three-year WMAP data [19], shown by solid data points, together with older one-year WMAP data
(light grey data points). Different lines correspond to slightly different best-fit cosmological models.

where Pl are the Legendre polynomials, and θ is the angle between ~n 1 and ~n2 . In particular, for mean
fluctuation of the temperature one has
X 2l + 1 Z
2 2 l(l + 1) 2
hδT i = Cl ≈ Cl d ln l .
4π 2π
l

Hence the quantity r

l(l + 1)
δT (l) ≡ Cl
2π
is a measure of the contribution coming from angular momenta in a decimal interval of l. It is this
quantity that is shown in Fig. 6, while Fig. 7 shows the square of this quantity.

15
211
 V. A. RUBAKOV

The CMB anisotropy, and also CMB polarization, are sensitive to most of the properties of the
Universe, such as the primordial spectrum of density perturbations, relic gravity waves, matter content
of the Universe, its expansion history, spatial geometry, etc.
It is clear from Fig. 7 that the anisotropy as a function of l has a fairly complex behaviour. There
are peaks (at least two of them are clearly visible) and dips. The physics beyond these features is roughly
as follows. The CMB anisotropy has its origin in the density perturbations 8 , i.e., sound waves of all
possible wavelengths. Waves of longer wavelengths are seen in the photographic picture at larger angles;
this is the small-l region in Fig. 6. Conversely, the region of large l corresponds to shorter wavelengths of
the sound waves. The properties of these waves are different for different modes. The modes with long
wavelengths, and hence small frequencies had no time to oscillate by the time of recombination. Their
amplitudes are determined by the initial spectrum of density perturbations, which, as the data show, is in
some sense flat, i.e., independent of the wavelength. The effect of these modes on temperature is mostly
due to the gravitational potential they produce (Sachs–Wolfe effect): roughly speaking, light from denser
regions with stronger negative gravitational potentials has to climb up the gravitational well, so it gets
more red-shifted as compared to light from less dense regions. This is the flat region of small l in Fig. 6.
For shorter waves there is another effect: the waves oscillate, i.e., the particles in plasma move; this
causes the Doppler effect leading to the CMB anisotropy. The waves of ‘just right’ frequency (and hence
wavelength) are in the phase of maximum motion at recombination. These are the waves seen at an angle
of 0.7 degrees (l ∼ 200), where Cl has the first peak. Further peaks correspond to higher harmonics.
All physical processes involved—the expansion of the Universe during its first 300 thousand years,
the evolution of density perturbations and the recombination itself—are very well understood, so the
calculations of the CMB anisotropy are very reliable. The predictions depend, of course, on a number of
parameters characterizing the early and present Universe, so the CMB data are used for extracting these
parameters.
Let us point out two specific results coming from the analysis of the CMB anisotropy. The first
one is that our Universe is spatially flat to rather high precision. The positions of the peaks in C l (in
particular, of the first peak measured with good accuracy) are sensitive to the spatial curvature: the abso-
lute wavelength of the corresponding sound waves is reliably calculable, while its angular size strongly
depends on whether space is a 3-sphere, 3-hyperboloid or 3-plane. Quantitatively, the result is usually
expressed in terms of the contribution of the curvature term to the right-hand side of the Friedmann equa-
tion (22): this contribution is less than about two per cent [we mentioned this already, see Eq. (27)]. In
more physical terms, this means that the radius of the spatial curvature of our Universe a is quite a bit
greater than the length of the visible part (size of the cosmological horizon) l H,0 ,

a > 4lH,0 . (38)

Another way to phrase this is to say that if the Universe were 3-sphere, its volume would still be a lot
larger than the volume we can observe,

Vtot 2π 2 a3
= 4 3 > 100 .
Vobs 3 πlH,0

Hence even if the Universe has finite volume, we know from observations that we are able to observe
not more than one per cent of it. It is worth stressing that Eq. (38) is an observational bound; it is likely
that the actual radius of the spatial curvature of the Universe is many orders of magnitude greater than
the horizon size.
The second result concerns the baryonic content of the Universe. The height of the second peak
in Cl is sensitive to the number density of electrons, which is equal to the number density of protons by
8
Another possible source is gravity waves; their effect, if any, is small, see Section 8.3.

16
212
C OSMOLOGY

electric neutrality. The analysis gives for the present number density of baryons

nB,0 = 2.5 · 10−7 cm−3 (39)

with precision better than 10 per cent. In terms of time-independent parameter η defined in Eq. (18) this
corresponds to
η = 6 · 10−10 . (40)
It is remarkable that the same value of this parameter comes from an entirely different set of observations,
which we are about to discuss in brief.
Before doing that, let us stress that the information extracted from CMB data is by no means
limited by these two examples. These data provide independent determinations of the present value of
the Hubble parameter, the rate of acceleration of the cosmological expansion, the power spectrum of the
primordial density perturbations, the amount of dark matter in the Universe, the bound on the primordial
gravitational waves, etc. It is important that these data be consistent with virtually all other cosmological
measurements.

3.4 Big Bang Nucleosynthesis

Theory of Big Bang Nucleosynthesis (BBN) and observations of primordial abundances of light elements
probe the earliest epoch of the evolution of the Universe, accessible to observations today. This epoch
corresponds to temperatures ranging from 1 MeV to (a few) · 10 keV, and the age of the Universe from
1 to 200 s. At temperatures above 1 MeV, there is thermal equilibrium with respect to reactions

p + e− ↔ n + νe . (41)

As the Universe cools down below T ≈ 1 MeV, neutrons are no longer produced or destroyed; their
concentration (relative to protons) ‘freezes out’. At temperatures of about 100 keV and somewhat lower,
these neutrons combine with protons into light nuclei, mostly 4 He, but also deuterium 2 H, lithium 7 Li
and others. These elements remain in the Universe, so their primordial abundances are measurable today.
The calculations of the thermonuclear reactions are again based on well-known physics, and the results
are sensitive to the only unknown parameter 9 η. The results of the calculations and data are shown in
Fig. 8.
It is worth pointing out that Big Bang Nucleosynthesis serves also as a source of constraints on
particle physics exotica. The very fact that the temperature of the Universe reached at least 1 MeV
or so, and that the expansion was described by known physics at that stage, constrains significantly
some extensions of the Standard Model, like models with large extra dimensions [23] (for a review see
Ref. [24]). More generally, constraints from BBN are important in models with stable or long-living
new particles: these are produced at earlier stages and may contribute too much to the energy density
at the nucleosynthesis epoch, modifying in this way the expansion rate, and hence the predictions of
the BBN theory. Another example are particles that decay at the BBN epoch: these may destroy thermal
equilibrium and therefore affect BBN; a situation of this sort occurs in some theories with light gravitinos.
Any extension of the Standard Model has to be checked against cosmology, in particular against Big Bang
Nucleosynthesis.
There are many other data of cosmological significance, like measurements of mass distributions
in galaxies and galactic clusters. We shall briefly discuss some of them in appropriate places, and now
proceed to immediate consequences.
9
Assuming the Standard Model particle content, see next paragraph.

17
213
 V. A. RUBAKOV

Baryon density Ωb h2
0.005 0.01 0.02 0.03
0.26
4He
0.25

0.24
Y
___ 0.23
D
H
0.22

10 −3
3He
___ CMB
H D/H p
10 − 4

3He/H
p
10 − 5

10 − 9

5
7Li/H
p
2

10 − 10

1 2 3 4 5 6 7 8 9 10
Baryon-to-photon ratio η10

Fig. 8: Big Bang Nucleosynthesis [22]: theoretical predictions (lines) versus observations (smaller boxes: 2σ
errors, statistical only; larger boxes: 2σ with systematic uncertainties included). The vertical lines are the results
from CMB anisotropy (the widths of bands depend on priors on other cosmological parameters).

4 Composition of the present Universe

As we shall discuss in this section, the cosmological data require a very surprising composition of the
Universe.
Before proceeding, let us introduce a notion traditional in the analysis of the composition of the
present Universe. For every type of matter i with the present energy density ρ i,0 , one defines a parameter
ρi,0
Ωi = . (42)
ρc
Then Eq. (27) tells us that X
Ωi = 1 ± 0.02 (43)
i

where the sum runs over all forms of energy. Let us now discuss contributions of different species to this
sum.
We begin with baryons. The result (39), together with (13) gives
GeV
ρB, 0 = mB · nB, 0 ≈ 2.5 · 10−7 . (44)
cm3
Comparing this result with the value of ρ c given in (29), one finds

ΩB = 0.05 . (45)

18
214
C OSMOLOGY

Thus baryons constitute a rather small fraction of the present energy density in the Universe. One point
to note is that most of the baryons in our Universe are dark: direct measurements of the mass density of
stars give an estimate
Ωstars ∼ 0.005
which is about an order of magnitude smaller than Ω B . There is nothing particularly dramatic about this
fact: baryons hide in dust and gas clouds (most likely), brown dwarfs and other non-luminous objects,
etc.
Photons contribute an even smaller fraction, as is clear from (14):

Ωγ ≈ 6 · 10−4 . (46)

From electric neutrality, the number density of electrons is about 10 the same as that of baryons, so
electrons contribute a negligible fraction to the total mass density. The remaining known stable particles
are neutrinos. Their number density is calculable in Hot Big Bang theory and these calculations are
nicely confirmed by Big Bang Nucleosynthesis. The number density of each type of neutrinos is
1
nνα = 115
cm3
where να = νe , νµ , ντ . A direct limit on the mass of the electron neutrino, m νe < 2.6 eV, together with the
observations of neutrino oscillations suggest that every type of neutrino has a mass smaller than 2.6 eV
(neutrinos with masses above 0.03 eV must be degenerate in mass, according to neutrino oscillation
data). The energy density of all types of neutrinos is thus smaller than ρ c :
X 1 GeV
ρν,total = mνα nνα < 3 · 2.6 eV · 115 ∼ 8 · 10−7
α
cm 3 cm3

which means
Ων,total < 0.16 .
This estimate does not make use of any cosmological data. In fact, cosmological observations give a
stronger bound
Ων,total < 0.01 . (47)
This bound is mostly due to the analysis of the structures at small length scales, and has to do with
streaming of neutrinos from the gravitational potential wells at early times when neutrinos were ultra-
relativistic. In terms of the neutrino masses the bound (47) reads [25]
X
mνα < 0.4 eV

so every neutrino must be lighter than 0.14 eV. On the other hand, atmospheric neutrino data and the
K2K experiment tell us that the mass of at least one neutrino must be larger than 0.04 eV. Comparing
these numbers, one sees that it may be feasible to measure neutrino masses by cosmological observations
(!).
Coming back to our main topic here, we conclude that not more than 6% of the energy density
in the present Universe is in the form of known particles; most energy in the present Universe must be
in ‘something unknown’. Furthermore, there is strong evidence that this ‘something unknown’ has two
components: clustered (dark matter) and unclustered (dark energy).
Clustered dark matter consists presumably of new stable massive particles. These make clumps
of energy density which count for much of the mass of galaxies and most of the mass of galactic clusters.
There are a number of ways of estimating the contribution of non-baryonic dark matter to the total energy
density of the Universe (see Ref. [26] for details):
10
There are neutrons, whose number is smaller than the number of protons by roughly a factor of 7.

19
215
 V. A. RUBAKOV

– The composition of the Universe affects the angular anisotropy of the cosmic microwave back-
ground. Quite accurate measurements of the CMB anisotropy, available today, enable one to esti-
mate the total mass density of dark matter.
– The composition of the Universe, and especially the density of non-baryonic dark matter, is crucial
for structure formation of the Universe. Comparison of the results of numerical simulations of
structure formation with observational data gives a reliable estimate of the mass density of non-
baryonic clustered dark matter.

The bottom line is that the non-relativistic component constitutes about 30 per cent of the total
present energy density, which means that non-baryonic ‘cold dark matter’ contributes

ΩCDM ≈ 0.25 . (48)

There is direct evidence that dark matter exists in the largest gravitationally bound objects—
clusters of galaxies. There are various methods of determining the gravitating mass of a cluster, and
even mass distribution in a cluster, which give consistent results. To name a few:

– One measures velocities of galaxies in galactic clusters, and makes use of the gravitational virial
theorem,
1
Kinetic energy = Potential energy .
2
In this way one obtains the gravitational potential, and thus the distribution of the total mass in a
cluster.
– Another measurement of masses of clusters makes use of intracluster gas. Its temperature obtained
from X-ray measurements is also related to the gravitational potential through the virial theorem.
– Fairly accurate reconstruction of mass distributions in clusters is obtained from the observations
of gravitational lensing of background galaxies by clusters.

These methods enable one to measure the mass-to-light ratio in clusters of galaxies. Assuming
that this ratio applies to all matter in the Universe 11 , one arrives at the estimate for the mass density of
clumped matter in the present Universe. Remarkably, this estimate coincides with (48).
Finally, dark matter exists also in galaxies. Its distribution is measured by the observations of
rotation velocities of distant stars and gas clouds around a galaxy. An example is shown in Fig. 9.
Thus cosmologists are confident that much of the energy density in our Universe consists of new
stable particles. We shall see that natural candidates are particles which participate in weak interactions
(or, more generally, particles whose annihilation cross-section is determined by a scale of the order of
the electroweak scale, MEW ∼ 100 GeV). Of course, this is only a hypothesis for the time being, and
there are many other candidates for dark matter species.
Non-baryonic clustered dark matter is not the whole story. Besides the known and dark matter
particles there is also unclustered dark energy, the most mysterious substance in the Universe. Making
use of the above estimates, one obtains an estimate for the energy density of all particles,

Ωγ + ΩB + Ων,total + ΩCDM ≈ 0.3 .

We note in passing that the contribution of photons and possible massless neutrinos is very small here,
so the left-hand side is the contribution of all non-relativistic matter, and is often denoted by Ω M . Thus

ΩM = ΩB + ΩCDM ≈ 0.3 .
11
This is a strong assumption, since only about 10 per cent of galaxies are in clusters.

20
216
C OSMOLOGY

Fig. 9: Rotation velocities of gas clouds for galaxy NGC 6503 [27]. Curves show contributions of different
components of the galaxy; the ‘halo’ is dark.

Equation (43) implies then that 70 per cent of the energy density is unclustered.
All this fits nicely with the observations of SNe 1a. Indeed, we have seen that neither relativistic,
nor non-relativistic matter can lead to the accelerated expansion of the Universe. So, the accelerated
expansion requires energy stored in something dramatically different from conventional particles. Fur-
thermore, we have seen that this ‘something’—dark energy—must have negative pressure. In fact the
analysis of the entire set of cosmological data [19,25] in terms of dark energy with the phenomenological
equation of state12
p = wρ , w = const
gives
ΩΛ = 0.72 ± 0.02
(here subscript Λ traditionally refers to dark energy) and, see Fig. 10,

−1.1 < w < −0.9 .

It is worth noting that the vacuum value, w = −1 is right in the middle of the allowed region.
To conclude, the composition of the present Universe is fairly complex. Most of the energy density
comes from species with which particle physicists are unfamiliar: vacuum or vacuum-like dark energy
and non-baryonic clumped dark matter (presumably, non-relativistic, weakly interacting particles). This
poses serious problems for both fundamental physics and cosmology:

– What are the particles of non-baryonic dark matter?

This appears to be a less difficult problem, as compared to some others listed below. Currently,
a popular option is the lightest supersymmetric particle, which is stable in many supersymmetric
extensions of the Standard Model. Indeed, we shall estimate in what follows that the present mass
density of such particles is naturally predicted to be in the right ballpark. Of course there are
12
The data are consistent with time-independent w, although allow for some time variation.

21
217
 V. A. RUBAKOV

0.0

–0.5

w
–1.0

WMAP WMAP
WMAP+SDSS WMAP+2dF
–1.5

0.0

–0.5
w

–1.0

WMAP WMAP
WMAP+SN(HST/GOODS) WMAP+SN(SNLS)
–1.5
0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8

Fig. 10: Contour plots [19] in the (ΩM , w) plane (confidence levels 68% and 95%). Solid lines: WMAP data
only; filled regions: WMAP data combined with galaxy surveys (upper panels) and SNe 1a measurements (lower
panels). Note the overall consistency of the data and the strong effect of SNe 1a observations on the error bars.

many other options, such as axions, gravitinos, Q-balls, to name a few. In any case, experimental
discovery of the dark matter particle would be a great achievement of both particle physics and
cosmology. This discovery may come either from experiments attempting to detect dark matter or
from collider searches, or both.
– Why are there baryons, and no anti-baryons in our Universe?
In other words, what is the origin of the matter–antimatter asymmetry of the Universe? This also
appears to be a less difficult problem; we shall discuss this issue later in these lectures. Here we
note only that any solution to this problem requires an extension of the Standard Model.
– Why is the mass density of the non-baryonic dark matter so similar (within less than an order of
magnitude) to the mass density of baryons?
Both these densities scale as a−3 (t), so their ratio stays constant during most of the evolution
of the Universe. It is not inconceivable that mechanisms which created baryons and dark matter
particles in the early Universe are related to each other, so that the approximate equality of the
mass densities is not a mere coincidence. It is, however, difficult to construct a corresponding
particle physics model, and it is fair to say that existing attempts are far from being compelling.
– What is the origin of dark energy? If this is vacuum, why does the vacuum have non-zero energy
density, which, however, is very small by particle physics standards?
This is a very fundamental problem of microscopic physics. In natural units, the vacuum energy
density is about ρc ∼ 10−46 GeV4 while on dimensional grounds one would expect values like
1 GeV4 (QCD scale) or 108 GeV4 (electroweak scale). This enormous discrepancy cries out for
an explanation, but despite numerous attempts it remains essentially an open problem. It may
very well be that the solution of this ‘cosmological constant problem’ will lead to an entirely
new concept of physics at ultra-large distances, or an entirely new concept of the history of the
Universe, or both. On the other hand, a very small value of the cosmological constant is crucial
for our existence, so the ‘solution’ of the cosmological constant problem may well be anthropic,
see below.
– Why now?
The energy densities of non-relativistic dark matter and dark energy scale differently: the for-

22
218
C OSMOLOGY

mer scales like a−3 (t) while the latter stays approximately constant. Hence, at small a(t) (early
Universe) the energy density of non-relativistic matter exceeded by far the dark energy density.
Conversely, future expansion of the Universe will be dominated by dark energy. Yet these energy
densities are of the same order of magnitude today. Why is this the case? What is special about
the present epoch of the evolution of the Universe?

5 Dark energy
It appears that by far the most difficult problem is the origin of dark energy. The most disappointing pos-
sibility would be that the carrier of dark energy is vacuum; in that case we shall hardly ever be confident
of a mechanism responsible for tiny, but non-zero vacuum energy density. As a last resort, we shall pos-
sibly have to rely upon anthropic considerations [28, 29], which are based on the observation that if the
vacuum energy density were substantially larger (in absolute value) than observed, the Universe would
not be suitable for the existence of observers like us. Indeed, a large negative cosmological constant
would give rise to early recollapse of the Universe, while with a large positive value, the accelerated
expansion would start much earlier. In either case there would not be enough time for stars and galaxies
to form, hence life would not develop. These considerations in fact provide rather strong bounds on the
vacuum energy density, within two orders of magnitude of the observed value. The idea is then that
there may be infinitely many regions in the Universe (or even infinitely many universes) where funda-
mental parameters like vacuum energy density are different, and span an entire range (−∞, +∞). The
observers like us can only find themselves in a region where the values of these parameters are suitable
for their existence. There have been various suggestions for how such a picture can occur, ranging from
wormholes/branching universes to ‘eternal inflation’ and, most recently, to string theory landscape [30].
The problem is that it will hardly be possible to check this picture experimentally even in the distant
future.
Another option, more promising from an observational viewpoint, is that dark energy is due to
a light field. In many models, this is a scalar field [31–33], dubbed quintessence. This scalar field
must have a very flat scalar potential. We shall see towards the end of these lectures that under some
circumstances, the energy density of a scalar field can indeed evolve very slowly, thus giving rise to the
accelerated expansion of the Universe. In this set of models the energy density decreases in time, so that
the effective w parameter is negative,
quintessence : p > −ρ , w > −1 .
The opposite case is often called ‘phantom energy’; for phantom, the energy density increases in time,
and
phantom : p < −ρ , w < −1 .
Phantom energy is way more theoretically exotic than quintessence; it is most likely that it exists only
in theories with spontaneously broken Lorentz invariance. In either case, the small value of the energy
density today requires extremely small values of the parameters of the quintessence or phantom models
(masses and coupling constants), thus making these models quite unnatural.
Yet another option for explaining the accelerated expansion of our Universe is that gravity devi-
ates from General Relativity at cosmological distances and time scales, so that the Friedmann equation
(22) merely is not valid at the present epoch. This option would probably have a similar naturalness
problem as the quintessence proposal, but even before that one meets a serious problem of constructing a
theoretically consistent and phenomenologically acceptable theory which would reduce to General Rel-
ativity at distances from a fraction of a millimetre (down to which gravity is experimentally known to
obey Newton’s law) to at least tens of megaparsecs, and deviate from General Relativity at cosmological
scales. In known Lorentz-invariant examples of such a theory 13 , there either exist ghosts (fields with
13
See, however, Ref. [34].

23
219
 V. A. RUBAKOV

negative energy unbounded from below) or gravity becomes strongly coupled at the quantum level (and
hence not tractable) below an unacceptably small ‘ultraviolet’ energy scale corresponding to a distance
of order 1000 km. A consistent theory of this sort would probably require a ‘gravitational Higgs mecha-
nism’ and violation of Lorentz invariance, but even this, rather exotic idea, has not yet led to a consistent
model capable of explaining the accelerated expansion of the Universe.
Quintessence/phantom models (and most likely models with infrared-modified gravity, if the latter
exist) imply that the effective dark energy density is not constant in time. Needless to say, an observa-
tional evidence for the time-dependence of ρ Λ would have enormous impact on fundamental physics. On
the other hand, it is hard to foresee any way of probing dark energy in a laboratory, so we have to rely
solely on cosmological observations when trying to reveal the origin of dark energy.

6 Hot Big Bang

The usual matter and dark matter have their origin in the early Universe. Before discussing plausible
scenarios of their generation, let us give some details of the evolution of the Universe in its hot stage.
One assumes that it makes sense to extrapolate the evolution of the Universe back in time by
making use of known laws of microscopic physics (nuclear physics, quantum field theory, including
electrodynamics, QCD and electroweak physics) and General Relativity. This theory is called the ‘Hot
Big Bang’ theory. As we have seen, the Universe was hotter at earlier stages (i.e., at smaller values of
the scale factor a(t)), as the temperature scales as a −1 (t). Also, the Universe was denser: the particle
number densities scale as a−3 (t) both for relativistic and non-relativistic particles. At high enough
temperature the Universe was quite different from what we observe today: instead of almost empty space
with galaxies here and there, there was hot, dense and almost homogeneous plasma filling the Universe.
This is why microscopic physics played a role in the early Universe: at temperatures of the order of the
nuclear physics scale, roughly a few MeV, one has to deal with nuclear reactions; at temperatures of
the order of the strong interaction scale, (a few) · 100 MeV, the relevant microscopic theory was largely
QCD, etc.

6.1 Expansion of the Universe

We begin with the analysis of the evolution of the Universe, i.e., the behaviour of the scale factor a(t)
as a function of time. As we already discussed in Section 2.6, at early times the Universe was radiation
dominated, then matter dominated, and currently dark-energy dominated, while the curvature term κ/a 2
was never important.

6.1.1 Deceleration to acceleration

Owing to the dominance of dark energy, the Universe accelerates today. When matter was dominating,
the Universe was decelerating. To work out when the change in the regime occurred, we write down the
Friedmann equation in the following form (assuming dark matter equation of state p = −ρ, neglecting
spatial curvature and also neglecting ultra-relativistic matter for the moment; we shall see that ultra-
relativistic matter was important for the expansion at a much earlier stage),
 2  
ȧ 8πG a30
= ρc ΩM 3 + Ω Λ
a 3 a
where a0 , ΩM , ΩΛ and ρc are the present values, thus time-independent constants. Therefore,
 
2 8πG a30 2
ȧ = ρc ΩM + Ω Λ a
3 a
and ä is equal to zero when
a30 2ΩΛ
3
≡ (1 + z)3 = .
a ΩM

24
220
C OSMOLOGY

With ΩΛ = 0.7 and ΩM = 0.3, we have

deceleration → acceleration : z ≈ 0.7 .

The Universe was decelerating until fairly recently. Before z ≈ 0.7, the expansion was dominated by
non-relativistic matter.

6.1.2 Radiation domination to matter domination

Since energy densities of non-relativistic and ultra-relativistic matter (‘radiation’) scale as a −3 and a−4 ,
respectively, the dominant contribution to the energy density of the Universe at very small a, i.e., at a
very early epoch, came from ultra-relativistic matter. To estimate z eq at which the equilibrium between
matter and radiation occurred, i.e., at which the expansion regime changed from the dominance of ultra-
relativistic particles to the dominance of non-relativistic matter, we write
 
ρM (t) ρM a(t)
=
ρrad (t) ρrad 0 a0

where the subscript 0 still refers to the present values. Equilibrium occurs at

ρM (teq )
≈1
ρrad (teq )

which gives  
a0 ρM ΩM
≡ 1 + zeq ≈ = .
a(teq ) ρrad 0 Ωrad
We already know the energy density of relic photons; massless neutrinos 14 of all three types have ρν,0 ≈
0.7ργ,0 . Thus, Ωrad ≈ 10−4 , see Eq. (46). With ΩM = 0.3 we have

radiation domination → matter domination : z eq ≈ 3000 .

The corresponding temperature is

Teq = T0 (1 + zeq ) ≈ 104 K ≈ 1 eV . (49)

At higher temperatures, the expansion of the Universe was dominated by ultra-relativistic matter.
It is important for the theory of structure formation that during much of its lifetime, the Uni-
verse was dominated by non-relativistic matter. The expansion rate at both radiation-dominated and
vacuum-dominated stages is such that gravitational perturbations grow slowly, and only during the
matter-dominated stage is their growth fast enough to account for the existing structures in the Uni-
verse. The bottom line is that the present composition of the Universe plus simple extrapolations back to
the past are consistent with the theory of structure formation. Various ingredients of standard cosmology
fit nicely together.

6.2 Epochs in the early Universe

We have already mentioned two important epochs in the evolution of the Universe: the recombination
epoch (transition from plasma to neutral gas that occurred at T ∼ 3000 K, t ∼ 3 · 10 5 yrs and lasted
much less than Hubble time) and the nucleosynthesis epoch (T = 1 MeV to (a few) · 10 keV). Another
‘event’ is neutrino decoupling. At high temperatures, weakly interacting particles, including neutrinos,
were in thermal equilibrium with the rest of cosmic plasma. The plasma became effectively transparent
14
Whether or not neutrinos are exactly massless is inessential for the estimate of zeq : the estimate remains valid if every
neutrino has mass smaller than 1 eV.

25
221
 V. A. RUBAKOV

for neutrinos at a temperature of about 1 MeV. The temperature of decoupling of neutrinos from cosmic
plasma is of importance for nucleosynthesis, as it affects the neutron–proton ratio just before nucleosyn-
thesis (and hence the abundances of light elements, which need neutrons for their formation), and also
the expansion of the Universe at the nucleosynthesis epoch. The fact that neutrinos decoupled much
earlier than photons implies that the present neutrino-to-photon ratio is less than one 15 .
As we move further back in time, the cosmic plasma has more and more components. At temper-
atures above roughly 0.5 MeV (set by the mass of the electron), there are lots of electrons and positrons
which frequently are pair created and annihilate; at T > 100 MeV the plasma contains muons and pions,
etc. Simple estimates given in the next subsection show that the plasma remains in thermal equilibrium
except possibly for phase transitions.

– QCD phase transition.

At temperatures well above 100 MeV (QCD scale), strongly interacting particles are dissolved into
quarks and gluons. This quark–gluon matter converts into hadronic matter (mostly pions) during
the quark–hadron phase transition. Theoretical estimates and lattice simulations in QCD suggest
that the temperature of this phase transition is about 170 MeV.
– Electroweak transition.
Loosely speaking, at temperatures well above 100 GeV, electroweak symmetry is unbroken, the
Higgs expectation value vanishes, and W and Z bosons are massless. At T ∼ 100 GeV, the phase
transition of the electroweak symmetry breaking takes place. This picture needs qualification,
however: there is no local, gauge-invariant order parameter in the standard electroweak theory,
and the electroweak transition is similar to vapour–liquid transition. For some values of parame-
ters, there is the first-order phase transition, whereas for other values, a smooth cross-over takes
place instead. With existing constraints on the Higgs boson mass, the Standard Model predicts
cross-over; what actually happened in the early Universe depends on what exactly is the exten-
sion of the Standard Model. Uncovering physics in the 100 GeV–1 TeV energy range will thus
allow cosmologists to develop a quantitative theory of a quite early epoch of the evolution of the
Universe.
– GUT transition.
Extrapolating further back is dangerous, but if we do so, we come to the Grand Unification epoch,
whose temperature is set by the GUT scale, T GUT ∼ 1016 GeV. At this temperature, one expects
Grand Unified phase transition to occur. However, most models of inflation suggest that the Uni-
verse never had such a high temperature after inflation.
– Even more interesting are the epoch of the generation of dark matter and the epoch of the genera-
tion of baryon asymmetry. We can only make guesses about these epochs, and some of the guesses
are the subject of these lectures. Before turning to them, it is convenient to consider the expansion
of the Universe at early times in a little more detail.

6.3 Expansion rate and lifetime at radiation domination

Before proceeding further, let us consider in a little more detail the expansion of the Universe at the
radiation-dominated stage, assuming thermal equilibrium of all ultra-relativistic species 16 . The energy
density of all ultra-relativistic species, which enters the Friedmann equation, is

π2
ρ= g∗ T 4
30
15
This is because photons are additionally heated, after neutrino decoupling, owing to annihilations of e+ and e− .
16
The assumption of thermal equilibrium is in fact not valid for neutrinos at temperatures below 1 MeV. The corresponding
modification of our discussion is straightforward, however.

26
222
C OSMOLOGY

where g∗ is the effective number of massless degrees of freedom at temperature T . The contribution
of bosons into g∗ is equal to the number of spin states (e.g., for photons g γ = 2, while for W bosons
at temperatures above 100 GeV, gW = 6 because of two charges and three projections of spin), while
fermions contribute 7/8 of the number of spin states (electrons plus positrons contribute 4 · 7/8, each
type of left-handed neutrino plus its antineutrino gives 2 · 7/8, etc.). The parameter g ∗ is the sum of
contributions of all ultra-relativistic species; it slightly depends on time because at higher temperatures,
more species are ultra-relativistic (say, electrons contribute at T > 0.5 MeV and do not contribute at
lower temperatures).
It is convenient to introduce the effective Planck mass
MPl
MPl∗ = √ .
1.66 g∗

This parameter slightly depends on temperature, and numerically is of order

MPl∗ = (a few) · 1018 GeV .

With this notation, the Friedmann equation tells us that the expansion rate is related to temperature in a
simple way,
T 2 (t)
H(t) = .
MPl∗
One recalls that the expansion law at the radiation-dominated stage (neglecting the dependence of g ∗ on
temperature) is √
a(t) = const · t
so the Hubble parameter is related to the lifetime as follows:
ȧ 1
H≡ = .
a 2t
We immediately deduce the relation between lifetime and temperature,
MPl∗
t= .
2T 2
Let us make use of the latter formula to estimate the age of the Universe at different epochs:

– Nucleosynthesis.
The temperatures relevant for BBN range from a few MeV to about 70 keV. From these we obtain
that the earliest time directly probed by observations is about

1018 GeV
t∼ ∼1s
10−6 GeV2
whereas BBN ends at t ∼ 200 s ∼ 3 min. We do have an idea of what the Universe was like at one
second old!
– Earlier epochs:

QCD phase transition : T ∼ 200 MeV , t ∼ 3 · 10 −5 s

Electroweak epoch : T ∼ 100 GeV , t ∼ 10 −10 s . (50)

One may wonder whether equilibrium thermodynamics, which we use throughout, is applicable
at these early times, when the Universe expands so rapidly. To see that this is indeed the case, let
us consider, as an example, electromagnetic scattering of light particles at T > 1 MeV. It is clear on

27
223
 V. A. RUBAKOV

dimensional grounds that the mean free time of a relativistic charged particle at temperature T exceeding
its mass is
τ ∼ (α2 T )−1
(the electromagnetic cross-section is proportional to α 2 ). For thermal equilibrium with respect to the
electromagnetic interactions to be established, the interaction rate τ −1 must be smaller than the expansion
rate of the Universe H which gives
T2
α2 T  ∗ .
MPl
This inequality is indeed valid at T  α 2 MPl∗ ∼ 1014 GeV, so electromagnetic (and many other) micro-
scopic processes are in thermal equilibrium at all temperatures of interest to us.
Thermal equilibrium is not a particularly interesting state of the Universe. What we are going to
discuss in the rest of this section are in fact inequilibrium phenomena. It is these phenomena that may
leave relics behind, and hence may have observable consequences.

6.4 Heavy relic: best guess for cold dark matter

As we discussed above, the observational data strongly suggest that a good part of the energy density
in the present Universe is due to new stable (or practically stable) particles. At least in some cases, the
Hot Big Bang theory is capable of predicting the density of such particles in terms of their interaction
cross-sections and masses. Here we present the corresponding estimate in the simplest possible scenario;
needless to say, this estimate can be (and has been) refined by more careful calculations.
Let us assume that there exists a heavy stable particle Y (weakly interacting massive particle,
WIMP) and its anti-particle Y. Let us assume for definiteness that the dominant process in which these
particles can be destroyed or created is their pair-annihilation or creation, with annihilation products
being the particles of the Standard Model (the analysis for a neutral particle Y, which coincides with its
own anti-particle, is very similar, provided that these particles are created and annihilate in pairs only,
as is the case for the neutralino). Let us further assume that there is no asymmetry between Y and Y
in the early Universe, i.e., the densities of Y and Y are equal to each other 17 . We shall see that the
overall cosmological behaviour of these particles is as follows. At high temperatures, T  m Y , the Y
particles are in thermal equilibrium with the rest of the cosmic plasma; there are lots of Y–Y pairs in the
plasma, which are copiously created and annihilate. As the temperature drops below m Y , the equilibrium
number density decreases. At some ‘freeze-out’ temperature T f the number density becomes so small
that Y and Y can no longer meet each other during the Hubble time, and their annihilation terminates.
The pair-creation of Y and Y terminates too, as there are too few Standard Model particles in plasma
whose energy exceeds mY . After that the number density of survived Y and Y decreases like a −3 , and
these relic particles contribute to the mass density in the present Universe. Our purpose is to estimate
the range of properties of Y particles, in which their present mass density is of the order of the critical
density ρc , so that Y particles serve as dark matter candidates.
Assuming thermal equilibrium, elementary considerations of the mean free path of a particle with
respect to the annihilation give for the lifetime of a non-relativistic Y particle in cosmic plasma, τ ann , the
following relation,
σann · v · τann · nY ∼ 1
where v is the velocity of the Y particle, σ ann is the annihilation cross-section at this velocity, and
17
This is actually a strong assumption. It is valid in many, but not all, realistic extensions of the Standard Model. In fact,
an alternative scenario with the generation of Y asymmetry is appealing too, because it might relate baryon asymmetry to the
density of dark matter [35].

28
224
C OSMOLOGY

nY = nY is the equilibrium number density (Boltzmann law at zero chemical potential, i.e., at nY = nY )
 3/2
mY T mY
nY = g Y · e− T .
2π
Let us assume for definiteness that the annihilation occurs in s-wave (other cases give similar, but not
exactly the same results), so at non-relativistic velocities
σ0
σann = ,
v
where σ0 is a constant about which we shall have more to say later. One should compare the lifetime with
the Hubble time, or annihilation rate Γ ann ≡ τann
−1 with the expansion rate H = T 2 /M ∗ . At T ∼ m ,
Pl Y
the equilibrium density is of order n Y ∼ T , and Γann  H for not too small σ0 . This means that the
3

annihilation (and, by reciprocity, creation) of YY pairs is indeed rapid, and Y particles are indeed in
thermal equilibrium with the plasma. At very low temperature, on the other hand, the number density
nY is exponentially small, and Γann  H. At low temperatures we cannot, of course, make use of
equilibrium formulas: Y particles no longer annihilate (and are no longer created), there is no thermal
equilibrium with respect to creation–annihilation processes, and the number density n Y gets diluted only
because of the cosmological expansion.
The freeze-out temperature Tf is determined by the relation
−1
τann ≡ Γann ∼ H

where we can still use the equilibrium formulas, as Yparticles are in thermal equilibrium (with respect
to annihilation and creation) just before freeze-out. We find
Tf2
σ0 · nY (Tf ) ∼ (51)
MPl∗
or  3/2
mY Tf −
mY
Tf2
σ0 · g Y · e Tf
∼ .
2π MPl∗
The latter equation gives the freeze-out temperature, which, up to loglog terms, is
mY
Tf ≈ .
ln(MPl∗ mY σ0 )
Note that this temperature is quite a bit smaller than m Y , if the relevant microscopic mass scale is much
below MPl . This means that Y particles freeze out when they are indeed non-relativistic, hence the term
‘cold dark matter’. The fact that the annihilation and creation of Y particles terminate at relatively low
temperature has to do with a rather slow expansion of the Universe, which should be compensated for by
the smallness of the number density n Y .
At the freeze-out temperature, we make use of Eq. (51) and obtain
Tf2
nY (Tf ) = .
MPl∗ σ0
Note that this density is inversely proportional to the annihilation cross-section (up to logarithms). The
reason is that for a higher annihilation cross-section, the creation–annihilation processes are no longer in
equilibrium, and fewer Y particles survive.
To estimate the present density of Y particles, it is convenient to consider the ratio n Y /s where s
is the entropy density,
2π 2
s= g∗ T 3 .
45

29
225
 V. A. RUBAKOV

The point is that during the adiabatic expansion after freeze-out, both entropy density and n Y behave as
a−3 , so this ratio stays constant. Up to a factor of order 1, this ratio at freeze-out is
nY 1
∼ .
s g∗ (Tf )MPl∗ Tf σ0

At late times, the entropy density, again up to a factor of order 1, is equal to the number density of
photons, so the present number density of Y particles is of order
n 
Y
nY,0 ∼ nγ,0 ·
s freeze-out
and the mass density is

ρY,0 = mY nY,0
ln(MPl∗ mY σ0 )
∼ nγ,0 · . (52)
g∗ (Tf )MPl∗ σ0

This formula is remarkable. The mass density depends mostly on one parameter, the annihilation cross-
section σ0 . The dependence on the mass of the Y particle is through the logarithm and g ∗ (Tf ), and is
very mild. From this formula we immediately derive the condition ensuring that Y particles are dark
matter candidates, i.e., their present mass density is of the order of ρ DM = 0.25ρc ,
nγ,0
σ0 ∼ ln(MPl∗ mY σ0 ) .
g∗ (Tf )MPl∗ ρDM

The value of the logarithm here is between 20 and 40, depending on parameters (this means, in particular,
that freeze-out occurs when the temperature drops 20 to 40 times below the mass of the Y particle).
Plugging in other numerical values (ρ DM ∼ 10−6 GeV cm−3 , nγ,0 ∼ 400 cm−3 , g∗ (Tf ) ∼ 100, MPl∗ ∼
1018 GeV), we obtain an estimate
σ0 ∼ 10−10 GeV−2
which already crudely tells us what the relevant range of mass scales is. In fact, the annihilation cross-
section may be parametrized as
α2
σ0 =
M2
where α is some coupling constant, and M is the mass scale. This parametrization is suggested by the
picture of Y–Y annihilation via exchange of another particle of mass M, which may be somewhat higher
than mY . With α ∼ 1/30 (SU(2)W gauge coupling) the estimate for the mass scale is roughly

M ∼ 1 TeV .

Thus with very mild assumptions, we find that the non-baryonic dark matter may naturally originate from
the TeV-scale physics!
In supersymmetric extensions of the Standard Model, the lightest supersymmetric particle (LSP,
most likely, neutralino—a mixture of superpartners of the photon, Z boson, and neutral Higgs bosons) is
often stable, and its annihilation cross-section is automatically in the right ballpark. This is illustrated in
Fig. 11.
Naturally, the search for both direct and indirect signals from neutralino dark matter (and more
generally, WIMPs) is an active area of experimental research. For discussions of the potential of existing
and future experiments, see, for example, Refs. [36, 37].
If dark matter particles are indeed WIMPs, and the relevant energy scale is of order 1 TeV, then the
Hot Big Bang theory will be probed experimentally up to temperatures of (a few) · 10 GeV and down to

30
226
C OSMOLOGY

Fig. 11: Present mass density of relic neutralino LSPs (denoted by χ), as function of the neutralino mass [36]
in the Minimal Supersymmetric Standard Model (MSSM). Each point corresponds to a randomly chosen set of
MSSM parameters. Crosses, dots and circles correspond to different neutralino content. Note that observational
data suggest Ωχ ≈ 0.25, h ≈ 0.7, so that Ωχ h2 ≈ 0.12.

age 10−9 s in the relatively near future (compare to 1 MeV and 1 s accessible today). With microscopic
physics to be known from collider experiments, the WIMP density will be reliably calculated and checked
against data from observational cosmology. Thus the WIMP scenario (and also some others) offers a
window to a very early stage of the evolution of the Universe.
The mechanism discussed here is by no means the only mechanism capable of producing cold
dark matter, and WIMPs are by no means the only candidates for dark-matter particles. Other dark-
matter candidates include very heavy relics produced towards the end of inflation, axions, gravitinos,
Q-balls, massive gravitons, etc. Unlike WIMPs, however, the right dark-matter density in other scenarios
comes out by tuning either microscopic parameters (masses, coupling constants of new particles) or the
parameters of the early Universe (e.g., maximum temperature reached after inflation), or both. Because
of this, the WIMP scenario may be considered as the best guess for the origin of dark matter.

6.5 Baryon asymmetry of the Universe

In the present Universe, there are baryons and almost no anti-baryons. The number density of baryons
today is characterized by the ratio η, see Eq. (18). In the early Universe, the appropriate quantity is
nB − nB̄
∆B =
s

31
227
 V. A. RUBAKOV

where nB̄ is the number density of anti-baryons, and s is the entropy density. If the baryon number is
conserved, and the Universe expands adiabatically, ∆ B is constant, and its value is, up to a numerical
factor, equal to η, so that
∆B ≈ 10−10 .
Back at early times, at temperatures well above 100 MeV, cosmic plasma contained many quark–antiquark
pairs, whose number density was of the order of the entropy density,

nq + nq ∼ s

while baryon number density was related to densities of quarks and antiquarks as follows (baryon number
of a quark equals 1/3),
1
nB = (nq − nq ) .
3
Hence in terms of quantities characterizing the very early epoch, the baryon asymmetry may be expressed
as
n q − nq
∆B ∼ .
nq + nq
We see that there was one extra quark per about 10 billion quark–antiquark pairs! It is this tiny excess
that is responsible for the entire baryonic matter in the present Universe.
There is no logical contradiction to suppose that the tiny excess of quarks over antiquarks was
built in as an initial condition. This is not at all satisfactory for a physicist, however. Furthermore, the
inflationary scenario does not provide such an initial condition for the Hot Big Bang; rather, inflation
theory predicts that the Universe was baryon-symmetric just after inflation. Hence, one would like to
explain the baryon asymmetry dynamically.
The baryon asymmetry may be generated from an initially symmetric state only if three necessary
conditions, dubbed Sakharov’s conditions, are satisfied. These are
(i) baryon number non-conservation;
(ii) C- and CP-violation;
(iii) deviation from thermal equilibrium.
All three conditions are easily understood. (i) If the baryon number were conserved, and assuming
the initial net baryon number in the Universe equal to zero, the Universe today would be symmetric
rather than asymmetric. (ii) If C or CP were conserved, then the rates of reactions with particles would
be the same as the rates of reactions with antiparticles. In other words, if the initial state of the Universe
was C- and CP-symmetric, then the asymmetry between particles and antiparticles may develop only if C
and CP are violated. (iii) Thermal equilibrium means that the system is stationary (no time dependence
at all). Hence, if the initial baryon number is zero, it is zero forever, unless there are deviations from
thermal equilibrium.
There are two well-understood mechanisms of baryon number non-conservation. One of them
emerges in Grand Unified Theories and is due to the exchange of super-massive particles. It is very
similar, say, to the mechanism of charm non-conservation in weak interactions which occurs via the
exchange of heavy W bosons. The scale of these new, baryon-number-violating interactions is the Grand
Unification scale, presumably of order 10 16 GeV.
Another mechanism is non-perturbative and exists already in the Standard Model, and, possibly
with slight modifications, operates in all of its extensions (see Ref. [38] for details). Baryon and lepton
currents jµB and jµL have electroweak anomalies, which arise due to quantum effects,
αW µνλρ a a
∂µ j B µ
= ∂µ j L µ = 3  Fµν Fλρ
8π

32
228
C OSMOLOGY

where αW = gW 2 /(4π) is the SU(2) gauge coupling, F a is the SU(2) gauge field (a = 1, 2, 3) and
W µν W
 µνλρ is the antisymmetric tensor. Integrating these relations over space–time, one finds that the changes
of baryon and lepton numbers are related to a space–time integral involving the gauge field,
Z
αW
∆B = ∆L = 3 d 4 x µνλρ Fµν
a
Fλa ρ . (53)
8π

The right-hand side of this equation may be non-zero for fluctuations of the gauge fields 18 whose field
strengths are large
a 1
Fµν ∝ .
gW
Obviously, such fluctuations, being inversly proportional to the coupling constant, cannot be described
within the perturbation theory; electroweak baryon number non-conservation is indeed invisible in Feyn-
man diagrams. In vacuo, these fluctuations are instantons; the probability for them to happen is extremely
small, so electroweak processes violating B and L practically do not take place. At high temperatures
there are thermal fluctuations (a keyword here is sphaleron), and they are quite probable. By compar-
ing the rate of electroweak baryon-number violating processes with the Hubble rate one sees that these
processes are fast in the early Universe at temperatures T obeying

T < 1011 GeV

and
T > v(T ) (54)
where v(T ) is the expectation value of the Higgs field, which in turn depends on temperature. At T ≈
v(T ) electroweak baryon number violation turns off in the cosmic plasma.
As seen from Eq. (53), electroweak processes do not conserve B and L separately, but their com-
bination is not violated,
B − L = conserved .
In fact, the complete set of conserved quantum numbers is (B − 3Le), (B − 3L ν) and (B − 3L τ), but
we shall not need this qualification in what follows.
Realistic mechanisms of baryon number non-conservation are thus not numerous, yet there are
several ways the baryon asymmetry could have been generated. They differ by the characteristic temper-
ature at which the asymmetry is produced.
(i) Grand Unification mechanisms operate at extremely high temperatures, T ∼ M GUT ∼ 1015 −
1016 GeV. The most commonly discussed source of the baryon asymmetry in this context are B- and
CP-violating decays of ultra-heavy particles. At later times, the baryon number is violated by anomalous
electroweak processes whose effect is basically to wash out (B + L). They would therefore reprocess
part of the baryon asymmetry, but if non-zero (B − L) is generated at GUT temperatures, then this
(B − L) would survive until the present epoch (provided there are no strong lepton-number-violating
interactions at intermediate temperatures, 100 GeV < T < 10 11 GeV, otherwise all fermion quantum
numbers would be violated at those temperatures, and no asymmetry would survive). Part of this (B −L)
would be carried by baryons.
It is unlikely that GUT baryogenesis operated at the hot stage of the evolution of the Universe,
after post-inflationary reheating. In the vast majority of inflationary models, the reheat temperature is
well below MGUT ∼ 1015 –1016 GeV. However, it is not unrealistic that GUT baryogenesis operated
during the reheating epoch, which occurred between inflation and the hot stage. Unfortunately, there
does not seem to exist any direct or indirect observational test of this scenario.
18
It is important here that the SU(2)W gauge theory is non-Abelian, and that only left-handed fermions participate in the
SU(2)W interactions.

33
229
 V. A. RUBAKOV

a) b)
Veff (φ) Veff (φ)

T &v
T =0

0 0
v φ φ
Fig. 12: Effective potential for the Higgs field at zero and high temperatures, panels (a) and (b), respectively

a) b)
hφiT hφiT

0 0
T T
Fig. 13: Higgs expectation value hφiT as a function of temperature in the cases of the first-order and second-order
phase transitions, panels (a) and (b), respectively

(ii) Electroweak baryogenesis is a scenario in which the baryon asymmetry is generated entirely
due to the anomalous electroweak processes (53). Its generation would occur at a temperature of order
100 GeV, at which these anomalous processes are switching off.
Since the Universe expands slowly at the electroweak epoch (as compared to the rates of mi-
croscopic interactions), considerable departure from thermal equilibrium is possible due only to the
first-order phase transition.
At low temperatures, the scalar potential has the shape shown in Fig. 12(a), so the Higgs field has
a non-vanishing expectation value. At high temperatures, the relevant quantity is the effective potential
Veff (φ, T ), which is defined as the free energy density at temperature T in the presence of the homo-
geneous scalar field φ. As often happens in condensed matter systems, symmetry is restored at high
temperatures, i.e., the effective potential has the shape shown in Fig. 12(b).
Symmetry breaking in principle can proceed via phase transitions of the first and second order,
or may occur as a smooth cross-over19 . In the case of the first-order phase transition, the value of
v(T ) ≡ hφiT jumps at a certain critical temperature, as shown in Fig. 13(a). This distinguishes the
first-order phase transition from the second-order phase transition, Fig. 13(b). The reason behind this
difference is that the effective potential changes with temperature as shown in Fig. 14: in the case of the
19
In the latter case there is no order parameter discriminating the phases, so one cannot, in fact, talk about symmetry breaking.
A well-known condensed matter example is the liquid–gas transition which is first order at low pressure and a smooth cross-over
at high pressure.

34
230
C OSMOLOGY

a) b)
Veff (φ) Veff (φ)

φ φ
Fig. 14: Effective potentials at different temperatures for the first- and second-order phase transition, panels (a)
and (b), respectively. Upper curves correspond to higher temperatures. Black circles show the field values hφi T ≡
v(T ).

φ=0

φ=0
φ = hφiT

φ = hφiT
φ = hφiT

φ=0

φ=0

Fig. 15: Bubbles of the new phase in the old phase at the first-order phase transition

first-order phase transition, in some range of temperatures there are two minima of the effective potential,
the minimum with v 6= 0 being deeper below the critical temperature. In realistic situations (including
the early Universe), the system supercools, i.e., the transition actually occurs when the depths of the two
minima are not equal to each other, as indicated in Fig. 14(a). For the second-order phase transition the
situation is different. The minimum of the effective potential continuously moves away from φ = 0,
there is nothing like supercooling, the system evolves smoothly.
The first-order phase transition cannot occur simultaneously everywhere in space: in that case the

35
231
 V. A. RUBAKOV

system would have to climb the infinitely high potential barrier (recall that V eff is free energy density, so
the free energy of the homogeneous scalar field away from the minimum of V eff is proportional to the
total volume of the system). Instead, the transition proceeds via thermal nucleation of bubbles of the new
phase, their growth and percolation, as sketched in Fig. 15. Obviously this is a highly non-equilibrium
process.
To generate the baryon asymmetry during the phase transition, the following necessary require-
ments have to be fulfilled:

(a) The baryon numbers have to be efficiently violated before the phase transition.
(b) The baryon number violation has to be switched off after the phase transition (otherwise the baryon
asymmetry generated during the transition would be washed out at later times when cosmic plasma
is again close to the thermal equilibrium); in other words, the inequality (54) has to be valid after
the phase transition.
(c) There should be enough CP-violation during the phase transition.

The requirement (a) is satisfied automatically (provided there is indeed the first-order phase tran-
sition). Whether requirements (b) and (c) are fulfilled depends on microscopic theory.
In the Standard Model, neither (b) nor (c) are satisfied. The strength of the phase transition in the
Standard Model depends on its only unknown parameter, the Higgs boson mass. With the current limit,
mH > 114 GeV, there is no electroweak phase transition at all—the Standard Model exhibits smooth
cross-over as the temperature decreases (see Ref. [39] and references therein). Also, CP-violation coming
from the Kobayashi–Maskawa mechanism is too weak at temperatures of order 100 GeV to produce
enough baryon asymmetry.
On the other hand, extensions of the Standard Model with additional relatively light scalars (in-
cluding a corner of MSSM with light stop) work fine in an appropriate range of parameters. The in-
teractions of the Higgs field with additional scalars often strengthen the phase transition, so that the
requirement (a) can be satisfied. An extra source of CP-violation, needed to satisfy (c), comes from CP
phases in the scalar sector. Interestingly, these phases induce electric dipole moments of the neutron and
electron, which are typically 1 to 3 orders of magnitude below the current limits, i.e., are in a potentially
observable range.
The electroweak generation of baryon asymmetry occurs when the extra particles are not very
heavy, otherwise they would not have any effect on the Higgs effective potential at T ∼ 100 GeV: the
masses of at least some of these particles have to be in the range of 100–200 GeV. Clearly, this range
is accessible at the LHC, so experiments in the near future will either confirm or reject the electroweak
mechanism.
(iii) A currently popular mechanism is leptogenesis. It may occur at some intermediate temper-
ature (the estimates range from 107 GeV to 1011 GeV) due to L- and CP-violating decays of heavy
Majorana neutrinos [40]. Of course, the generation of lepton asymmetry requires lepton number vio-
lation, i.e., extension of the Standard Model, but such an extension is favoured by neutrino oscillation
data anyway. The lepton asymmetry would then be partially reprocessed into baryon asymmetry by
anomalous electroweak processes. Interestingly, the range of Majorana masses of ordinary neutrinos
compatible with this mechanism is indeed consistent with the range inferred from neutrino oscillations.
Let us discuss the leptogenesis scenario in a little more detail.
Let us assume that ordinary neutrinos get their small masses via the see-saw mechanism. This
mechanism invokes heavy ‘sterile’ (neutral with respect to the Standard Model gauge interactions) neu-
trinos Ni , where i = 1, 2, 3 is the generation number. They have large Majorana masses; in an ap-
propriate basis the corresponding mass matrix is diagonal and real. They are assumed to have Yukawa
interactions with conventional lepton doublets L and the Standard Model Higgs field H. Thus, besides

36
232
C OSMOLOGY

li li li


 h li ¯lk
h
N1
 N1 
 N1
N
l
+ Nl +

  hli
h1i h∗1k h∗1k hlk
¯lk 
hlk h
h h h
Fig. 16: Diagrams for the decay N1 → li h. Note that no CP-asymmetry emerges at the tree level, so to the
lowest non-trivial order, this asymmetry is given by the interference of tree and one-loop diagrams. In the latter, a
summation over k and l is assumed.

the kinetic terms, the Lagrangian involving N i contains two terms

X X
L= Mi N̄ic Ni + ( hij L̄i Nj H̃ + h.c.)
i ij

where the fermions N are right, the Yukawa couplings h ij are in general complex, and H̃α = αβ Hβ∗ is a
weak doublet whose vacuum expectation value is (v, 0). Owing to this vacuum expectation value, there
is mixing between ordinary neutrinos and N ’s. Upon diagonalizing the mass matrix, one finds that there
are heavy states, predominantly N ’s, with masses M i , and light states with mass matrix
X v2
mij = h∗ik h∗jk . (55)
Mk
k

These masses are naturally small for large M i , which is the advantage of the see-saw mechanism. Obvi-
ously, the original Lagrangian does not conserve any of the lepton numbers.
Now, assume that at high temperatures, heavy neutrinos are in thermal equilibrium 20 . As the Uni-
verse cools down below T ∼ M1 (the smallest of Mi ), the lightest of N ’s (call it N1 ) starts decaying21 .
There is also an inverse decay process which tends to keep N 1 in thermal equilibrium. The simplest
version of the leptogenesis scenario assumes that N 1 is not in thermal equilibrium with respect to decay
and inverse decay at T ∼ M1 . This requires that its width be small enough,

Γ1 < H(T ∼ M1 ) . (56)

Now, since the Yukawa couplings contain CP-phases, there is an asymmetry in decays of N 1 ’s,

Γ(N1 → LH) − Γ(N1 → L̄H̄)

δ≡ 6= 0 .
Γ(N1 → LH) + Γ(N1 → L̄H̄)

This asymmetry occurs because of the interference of the tree diagram and one-loop diagram in which
all Ni run in intermediate states, Fig. 16. This ‘microscopic’ asymmetry is

1 1 X h i M 
† 2 1
δ= Im (h h)1k F
8π (h† h)11 Mk
k=2,3

20
This may be due either to some new interactions in which N ’s participate, or to the Yukawa interactions themselves. In
the latter case, thermal equilibrium is established with large enough Yukawa couplings only, i.e., the usual neutrinos must be
sufficiently heavy, mν > 3 · 10−3 eV. This is consistent with neutrino oscillation data.
21
Generically, the lightest of N ’s is most relevant, since any asymmetry produced in decays of heavier N ’s is washed out by
intractions involving the lightest one.

37
233
 V. A. RUBAKOV

where F (M1 /Mk ) ∼ M1 /Mk for M1  Mk , up to a factor of order 1. Provided the out-of-equilibrium
condition (56) is satisfied, the generated lepton asymmetry is of order
nL
∼ 0.01δ (57)
s
where the factor 0.01 is due to the large number of ultra-relativistic species at temperature T ∼ M 1 . A
good part of this asymmetry gets reprocessed into the baryon asymmetry by electroweak processes, so
the generated baryon asymmetry is roughly of the same order of magnitude as (57).
Let us first see what the out-of-equilibrium condition (56) means in terms of neutrino masses. Let
L1 be a linear combination of lepton fields to which N 1 is coupled (generally, this combination is neither
a flavour nor a mass eigenstate), X
L1 ∝ h∗k1 Lk .
k
After rotating into the lepton basis whose first basis vector is L 1 , the lightest N produces the only term
in the neutrino mass matrix,
m11 ν̄1c ν1
where
v2
m11 = (h∗11 )2 .
M1
The width of N1 is
1 |m11 |M12
Γ1 = |h11 |2 M1 = .
16π 16πv 2
Let us now recall that H(T ∼ M1 ) = M12 /MPl∗ , so the relation (56) gives

16πv 2
m11 < .
MPl∗

Inserting numbers, one finds

m11 < 3 · 10−3 eV .
Miraculously, this number is in the right ballpark of neutrino masses, suggested by neutrino oscillations!
In fact, the situation is even better. If the out-of-equilibrium condition (56) is not satisfied, the
resulting lepton (and hence baryon) asymmetry is suppressed roughly by a factor H(T = M 1 )/Γ1 (up
to logarithm), so that
3 · 10−3 eV
∆B ∼ ∆L ∼ 0.01 · δ · .
|m11 |
This means that the asymmetry is in the right ballpark even for m 11 ∼ 0.05 eV, the value corresponding
to ∆m2 determined by atmospheric neutrino oscillations.
The masses of heavy neutrinos must be quite large. For a crude estimate, let us assume that all
these masses are of the same order of magnitude; let us also assume that all Yukawa couplings are of the
same order. Then, assuming that CP phases are large, we have
1 2
δ∼ h .
8π
We need δ > 10−8 , so
1 2 mν M
h ∼ > 10−8 .
8π 8πv 2
This gives for mν ∼ 0.1 eV the following estimate,

M > 108 GeV

38
234
C OSMOLOGY

which is not unreasonable for the see-saw mechanism. The conclusion is that the mechanism described
can indeed produce the right amount of the baryon asymmetry. Unfortunately, the CP phases responsible
for the asymmetry are generally different from CP phases in the mass matrix of light neutrinos. So,
future observation of CP violation in neutrino oscillations, although it will add weight to the leptogenesis
scenario, it will not be proof of it.
The bottom line is that the observed baryon asymmetry may be explained by a number of mech-
anisms, all of which, however, exist in extensions of the Standard Model only. The problem is that,
with the notable exception of electroweak baryogenesis, direct proof that any given mechanism is in-
deed responsible for the baryon asymmetry does not seem possible in the near future. For reviews of
baryogenesis see, for example, Refs. [39, 41–43].

6.6 Remarks
We have seen in this section that extensions of the Standard Model can explain the dark matter and the
baryon asymmetry of the Universe. Let us stress again, however, that the mechanisms outlined here do
not explain the approximate equality between the baryon mass density and the dark-matter mass density
in the present Universe,
ρDM ∼ ρB , within a factor of 5 . (58)
Indeed, these mechanisms of the generation of baryon asymmetry and dark matter are completely unre-
lated to each other. The approximate relation (58) may well be a coincidence; still, a compelling model
that would explain this coincidence is more than welcome.
Another remark is that some of the most appealing mechanisms for the generation of dark energy
and baryon asymmetry operate at relatively low temperatures, T ∼ 10–100 GeV. The corresponding
extensions of the Standard Model contain new particles in the 100 GeV–1 TeV mass range. Hence at
least some of these ideas will be probed at the LHC. Phrased differently, there is a good chance that the
LHC will open a window to the Universe at temperature 10–100 GeV and age 10 −8 –10−10 s.
The final remark is that the mechanisms that we discussed here operate at the hot stage of the cos-
mological evolution. Therefore, the resulting composition of the cosmic plasma is the same everywhere
in the (visible part of the) Universe, i.e.,
ρB
= independent of x (59)
s
and
ρDM
= independent of x . (60)
s
The temperature itself (and hence entropy density), however, varies in space—these are density per-
turbations in the early Universe. Equations (59) and (60) set the initial data for the evolution of the
cosmological perturbations at the hot stage,

δρB (x) δρDM (x) δs(x)

= = .
ρB ρDM s
Perturbations with this property are called ‘adiabatic’; once baryon asymmetry and dark matter are gen-
erated at the hot stage, there is only the adiabatic mode in the density perturbations.
Off hand, there may be other modes, for example, with

δρB (x) 6= 0 , δT (x) = 0

or
δρDM (x) 6= 0 , δT (x) = 0 .

39
235
 V. A. RUBAKOV

These are called baryon and dark matter isocurvature modes, respectively; they correspond to perturba-
tions in the composition of cosmic plasma without perturbations in the energy density of the relativistic
component. Cosmological observations, notably CMB measurements and galaxy surveys, can discrimi-
nate between different types of modes. Currently, the observations are consistent with the adiabatic mode
only, while the limits on isocurvature modes are not very strong: crudely speaking,

δρisocurvature < 0.2 δρadiabatic .

We stress that the detection of any admixture of baryon and/or dark matter isocurvature modes would
have a profound effect on cosmology: this detection would mean that the baryon asymmetry and/or dark
matter were generated prior to the hot stage by a mechanism quite different from the family of mecha-
nisms discussed in this section. A good candidate would then be the Affleck–Dine mechanism [44].

7 Inflation
The Hot Big Bang theory, although very successful in many aspects, is not free of problems. These have
to do with the initial conditions for the cosmological evolution: the initial data required are very special,
and in several respects very unnatural. This situation is improved dramatically if the Universe had under-
gone an inflationary epoch before the Hot Big Bang stage. In this section we first discuss the motivation
for inflation, and then briefly study mechanisms of inflation and observational predictions of the infla-
tionary theory. A major success of inflation, from the observational point of view, is that it provides a
mechanism of the generation of primordial density perturbations in the early Universe, whose spectrum
is almost flat. The approximate flatness of the spectrum has been confirmed by the measurements of the
angular anisotropy of cosmic microwave background radiation, while many alternative mechanisms of
the generation of density perturbations, like topological defects, are ruled out by the CMB data. Inflation
also predicts a certain spectrum of primordial gravitational waves in our Universe, which, in principle, is
observable through CMB. In these lectures we shall illustrate the basic mechanism of the generation of
density perturbations and gravitational waves at inflation. Of course, the corresponding theory is rather
involved, so our discussion here will be fairly qualitative. For reviews on inflationary cosmology, see
Refs. [45–47].

7.1 Problems of Hot Big Bang theory

Within Hot Big Bang theory, the Universe started its expansion from the singularity. Of course, the
singularity is a property of classical General Relativity, and it may be replaced by something else in full
quantum theory of gravity and matter. In other words, it is not legitimate to extrapolate the evolution,
by making use of the classical Einstein equations, to curvatures and energy densities of the order of the
Planck values or higher. Still, the classical theory is applicable soon after the Planck epoch, and we may
ask what were the properties of the Universe at that time. This is what we mean by initial conditions in
the Hot Big Bang theory.
Off hand, one would think that these initial conditions are more or less random: different parts of
the Universe would have different properties after the Planck epoch, both spatial curvature and energy
density would be close to the Planck values, etc. It is straightforward to see, however, that with such
random initial conditions, Hot Big Bang theory alone would not be able to explain the high homogeneity
and isotropy of the Universe that we see today. When quantifying this statement, one arrives at several
‘problems of Hot Big Bang theory’.

– Horizon problem.
As we know, relic photons were emitted/last scattered when the Universe was rather young, t rec ≈
3 · 105 yrs. In the Hot Big Bang theory, the horizon size at that time was about l hor, rec = 3trec ≈
106 light yrs. In the Hot Big Bang theory, there was no cross-talk, by the time of recombination,

40
236
C OSMOLOGY

between regions separated by a distance larger than l hor, rec , i.e., these regions were not in causal
contact with each other. Hence, there is no reason for such regions to have the same properties,
e.g., the same temperature; CMB photons coming from different regions should a priori have quite
different temperatures.
The present size of the region whose size at recombination was l hor, rec is
today
lhor, rec = lhor, rec · (1 + zrec )
≈ 106 · 103 light yrs
≈ 300 Mpc (61)

while the present horizon size is lH,0 ≈ 104 Mpc. This means that the present angular size of the
horizon at recombination is
ltoday
θrec = hor, rec ≈ 0.03 ≈ 2◦ . (62)
lH, 0
We conclude that there is no reason for angular isotropy of CMB temperature at angular scales
greater than 2◦ . Yet CMB is isotropic to better than 10 −4 ! Why is this so? Why are the initial
conditions for the Hot Big Bang so homogeneous and isotropic even over causally disconnected
regions of space? This is the horizon problem, which cannot be solved in the context of the Hot
Big Bang theory.
– Flatness problem.
The Universe today is almost spatially flat. Quantitatively,
1 1
|Ωcurv | ≡ 8π < 0.02 .
a2 3 Gρc

The curvature contribution into the Friedmann equation,

1 1
|ρcurv | ≡
a2 8π
3 G

scales as 1/a2 , while matter and radiation contributions scale as 1/a 3 and 1/a4 , respectively.
Thus, the curvature term in the Friedmann equation was even less important at earlier epochs, for
example
ρcurv
nucleosynthesis epoch : |Ωcurv | = < 10−16 ,
ρtot
electroweak epoch : |Ωcurv | < 10−26 .
Thus the spatial curvature of the Universe was tiny at the beginning. Why was space so flat
initially? In other words, one can compare the radius of spatial curvature a with the radius of
space–time curvature, the latter being of the order of the inverse Hubble parameter. One obtains,
for example, at the electroweak epoch,

electroweak epoch : a > 1013 H −1 .

Why are the initial conditions such that the radius of the Universe is so large? This ‘flatness
problem’ again cannot be solved within Hot Big Bang theory.
– Entropy problem.
Let us estimate the entropy of the visible part of the Universe, i.e., entropy inside a sphere of size
lH,0 . This entropy is of the order of the number of photons inside this sphere,
3
S ∼ Nγ ∼ nγ lH,0

41
237
 V. A. RUBAKOV

which gives
S ∼ 1088 .
In the Hot Big Bang theory, the expansion of the Universe is almost adiabatic, so this huge entropy
should be built in as an initial condition. Certainly, this initial condition is very special: off hand,
one would rather expect that all dimensionless quantities are roughly of order 1 at the beginning
of the Universe.
– Besides these problems which basically mean that the Hot Big Bang theory does not explain why
our Universe is so large, hot and homogeneous, there is another problem of a different kind. This
is the problem of primordial perturbations. At early times (e.g., at the recombination epoch), the
Universe was not exactly homogeneous: there were density perturbations at the level δρ/ρ ∼ 10 −5 .
These density perturbations grew up, and finally gave rise to structures in our Universe (galaxies,
galactic clusters, etc.). The problem is that in the Hot Big Bang theory, the density perturbations
are to be built in as initial conditions, and there is no way to explain their origin.

Inflation is a dynamical mechanism that makes the Universe large, homogeneous, flat, and hot.
As a bonus, it provides a mechanism for the generation of primordial density perturbations (and also
gravitational waves). These perturbations originate from vacuum fluctuations of quantum fields, which
get enhanced during the inflationary epoch.

7.2 Basic picture

The idea of inflation is that before the Hot Big Bang (but after the Planck era), the Universe was in a
vacuum-like state and underwent the exponential expansion 22 ,
R
Hinfl dt
a(t) = const · e

where Hinfl is almost constant in time. Owing to the exponential expansion, a small patch of the Universe
expands to great size. Say, if the duration of inflation t infl exceeds 140 Hubble times23 ,
140
tinfl >
Hinfl

then a patch of initial Planck size l Pl = 1/MPl ∼ 10−33 cm expands to the size exceeding the present
horizon size lH,0 ∼ 1028 cm. Obviously, the Universe flattens out, any initial inhomogeneities get
diluted and, by the end of inflation, the Universe becomes spatially flat, homogeneous and isotropic at
exponentially large spatial scales. This solves the horizon and flatness problems.
A natural and most popular way to ensure that the Universe expands exponentially is to assume
that the matter at the inflationary stage is in the vacuum-like state characterized by energy density ρ infl
which is almost constant in time. At some point, however, this energy density should transform into the
conventional energy density of hot plasma. This transformation is called reheating, and after reheating
the Hot Big Bang era begins. During reheating, huge entropy is released, and this solves the entropy
problem.
This scenario automatically solves three problems mentioned above, which have to do with hori-
zon, flatness, and entropy. It is not at all obvious that inflation solves the fourth problem of primordial
perturbations, but it does!
22
It is not absolutely necessary for the expansion to be exponential. What is needed is that by the end of inflation, the size of
the cosmological horizon be very large. As an example, power-law behaviour of the scale factor (34) with α > 1 would also
correspond to inflation.
23
In fact, the number of e-foldings required is smaller than 140, since the Universe expanded considerably after the end
of inflation. The minimum number of e-foldings is 50–70, depending on the Hubble parameter at inflation, duration of the
reheating stage, etc.

42
238
C OSMOLOGY

7.3 Slowly rolling inflaton

Currently, the most popular models of inflation invoke a new scalar field—inflaton, which drives inflation
and then automatically ensures an exit from the inflationary stage. Depending on the scalar potential,
several versions of the inflationary scenario have been designed. Their common feature is inflation, the
slow-roll regime, in which the inflaton field slowly changes in time.
Let us assume that ‘at the beginning’ there is a sufficiently large patch in the Universe where the
field φ(x) is reasonably homogeneous. We should stress that these assumptions are rather mild: if ‘the
beginning’ is just after the Planck era, then ‘sufficiently large’ means ‘somewhat larger than the Planck
size’, and ‘reasonably homogeneous’ means that the gradient term in energy density is somewhat smaller
than the potential term. Under these assumptions, one may consider both the metric and scalar field as
homogeneous and isotropic, which means that the metric has the FRW form, and the scalar field does
not depend on spatial coordinates. With non-trivial scalar potential, the action (3) for the scalar field in
the expanding Universe is modified as follows:
Z  
3 3 1 2 1 2
S= dt d x a (t) φ̇ − 2 (∂i φ) − V (φ) .
2 2a (t)

Neglecting the gradient term and then varying the action, one obtains the field equation

ȧ ∂V
φ̈ + 3 φ̇ = − . (63)
a ∂φ
It is worth noting that this equation formally coincides with the equation of classical mechanics for a
‘particle’ with coordinate φ moving with friction in the potential V (φ), the friction coefficient being
3H(t) ≡ 3ȧ/a.
To complete the system of equations, we have to write the Friedmann equation. For the homoge-
neous scalar field, the energy density is
1
ρ = φ̇2 + V (φ) .
2
Hence the Friedmann equation has the following form:
 2  
2 ȧ 8π 1 2
H ≡ = φ̇ + V (φ) . (64)
a 3MPl2 2

The vacuum-like state occurs when the energy density is almost constant in time. This is possible when
the friction in Eq. (63) is so strong that the scalar field barely evolves. Its kinetic energy (1/2) φ̇2 is
small compared to the potential energy V (φ), the latter stays almost constant in time. This is called the
slow-roll regime. The conditions for slow roll are

H φ̇  φ̈ (65)

and
V (φ)  φ̇2 . (66)
If these conditions are satisfied, then the system of equations (63) and (64) simplifies; one has instead

∂V
3H φ̇ = − (67)
∂φ
and
8π
H2 = V (φ) . (68)
3MPl2

43
239
 V. A. RUBAKOV

It follows from Eq. (68) that at large enough V (φ), the Hubble parameter H ≡ ȧ/a is large, and then
it follows from Eq. (67) that the field φ indeed rolls down slowly. The potential V (φ) indeed remains
almost constant, and the Universe expands exponentially. Thus, once the slow-roll regime is ensured,
inflation occurs automatically.
The slow-roll regime terminates when the Hubble parameter H becomes relatively small, so that
the friction term in (63) no longer prevents the scalar field from evolving quickly. Then the inflaton field
quickly rolls down to the minimum of its potential, and possibly oscillates about this minimum. This
produces a scalar field background rapidly varying in time, in which particles (either inflatons themselves,
or particles of other fields) are copiously produced. The latter is the reheating process, which eventually
leads to thermal equilibrium. Of course, one needs to introduce interactions between the inflaton and
conventional matter to have the standard hot era in the end.

7.4 Simple model: one-field inflation

Let us study whether the slow-roll conditions can be satisfied. We consider the simplest potential of the
form shown in Fig. 17. Let the potential at large φ have the power-law behaviour, e.g.,

m2 2
V (φ) = φ or V (φ) = λφ4 .
2
Making use of Eq. (67), and then Eq. (68), and dropping numerical factors of order one, we write
 
2 ∂V 2 1
φ̇ ∼ · 2
∂φ H
 2 2
∂V MPl
∼ . (69)
∂φ V
Therefore, the slow-roll condition (66) takes the form
∂V V
 .
∂φ MPl
For potentials which have power-law behaviour at large φ one has
∂V V
∼
∂φ φ
so the slow-roll condition (66) is satisfied at

φ  MPl .

It is straightforward to see that the first slow-roll condition (65) is also satisfied for power-law potentials
at φ  MPl . Inflation occurs whenever the value of the scalar field is larger than the Planck mass.
It is worth noting that when considering the field values of order of and larger than the Planck
mass one makes quite an extrapolation. One may suspect that there may be contributions to the scalar
potential which are generated by gravitational effects and have the form
X φN +4
CN (70)
N
MPlN

with coefficients CN of the order of one. Such a behaviour would destroy the slow roll; in fact, one can
place strong bounds on CN from the analysis of density perturbations (see below): we shall see that the
correct amplitude of density perturbations is obtained for very flat scalar potentials. This brings up the
issue of the nature of the inflaton field and a mechanism forbidding contributions like (70).

44
240
C OSMOLOGY

Fig. 17: Evolution of the inflaton field

Also, one may worry that even with sufficiently flat scalar potentials, the classical analysis of the
evolution of the Universe is not applicable at φ  M Pl . This is not the case, since, at such large values
of φ, energy density may still be well below the Planck value M Pl4 . For example, consider the quartic
potential, V (φ) = λφ4 , where λ is a dimensionless coupling constant. The energy density at the Planck
value of the scalar field is smaller than the Planck energy density, provided λ  1,

V (φ ∼ MPl ) ∼ λMPl4  MPl4 , for λ  1 .

We shall see that the correct amplitude of density perturbations is obtained when λ is very small indeed,
λ < 10−10 . Taking this value, we see that inflation occurs well below the Planck energies, and our
classical analysis makes sense. Furthermore, we shall point out that there is direct observational evidence
that towards the end of inflation, its energy scale is well below the Planck scale,

H < 10−5 MPl , end of inflation . (71)

This bound has to do with the generation of gravitational waves at inflation and their effect on CMB
anisotropy.
One’s best guess about the initial value of the scalar field φ b is that the energy density at the
beginning of inflation is of the order of the Planck energy density,

V (φb ) ∼ MPl4 .

For the quartic potential this means

φb ∼ λ−1/4 MPl  MPl . (72)

Starting from this value, the field slowly rolls down the potential, until it reaches the value φ e ∼ MPl .
Inflation occurs during all this period of time. Once the field becomes of order φ e ∼ MPl , the slow-roll
regime terminates, the field quickly rolls down to φ ≈ 0, and then oscillates around its minimum φ = 0.
Field oscillations generally lead to particle creation from vacuum, so the oscillations get damped, and
the Universe begins to be filled with particles. This is a reheating process, which ends when the coherent
oscillations of the scalar field terminate, the classical scalar field settles down to the minimum of the
scalar potential, and particles created by the oscillations get in thermal equilibrium. After the system
thermalizes, it is described by the Hot Big Bang theory.
The reheating process is quite complex, and may occur in several stages. We shall not discuss
it here; on general grounds it is clear that the outcome of this process is a thermal state anyway. The

45
241
 V. A. RUBAKOV

end of reheating is at the same time the beginning of the Hot Big Bang. A most naive (and, in fact,
unrealistic) estimate of the temperature of the Universe after reheating is obtained if one assumes that
reheating takes of order one Hubble time. Under this assumption, all energy density of the inflaton field
V (φe ) transforms into heat, and the energy density is not substantially diluted, because of the expansion
of the Universe, during reheating. This picture implies that the Hubble parameters at the end of inflation
and in the beginning of the Hot Big Bang are of the same order,

T2
Hend of inflation ∼
MPl∗

where we made use of the standard formula for the Hubble parameter at the radiation-dominated stage.
According to this estimate, the temperature of the Universe at the beginning of the Hot Big Bang may be
quite high: a model-independent bound comes from Eq. (71), which gives
p
T < 3 · 10−3 MPl MPl∗ ∼ 1016 GeV .

More realistic estimates give the reheat temperature which is several orders of magnitude lower, because
reheating takes more than one Hubble time. In fact, one can design inflationary models with arbitrarily
low reheat temperature.
Let us see that the inflationary stage naturally lasts long, so that the scale factor increases a lot
during inflation. In this way we make sure that the three problems of the Hot Big Bang theory (horizon,
flatness, and entropy) are naturally solved. During inflation, the scale factor increases by
ae
= eNe-folds
ab
Z te
Ne-folds = Hdt (73)
tb

where subscripts b and e refer to the beginning and end of inflation. We obtain from Eqs. (67) and (68)
that the number of e-foldings may be written as follows (again omitting factors of order one),
Z φb
H2
Ne-folds ∼ dφ
φe ∂V /∂φ
Z φb
V (φ)
∼ dφ . (74)
φe MPl2 (∂V /∂φ)

For power-law potentials, and φb  φe one estimates

Z φb
φdφ
Ne-folds ∼
φe MPl2
φ2b
∼ . (75)
MPl2

We have seen that at the beginning of inflation, the scalar field is naturally very large, φ b  MPl . There-
fore, the number of e-foldings is indeed naturally very large. As an example, for the quartic potential we
have an estimate (72), and taking λ ∼ 10 −10 we estimate

Ne-folds ∼ 105 .

The size of the Universe after inflation in this case is of order

a ∼ eNe-folds ∼ 10100 000 (76)

46
242
C OSMOLOGY

(it does not matter in which units!); this is more than enough to solve the Hot Big Bang problems.
One remark is in order. When discussing the Hot Big Bang theory, we saw that the horizon size
was of the order of the Hubble distance, H −1 . If there was inflation before the Hot Big Bang, this
estimate is no longer valid: the actual horizon size is much larger, as the size of the causally connected
region of space increased during inflation by a factor e Ne-folds . In particular, the present Hubble volume of
the size of 104 Mpc makes only a small fraction of the true horizon volume. Yet it is the distance of order
104 Mpc from which the earliest electromagnetic radiation—relic photons—reaches us. In this sense the
distance 104 Mpc is still the size of the visible part of the Universe even in theories with an inflationary
stage.
One more point is that the estimates like (76) show that inflation naturally predicts that the spatial
curvature of the Universe is extremely small today, i.e., ρ = ρ c with extremely high precision. At some
point many cosmologists believed that the Universe was open, with ρ ≈ 0.3ρ c . This was a problem for
most inflationary models. The situation was cleared up with the data on CMB anisotropy, which tell us
that ρ = ρc within 2%, and show no evidence for spatial curvature.
Finally, we stress that one-field inflation studied in this subsection is not at all the only model of
inflation (and it was not the first historically). Gross features of an inflationary scenario are quite similar
in various models. Their detailed predictions for density perturbations and gravitational waves differ,
however, so detailed measurements of the properties of CMB and large-scale structure make it possible
to discriminate between various inflationary models.

8 Generation of primordial perturbations at inflation

8.1 Fluctuations of the inflaton field
In the inflationary scenario, density perturbations are generated from vacuum fluctuations of the inflaton
field, for reviews see Refs. [48–50]. In this subsection, we illustrate the main idea and obtain the basic
formulas for the scalar field perturbations. In effect, we shall be dealing with quantum field theory of the
free scalar field in time-dependent background, and study pair creation. It is amazing that such a simple
mechanism is likely to be the origin of all the structure of our Universe.
To warm up, let us estimate the amplitudes of vacuum fluctuations of the massless scalar field ϕ
in Minkowski space–time. The energy functional reads
Z  
1 1
E= d3 x ϕ̇2 + (∇ϕ)2 . (77)
2 2
Let us denote by δϕvac the vacuum amplitude at the momentum scale p (length scale λ ∼ p −1 ). To
estimate this amplitude, we note that the relevant frequency is ω p = p, and the typical volume is λ3 ∼
p−3 . We require that the energy (77) of a fluctuation in this volume be of the order of the zero-point
energy of a field oscillator,
1
E ∼ ωp .
2
This gives the relation
λ3 · ωp2 · (δϕvac )2 ∼ ωp .
In this way we obtain the estimate
δϕvac ∼ ωp = p . (78)
Small-frequency, long-wavelength vacuum fluctuations have small amplitudes, and vice versa.
To substantiate this estimate, let us make use of the standard expression for the free-field operator
in Minkowski space–time,
Z  
d3 p iωp t−ipi xi † −iωp t+ipi xi
ϕ(t, x) = p e A p + e A p (79)
(2π)3/2 2ωp

47
243
 V. A. RUBAKOV

where A†p and Ap are the creation and annihilation operators obeying

[Ap , A†p0 ] = δ(p − p0 ) . (80)

It is straightforward to calculate the variance of the field in vacuo,

Z
2 d3 p
hϕ (x)i =
(2π)3 2ωp
Z ∞
1 dp
= 2
p2 . (81)
(2π) 0 p
We see that the contribution of a decimal place in momenta into the variance of ϕ is indeed estimated as
(δϕ)2 where δϕ is given by (78), more precisely,
p
δϕvac = . (82)
2π
Needless to say, these amplitudes are tiny at cosmologically interesting scales, p −1 ∼ Mpc or so.
One remark is in order. From (79) it follows that the vacum fluctuations make up a Gaussian field:
the multi-point correlation functions factorize into products of two-point functions; as an example, for
equal-time correlators one has

hϕ(x1 )ϕ(x2 )ϕ(x3 )ϕ(x4 )i = hϕ(x1 )ϕ(x2 )ihϕ(x3 )ϕ(x4 )i + permutations .

In quantum field theory this property is known as Wick’s theorem.

Let us now turn to perturbations of the inflaton field at the inflationary epoch. We begin with
dividing the quantum inflaton field φ̂ into its classical part and quantum fluctuation operator,

φ̂(x, t) = φ(t) + ϕ(x, t)

where the classical part φ(t) is precisely the object we discussed in the previous subsection, and ϕ is the
operator describing small perturbations. We are going to develop linear theory in ϕ. Furthermore, we
are going to neglect the curvature of the scalar potential, as it is small at the inflationary stage. Finally,
for the time being we approximate the Hubble parameter H at inflation by a constant, i.e., neglect its
dependence on time. Introducing the weak dependence of H on time is not difficult; we shall comment
in appropriate places on the effects coming from the time dependence of H.
Under these assumptions, the action for perturbations coincides with the action (3), and the field
equation is precisely (4). Let us write those formulas here again,
Z  
3 3 1 2 1 2
S= dt d x a (t) (ϕ̇) − 2 (∂i ϕ) (83)
2 2a (t)
and, in momentum representation, the field equation is

k2
ϕ̈k + 3H ϕ̇k − ϕk = 0 . (84)
a2
Recall that k is the time-independent coordinate momentum, the physical momentum being
k
p(t) = .
a(t)
Properties of solutions to this equation depend on whether the physical momentum is greater or smaller
than H. Modes with
p(t)  H

48
244
C OSMOLOGY

are called subhorizon, as their wavelengths λ = 2π/p are shorter than the Hubble length (horizon 24 size)
H −1 . Modes with
p(t)  H
are superhorizon. The WKB solution (6),
const i R ω(t)dt
ϕk (t) = e (85)
a(t)
applies to subhorizon modes only. For superhorizon modes, the friction is so strong that they stay con-
stant in time25 ,
ϕk = const , p  H .
The Hubble parameter stays (almost) constant at inflation, while the physical momentum gets red-shifted.
Thus a mode of given k is first subhorizon and then superhorizon. This is what inflation is about: short
scales get stretched beyond the Hubble radius.
It is worth noting that this sequence of events for a mode with fixed k is specific to inflation. In a
radiation-dominated and a matter-dominated Universe, the scale factor behaves as a ∝ t 1/2 and a ∝ t2/3 ,
respectively. Thus, the physical momentum red-shifts as
1 1
p∝ and p ∝
t1/2 t2/3
while the Hubble parameter behaves as
1
. H∝
t
Thus at the radiation- or matter-dominated stage, a mode with given k is first a superhorizon, and then
becomes a subhorizon. This is just the opposite to the situation at inflation.
Coming back to the inflationary stage, we see that for modes of any given k, the expansion of the
Universe is effectively adiabatic at early times, so these modes experience vacuum fluctuations like in
Minkowski space–time, and their amplitudes are of order ϕ ∼ p(t). When a mode of a given k is still
a superhorizon, its amplitude decreases in time because of the redshift. These fluctuations get frozen in
at the time when they exit the horizon. At the radiation- or matter-dominated epoch these modes would
re-enter the horizon and would start to oscillate again; at this moment their amplitudes would be much
greater than the amplitudes of vacuum fluctuations at the same frequency 26 .
As the modes cross out the horizon when ω(t) = p(t) ∼ H, the field amplitude freezes in at the
inflationary stage when ϕ ∼ p ∼ H, and no longer decreases, even though the physical wavelength
continues to increase. Relative to vacuum, the amplitude increases like a(t), and is thus enhanced by
a huge factor. We immediately obtain an estimate for the amplitudes of perturbations of superhorizon
modes, created from vacuum,
ϕ∼H, pH
and infer that these amplitudes are (almost) independent of wavelengths (flat, Harrison–Zeldovich spec-
trum).
To obtain quantitative estimates, let us write the WKB solution for the operator of the quantum
field ϕ at the early stage of inflation as follows,
Z !
d3 k 1 1 R i
ϕ(t, x) = p ei ω(t)dt−iki x A†k + h.c. (86)
(2π)3/2 a3/2 2ω(t)
24
The term ‘horizon’ here refers to the de Sitter horizon, not to be confused with the cosmological horizon. In fact, what
matters here is the length scale H −1 .
25
Another superhorizon solution rapidly decays in time, and is therefore irrelevant.
26
In fact, our analysis may not apply to the period after inflation, as the classical part φ is small at that time, and the curvature
of the scalar potential may not be negligible. However, we will need the behaviour of the fluctuations at the inflationary stage
only.

49
245
 V. A. RUBAKOV

where ω(t) = |p(t)| = k/a(t). To write this solution, we made use of two points:
(i) interesting modes are subhorizon at early stages of inflation, so they behave as given in (85):
√
the correspondence follows from the fact that a 3/2 ω ∝ a;
(ii) prefactor a−3/2 compensates for the volume factor in the action (83); with this prefactor,
the action coincides with the usual action for the scalar field in Minkowski background, modulo slow
dependence of the physical momentum (and hence frequency) on time.
Now, the solution (86) is valid only for subhorizon modes. At the moment t k of horizon crossing,
k
= H(tk ) (87)
a(tk )
the mode of the coordinate momentum k gets frozen in, so after that it stays constant. This constant is
obtained by matching the solution (86) to a constant solution at t = t k . In this way we find that after the
mode becomes superhorizon, it is equal to (up to a phase)
1 1 1
ϕk = 3/2 3/2
p A†k + h.c.
(2π) a (tk ) 2H(tk )
Eliminating a(tk ) via (87), we end up with the following expression for the contribution of superhorizon
modes into the field operator (again up to an unimportant phase in the integrand),
Z
d3 k H(tk ) †
ϕ(x) = √ (Ak + h.c.) . (88)
(2πk)3/2 2
Making use of this expression, one can calculate all correlation functions involving superhorizon modes.
One finds that these are the Gaussian fluctuations with an almost flat (i.e., scale-independent) spectrum.
For example, for time-independent H one finds
Z
2 H2 d3 k
hϕ i =
2 (2π)3 k 3
2 Z
H dk
= 2
. (89)
(2π) k
This corresponds to fluctuations of scale-independent amplitude H/(2π) in a decimal interval of wave-
lengths.
Two remarks are in order. First, the Hubble parameter entering the fluctuation spectrum (89) is
the Hubble parameter at inflation. In fact, this is the Hubble parameter towards the end of inflation, as
relevant fluctuations of the scalar field cross out the horizon towards the end of inflation. The spectrum is
exactly flat only if the Hubble parameter is constant in time; if this is not so, the spectrum is slightly tilted
(there is extra dependence on k). The reason is that fluctuations of different momenta k cross out the
horizon at different times, and the relevant value of the Hubble parameter is the value at the time when
a fluctuation of a given wavenumber crosses out the horizon, see (87). This tilt is different in different
models of inflation; we shall come to this point later.
Second, when writing (86) we made an implicit assumption that the modes are described by the
usual quantum field theory at very early times. This is certainly questionable, if ‘very early times’ means,
say, the beginning of inflation, tb . Indeed, at that time the physical wavelength of every interesting mode
was extremely short,
a(tb )
λ(tb ) = λ0
a0
(superscript 0 still means ‘present’). In view of estimates like (76), the wavelength λ(t b ) is naturally
many orders of magnitude smaller than the Planck length, for example. It is certainly of interest to
understand how robust the predictions of inflation are with respect to possible new effects at such short
distances. For a review of the work in this direction see, for example, Ref. [49].

50
246
C OSMOLOGY

8.2 Density perturbations

How does the above discussion relate to density perturbations? An intuitive way to understand the
creation of density perturbations by fluctuations of the inflaton field is as follows. Let us consider a
region of the Hubble size towards the end of inflation. In this region, the actual value of the scalar field
is
φact = φ + ϕ
where φ is the unperturbed value determined by the classical field equations; φ is homogeneous over the
whole inflating Universe. Now, ϕ is a linear combination of scalar field fluctuations of superhorizon size.
Because of the second term, the scalar field is different in different regions of the Hubble size; if ϕ > 0,
the field is larger than the average over the whole Universe, and vice versa. The Hubble regions evolve
in time independently, so a region where ϕ > 0 exits from inflation later than on average. The delay time
is
ϕ
δt = .
φ̇
This delay leads to higher energy density, after inflation, in the region we look at: in other regions there
is more time for the density to get diluted due to the expansion. We obtain the density perturbation

δρ = ρ̇δt

where
ρ̇ ∼ Hρ .
Combining all factors, we arrive at scales which at the end of inflation exceed the inflationary Hubble
size
δρ H
(x) = ϕ(x) .
ρ φ̇
This means that primordial density perturbations are proportional to fluctuations of the inflaton field; they
form a random (Gaussian) field whose amplitude in a decimal interval of wavelengths is [see Eq. (89)]

δρ H2
∼ . (90)
ρ φ̇

If H and φ̇ were constant in time, this would be a flat, Harrison–Zeldovich spectrum with no preferred
length scale; as we discussed at the end of the previous subsection, the spectrum is in fact slightly tilted,
the tilt being model dependent.
To obtain the correct magnitude for primordial density perturbations,
δρ
∼ 10−5 , (91)
ρ
one has to tune the parameters of the inflaton potential. As an example, let us consider one-field inflation
with quartic potential, discussed in the previous section. In the slow-roll approximation, one has
8π
H2 = V (φ)
3MPl2

and
1 ∂V
φ̇ = − .
3H ∂φ
This gives " #
δρ V 3/2
∼ .
ρ MPl3 |∂V /∂φ|
end of inflation

51
247
 V. A. RUBAKOV

Inflation ends when φ is about MPl , so for the quartic potential V = λφ4 one has a crude estimate
δρ √
∼ λ.
ρ
The right magnitude of the density perturbations, Eq. (91), is, roughly speaking, obtained for

λ ∼ 10−10 .

In fact, one should take into account the fact that the modes which currently have the cosmologically in-
teresting wavelengths crossed out the horizon some time (50 to 70 e-foldings) before the end of inflation.
This gives an improved estimate,
λ ∼ 10−12 .
This tiny value of the self-coupling is not a peculiarity of the quartic potential: generally speaking, small
density perturbations require a flat inflaton potential.
Coming back to the tilt, the traditional way to parametrize the spectrum is 27
δρ ns −1
(k) = As k 2
ρ
where As and ns are the amplitude and spectral index. The flat spectrum corresponds to n s = 1, the
blue and red tilted spectra have ns > 1 and ns < 1, respectively. Let us again consider an example of
one-field inflation. Using the slow-roll equations, one finds from Eq. (90)
δρ V 3/2
∝ . (92)
ρ V0
For a mode of momentum k the quantities entering the right-hand side of (92) are to be evaluated at the
time it exits the horizon, which is determined by (87). Modes with higher momenta exit the horizon
later, at later times the value of φ is smaller, hence for power-law potentials the right-hand side of (92)
is smaller. Thus, one-field models predict less power at high momenta, i.e., a slightly red spectrum of
primordial density perturbations. Other models may give rise to a slightly blue spectrum; in any case, a
non-vanishing tilt is an interesting and quite generic prediction of inflationary models.
The primordial density perturbations stay constant until they re-enter the horizon at the radiation-
dominated or matter-dominated stage. After that they make sound waves. This picture, together with
the flatness of the spectrum of density perturbations, immediately implies that small structures (galaxies)
form earlier than larger structures (clusters): shorter wavelengths re-enter the horizon earlier, and hence
smaller structures start to develop earlier. This general prediction is in accord with observational data on
the structure in the Universe.

8.3 Gravity waves: the scale of inflation

Inflation creates not only scalar field perturbations, but also gravitational waves (cosmologists often call
them tensor perturbations, while density perturbations are scalar in the three-dimensional sense). The
mechanism is precisely the same as that outlined in Section 8.1, and, in fact, the equation for gravita-
tional waves (tensor perturbations) is literally the same as the equation for perturbations of the massless
scalar field. The action for metric perturbations, however, contains an extra factor 1/G = M Pl2 , so the
canonically normalized field differs from metric perturbation by the factor M Pl . With this qualifica-
tion, we make use of Eq. (89) to obtain the amplitude of gravitational waves in a decimal interval of
wavelengths,
H
h= (93)
2πMPl
fi“ ”2 fl
27
R
In the sense that δρ
ρ
= dk 2 (ns −1)
k
As k .

52
248
C OSMOLOGY

where H is the Hubble parameter towards the end of inflation. The primordial spectrum of the gravita-
tional waves is almost flat, like the spectrum of the scalar field fluctuations, see Eq. (89).
Primordial gravitational waves make a contribution to the CMB anisotropy. This contribution is
potentially important at large angular scales, ∆θ > (a few) ◦ . At these scales,
 
δT
∼h.
T grav. waves

The very fact that the CMB anisotropy δT /T does not exceed 10 −5 at large angular scales implies the
bound on the Hubble parameter towards the end of inflation,
H < 10−5 MPl , end of inflation .
We have already discussed the significance of this bound in Section 7.
1.0
WMAP WMAP + SDSS
0.8

N= N=
0.6 HZ HZ
N= 60 N= 60
r 0.002

50 50
0.4

0.2

0.0
1.0
WMAP + 2dF WMAP + CBI + VSA
0.8

N= HZ N= HZ
0.6
N= 60 N= 60
r 0.002

50 50
0.4

0.2

0.0
0.90 0.95 1.00 1.05 0.90 0.95 1.00 1.05
ns ns

Fig. 18: Contour plots (68% and 95%) allowed by WMAP (left upper panel) and WMAP combined with other
measurements, in the plane (ns , r) where r is the tensor-to-scalar ratio. Solid and dashed lines are predictions of
one-field inflationary models for different durations of the preheating stage. Open and filled circles are predictions
of m2 φ2 and λφ4 models, respectively. The rectangle denotes the pure Harrison–Zeldovich spectrum, n s = 1,
r = 0 (flat spectrum of primordial density perturbations, no gravity waves).

In most inflationary models, the contribution of gravitational waves into the CMB anisotropy is
smaller (sometimes much smaller) than the contribution of density perturbations, even at large angular
scales. Indeed, let us introduce the tensor-to-scalar ratio
 
h 2
r= .
δρ/ρ
Making use of (90) and (93) we obtain
" #2
φ̇
r ∼
HMPl

53
249
 V. A. RUBAKOV

φ̇2
∼
V

where we used the slow-roll equations to obtain the last estimate. At slow roll, φ̇2  V , so a generic
prediction of inflation is that
r<1.
Often, however, r is not extremely small (it can easily be at the 10% level), as the modes of interest cross
out the horizon towards the end of inflation, when the slow-roll conditions are about to be violated.
There is definitely a chance to detect the contribution of gravitational waves, and discriminate it
from the contribution of density perturbations, in future measurements of the CMB polarization. In this
way one would be able to find the scale of inflation, and possibly even reconstruct part of the inflaton
potential [51].
Recent WMAP data [19] seem to provide the first evidence for inflation. The point is that these
data appear to require either a tilt in the spectrum of scalar perturbations, or a contribution of gravity
waves, or both, see Fig. 18. If confirmed, this result would be extremely exciting for cosmology, and for
physics in general.

9 Conclusions
At first sight, our Universe appears infinitely complex. Yet, with not so many parameters, a coherent
picture of the present and past Universe emerges, which has already passed precision tests of CMB
anisotropy, Big Bang Nucleosynthesis, structure formation, etc. Even more precise measurements are
due to come, which makes the whole field lively and fascinating.
Even the gross features of cosmology are ‘orthogonal’ to the Standard Model of particle physics:

– Most of the energy in the present Universe is in unknown forms. Furthermore, cosmology requires
the existence of both new stable particles (clumped non-baryonic dark matter) and dark energy.
The latter is the most fundamental and mysterious of all aspects of cosmology, as we know it today.
– Inflation needs an inflaton, a new scalar field with a very flat potential. No such field exists in the
Standard Model, nor does it emerge naturally in the simplest extensions of the Standard Model.
– Baryogenesis needs new sources of CP-violation and mechanisms of baryon and/or lepton number
violation.

Cosmology certainly has its own intrinsic problems, some of which have been mentioned in these
lectures. We all know that the Standard Model has its intrinsic problems too. Experiments and theory in
particle physics and cosmology still have a lot to tell us about the micro and the macro world, as well as
about the interconnections between them.

References
[1] A. D. Linde, Particle Physics And Inflationary Cosmology (Harwood, Chur, 1990), Contemporary
concepts in physics, 5.
[2] E. W. Kolb and M. S. Turner, The Early Universe (Addison-Wesley, Redwood City, 1990), Frontiers
in physics, 69.
[3] G. G. Raffelt, Stars as Laboratories For Fundamental Physics: The Astrophysics Of Neutrinos,
Axions, And Other Weakly Interacting Particles (Chicago University Press, 1996).
[4] J. A. Peacock, Cosmological Physics (Cambridge University Press, 1999).
[5] S. Dodelson, Modern Cosmology (Academic Press, Amsterdam, 2003).
[6] V. Mukhanov, Physical Foundations of Cosmology (Cambridge University Press, 2005).

54
250
C OSMOLOGY

[7] M. S. Turner and J. A. Tyson, Rev. Mod. Phys. 71 (1999) S145 [arXiv:astro-ph/9901113].
[8] M. Shaposhnikov, Cosmology and astrophysics, Proc. 2000 School of High-Energy Physics, Cara-
mulo, Portugal, CERN–2001–003, J.D. March-Russell and N. Ellis (Eds.).
[9] P. J. E. Peebles and B. Ratra, Rev. Mod. Phys. 75 (2003) 559 [arXiv:astro-ph/0207347].
[10] W. L. Freedman and M. S. Turner, Rev. Mod. Phys. 75 (2003) 1433 [arXiv:astro-ph/0308418].
[11] L. M. Krauss, The Standard Model, dark matter, and dark energy: from the sublime to the ridicu-
lous, arXiv:astro-ph/0406673.
[12] I. I. Tkachev, Astroparticle physics, arXiv:hep-ph/0405168.
[13] A. de Oliveira-Costa, G. F. Smoot and A. A. Starobinsky, Constraining topology with the CMB,
arXiv:astro-ph/9705125;
W. N. Colley and J. R. I. Gott, Mon. Not. Roy. Astron. Soc. 344 (2003) 686 [arXiv:astro-
ph/0303020];
N. G. Phillips and A. Kogut, Constraints on the topology of the Universe from the WMAP first-year
sky maps, arXiv:astro-ph/0404400.
[14] D. P. Schneider et al. (SDSS Collaboration), Astron. J. 123 (2002) 567,
http:www.sdss.org/dr1/algorithms/edrpaper.html#fig-ZhistQso
[15] A. V. Filippenko and A. G. Riess, Phys. Rep. 307 (1998) 31 [arXiv:astro-ph/9807008].
[16] A. V. Filippenko and A. G. Riess, Evidence from type Ia supernovae for an accelerating Universe,
arXiv:astro-ph/0008057.
[17] A. G. Riess et al. (Supernova Search Team Collaboration), Astrophys. J. 607 (2004) 665
[arXiv:astro-ph/0402512].
[18] D. N. Spergel et al., Astrophys. J. Suppl. 148 (2003) 175.
[19] D. N. Spergel et al., Wilkinson Microwave Anisotropy Probe (WMAP) three year results: implica-
tions for cosmology, arXiv:astro-ph/0603449.
[20] T. J. Pearson et al., Astrophys. J. 591 (2003) 556.
[21] C. l. Kuo et al. (ACBAR Collaboration), Astrophys. J. 600 (2004) 32.
[22] S. Eidelman et al. (Particle Data Group), Phys. Lett. B 592 (2004) 1.
[23] N. Arkani-Hamed, S. Dimopoulos and G. R. Dvali, Phys. Rev. D 59 (1999) 086004 [arXiv:hep-
ph/9807344].
[24] V. A. Rubakov, Usp. Fiz. Nauk 171 (2001) 913 [arXiv:hep-ph/0104152.]
[25] U. Seljak et al., Cosmological parameter analysis including SDSS Ly-alpha forest and galaxy bias:
constraints on the primordial spectrum of fluctuations, neutrino mass, and dark energy, arXiv:astro-
ph/0407372.
[26] V. Sahni and A. Starobinsky, Int. J. Mod. Phys. D 9 (2000) 373 [arXiv:astro-ph/9904398].
[27] K. G. Begeman, A. H. Broeils and R. H. Sanders, Mon. Not. Roy. Astron. Soc. 249 (1991) 523.
[28] S. Weinberg, Rev. Mod. Phys. 61 (1989) 1.
[29] S. Weinberg, The cosmological constant problems, arXiv:astro-ph/0005265.
[30] L. Susskind, The anthropic landscape of string theory, arXiv:hep-th/0302219.
[31] C. Wetterich, Nucl. Phys. B 302 (1988) 645.
[32] P. J. E. Peebles and B. Ratra, Astrophys. J. 325 (1988) L17.
[33] B. Ratra and P. J. E. Peebles, Phys. Rev. D 37 (1988) 3406.
[34] A. Nicolis and R. Rattazzi, JHEP 0406 (2004) 059 [arXiv:hep-th/0404159].
[35] V. A. Kuzmin, Phys. Part. Nucl. 29 (1998) 257 [Fiz. Elem. Chast. Atom. Yadra 29 (1998) 637]
[arXiv:hep-ph/9701269].
[36] A. Bottino and N. Fornengo, Dark matter and its particle candidates, arXiv:hep-ph/9904469.

55
251
 V. A. RUBAKOV

[37] J. R. Ellis, J. L. Feng, A. Ferstl, K. T. Matchev and K. A. Olive, Prospects for detecting supersym-
metric dark matter at post-LEP benchmark points, arXiv:astro-ph/0110225.
[38] V. A. Rubakov, Classical Theory of Gauge Fields (Princeton University Press, 2002).
[39] V. A. Rubakov and M. E. Shaposhnikov, Usp. Fiz. Nauk 166 (1996) 493 [Phys. Usp. 39 (1996) 461]
[arXiv:hep-ph/9603208].
[40] M. Fukugita and T. Yanagida, Phys. Lett. B 174 (1986) 45.
[41] A. D. Dolgov, Baryogenesis, 30 years after, arXiv:hep-ph/9707419.
[42] A. Riotto and M. Trodden, Annu. Rev. Nucl. Part. Sci. 49 (1999) 35 [arXiv:hep-ph/9901362].
[43] W. Buchmuller and M. Plumacher, Neutrino masses and the baryon asymmetry, arXiv:hep-
ph/0007176.
[44] I. Affleck and M. Dine, Nucl. Phys. B249 (1985) 361.
[45] A. D. Linde, Phys. Rep. 333 (2000) 575.
[46] A. Albrecht, Cosmic inflation, arXiv:astro-ph/0007247.
[47] G. Lazarides, Inflationary cosmology, arXiv:hep-ph/0111328.
[48] V. F. Mukhanov, H. A. Feldman and R. H. Brandenberger, Phys. Rep. 215 (1992) 203.
[49] R. H. Brandenberger, Theory of cosmological perturbations and applications to superstring cos-
mology, arXiv:hep-th/0501033.
[50] R. H. Brandenberger, Particle physics aspects of modern cosmology, arXiv:hep-ph/9701276.
[51] J. E. Lidsey, A. R. Liddle, E. W. Kolb, E. J. Copeland, T. Barreiro and M. Abney, Rev. Mod. Phys.
69 (1997) 373 [arXiv:astro-ph/9508078].

56
252