Sie sind auf Seite 1von 154

Lecture Notes in Cosmology

Armen Sedrakian
Institute for Theoretical Physics,
Frankfurt University,
D-60438, Frankfurt am Main, Germany
February 9, 2011
2
Chapter 1
Expanding Universe
Cosmology studies the large-scale nature of the material world around us by methods of natural sciences
(i.e., astronomy, physics, mathematics, etc). We can say cosmology studies the universe itself.
We need to realize that the human race occupies a small fraction of the time and space of the
universe. Therefore, we are able to obtain information only on the part of the universe. Obviously, to
describe the entire universe, with all its parts that are not accessible to us, we need hypothesis which
will allow us to extrapolate from what we know about our space-time to the entire universe. We will
start with the review of the empirical data that is available on the universe.
1.1 Hubble law
The most important feature of our universe is that, in a rst approximation, the universe is homogeneous
and isotropic. This statement is known as the cosmological principle. It implies that observations made
on the Earth can be, in fact, used to test cosmological models. We observe about 3000 Mpc of the
Universe. The units of length that we will use are related to each other by
1 Mpc = 3.26 10
6
light years 3.08 10
24
cm.
Observations of this segment of our universe suggest that the cosmological principle is true when av-
eraging is carried out over the scales in the range 100 l 3000 Mpc. On smaller scales there exist
large inhomogeneities - galaxies, clusters of galaxies, and superclusters of galaxies. Within a galaxy the
matter is distributed very inhomogeneously, as is evident from observations of stars, planets, and other
objects in our galaxy.
The next most important feature of our universe is that it expands. This expansion is according
to the Hubble law. Let us state this law in the Newtonian theory. Suppose an observer is at rest at
the origin of the system of coordinates K. According to Hubbles law the velocity of any point B with
respect to the observer is given by
v
B
= H(t)r
B
, (1.1)
where r
B
is the radius vector of the point B and the parameter H(t) is called the Hubble parameter;
sometimes it is called Hubble constant to stress that it is independent of the spatial coordinates; note,
however, that it is time-dependent. If we take another point in the same system of coordinates, say
C, then v
C
= H(t)r
C
. Consider now a second system of coordinates K

which moves with respect to


the rst one with the velocity v
B
(i.e. in which the particle B is at rest). According to the Galilean
transformations
v
C
= v

C
+v
B
(1.2)
3
4 CHAPTER 1. EXPANDING UNIVERSE
where the prime indicates that the quantity is measured in the reference frame K

. Then we nd
v

C
= v
C
v
B
= H(t)(r
C
r
B
) = H(t)r

C
. (1.3)
Thus, we see that the expansion law is independent of the reference frame, i.e., from both reference
frames we see the same expansion according to the Hubble law. A useful way to envision the Hubble
Figure 1.1: Galaxy distribution in the nearby universe (gure from Harvard Smithsonian Center for
Astrophysics Redshift survey). Each point is a galaxy. The three dimensional volume which is 900
millions light years deep is split into 3 slices. Each slice covers a region 300 300 Mpc.
expansion is the two-dimensional surface of a sphere. Suppose we have a sphere, whose radius a(t) is
increasing with time. The angle between two xed points A and B on the sphere, which we denote as

AB
, remains obviously constant. The distance between the two points on the sphere is given by
r
AB
(t) = a(t)
AB
. (1.4)
The relative velocity at which the points A and B move is then given by
v
AB
= r
AB
(t) = a(t)
AB
. (1.5)
Eliminating the angle from the last equation using (1.4) we obtain
v
AB
(t) =
a(t)
a(t)
r
AB
(t) = H(t)r
AB
(t). (1.6)
We see that the Hubble law emerges naturally upon identication H = a/a.
1.2. DYNAMICS OF DUST IN NEWTONIAN COSMOLOGY 5
Let us go back to the Hubble law in the general 3d case. We see that
v = H(t)r,

r = H(t)r,
dr
r
= H(t)dt, ln r =
_
H(t)dt + C, (1.7)
where C is the integration constant. Thus we obtain
r = exp
__
H(t)dt
_
e
C
= a(t) (1.8)
where a(t) is called scale factor (which is the analogue of the radius of the 2d sphere in the example
above) and = e
C
is the integration constant which is xed by the distance between two points at any
given time. It is called the Lagrangian coordinate (or comoving coordinate of point B in the reference
frame xed with A).
We want to stress that the Hubble law is valid over the scales where the universe is homogeneous.
On smaller scales the motion is dominated by the inhomogeneities in the gravitational eld; for example
the local motion can have orbital nature. The velocity of an object relative to the comoving coordinates
are called peculiar velocity.
How the Hubble constant can be measured? If the peculiar velocities of two objects are small, we
can measure the recession velocity via Doppler shift in spectral lines. If we have also a reliable measure
of the distance we can then measure the Hubble constant (more precisely its current value) via the
Hubble law Eq. (1.1). There are two methods that allow for accurate measurements of the distances
in the space and hence the Hubble constant; these are based on the so called standard candles and
standard rulers.
Standard candles are astrophysical objects which all have about the same luminosity. There are
two prominent examples: (1) the Cepheid variable stars, which pulse at periodic rate; (2) Type-IA
supernovae, which are bright, exploding stars, which show characteristic spectral pattern. Type-IA are
better standard candles because they can be observed to much greater distances than Cepheids.
Furthermore, the distance to nearby objects are measured directly by parallax. The inverse square
law related the apparent luminosity of distant objects to that of nearby one whose distance is known.
The method based on standard rulers is the same but one identies a class of objects with the same
size instead of luminosity.
The current value of the Hubble constant is H 65 80 km s
1
Mpc
1
. With this value we
can estimate the age of the Universe if we neglect the eects of gravity and assume that the velocity
remains constant over time. Then, if two points are separated by r today, they were coincident at
t
0
= [r[/v = 1/H(t) 1/H time ago, which gives t
0
1.5 10
10
yr.
1.2 Dynamics of dust in Newtonian cosmology
Consider innite, expanding, homogeneous and isotropic universe. We shall assume that the universe
is lled with dust, which means that its pressure p , where is the energy density. Our initial
study will be based on Newtonian dynamics, i.e., all the velocities are assumed to be non-relativistic.
Consider a sphere within such universe. Its radius changes in time (expands) according to
R(t) = a(t)
com
, (1.9)
where a(t) is the scale factor,
com
is the comoving coordinate. The total mass within the sphere is
conserved M = const (i.e., is independent of time). Thus we can write
(t) =
M
(4/3)R
3
(t)
=
M
(4/3)a
3
(t)
3
com
=
M
(4/3)a
3
0
(t)
3
com
_
a
3
0
(t)
a
3
(t)
_
=
0
_
a
3
0
(t)
a
3
(t)
_
, (1.10)
6 CHAPTER 1. EXPANDING UNIVERSE
where
0
is obviously the energy density at t = 0 and we used short-hand notation a(0) = a
0
. Now we
take the time derivative of Eq. (1.10)
(t) = 3
0
a(t)
4
a(t) = 3
0
_
a
0
a
_
3
a
a
= 3(t)
a
a
= 3(t)H(t), (1.11)
i.e.,
(t) + 3(t)H(t) = 0. (1.12)
We can compare the last relation with the energy conservation, which quite generally can be written in
the following form
d
dt
=

t
+

(v) = 0. (1.13)
We note that (r, t) = (t) and that the velocity is given by the Hubble law (1.2). Then the second
term in Eq. (1.13) is transformed as follows:

(v) =

(v) = H(t)

r = 3H(t), which conrms that


Eq. (1.12) is nothing else as the conservation of the energy.
Figure 1.2:
1.2.1 Acceleration equation
Gravitational forces among masses are attractive. Therefore, matter masses are self-attractive. As a
result the expansion of matter is slowed down by gravity. Let us return to the example of the 3d sphere
from the previous section. Consider a probe of mass m on the surface of the sphere. Particles outside
the sphere have no eect on the test mass; only particles inside the sphere are relevant. The equation
of motion then reads:
m

R =
GmM
R
2
=
4
3
Gm
M
(4/3)R
3
. .
(t)
R (1.14)
Eliminating the radius in favor of the scale factor we obtain
m a
com
=
4
3
Gm(t)a
com
. (1.15)
1.2. DYNAMICS OF DUST IN NEWTONIAN COSMOLOGY 7
We see that the mass m and the comoving coordinate drop out of this equation, which in combination
with the energy conservation (1.13) determines the evolution of the system
a =
4
3
G(t)a. (1.16)
The equations (1.13) and (1.16) completely determine the evolution of the system. These equations
exactly coincide with their counterparts in general relativity. The reason is that the analysis above
should be valid for innitesimally small spheres, in which case the velocities and masses involved are
small.
Now let us nd the solutions of these equations. To eliminate the energy density from the acceleration
equation with substitute Eq. (1.10) in Eq. (1.16)
a =
4
3
G
_
a
0
a(t)
_
3

0
a. (1.17)
Next we multiply the both sides of this equation by a and integrate; explicitly
a a =
4
3
Ga
3
0

0
a
a
2
,
1
2
a
2
=
4
3a
Ga
3
0

0
. .
V (a)
+E, (1.18)
where E is an integration constant. This equation is analogous to the problem which one studies during
the course of classical mechanics, namely that of the rocket launched from the Earth. This is apparent
if we rewrite Eq. (1.18) as
1
2
a
2
+ V (a) = E. (1.19)
The rst term is the kinetic energy, the second term the potential energy (gravitational pull of the Earth
on the rocket). If the kinetic energy is larger than the potential energy, i.e., E > 0 the rocket leaves
the Earth, otherwise, i.e. when E < 0 the rocket falls back to Earth. In analogy the dust dominated
Universe expands forever in the case of positive E and recollapses in the case of negative E. Let us
rewrite this condition as follows
a
2
+
8
a
Ga
3
0

0
a
= 2E,
_
a
a
_
2
. .
H
2

2E
a
2
=
8G
3
_

a
3
0
a
3
_

0
. .
(t)
(1.20)
thus,
H
2

2E
a
2
=
8G
3
(t). (1.21)
The fate of the dust dominated universe depends on the sign of E. The following cases are possible:
E > 0, i.e., when the kinetic energy dominates, a/2 > V (a), and we have an expanding universe
E < 0, i.e., when the potential energy dominates, a/2 < V (a), and we have an recollapsing
universe
8 CHAPTER 1. EXPANDING UNIVERSE
The limiting case E = 0 denes the critical density

cr
=
3H
2
8G
. (1.22)
Let us eliminate in Eq.(1.16) the Hubble constant in favor of the critical density. We obtain
8G
3
[
cr
(t)] =
2E
a
2
, (1.23)
or, alternatively,
E =
4G
3
a
2

cr
[1 (t)] . (1.24)
Here we have introduced a new quantity
(t) =
(t)

cr
, (1.25)
which is called the cosmological parameter. In general the sign of E is xed, therefore the quantity
1 (t) does not change the sign. This does not mean that the function does not change with time;
indeed since the Hubble parameter is time dependent, so is the cosmological parameter.
We shall see that the sign of E determines the spatial geometry of the universe in General Relativity.
In a dust dominated universe we have the following cases:
if
0
=
0
/
cr
0
> 1 (the index 0 refers to the current value of the parameter), then E is negative
and the spatial curvature of the universe is positive. This corresponds to closed universe. During
its expansion the universe reaches some maximal value of the scale factor and then the universe
recollapses.
if
0
=
0
/
cr
0
< 1, then E is positive and the spatial curvature of the universe is negative. This
corresponds to an open universe, which expands hyperbolically.
if
0
= 1, then E = 0, which corresponds to parabolic expansion of the universe with at spatial
geometry.
These cases are illustrated in Fig. 1.4. In the cases when universe is at or open, it expands forever (the
scale factor increases forever). In the case of closed universe the scale factor eventually decreases to
zero, i.e., the universe returns to the initial singularity. It should be stressed that our example concerns
the dust dominated universe. In general the evolution of the Universe depends on its matter content,
so that, for example, we may have a Universe that is closed, but does not recollaps at asymptotically
large times.
Consider the special case of at universe, E = 0. We would like to nd an explicit solution to the
scale factor from Eq. (1.18); in that case we have
1
2
a
2
+ V (a) = 0,
a
2
2

4G
0
a
3
0
3a
= 0,
a a
2
2
=
4G
0
a
3
0
3
= Const. (1.26)
The left-hand side of the last relation can be written
a a
2
=
4
9
_
da
3/2
dt
_
2
= Const. a t
2/3
. (1.27)
It follows immediately that in this case H(t) = 2/3t and that current age of the universe is t
0
= 2/3H
0
,
which diers from our previous estimate on page 5 by a factor of order unity.
1.3. GEOMETRY OF HOMOGENEOUS AND ISOTROPIC SPACE 9
Figure 1.3: Dependence of the scale factor on time for open, at and closed universes.
1.3 Geometry of homogeneous and isotropic space
The evolution of our universe in time can be represented by a sequence of three-dimensional spatial
hypersurfaces, each of which by assumption are isotropic and homogeneous. Note that isotropy at every
point implies homogeneity, but homogeneity does not imply isotropy. Homogeneous matter can execute
motions that are dependent of directions (e.g. expand in one direction and contract in two others; such
motions will yield physics which is anisotropic, i.e., depends on directions).
The symmetry group of homogeneous and isotropic spaces include three independent translations
and three rotations. There are three types of such spaces (assuming that the topology of the spaces
are simple): 1) a three dimensional sphere of constant and positive curvature; 2) at space; 3) a three-
dimensional hyperbolic space of constant negative curvature.
There are two well-known examples of embeddings of 2d surfaces in the Euclidean space: at plane
and 2d sphere. These will help us to visualize the isotropic and homogeneous spaces; the extension to
higher spatial dimensions is straightforward. The metric of 3d Euclidean space is given by
dl
2
= dx
2
+ dy
2
+ dz
2
. (1.28)
The embedding of a 2d sphere in this 3d space is given by
x
2
+ y
2
+ z
2
= a
2
, (1.29)
where a is the radius of the sphere. From this last equation it follows that
dz =
xdx +ydy
z
=
xdx + ydy
_
a
2
x
2
y
2
. (1.30)
and, therefore, the metric (1.28) can be written as
dl
2
= dx
2
+ dy
2
+
(xdx + ydy)
2
a
2
x
2
y
2
. (1.31)
10 CHAPTER 1. EXPANDING UNIVERSE
Figure 1.4: Illustration of the Lobachevsky space in two dimensions.
This natural result shows that the distance between two points located on the sphere can be expressed
in terms of two independent variables. Let us go over to the spherical (cylindrical) coordinates:
x = cos , y = sin . (1.32)
where
x
2
+ y
2
=
2
, xdx + ydy = d. (1.33)
On the other hand
dx
2
+ dy
2
= d
2
+
2
d
2
(1.34)
Substituting these in Eq. (1.31) we obtain for the metric
dl
2
=
_
1 +

2
a
2

2
_
d
2
+
2
d
2
=
d
2
1
2
/a
2
+
2
d
2
. (1.35)
We can now distinguish three cases: (1) a
2
> 0, two-dimensional space with positive curvature; (2)
a , which corresponds to a 2d plane, i.e., at space; (3) a
2
< 0, two-dimensional space with negative
curvature; this space is called Lobachevsky space. It cannot be embedded into the Euclidean space
because the radius of the sphere is imaginary (sometimes this space is called pseudo-sphere or hyperbolic
space). However, Lobachevsky space arises from the embedding of the surface x
2
+y
2
z
2
= a
2
with
a >= 0 in the space with metric dl
2
= dx
2
+ dy
2
dz
2
(see Figure 1.5 for an illustration). One can
write the metric (1.35) in a dierent form by introducing = r
_
[a
2
[
dl
2
= [a
2
[
_
dr
2
1 kr
2
+r
2
d
2
_
. (1.36)
where k = +1 corresponds to the sphere, k = 1 - pseudo-sphere, and k = 0 to the plane. The meaning
of [a[ is to quantify the curvature of the space; in the case of the at space it does not have any physical
meaning, i.e., when k = 0 it can be absorbed in the denition of r.
1.3. GEOMETRY OF HOMOGENEOUS AND ISOTROPIC SPACE 11
Similar to the discussion above we can go up in the number of spatial dimensions and obtain the 3d
embedding of a sphere (pseudo-sphere) in 4d Euclidean (Lorentzian) space. In that case we obtain
dl
2
= [a
2
[
_
dr
2
1 kr
2
+r
2
(d
2
+ sin
2
d
2
)
_
. (1.37)
where as before k = 0, 1. Sometime we will work with the coordinate dened via the relations
r =
_
_
_
sinh , k = 1
, k = 0
sin , k = +1
and
d
2
=
dr
2
1 kr
2
(1.38)
Then the metric takes the simple form
dl
2
= a
2
(d
2
+
2
()d
2
), d
2
= d
2
+ sin
2
d
2
, (1.39)
and () = [0, ] for k = 0, () = sin [0; ] for k = 1, and () = sinh [0; ] for
k = 1.
Next let us consider the embedding of a 2d sphere in a three dimensional space with positive
curvature. The distance element on the sphere of xed radius is given by
dl
2
= a
2
sin
2
d
2
. (1.40)
This expression is mathematically identical to that for the sphere in a at space with radius R =
a
2
sin
2
. The surface area, which is just the integral over the solid angle, follows immediately as
S = 4R
2
= 4a
2
sin
2
. (1.41)
We see that the surface area is non-monotonic as [0; ]. In particular it vanishes at the end points
of this domain. (A straightforward analogy is the circumference of a circle of constant latitude on a
usual sphere. When one moves from the north pole to the equator and then to the south pole; the
magnitude of the circumference is non-monotonic.) Because the area remains always bound, we expect
that the volume of the space with positive curvature is also bound. A characteristic feature of the space
of positive curvature is that the sum of the angles of a triangle formed by geodesics (curves of minimal
length) is lager than 180 degree.
The same arguments as above applies to the space of negative curvature k = 1 with metric of the
surface of 2d pseudosphere given by
dl
2
= a
2
sinh
2
d
2
, (1.42)
which leads to the surface area
S
2d
= 4a
2
sinh
2
. (1.43)
The surface area and the volume of such space is unbound. Furthermore, the sum of the angles of a
triangle formed by geodesics is smaller than 180 degree.
The volumes of these spaces can be calculated from
dV = S
2d
ad. (1.44)
12 CHAPTER 1. EXPANDING UNIVERSE
Figure 1.5: The surface area of 2d sphere in three cases of open, at and closed universes.
1.4 Einsteins equations
The general theory of relativity provides a consistent framework for the description of the universe lled
in with matter with arbitrary equation of state. In this section we will extend our previous discussion
of dynamics of Newtonian universe to general relativity.
We will use Einsteins convention for summation over repeated indices:
g

dx

dx

dx

dx

. (1.45)
Greek indices will run over 0, 1, 2, 3, while the Latin indices will run over 1, 2, 3.
The dynamics of space-time continuum in General Relativity is described by the metric functions
g

(x

), which obey the Einstein equations:


G

= R


1
2

= 8GT

. (1.46)
Here R

is the Ricci tensor which can be written in terms of the metric tensor and Christoel symbols
as
R

= g

_
, (1.47)
where g

and the unit tensor is dened as

= 1 if = and 0 otherwise. The Christoel


symbols are dened entirely in terms of the metric functions

=
1
2
g

_
g

+
g

_
. (1.48)
Further R = R

is the scalar curvature and = const is the cosmological constant. On the left hand
side of Eq. (1.46) stands the energy-momentum tensor T

of matter, which is symmetric in its indecies,


i.e.,
T

= g

= T

. (1.49)
1.5. FRIEDMAN EQUATIONS 13
The energy momentum tensor is deed by the conservation equation (or equation of motion)
T

= 0, (1.50)
in Minkowski spacetime. In cured spacetime the equation is modied to
T

;
=
T

= 0. (1.51)
The equation (1.51) follows from the Einstein equations as consequence of the Bianchi identities satised
by the Einstein tensor:
G

;
= 0. (1.52)
For most of the application the matter can be assumed to be perfect uid with th energy and momentum
tensor given by
T

= ( +p)u

, (1.53)
where p = p() is the equation of state of matter, which needs to be specied at each particular case, u

is the four velocity of the uid, p is the pressure, - the energy density. For ultrarelativistic particles
p = /3; in cosmology we will nd often that p = w, where w is a constant.
In relativistic theory the invariant with respect to coordinate transformation is the innitesimal
spacetime interval between events. It is convenient to take the interval in the following form
ds
2
= dt
2
dl
2
= dt
2
a(t)
2
_
dr
2
1 kr
2
+ r
2
d
2
_
g

dx

dx

; (1.54)
the corresponding metric is called the Robertson-Walker metric. Note that the spatial coordinates are
comoving, which means that every object with zero peculiar velocity has constant coordinates r, and
. The distance between two comoving observers is a(t).
1.5 Friedman equations
We have already derived the equations for cosmological evolution in Newtonian theory; these are given
by (1.12), (1.16) and (1.21). Einstein equations determine the evolution of the universe in general
relativity; we shall discuss the formal derivation somewhat later; for the moment we can heuristically
derive the general relativistic equation by modifying the Newtonian equations. The main modication
comes from the fact that we need to relax the assumption that we are dealing with a dust and include
the contribution of pressure. The change in the energy due to the work done by the pressure is given
by (we assume that the entropy of the system is constant, i.e., dissipative processes are negligible)
dE = pdV. (1.55)
Since E = V and V a
3
, we obtain V d + dV = pdV or d = V
1
(p + )dV . Then we nd that
dV/V = 3a
2
da/a
3
= 3da/a = 3d ln a. Thus,
d = 3( + p)d ln a + 3( + p)H = 0, (1.56)
where we substituted H = a/a. This equation is the relativistic version of Eq. (1.12) and it reects the
energy conservation in an isotropic and homogeneous universe, T

0;
= 0. Thus, we see that the general
14 CHAPTER 1. EXPANDING UNIVERSE
relativistic counterparts of the Newtonian equations are obtained by addition of a term arising from
pressure. The acceleration equation is now modied from (1.16) to
a =
4
3
G( + 3p)a. (1.57)
The form of the pressure term arises from the sum of the diagonal terms in the spatial part of the
Einstein equations. Equation (1.56) is the rst Friedman equation. To obtain the second Friedman
equation we use the second Eq. (1.56) to eliminate pressure from Eq. (1.57) and multiply the resulting
equation by a to bring it to a form suitable for integration. The details are reected in the following
chain (note that as before =
0
a
3
0
/a
3
)
a a =
4G
3
_
2

H
_
a a =
4G
3

0
_

2a
3
0
a
3
+ 3
a
3
0
a
a
2
H
_
a a =
4Ga
3
0
3

0
a
a
3
. (1.58)
Now we can integrate this equation to obtain
a
2
2
=
4Ga
3
0
3

0
1
a
2
+ E. (1.59)
Upon reintroducing the Hubble constant H = a/a and dening k = 2E, simple algebraic manipula-
tions lead us to
H
2
+
k
a
2
=
8G
3
(t). (1.60)
This is the second Friedman equation. The dierence to the non-relativistic counterpart (1.21) is that
this equation applies for arbitrary equation of state. Another important point is that k here has the
meaning of the curvature, as introduced above, i.e., k = 0, 1. This identication is possible if we
derive the Friedman equations from the 0 0 component of Einsteins equations. Now we can rewrite
(1.24) as
k =
2G
3
a
2

cr
[1 (t)] . (1.61)
Thus we see that the value of the cosmological parameter = /
cr
determines the geometry of the
universe; specically, when > 1, the Universe is closed and has a geometry of three-dimensional
sphere k = +1; = 1 corresponds to the at universe (k = 0); and < 1 corresponds to open universe
with hyperbolic geometry (k = 1).
The two Friedman equations (1.57) and (1.60) form a closed system of equations if the equation of
state p = p() is specied. Note that Eq. (1.57) can be replaced by the second equation of (1.56), i.e.,
the conservation law.
1.5.1 Derivation from Einsteins equations
To derive the Friedman equations from Einsteins equations and to nd particular solutions to them it
is more convenient to dene the conformal time as
dt = a(t)d,
_
dt
a(t)
. (1.62)
To be specic let us rst consider the universe with positive curvature, i.e., k = +1. In terms of
conformal time the metric is written as
ds
2
= a
2
()[d
2
d
2
sin
2
(d
2
+ sin
2
d
2
)]. (1.63)
1.5. FRIEDMAN EQUATIONS 15
Now we would like to write down the Einsteins equations for this metric. For that purpose we identify
the non-vanishing metric functions by comparing Eq. (1.63) with ds
2
= g

dx

dx

. This gives (note


that the coordinates x
0
, x
1
, x
2
, x
3
here should be identied with , , , ):
g
00
= a
2
, g
11
= a
2
, g
22
= a
2
sin
2
, g
33
= a
2
sin
2
sin
2
. (1.64)
Note that because of isotropy of the space (i.e., equivalence of all directions, the components of metric
tensor g
0
must be zero in the system of coordinates we have chosen). Now we use the denition of
Christoel symbols (1.48) to compute the components

0
00
=
a

a
,
0

=
a

,
0
0
=

00
= 0, (1.65)
where prime means dierentiation with respect to . Now the components of the Ricci tensor can be
computed via Eq. (1.47). We obtain
R
0
0
=
3
a
4
(a

2
aa

). (1.66)
From symmetry considerations (isotropy of the space) the components R
0
= 0. The remaining com-
ponents are
R

=
1
a
4
(2a
2
+ a

2
+ aa

. (1.67)
The Ricci scalar is then obtained as
R = R
0
0
+ R

=
6
a
3
(a + a

). (1.68)
Finally, we need to specify the form of the energy-momentum tensor; in the comoving system of coor-
dinates u

= 0, and u
0
= 1/a, therefore from the denition
T
ik
= (p + )u
i
u
k
pg
ik
, (1.69)
it follows that T
0
0
= . These results should be substituted in the 00 component of Einsteins equations,
i.e.,
R
0
0

1
2
R = 8GT
0
0
, (1.70)
which gives
a
2
+ a

2
a
4
=
8G
3
. (1.71)
A more convenient form of this equation is obtained upon dierentiating it with respect to . To
compute the derivative note that from the rst equation of (1.56) we have
d
d
= 3( + p)
d
d
ln a = 3( + p)
a

a
. (1.72)
Then we nd that
2a

+ 2kaa

=
8G
3
[a
3
a

( 3p)], a

+ka =
4G
3
( 3p)a
3
. (1.73)
16 CHAPTER 1. EXPANDING UNIVERSE
Next we wish to compare the results that we have obtained with those derived heuristically. For that
purpose, we go over from derivative with respect to the conformal time to the ordinary time (which
induces an extra factor a). We obtain
1
a
2
+
a
2
a
2
= H
2
+
1
a
2
=
8G
3
, (1.74)
which coincides with the second Friedman equation (1.60) for k = +1. The second equation (which
closes the system of equations for the unknowns a and ) can be chosen one of those in Eq. (1.56), e.g,
d = 3( + p)d ln a. (1.75)
How the calculations are modied in the case of open universe with negative curvature? The metric in
this case is given by
ds
2
= a
2
()[d
2
d
2
sinh
2
(d
2
+ sin
2
d
2
)]. (1.76)
This can be obtained formally from Eq. (1.63) by replacements i, i, and a ia. Equation
(1.75) will not change its form, whereas Eq. (1.74) becomes
H
2

1
a
2
=
8G
3
, (1.77)
which is consistent with Eq. (1.60).
1.6 Solutions of Friedman equations
To analyze the solutions of the Friedman equations we start with Eq. (1.73)
a

+ka =
4G
3
( 3p)a
3
, (1.78)
which can be solved provided the equation of state p() is given. We will also use Eq. (1.60) which we
rewrite in terms of conformal time
H
2
+
k
a
2
=
8G
3
, a
2
+ k =
8G
3
a
2
, a
2
+ ka
2
=
8G
3
a
4
. (1.79)
In the following we will consider solutions of the Frideman equations for specic equations of state.
1.6.1 Radiation dominated universe, p = /3, k = 0, 1
This case is particularly simple since the right hand side of (1.78) vanishes, i.e.
a

+ ka = 0. (1.80)
The solution of this dierential equation is
a() = a
m
_
_
_
sinh , k = 1
, k = 0
sin . k = +1
1.6. SOLUTIONS OF FRIEDMAN EQUATIONS 17
where a
m
is one of the integration constants, the other is xed from the condition a( = 0) = 0. The
physical time is the integral of a()d. Integration in Eq. (1.81) gives
t = a
m
_
_
_
cosh 1, k = 1

2
/2, k = 0
1 cos . k = +1
One conclusion that can be drawn from these equations is that in the case of at universe a t
1/2
,
form which it follows that H = 1/2t. We can also nd the energy density from Eq. (1.60) with k = 0:
=
3
32Gt
2
a
4
. (1.81)
The same can be found from the Eq. (1.56)
d = 4d ln a d ln = d ln a
4
+ C a
4
. (1.82)
This means that the energies of particles (whose number is conserved) is reduced proportional to a
1
.
1.6.2 Dust dominated universe p = 0, k = 0, 1
First we note that Eq. (1.56) in this case implies
d = 3d ln a d ln = d ln a
4
+ C a
3
. (1.83)
Therefore a
3
= const and
a

+ka = C. (1.84)
The solutions of this equation read
a() = a
m
_
_
_
cosh 1, k = 1

2
, k = 0
1 cos , k = +1
and the physical time is given via integration
t = a
m
_
_
_
sinh , k = 1

3
/3, k = 0
sin . k = +1
Since the scale factor varies in the range 0 a we see that the conformal time varies in the
range 0 in the cases of open or at universes independent whether it is radiation or matter
dominated. In a closed universe the conformal time is bounded and assumes values in the range [0; ]
for radiation dominated universe and [0; 2] for matter dominated universe.
1.6.3 Strongly interacting universe p
We consider a concrete case of at universe k = 0. We already saw that the energy density for radiation
scales as
R
a
4
and for matter scales as
M
a
3
. The total energy density is the sum of both,
which we write as
=
R
+
M
= C
R
a
4
+C
M
a
3
, (1.85)
18 CHAPTER 1. EXPANDING UNIVERSE
where C
M
and C
R
are constants. It will be convenient to work with the energy of universe when these
two contributions are equal; we denote the scale factor and the total energy density in these cases as a
and . By denition
= C
R
a
4
+ C
M
a
3
, C
R
a
4
= C
M
a
3
C
R
a
1
= C
M
(1.86)
Substituting the last relation in Eq. (1.86) in the rst one we nd
= C
R
a
4
+ C
R
a
4
, C
R
=

2
a
4
, C
M
=

2
a
3
. (1.87)
Thus the total energy density is given by
=

2
_
a
4
a
4
+ a
3
a
3

. (1.88)
Taking a derivative of Eq. (1.79) we obtain
a

=
2G
3
a
3
. (1.89)
Because the right hand side is a constant, we can straightforwardly integrate this equation to obtain
a() =
G
3
a
3

2
+ C
1
+ C
2
. (1.90)
From the condition a( = 0) = 0 it follows that C
2
= 0. To obtain the second integration constant we
substitute in Eq. (1.79) written for the case k = 0
a
2
=
8G
3
a
4
, (1.91)
the solution (1.90):
_
G
3
a
3
+ C
1
_
2
=
4G
3

_
a
4
+a a
3

, (1.92)
where in the right hand side we substituted Eq. (1.88). Note that the rst term in the right hand side
is due to radiation and the second due to the matter. Substituting the initial condition a( = 0) = 0
once more we obtain
C
2
1
=
4G
3
a
4
, C
1
= +
_
4G
3
a
4
_
1/2
. (1.93)
We see that the term C
1
is controlled by the radiation contribution, since the matter contribution
dropped out upon imposing the initial conditions. The plus sign in the last relation follows from the
requirement of positivity of the scale factor. Then the solution (1.90) can be written in an elegant form
a() = a
_

+ 2

_
, (1.94)
where

=
_
G
3
a
2
_
1/2
. (1.95)
1.7. MILNE UNIVERSE 19
In Eq. (1.94) the rst term can be identied with the matter contribution, while the second term with
the radiation contribution. When a = a the conformal time satises the equation

+ 2

= 1, (1.96)
whose solution (which we will denote by ) is given by
=

2 1). (1.97)
Note, however, that for the purpose of order of magnitude estimates we can use both s, since

.
We see that if the linear term, representing the radiation contribution, in Eq. (1.94) dominates.
The scale factor grows linearly with . Conversely, when the matter contribution dominates and
a()
2
.
1.7 Milne universe
An interesting special case is the case of Milne universe, which is dened as an open universe (k = 1)
with zero matter energy density. In that case from Eq. (1.79) we obtain
a
2
(t) = 1, a(t) = t. (1.98)
Then the Robertson-Walker (RW) metric becomes
ds
2
= dt
2
a(t)
2
_
dr
2
1 kr
2
+ r
2
d
2
_
= dt
2
t
2
_
dr
2
1 + r
2
+ r
2
d
2
_
= dt
2
t
2
_
d
2
+ sinh
2
d
2
_
,
(1.99)
where the coordinate is dened via Eq. (1.38). Since the Milne universe does not have any matter
content we can expect that its metric should be related to the Minkowski metric which corresponds to
the metric of empty space. It can be written
ds
2
= d
2
dr
2
r
2
d
2
. (1.100)
(The Minkowski coordinate r should not be confused with the one appearing in the RW-metric). To
map the Minkowski metric (1.100) onto the RW metric we dene new coordinates
= t cosh, r = t sinh. (1.101)
We can compute further the dierentials
d = dt cosh + t sinhd, dr = dt sinh + t coshd, (1.102)
from which we see that
d
2
dr
2
= dt
2
t
2
d
2
. (1.103)
We conclude that (1.99) and (1.100) are equivalent. However, the Milne coordinates do not cover
the entire Minkowski space. To see this consider a particle which has xed comoving coordinates; its
velocity is equal
[v[ =
r

= tanh < 1. (1.104)


20 CHAPTER 1. EXPANDING UNIVERSE
Figure 1.6: The constant t hyperboloids in the Minkowski space cover the Milne universe.
The proper time of the particle is given by
_
1 [v[
2
= t. (1.105)
Note that on the left hand side of these equation are the Minkowski coordinates, on the right hand
sides the Milne coordinates. Now we nd the hypersurfaces of constant time t which in the Minkowski
coordinates are given expressed
t =

2
r
2
= Const. (1.106)
These can be plotted in the Minkowski space (i.e. as a function of the coordinates and r). The result
is shown in gure Fig. (1.6). We see that the hypersurface t = 0 is simply the light cone; the surfaces
of constant t > are hyperboloids in the Minkowski coordinates which are within (half of) the forward
light cone. Thus the Milne universe covers only 1/4 of the entire Minkowski space (the uncovered parts
include the backward light-cone, a half of the forward light cone). Some remarks are in order
The Milne universe has a center, which is apparent from its coverage shown in the gure.
The Milne universe is an example of a universe which is at in the 4d space, but has a curvature
in the 3d space (since k = 1).
In the universes with energy content the correct foliation of the space is given by the constant
energy hypersurfaces. Obviously in the Milne universe we have a translational symmetry in time
- all the hypersurface have the same zero energy.
1.8 De Sitter universe
The de Sitter universe is a space-time void of matter and radiation with positive constant 4-curvature
which is isotropic and homogeneous. Since it is a space-time, it can be derived from purely geometrical
considerations. (We will return to more physical derivation from Friedman equations later on). We have
seen that in the case of purely spatial spaces we were able to represent curved spaces as embeddings of
various (hyper)surfaces in higher dimensional at spaces. To make the discussion intuitively accessible
1.8. DE SITTER UNIVERSE 21
Figure 1.7: de Sitter space-time in two dimensions.
we will consider only two spatial dimensions. Addition of third dimension will be trivial. Consider
three-dimensional Minkowski metric given by
ds
2
= dz
2
dx
2
dy
2
. (only spatial) (1.107)
In this space we consider an embedding (hyperboloid) dened as
H

2
= z
2
x
2
y
2
. (1.108)
The hyperboloid will be parameterized in terms of the coordinates x and y. Since zdz = xdx+ydy and
z =
_
x
2
+ y
2
H

1/2
the metric can be expressed entirely in terms of the variable x and y
ds
2
=
(xdx + ydy)
2
x
2
+ y
2
H

2
dx
2
dy
2
. (1.109)
where according to (1.107) x
2
+y
2
> H

2
. This metric describes the de Sitter space in two-dimensions.
By coordinate transformation this metric can be brought in a more compact form (which also make
apparent its symmetries) by, for example, transformation
x = H

1
cosh(H

t) cos , y = H

1
cosh(H

t) sin . (1.110)
Where t and [0; 2]. Then, the metric takes the form
ds
2
= dt
2
H

2
cosh
2
(H

t)d
2
. (1.111)
This space-time is described by the hyperboloid shown in Fig. 1.7. In four dimensions this metric
corresponds to closed universe with positive curvature.
Now we consider the four-dimensional de Sitter universe, which is described by the equation of
state p

. The index indicates that the pressure originates from cosmological constant in the
Einsteins equations. The general equation (1.56) in this case shows that
d = 3( +p)d ln a d

= 3(

+ p

)d ln a = 0,

= Const. (1.112)
22 CHAPTER 1. EXPANDING UNIVERSE
Figure 1.8: Scale factor in de Sitter universe as a function of time. Note the exponential expansion for
t .
Then the acceleration equation (1.57) reads
a =
4
3
G( + 3p)a a
8
3
G

a = 0, a H

2
a = 0, (1.113)
where H

= [8G

/3]
1/2
. The solution of Eq. (1.113) is given by
a(t) = C
1
exp(H

t) + C
2
exp(H

t), (1.114)
with C
1
and C
2
being the integration constants. To x the integration constants we go back to the
second Friedman equation (1.60) and write it as
H
2
+
k
a
2
=
8G
3

H
2

. (1.115)
Inserting H = a/a in this equation, we rewrite it as
H
2

a
2
a
2
= k. (1.116)
We substitute the solution (1.114) in this equation to obtain
C
1
C
2
=
k
4H
2

. (1.117)
Now let us consider separately the following cases:
Flat universe k = 0; either C
1
= 0 or C
2
= 0. If C
2
= 0 we obtain exponentially expanding
universe.
1.8. DE SITTER UNIVERSE 23
For closed universe k = 1 and we can choose [C
1
[ = [C
2
[ to obtain C
1
= C
2
= (2H

)
1
. In this
case the scale factor becomes
a(t) = H
1

_
exp(H

t) + exp(H

t)
2
_
= H
1

sinh(H

t). (1.118)
For open universe k = 1 and we can choose [C
1
[ = [C
2
[, i.e., C
1
= C
2
= (2H

)
1
. In this case
the scale factor reads
a(t) = H
1

_
exp(H

t) exp(H

t)
2
_
= H
1

cosh(H

t). (1.119)
The metric for these three cases can be written (upon substituting the scale factor a(t)) as
ds
2
= dt
2
H
2

_
_
sinh
2
(H

t)
exp(2H

t)
cosh
2
(H

t)
_
_
_
_
d
2
+
_
_
sinh
2

2
sin
2

_
_
d
2
_
_
.
24 CHAPTER 1. EXPANDING UNIVERSE
Chapter 2
Light propagation in expanding universe
Most of the universe is seen by us in the electromagnetic spectrum. Although the existence of the
gravitational waves has been indirectly proven in the binary systems involving two compact stars, their
direct detection has not been achieved yet. Thus vast astronomical information reaches us in form
of photons of dierent wave-lengths. (Neutrinos and high energy cosmic rays are other sources that
complement our information on the universe).
The fact (known from the special relativity) that the speed of any particle is limited to that of the
light limits the amount of information that can be transmitted. Indeed since no information can travel
faster than light their exist horizons beyond which we can not obtain information about the universe.
2.1 Light trajectories
Consider a massless particle propagating in a at universe at the speed of light (e.g. photon). Its
trajectory is given by the geodesics that satises the condition
ds
2
= 0. (2.1)
We know that in the curved universes we can always choose a local frame where physics corresponds to
a at universe so that the condition (2.1) remains intact.
Consider isotropic universe, with light propagating radially in a coordinate system where observer
is at the origin. To describe the trajectories we use the conformal time
=
_
dt
a(t)
. (2.2)
The metric is written as
ds
2
= a
2
()
_
d
2
d
2

2
()(d
2
+ sin
2
d
2
)

, (2.3)
where

2
() =
_
_
sinh
2
k = 1

2
k = 0
sin
2
k = +1
_
_
.
Since we have spherical symmetry the radial trajectories with , = const are geodesics (i.e. d = 0 =
d). Then, from the requirement ds
2
= 0 it follows
d
2
d
2
= 0. (2.4)
25
26 CHAPTER 2. LIGHT PROPAGATION IN EXPANDING UNIVERSE
The solution of this equation is
() = + C. (2.5)
We see that the trajectories are simply straight lines at 45 degree in the plane.
2.2 Horizons
The information that we receive about the universe around us is limited by the speed of light and the
fact that the universe has a nite age. We can receive only information about parts of the universe
from which the light could have traveled within the time equal or less than the age of the universe.
The volume from which we can obtain information is the bounded; its boundary is called the particle
horizon.
The age of the universe is 1.5 10
10
yr, so that the size of particle horizon is about 1.5 10
10
light
yr. Let us examine now the distance that light can propagate in an expanding universe. According to
(2.5)

p
() =
i
=
_
t
t
1
dt
a
. (2.6)
where t
1
is the initial time (start of the universe). Therefore, at time an event with >
p
() is
inaccessible to the observer located at the origin of the coordinate system = 0. Note that we cannot
always put
i
= 0 = t
i
; this is only valid for the universes that are singular at the beginning. There are
example of other (e.g. de Sitter) universes where the initial state is non-singular. The physical distance
is obtained upon multiplying the the comoving coordinate with the scale factor
d
p
(t) = a(t)
p
= a(t)
_
t
t

dt
a
. (2.7)
Now we need to take into account the fact that the early universe was dense and photons where in termal
equilibrium with matter. Their decoupling from matter occured at the time of hydrogen recombination.
At that time the universe was about 1000 time smaller than now. Therefore, it is easy to see that
the information that we obtain is, in fact, limited by the distance that the light can travel since the
recombination time. Eectively the initial time is replaced by the recombination time
d
opt
(t) = a(t)
p
= a(t)
_
t
t

dt
a
, (2.8)
where t

is the time of recombination. This horizon is clearly less distant that the particle horizon. It
is called optical horizon. Only primordial neutrinos or gravitational wave can allow us to see beyond
the optical horizon (but only up to the particle horizon).
For any concrete model of the universe for which the scale factor is know we can compute the
horizons. As an example let us take a at universe and assume that radiation dominates; then a(t) t
1/2
and we nd that d
p
= 2t (note that c = 1); in the case of matter dominated universe a(t) t
2/3
and
d
p
= 3t.
There is another scale in the universe that sometimes is called horizon. Evidently we have the
length scale 1/H (again c = 1). This is called curvature horizon and Hubble horizon. In some models
curvature horizon could be of the same order as the particle horizon. However there are counterexamples.
One such counterexample is the at de Sitter space where a(t) exp(H

t) and therefore
d
p
(t) = exp(H

t)
_
t
t

exp(H

t)dt = H

1
exp[H

(t t

)] 1 . (2.9)
2.3. REDSHIFT 27
For times satisfying tt

H
1
, d
p
grows exponentially, i.e., the particle horizon is very large compared
with the Hubble horizon H
1
which is constant.
One can also dene an event horizon. The event horizon is dened with respect to a given time (in
terms of conformal time ). It is the boundary of the volume from which no signal can be obtained by
the observer in the future. The coordinates of these points are dened
>
e
() =
_

max

d =
max
. (2.10)
where
max
refers to the nal time. The physical location of the horizon at the time t is given by
d
e
(t) = a(t)
_
t
max
t
dt
a
,
max
=
_
t
max
t
dt
a(t)
. (2.11)
If universe expands forever then by denition the time t
max
. The value of
max
and d
e
(t) obviously
depends on the convergence of the integral dening these quantities. Let is consider several examples.
Open or at universe; since t
max
and
max
are innite there are two possibilities:
the integrand in Eq. (2.11) is convergent, i.e., there is a nite event horizon
this integrand is divergent, i.e., there is no event horizon.
The convergence of the integrand depends on the function a(t).
For linear or slower growing a(t) the integrand is divergent (in the linear case logarithmically).
This corresponds to decelerating universe. Mathematically, this means that a 0.
If a(t) grows faster than linear the integrand is convergent. This corresponds to accelerating
universe ( a > 0) (gebremstes Universum).
de Sitter universe. For the event horizon we obtain
d
e
(t) = exp(H

t)
_

t
exp(H

t)dt = H

1
. (2.12)
This means that the event horizon is equal to the curvature scale. All the events that occur at a
given moment at a distance greater than 1/H have no eect on an observer because the universe
is expanding too fast. In closed and decelerating universes the t
max
is nite and therefore there
always exists an event horizon and a particle horizon.
In the closed universe there is always an event horizon because the integration range is limited
from above.
2.3 Redshift
Imagine a photon emitted from a distance galaxy. If we (observers) were static with respect to that
galaxy, then the wave-length of the photon will be the same for us as it was at the moment of emission.
This is not true in an expanding universe. The wave-length of the photon will undergo a redshift because
the universe is expanding.
28 CHAPTER 2. LIGHT PROPAGATION IN EXPANDING UNIVERSE
Figure 2.1: Illustration of light propagation from the source to the observer in an expanding universe.
Consider a source or radiation at
em
and t
em
. Suppose it emits a signal with a duration . The
equation dening the light trajectory is then [see Eq. (2.5), where we take the sign

]
() = + C. (2.13)
The integration constant is xed from the initial condition (
em
) =
em
+ C or C =
em
+
em
.
Substituting this constant in the above equation we obtain
() =
em
(
em
), (2.14)
which denes the trajectory of a photon. The photon reaches the observer at the location
obs
= 0 at
time
obs
=
em
+
em
. The physical intervals for the emission of light at the point of emission and
detection are dierent and are given by
t
em
= a(
em
)
. .
emission point
, t
obs
= a(
obs
)
. .
observation point
. (2.15)
The wave-length and the period of the photonic wave are related simply by (c = 1)
em
= t
em
If t
is the period of the light wave, the light is emitted with the wavelength
em
= t
em
; similar relation
holds for the observed wavelength. Their ratio is then given by

obs

em
=
t
obs
t
em
=
a
obs
a
em
. (2.16)
We see that the expansion of the universe leads to the scaling (t) a(t). Then the frequency (t)
1/(t) 1/a(t).
If all the photons have their frequencies shifted by /a the shape of the Planck distribution
does not change. The temperature, which is determined from the maximum of the Planck distribution
is therefore shifted by the amount 1/a. Therefore the energy radiation (photon luminosity T
4
) scales
as 1/a
4
. Because the number density of photons scales as n

T
3
a
3
, while the volume scales as
V a
3
, the total number photons N

= n

V is conserved.
2.4. DOPPLER SHIFT 29
2.4 Doppler shift
The redshift of photons can be interpreted as a Doppler shift due to the relative expansion of two
galaxies. Let us denote the observed frequency as
o
and source frequency as
s
. The relativistic
Doppler eect states that

o
= (1 )
s
, =
v
c
, =
1
_
1
2
. (2.17)
If the velocities are non-relativistic 1, then = 1 and we can write the Doppler law

o
=
s
. (2.18)
Suppose we have two galaxies (1 and 2) at a distance l H
1
. At these small scales we can identify a
local inertial frame in which the eects of general relativity can be neglected and space can be assumed
to be at (i.e., we can use the Special Theory of Relativity). Hubble law tells us that the galaxies are
moving apart with the velocity v = Hl.
Consider now a photon which has a frequency
1
in (source) galaxy 1 at the moment t
1
. Its frequency
is measured at a later time t
2
when it reaches the (observation) galaxy 2. The shift in the frequency is
given then by the Doppler law (2.18)
= (t
1
) (t
2
) (t
1
)v = (t
1
)H(t)l. (2.19)
Since t = t
2
t
1
= l we can rewrite this as

t
= = H(t) (2.20)
which has the solution

1
a
. (2.21)
Applying this argument to small pieces of photon trajectories we conclude that our solution is valid in
general relativistic case as well.
2.5 Peculiar velocity induced redshifts
Motions that are superimposed on the general expansion of the universe were dened as peculiar ve-
locities. We can expect redshifts originating from these velocities. Suppose we have two observers (1
and 2). The observer 1 is located in the lab frame; the observer 2 is located at the origin of a frame
which moves with respect to the lab frame according to the Hubble law with the velocity v = H(t)l,
where l is the distance between the origins of the two reference frames. The velocity measured by
the observer 1 is denoted by w
1
(t
1
) and that by the observer 2 by w
2
(t
2
). Then applying the Galilean
transformation to connect the velocities w(t) in both frames we obtain
w(t
1
) w(t
2
) v = H(t)l. (2.22)
We assume that the times t
1
and t
2
are innitesimally close; in fact we will take the limit t
1
t
2
shortly.
The time that will take a particle to move from observer 1 to observer 2 is t = t
2
t
1
= l/w. Then
from Eq. (2.22) w(t
1
) w(t
2
) = H(t)wt. Taking the limit t 0 we obtain
w = H(t)w w
1
a
. (2.23)
30 CHAPTER 2. LIGHT PROPAGATION IN EXPANDING UNIVERSE
We see that the peculiar velocity of a particle decays inversely as the scale factor increases. We conclude
that the overall expansion of the universe leads to a damping of the peculiar motions in the comoving
frame!
Consider as an example now a non-relativistic non-interacting gas. Its temperature is proportional
to their kinetic energy of particles, i.e., their peculiar velocity squared
T
nonrel.gas
w
2
a
2
. (2.24)
This behavior diers from that for photons (radiation) for which T

a
1
. We see that if radiation
and (non-relativistic) matter are decoupled, then the gas cools faster than the radiation as the universe
expands. As in the case of radiation we can extend our argument to the case of General Relativity by
considering piece-wise tracks with locally inertial systems attached to them.
2.5.1 Redshift as a measure of time
Next we dene the redshift parameter (or simply redshift) through the following relation
z =

obs

em

em
, (2.25)
where
obs
is the wave-length of the photon observed currently on the Earth,
em
is the wave-length of
the photon emitted by a distant galaxy. However, according to Eq. (2.16)
obs
/
em
= a
obs
/a
em
and we
can write
1 + z =
a
obs
a(t
em
)
. (2.26)
We will take further the observation time t
obs
= t
0
, i.e., the contemporary time, and therefore a(t
obs
) =
a
0
.
Equation (2.26) establishes a one-to-one correspondence between the time and the redshift. There-
fore, we conclude that redshift can be used to parameterize the history of the universe, instead of time,
i.e. all the functions of time can be rewritten in terms of z. Applying Eq. (2.26) to the current time we
obtain
1 + z =
a
0
a(t
em
)
. (2.27)
This relation tells us that the universe was 1 + z times smaller at redshift z than now.
It is useful to rewrite some of our previous expressions in terms of the redshift. For example,
Eq. (1.56), which reads d = 3( + p)d ln a can be integrated to obtain
_
(z)

0
d
+ p()
= 3 ln
a(z)
a
0
,
_
(z)

0
d
+p()
= 3 ln(1 + z), (2.28)
where we inserted (2.26) for the ratio scale factor under the logarithm. Next, let us obtain the expression
for the Hubble parameter H in terms of z. We modify the second Friedman equation (1.74) as
H
2
+
k
a
2
H
2
+
k
a
2
0
(1 + z)
2
,
8G
3
=
8G
3

cr
. .
H
2
0

cr
..

0
(z)

0
, (2.29)
which gives us nally
H
2
(z) +
k
a
2
0
(1 +z)
2
=
0
H
2
0
(z)

0
. (2.30)
2.5. PECULIAR VELOCITY INDUCED REDSHIFTS 31
Note that z = 0 point corresponds to the current time (i.e. all the quantities have an index 0) and
z = corresponds to the initial time. If z = 0 then H(z) = H
0
and Eq. (2.30) reduces to
k
a
2
0
= (
0
1)H
2
0
. (2.31)
This is an interesting relation, since it expresses the current value of the scale factor a
0
in terms of
current values of Hubble constant H
0
and cosmological parameter
0
provided the universe is not at,
i.e., k = 1. Now let us substitute (2.31) back into (2.30) to eliminate the factor k/a
2
0
; then
H(z) = H
0
_
(1
0
)(1 + z)
2
+
0
(z)

0
_
1/2
. (2.32)
This is an alternative to Eq. (2.30) which expresses the time dependence of the Hubble parameter in
terms of
0
,
0
and (z).
Next we would like to express the time in terms of the redshift. Expression (2.26) 1 + z = a
0
/a(t)
provides an indirect relation. To obtain a direct expression let us dierentiate it
dz =
a
0
a
2
(t)
a(t)dt = (1 +z)H(t)dt, (2.33)
which, after integrating, gives
t =
_

z
dz

H(z

)(1 +z

)
. (2.34)
Thus given (z) we can obtain from Eq. (2.30) H(z) after which we can integrate (2.34) to obtain the
time. Note that the upper limit z = corresponds to the initial time (t = 0).
2.5.2 Redshift as a measure of distance
If we measure the redshift of light from a distant galaxy we can determine the distance to that galaxy,
i.e., the redshift can be used eectively to measure not only time but also distances. The comoving
distance is generally given by
=
0

em
=
_
t
0
t
em
dt
a(t)
=
1
a
0
_
t
0
t
em
dt(1 + z) (2.35)
where we have substituted a(t) = a
0
/(1 + z). According to Eq. (2.33) (1 + z)dt = dz/H(t) and we
nd
(z) =
1
a
0
_
z
0
dz

H(z

)
. (2.36)
If the universe is curved k ,= 0 then according to (2.31)
1
a
0
=
_
[
0
1[H
0
, (z) =
_
[
0
1[H
0
_
z
0
dz

H(z

)
. (2.37)
Now let us give an example, where we know the function (z). Such example is provided by the
dust-dominated universe where (z) =
0
(1 + z)
3
. Substituting this in Eq. (2.32)
H(z) = H
0
(1 + z)
_
1 +
0
z. (2.38)
32 CHAPTER 2. LIGHT PROPAGATION IN EXPANDING UNIVERSE
Figure 2.2: Illustration of geometry of light emission by a standard ruler.
Further assume that the universe is at
0
= 1. Then Eqs. (2.34) and (2.36) valid for this case can be
integrated analytically upon substituting (2.38) and we nd
t(z) =
2
3H
0
(1 + z)
3/2
, (z) =
2
a
0
H
0
_
1
1

1 + z
_
. (2.39)
In the limit z 0 (current time) we obtain = 0 and t = 2/3H
0
, as expected.
2.6 Determining cosmological parameters
We already mentioned that there are two ways to determine the cosmological parameters. First, one can
measure the angular size of certain objects and nd the evolution of the angular size with the redshift.
If these objects have about the same size (standard rulers) we can make conclusion about the dynamics
of the universe. Similarly, by using objects which have same total brightness (standard candles) we
can determine their apparent luminosities at dierent redshifts to x the cosmological parameters. By
using such observation one can study the recent expansion of the universe with dierent matter content
and distinguish between dierent models.
2.6.1 Standard rulers
We start by considering an object with the size l at distance
em
, where it emits light. Without loss
of generality we can assume that the light propagates along the geodesics, and therefore the ends of
this object have angular coordinates
0
,
0
and
0
,
0
+, (see Fig. 4.2) We are interested in photons
emitted by the endpoints of the standard ruler. The observer is located at comoving coordinate
0
= 0
and observes at t = t
0
. The size of the object is given by the interval
l =

s
2
= a(t
em
)(
em
), (2.40)
2.6. DETERMINING COSMOLOGICAL PARAMETERS 33
where we used the metric (2.3)
ds
2
= a
2
()
_
d
2
d
2

2
()(d
2
+ sin
2
d
2
)

, (2.41)
with = const for xed conformal time and comoving coordinate. The angle is then given by
=
l
a(t
em
)(
em
)
=
l
a(
0

em
)(
em
)
(2.42)
where we have replaced t
em

0

em
. Consider limiting cases:
the object is close to us
em

0
[i.e. the emission time
em

0
], therefore
a(
0

em
) a(
0
), (
em
)
em
, =
l
a(
0
)
em
=
l
d
, (2.43)
where d is the distance to the object and the second relation assumes approximately at universe.
The last expression is the one that we would expect from the Euclidean geometry.
the object is located far away (close to the particle horizon)
em
=
0

em

0
, then
a(
em
) a
0
, (
em
) (
p
) = Const, (2.44)
and the general formula for this case reads

l
a(
em
)
. (2.45)
We see that as a 0 the angle increases to innity and the object covers the whole horizon.
However, the luminosity of the object drops drastically with the distance (otherwise it will outshine
all the objects in the vicinity of the observer).
To compare with concrete predictions of cosmological models we rst express the angular size in
terms of redshift instead of emission time:
(z) = (1 + z)
l
a
0
(
em
(z))
,
em
(z) =
1
a
0
_
z
0
dz

H(z

)
, (2.46)
where we used that (1 + z)/a
0
= 1/a(t
em
) and (2.36) for
em
.
This rst formula can be applied to concrete cosmological models. Let us take the example of at,
dust dominated universe, in which case
em
is given by Eq. (2.39) and Eq. (2.4) implies (
em
) =
em
.
Then substituting
em
from Eq. (2.36) we obtain
(z) =
lH
0
2
(1 + z)
3/2
(1 + z)
1/2
1
. (2.47)
The behavior of this curve is as follows: z 1, (z) z
1
, it has a minimum at 5/4 and is z
for z 1. If we could experimentally determine the angular diameter for several redshifts, we could
test the cosmological models. However, such measurements are seldom because of the lack of standard
rulers. The only successful example is the measurement of the cosmic microwave background (i.e.
just one standard ruler). The measurements of acoustic peaks in the power spectrum of temperature
autocorrelation function allows to determine the cosmological parameter in an almost model independent
way from the rst acoustic peak. We will return to this problem in a later lecture.
34 CHAPTER 2. LIGHT PROPAGATION IN EXPANDING UNIVERSE
2.6.2 Standard candles
Consider a source located at the comoving coordinate
em
with total luminosity L. The energy output
at time t
em
within the conformal interval is given by
E
em
= Lt
em
= La(t
em
). (2.48)
We see that the energy of emitted photons depends on the scale factor, however the conformal width
of the shell from which the photons are emitted = const. Therefore, at the moment of observation
the energy output will be
E
obs
= E
em
a(t
em
)
a
0
= L
a
2
(t
em
)
a
0
. (2.49)
Generalizing Eqs. (1.41) and (1.43) for surface area of spheres in expanding universe we can write
S(t
0
) = 4a
2
0

2
(
em
), (2.50)
since
em
=
obs
= Const. The time interval needed for the shell to pass over the observers position is
t = l, where l is the width of the shell is given by
l = a
0
= a
0
. (2.51)
The measured energy per unit area per unit time is equal to
F =
E
obs
S(t
0
)t
=
L
S(t
0
)t
a
2
(t
em
)
a
0
=
L
4
2
(
em
)t
a
2
(t
em
)
a
3
0
=
L
4
2
(
em
)
a
2
(t
em
)
a
4
0
. (2.52)
But the ratio of the scale factors can be expressed via the redshift according to Eq. (2.26), so that
F =
L
4a
2
0

2
[
em
(z)](1 +z)
2
, (z) =
1
a
0
_
z
0
dz

a(z

)
2
a(z

)
2
. (2.53)
In astronomy we measure the bolometric magnitude
m
bol
= 2.5 log
10
F = 2.5 log
10
L + 2.5 log
10
(4a
2
0

2
[
em
(z)](1 + z)
2
)
= 5 log
10
(1 + z) + 5 log
10
([
em
(z)]) + 2.5 log
10
(4a
2
0
) 2.5 log
10
L. (2.54)
The last two terms are constants independent of z. Now let us expand this function in the limit z 1
(contemporary time). Since to leading order in small we have sin and sinh , then
independent of the curvature of the universe we can substitute [
em
(z)] =
em
(z). The expansion to
leading order gives
m
bol
(z) = 5 log
10
z +
2.5
ln 10
(1 q
0
)z + O(z
2
), q =
a
aH
2
. (2.55)
where q
0
is the deceleration parameter dened as
q
0
=
1
2

0
_
1 + 3
p

_
0
. (2.56)
By determining the function m
bol
(z) experimentally from observations and tting it by some choice of
parameters entering q
0
we can determine these parameters, i.e.,
0
, p
0
and
0
.
Type IA supernovae are suitable standard candles. Their observation led to the following conclusions:
2.6. DETERMINING COSMOLOGICAL PARAMETERS 35
Figure 2.3: Fitting the supernova data with models with various fractional contributions from matter
and cosmological constant.
The expansion of the Universe is accelerating.
From Friedman equation
a =
4
3
G( + 3p)a (2.57)
we see that the universe will be accelerating if a > 0, which means that the right hand side of
this equation must be positive, which in turn means that a substantial fraction of the total energy
density is dark energy with negative pressure (e.g. p = w, with w being negative).
What is the origin of the dark energy?
Cosmological constant, i.e., w = 1.
Dynamical eld, e.g., time-dependent scalar eld (called quintessence).
Following problems of modern cosmology are related to the acceleration of the universe:
Why the dark energy is dominating the universe at the current time (much later than the
birth of the universe)
The long term future of the universe can not be predicted since we can not predict the
behavior of the dark matter (e.g. the scalar eld can dissipate by creating matter and
radiation)
The expansion (2.55) is not valid for typical observed supernovae, since their redshifts are of order of
unity. One then needs to carry out a numerical study by using dierent models of the universe. For
example, assume that the universe is at and consists of cold dark matter and cosmological constant,
i.e.,
0
=

+
m
= 1. Then
[
em
(z)] =
em
(z) =
1
H
0
a
0
_
z
0
dz

m
(1 + z

)
3
+ (1
m
)
. (2.58)
36 CHAPTER 2. LIGHT PROPAGATION IN EXPANDING UNIVERSE
With this assumption we can compute the exact expression (2.55) for bolometric luminosity and compare
to observations. The data is best described by
m
= 0.3.
2.6.3 Further methods
Let us calculate the number of a galaxies at a given redshift. We will need to make some simplications;
we assume that the density is uniform and is given by n(z). The dierential number of the galaxies is
given by
N = n(z)V (z) (2.59)
where V (z) is an element of volume at a distance z. To calculate the latter quantity we note that a
surface element can be written
dS(z) = a(z)
2
((z))
2
d, (2.60)
where d is a element of a solid angle. This quantity should be multiplied with the third (physical)
dimension d = a(z), so that we obtain
V = a(z)
3

2
()(z). (2.61)
Next we use that z = aH
0
and 1 +z = a
0
/a(z) [Eq. (2.26)] to obtain
V = (1 + z)
3
a
3
0

2
()(a
0
H)
1
z = (1 + z)
3
a
2
0

2
()H
1
(z)z. (2.62)
To show a concrete example, we substitute the following expression for
2
(),
((z)) =
2
_
[
0
1[

2
0
(1 + z)

0
z + (
0
2)[(1 +
0
z)
1/2
1], (2.63)
which is valid for dust dominated universe (obtained in an Exercise)
N
z
=
4n(z)
H
3
0

4
0
_

0
z + (
0
2)[(1 +
0
z)
1/2
1]
_
2
(1 + z)
6
(1 +
0
z)
1/2
. (2.64)
Given the function n(z) measurements of N/z can be used to test cosmological models. This
method does not take into account the fact that the galaxies undergo some evolution (e.g. merge and
interact); however this is not a real problem if we can compute the galaxy dynamics.
An additional, higher order eect arises from the evolution of the redshift. Consider again the
example of light emitted at the conformal time
em
at a distance which is observed at the current
conformal time
0
. Then,
z(
0
) =
a
0
a
e
=
a(
0
)
a(
0

. .
emission t
)
. (2.65)
Now let us look at the time derivative of the redshift
dz(t)
dt
=
1
a(
0
)
dz()
d
0
=
1
a(
0
)
d
d
0
a(
0
)
a(
0
)
=
1
a(
0
)
_
a

(
0
)
a(
0
)

a(
0
)
a(
0
)
2
a

(
0
)
_
=
a
0
a
e

a
e
a
e
(2.66)
2.6. DETERMINING COSMOLOGICAL PARAMETERS 37
where prime denotes derivative with respect to conformal time and dot denotes derivative with respect
to physical time (dt = ad). Further,
z =
a
0
a
e

a
e
a
e
=
a
0
a
e
..
1+z
a
0
a
0

a
e
a
e
= (1 + z)H
0


H(z). (2.67)
Consider a concrete example of universe which consists of a mixture of matter and vacuum density such
that its energy density is given by
(z) =
cr
0
[

+
m
(1 + z)
3
]. (2.68)
We use the expression for the Hubble parameter (2.32)
H(z) = H
0
_
(1
0
)(1 + z)
2
+
0
(z)

0
_
1/2
. (2.69)
Substituting we obtain
z = (1 +z)H
0
_
1
_
(1
0
+
m
(1 +z) +

(1 + z)
2

1/2
_
. (2.70)
Suppose the universe is at:
0
= 1 and

= 1
m
. The change in the redshift is u = z/(1+z) =
zt/(1 + z) and thus
u = tH
0
_
_
(
m
(1 +z) + (1
m
)(1 + z)
2

1/2
1
_
. (2.71)
We see that when
m
1, u > 0 and when
m
0, u < 0. Suppose
m
= 1 and

= 0 and we
make an observation with an interval 1 yr; then
v 2(

1 + z 1) cm s
1
. (2.72)
Currently it is possible to measure shifts that are by several orders of magnitude larger than those that
are required. However, it is conceivable that in the future decades such a shift can be measured. Such
measurements will provide a direct probe of acceleration of the universe and should be a complement
to other tests discussed above.
38 CHAPTER 2. LIGHT PROPAGATION IN EXPANDING UNIVERSE
Chapter 3
Thermal evolution of the universe
This chapter deals with the evolution of the universe at energies below few MeV. Compared to the
earlier (higher-energy) stages the physics at this scales is well understood.
The main elements of the composition of the universe at this stage are:
Primordial radiation. This is the cosmic microwave radiation at temperature T = 2.73 K. We can
scale back the temperature (T a
1
) to see that it was very high in the early universes. The
current energy content of this radiation is negligibly small (

= 10
34
g cm
3
, which is only 10
5
fraction of the total energy density).
Baryonic matter. This is the component of matter from which ordinary objects are made. The
contribution of the baryonic matter to the total energy density is small, of the order of several
percent.
Dark matter. Experiments indicate that a large fraction of matter and energy of the universe
is dark, i.e., invisible in ordinary light. The total of experimental data indicates that the dark
matter is cold and can contribute to the gravitational instability and structure formation in the
early universe; cold dark matter contributes about 1/4 of the critical density. A good candidate
for such matter are weakly interacting massive particles (WIMPs) predicted by supersymmetric
theories.
Dark energy is non-clustered, i.e., it does not participate in the gravitational instability forma-
tion and has negative pressure, which is equivalent to a cosmological constant (p

). It
constitutes the signicant rest 3/4 of the energy density of the universe.
Primordial neutrinos are relicts of the early stage of the evolution of the universe. They have
negligible contribution to the energy density of the universe and their temperature (in the case of
only three neutrino families) is 1.9 K.
The energy densities of these components (actually, those that contribute substantially to the total
energy density) scale with the redshift as

D
=
cr
0

D
(1 + z)
3(1+w)
,
m
=
cr
0

m
(1 +z)
3
,

=
0

m
(1 +z)
4
, (3.1)
where
cr
0
= 3H
2
0
/8G is the current value of the critical density,
m
is the contribution of the baryonic
and cold dark matter to the cosmological parameter,
D
is the same for dark energy with w 1/3
(currently). From the large z asymptotics we see that initially dominated the radiation, then the
39
40 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
Figure 3.1: The thermal evolution of the universe. This chapter deals with the stages below the few
MeV temperatures, i.e., nucleosynthesis and recombination.
baryonic matter, and nally the dark energy. The transition from baryonic to dark energy dominated
universe occurs at modern times.
Comparing the redshifts at which the energy densities of these components become comparable we
obtain three basic epochs of evolution in terms of redshift:
radiation dominated epoch z
R
> 10
4
, ultra-relativistic matter p = /3 and a

t.
matter dominated epoch 10
4
z
M
0.33 1.33, p = 0 component (dark matter) determines the
evolution with a t
2/3
.
dark energy dominated epoch z < z
M
, the equation of state is p = w and a t
2/3
(1 + w), where
w is typically negative.
Let us take a closer look at the evolution of the universe and the main events that occur. If we start
from cold universe and go back in time to the hot universe, there are roughly the following stages in
times/energies:
10
16
10
17
s. Galaxy formation through gravitational instability.
3.1. PHYSICS AT LEPTON ERA 41
10
12
10
13
s. Free electrons and protons recombine to form neutral hydrogen. Last scattering
surface forms for photons - imprints are seen in the CMB (Hintergrundstrahlung) temperature
uctuations.
10
11
s (T eV). Matter - radiation equipartition.
200 300 s (T 0.05 MeV). Light elements are formed from free neutrons and protons through
the nuclear reactions. We can measure the abundances of the light elements from primordial
nucleosynthesis.
1 s (T 0.5 MeV) The temperature is of the order of the electron mass - electron-positron
annihilation takes place leaving behind photons. (Only a small fraction of electrons survives).
0.2 s (T 12 MeV) Weak interactions are out of equilibrium and primordial neutrinos decouple
from the rest of the matter. The ratio of neutrons to protons remains xed as there is no chemical
equilibrium among them.
10
5
s/200 MeV. The quark-gluon phase transition takes place.
10
10
10
14
s/100 GeV-10 TeV. The electroweak symmetry is restored and gauge bosons are
massless.
10
14
10
43
s/10 TeV-10
19
GeV. Up to these times/energies the classical general relativity
remains likely intact. Matter composition is unclear. Possibly supersymmetry plays a role by
requiring that the supersymmetric partners of the particles in the standard model be present.
10
43
s/10
19
GeV (the Planckian scale). Classical gravity is not valid and should be replaced by
quantum gravity. Several orders below this energy scale one expects a Grand Unication of all
forces in Nature.
3.1 Physics at lepton era
In this section we consider matter below the phase transition from free quarks to baryons, i.e., the
temperature T 100 MeV and the time t > 10
4
s. At these energies the matter composition is given
by
n, p e
+
, e

e
,

. (3.2)
Other components are
, , . . . (, negligible) (3.3)
where . . . stand for other mesons. The leptons, i.e., electrons, muons, tauons and neutrinos play a
crucial role in the reactions taking place at these times, therefore one calls this epoch lepton era.
For each conserved quantity we need one chemical potential; in the lepton era the conserved quan-
tities are electric charge, baryon number, and lepton number. Suppose the particles p
i
, i = 1, 2, 3, 4 are
in equilibrium via a reaction
p
1
+ p
2
= p
3
+ p
4
. (3.4)
Then the equilibrium means the equality of their chemical potentials

1
+
2
=
3
+
4
. (3.5)
42 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
Thus, in general we need to establish the equilibrium reactions among the particles, form which we
determine the chemical potentials for each conserved quantity:

Q
,
B
,
l
, (3.6)
i.e. we need three equations. The chemical potentials of anti-particles are equal to that of the particle
in magnitude but have an opposite sign.
Let us outline how this works on an example where we have the following constituents of matter
n, p, (baryons) (3.7)

0
,

, (mesons) (3.8)
e, , ,
e
,

(leptons). (3.9)
The reactions taking place in such a mixture and relation between the chemical potentials that follow
from application of the rules (3.5) are as follows:
n p +e

+
e
,
n
=
p
+
e
+

e
. (3.10)
The life time of the pion is short (in free space) so that

0
,

0 = 0 =

(3.11)
Heavy muon decays into light particles
e

+
e
+
e
,

=
e

. (3.12)
The same applies for lepton
e

+
e
+

=
e

. (3.13)
Finally


e
+ e

=
e
+

e
. (3.14)
Conservation of baryon number implies that the number of baryons minus the number of anti-baryons
scales as a
3
; if there is no dissipation in matter then the entropy scales also as a
3
and the entropy
per baryon remains constant:
B =
1
s
(n
p
n
p
+ n
n
n
n
) = const. (3.15)
Conservation of the total electric charge is
Q =
1
s

i=p,e,,

(n
i
n
i
). (3.16)
Finally the lepton number conservation for each lepton family f = e, , is given by
L
i
=
1
s

f=e,,
(n
f
n
f
). (3.17)
3.1. PHYSICS AT LEPTON ERA 43
The equilibrium distribution functions for fermionic and bosonic constituents are (we omit here the
degeneracy factors that take into account their internal degrees of freedom, but will recover them in
the following section)
f =
1
exp
_
E
T
_
+ 1
, g =
1
exp
_
E
T
_
1
, (3.18)
so that the densities dened as
n
f
=
_
d
3
p
(2)
3
f(E(p), , T), n
g
=
_
d
3
p
(2)
3
g(E(p), , T). (3.19)
The entropy conservation implies
d
dt
_
sa
3
_
= 0; (3.20)
note that s is the entropy density and should be multiplied with the volume to yield the entropy of
the system; since V a
3
, we obtain (3.20). The conservation laws above determine the six unknown
functions
T,
e
,
n
, and

f
(f = e, , ). (3.21)
The entire universe is electrically neutral, therefore
Q = 0. (3.22)
Baryon per entropy is established from observations to be in the range
B 10
10
10
9
. (3.23)
The lepton to entropy ratio is not well established. From the arguments that will become clear in the
later chapters, the numbers of leptons and baryons can not be very dierent; this implies that
[L
i
[ < 10
9
. (3.24)
When the temperature is much larger than the mass m of the particles - many particle and anti-particle
pairs are created from vacuum, i.e., the number of pairs are of the order of the number of photons, which
scales as n

T
3
. What happens when T < m? We might expect that the pairs annihilate until a
small number of particles are left. We will need to recover some basic relations from thermodynamics
in order to quantify the particle number excess.
3.1.1 A recourse in thermodynamics
Let us recall some basics thermodynamics. We have pointed out above that the equilibrium distribution
functions for fermions and bosons are given in terms of the Fermi-Dirac and Bose-Einstein distribution
functions as
f =
1
exp
_
E
T
_
+ 1
, g =
1
exp
_
E
T
_
1
. (3.25)
The density of the Fermi/Bose particles is then given as
n = g
_

0
d
3
p
(2)
3
1
exp
_
E
T
_
1
, (3.26)
44 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
where the upper sign corresponds to fermions and the lower sign to bosons, g is the degeneracy factor
(g = 2s+1, where s is the value of the spin of particles). Suppose now the particle spectrum is given by
E =
_
p
2
+ m
2
, which corresponds to an ideal, i.e., non-interacting gas of relativistic particles. Then
we have
EdE = pdp, 0 p , m E . (3.27)
Changing the integration in Eq. (3.3) from momentum to energy we obtain
n =
g
2
2
_

m
dE E

E
2
m
2
exp
_
E
T
_
1
, (3.28)
where we have carried out the angular integration over the phases d = sin dd = 4, since the
spectrum and, therefore, the integrand are independent of direction. The energy of the system can be
written as
= g
_

0
d
3
p
(2)
3
E
exp
_
E
T
_
1
=
g
2
2
_

m
dE E
2

E
2
m
2
exp
_
E
T
_
1
. (3.29)
The expression for the pressure is then given by
1
p =

3

m
2
g
6
2
_

m

E
2
m
2
exp
_
E
T
_
1
. (3.30)
Adding the contribution from anti-particles amounts to adding identical integrals, but with .
Thus, we see that to compute the thermodynamics quantities of non-interacting relativistic gas we need
integrals of the type
I
()

(, ) =
_

(x
2

2
)
/2
e
x
1
+ ( ), (3.31)
where we have introduced non-dimensional quantities = m/T, = /T and the variable x = E/T.
Using these denitions we can write the previous expression in a compact form. The energy becomes
+ =
gT
4
2
2
_
I
(3)

+
2
I
(1)

_
. (3.32)
For the pressure we obtain
p =
gT
4
6
2
I
(3)

. (3.33)
Finally, the excess of particles over the anti-particles can be written as
n n =
gT
3
6
2

I
(3)

. (3.34)
While the integrals (3.31) are easily calculated numerically it is instructive to consider several limiting
cases.
1
We will accept this expression without proof. See any text on thermodynamics for further detail.
3.1. PHYSICS AT LEPTON ERA 45
High temperature expansion of the integrals
Consider the limit where T m, we have , 1. and the corresponding expansions for fermions
and bosons
I
(1)

=
_

3
3


2
2

_

1
2

2
+ O(
4
,
2

2
),

2
6
+

2
2
+

2
2
+ O(
4
,
2

2
),
(3.35)
where C
F
= 0.577 is the Euler number and
=
_

4
+ C
E

1
2
_
. (3.36)
Further
I
(3)

=
_
2
4
15
+

2
2
(2
2

2
) + (
2

2
)
3/2


+ O(
6
,
4

2
),
7
4
60
+

2
4
(2
2

2
) +


3
4
(ln 2)
4
+ O(
4
,
4

2
),
(3.37)
where

=
1
8
_
2
4
6
2

2
3
4
ln
_
C
E

4e
3/4
__
. (3.38)
Low temperature expansion of the integrals
The low temperature limit is given by the conditions = m/T 1 and 1. In this limit we
obtain
I
(1)

2e

cosh
_
1 +
3
8
+ O(
2
)
_
, (3.39)
and
I
(3)

18
3
e

cosh
_
1 +
15
8
+ O(
2
)
_
. (3.40)
With this formulae we can compute the thermodynamic quantities of of non-relativistic particles in the
limit 1. It is clear that in this case the Fermi and Bose distribution functions reduce to the
Boltzmann distribution function, i.e., the thermodynamic parameters are the same for both statistics.
3.1.2 Thermodynamics of ultra-relativistic and non-relativistic gases
Now using the general formulae we can obtain explicit results for some specic examples that are of
interest in cosmology.
Ultrarelativistic particles: bosons
The form of the Bose distribution function implies that in order the distribution to be positive denite
the chemical potential of a boson can not exceed its mass m. We take the high-temperature
expansion for bosons (3.37) and evaluate the particle-anti-particle asymmetry from (3.34). The result
is
n n =
gT
3

3T
, ( < m). (3.41)
Note that this equation does not take into account the possibility of Bose-Einstein condensation, in
which case a macroscopically large number of particles can occupy the ground state with
= lim
p0
_
p
2
+ m
2
= m. (3.42)
46 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
For ultra-relativistic particles p; setting = 0 in Eq. (3.26) we obtain the estimate
n
(3)gT
3

2
. (3.43)
Note that Eqs. (3.41) and (3.43) can not be compared to each other since the latter is taken in the limit
of Bose-Einstein condensation while the former excludes that limit. We stress again that The number
of particles that can be accommodated by Bose condensate in its ground state can be macroscopically
large.
However, in the absence of Bose-Einstein condensation then n n n and particle and anti-particles
contribute equally to the thermodynamic quantities - energy density, pressure, and entropy
=
1
2
( + )

2
gT
4
30
, (3.44)
p =

3
, (3.45)
s =
4
3T
=
2
4
45(3)
n. (3.46)
Ultrarelativistic particles: fermions
For fermions the chemical potential is not bounded from above. We take the ultrarelativistic limit by
formally putting m = 0, i.e., = 0 in Eq. (3.37) to obtain
+ =
7
2
120
gT
4
_
1 +
30
2
7
2
+
15
4
7
4
_
, (3.47)
which is valid at arbitrary chemical potentials for ultrarelativistic gas. As before we have p = /3.
Using Eq. (3.34) we obtain for particle-antiparticle asymmetry
n n =
gT
3
6

_
1 +

2

2
_
. (3.48)
The entropy can be computed from a thermodynamical equation s = T
1
( + p n):
s + s =
7
2
90
gT
3
_
1 +
15
2
7
2
_
. (3.49)
As noted Eqs. (3.1.3)-(3.49) are valid at any . Let us consider now some limiting cases.
Large . This constitutes the degenerate Fermi gas, where T . In that case the rightmost terms
in these equations are dominant. The energy density and entropy of a degenerate Fermi gas to leading
order in /T is then given by
+ =
g
4
8
2
+
gT
2

2
4
, (3.50)
s + s =
g
2
T
4
. (3.51)
We see that to leading order the energy of degenerate gas is independent of temperature and is entirely
due to the degeneracy pressure (Pauli principle). The next to leading correction arises from the narrow
3.1. PHYSICS AT LEPTON ERA 47
shell around the Fermi surface. The entropy again is linear in temperature and vanishes in the zero
temperature limit (as it should).
Small . This corresponds to /T = 1, i.e., we can neglect the chemical potentials in (3.26).
The integral is the straightforward and gives
n =
3(3)gT
3
4
2
. (3.52)
At the same time from the general equation (3.48) we obtain in this limit
n n
gT
3
6
n, (3.53)
i.e., the particle over anti-particle densities are parametrically suppressed by compared to the particle
density. For the remaining parameters one nds
=
7
2
240
gT
4
, p =

3
, s =
4
3T
. (3.54)
If we have simultaneously bosons and fermions with the same number of internal degrees for freedom
(i.e. g is the same) then their entropies are related by
s
fermions
=
7
8
s
bosons
. (3.55)
Non-relativistic particles
Here we use the results obtained in the case of low-temperature expansion. We saw that it is valid in
the limit = (m )/T 1. In this case the exponential factor is larger than the unity and we
have Boltzmann distribution function. For this case we use again Eq. (3.34) to obtain
n n 2g
_
Tm
2
_
3/2
exp
_

m
T
_
sinh
_

T
_
_
1 +
15T
8m
_
. (3.56)
The number density of the particle is given by
n g
_
Tm
2
_
3/2
exp
_

m
T
_
sinh
_

T
_
_
1 +
15T
8m
_
. (3.57)
The number density of anti-particles is obtained via the replacement and is suppressed by a
factor exp(2)/T) (the second exponent arises from sinh function). Substituting (3.39) and (3.40)
into the expression for the energy (3.32) and pressure the energy (3.33) we obtain
= mn +
3
2
nT, p nT. (3.58)
The entropy is again computed from s = T
1
( + p n) to be
s
_
m
T
+
5
2
_
n. (3.59)
If the opposite condition is satised for fermions, i.e., ( m)/T 1 then we have a degenerate Fermi
gas. This case case is not of interest for our discussion.
48 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
3.1.3 Particle-antiparticle annihilation and neutrino decoupling
Having determined the general formulae for thermodynamic functions lets resume the study the lepton
era. We start with the eect of extinction of the particle-anti-particle pairs:
When the temperature is higher than the particle mass there are many particle-anti-particle pairs.
Once the temperature drops below the particle mass there is a substantial excess of particle over
anti-particles.
To quantify this eect consider the particle-anti-particle asymmetry
=
1
s
(n n). (3.60)
This equation can be solved by squaring it, i.e., n
2
2n n + n
2
s
2

2
= 0 and eliminating n
2
by using
the denition (3.60) n = n s and denoting n n = b we obtain
n
2
sn b = 0, n =
s +

s
2

2
+ 4b
2
, n =
s +

s
2

2
+ 4b
2
, (3.61)
In the case of non-relativistic particle we can use Eq. (3.57) to obtain the estimate
b = n n = g
2
_
Tm
2
_
3
exp
_

2m
T
_
. (3.62)
It is seen that the anti-particles will become extinct when

2

b
s
exp
_

m
T
_
. (3.63)
Since for m T the right-hand side is exponentially suppressed, we see that once the temperature
drops below the particle mass the anti-particles are strongly suppressed.
Baryon parameters
Now let us look at the baryon number again
B =
1
s
(n
p
n
p
+ n
n
n
n
+ n

) = const. (3.64)
where we have added the particles to our collection of baryons. Since the equilibrium tells us that

=
n
we see that
n

n
n
n
n
=
_
m

m
n
_
3/2
exp
_

m
n
T
_
exp
_

176 MeV
T
_
. (3.65)
We see that below the temperature of the order 100 MeV the particles do not contribute to the total
baryon number. Furthermore since the temperatures are well below the mass of the nucleons 938 MeV,
the contribution due to the anti-baryons can be neglected and the baryon number in the lepton era is
due to the neutrons and protons only; thus we obtain (assuming n
p
n
n
)
B =
1
s
n
p
_
1 +
n
n
n
p
_

1
s
n
p

g
s
_
Tm
p
2
_
3/2
exp
_

m
p

p
T
_
sinh
_

p
T
_
(3.66)
3.1. PHYSICS AT LEPTON ERA 49
or, equivalently,
m
p

p
T
ln
_
1
B
_
m
p
T
_
3/2
_
, s T
3
. (3.67)
The chemical potential of protons is estimated to be in the range 100
p
1000 MeV for temper-
atures in the range 40 T 1. The entropy of protons is given by
s
p
s
=
n
p
s
_
m
p

p
(t)
T
+
5
2
_
= Bln
_
1
B
_
m
p
T
_
3/2
_
10
8
, (3.68)
if we assume that B 10
10
.
To derive Eq. (3.66) we have assumed that n
n
n
p
. Let us clarify when this is the case. From
the -equilibrium between neutrons and protons we have the condition
n
=
p
+
e
+

, where

is
the chemical potential of anti-neutrinos and can be replaced by the chemical potential of neutrinos via

. Neglecting the dierence in the masses of neutrons and protons we obtain from Eq. (eq:3.122)
n
n
n
p
= exp
_

m
n
m
p
+

e
T
_
exp
_

Q
T
_
, (3.69)
where Q = m
n
m
p
= 1.293 MeV. We will show below that the chemical potentials of electrons and
neutrino can be neglected compared to Q.
Lepton parameters
In the lepton number conservation we shall keep only the electrons, and the three sorts of neutrinos;
the -on and -on are to heavy to be present at signicant amounts in matter. We rewrite the lepton
number conservation for these species as
L
e
=
1
s
(n
e
+ n

e
), L

=
1
s
n

, L

=
1
s
n

. (3.70)
We use the fact that [see Eq. (3.48)] and that s T
3
n

s

g

6T
L

,
n

s
n

6T
L

. (3.71)
Since L
i
= const., we conclude that the chemical potentials of leptons decrease inversely proportional to
the temperature as the universe expands. In the electronic sector we have
n
e
+ n

e

g
e

e
+ g

e
6T
L
e
= const. (3.72)
This does not allow us to estimate the chemical potential of electrons. This can be estimate from
the conservation of charge (electrons are responsible for compensating the positive charge of protons).
Indeed we have
n
e
s


e
T
=
n
p
s
B 10
10

e
10
10
T. (3.73)
50 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
Neutrino decoupling
Consider a universe where where we have photons, electrons, and the three neutrino species in equilib-
rium. All the species are ultrarelativistic. The energy density of such a mix follows from Eqs. () and ()
to leading order in


2
T
4
30
_
g
B
+
7
4
g
F
_
. (3.74)
where g
B/F
are the total number of internal degrees of freedom of bosons and fermions, respectively. The
only bosons in our mixture are photons, which have two polarizations, hence g
B
= 2. Electrons have to
spin orientations, whereas each of the three neutrino species have one orientation (since neutrinos are
left-handed). In total the contribution of fermionic particles is g
F
= 2+1+1+1 = 5. The anti-particles
double this value, so that nally g
F
= 10. Computing the numerical factor we obtain
= 3.54T
4
. (3.75)
For the at radiation dominated universe we had a

t and = 3/32Gt
2
[see Eq. (1.81)]. From
which it follows that
t
1

32
T
2
, t(sec) (T MeV)
2
. (3.76)
We see that the lepton era which corresponds to temperatures of the order of few MeV occurs at about
1 sec after the beginning of the Universe.
Lepton era is characterized by the following events:
The neutrino decoupling. This occurs when the mean free path of neutrino scattering o other
constituents of matter becomes of the order of the cosmological time given by Eq. (3.76). This
occurs when the temperature is of the order 1.5 MeV
The departure of baryons from chemical equilibrium via weak interactions. This occurs when the
temperature is of the order of 0.8 MeV.
The electron-positron annihilation at temperature T m
e
heats up the universe.
Now we concentrate on the rst eect. The key processes are
e
+
+ e

e
+
e
(electron positron annihilation) (3.77)
e

+
e
e

+
e
(electron/positron neutrino scattering) (3.78)
e

+
e
e

+
e
(electron/positron anti neutrino scattering) (3.79)
The diagram showing the scattering of the neutrino o the electron is shown in Fig. 3.3. The matrix
element for this weak neutral process is given by
/=
g
2
Z
8M
2
Z
_
u(3)

(1
5
)u(1)
_
u(4)

(c
V
c
A

5
)u(2)

, (3.80)
where c
V
and c
A
are the weak coupling constants and the particle spinors u(1) etc and antiparticle
spinors v(1) etc satisfy the Dirac equation
(,p m)u(p) = 0, (,p + m)v(p) = 0. (3.81)
3.1. PHYSICS AT LEPTON ERA 51
Figure 3.2: Key processes contributing to the thermalization of neutrinos. (a) Electron-positron annihi-
lation, (b) electron neutrino scattering via Z
0
boson exchange, electron-neutrino (exchange) scattering
with W

boson exchange.
p
p p
p
1
3
4
2

e
e
Figure 3.3: The Feynman diagram for neutrino-electron scattering.
The matrix element squared (which is an average over the initial spin states and sum over all possible
nal ones) is given by
[/
2
[ = 2
_
g
Z
4M
Z
_
4
Tr
_

(1
5
),p
1

(1
5
),p
3

Tr
_

(c
V
c
A

5
)(,p
2
+m)

(c
V
c
A

5
)(,p
4
+ m)

=
1
2
_
g
Z
M
Z
_
4
_
(c
V
+ c
A
)
2
(p
1
p
2
)(p
3
p
4
) + (c
V
c
A
)
2
(p
1
p
4
)(p
2
p
3
) m
2
(c
2
V
c
2
A
)(p
1
p
3
)

.
(3.82)
We shall assume that electrons are ultrarelativistic m = 0 and go to the center-of-mass frame. Then
[/
2
[ =
1
2
_
g
Z
M
Z
_
4
_
(c
V
+ c
A
)
2
(p
1
p
2
)(p
3
p
4
) + (c
V
c
A
)
2
(p
1
p
4
)(p
2
p
3
)

. (3.83)
The four scalar products are done according to the rule p
1
p
2
= E
1
E
2
p
1
p
2
and we nally obtain
[/
2
[ = 2
_
g
Z
M
Z
_
4
E
4
_
(c
V
+c
A
)
2
+ (c
V
c
A
)
2
cos
4

2
_
. (3.84)
Now we can dene the dierential cross-section as
d
e
d
=
1
64
2
[/
2
[
(E
1
+ E
2
)
2
[ p
f
[
[ p
i
[
=
2

2
_
g
Z
4M
Z
_
4
E
2
_
(c
V
+ c
A
)
2
+ (c
V
c
A
)
2
cos
4

2
_
. (3.85)
where p
i
and p
f
are the initial and nal momenta in a general center of mass frame and have equal
moduli since the scattering is elastic. For the order of magnitude estimate we will need only the total
cross section

e
=
2
3
_
g
Z
2M
Z
_
4
E
2
(c
2
V
+ c
2
A
+ c
V
c
A
) =
32
3
3

2
W
M
4
Z
E
2
(c
2
V
+ c
2
A
+ c
V
c
A
). (3.86)
52 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
where
W
= g
2
W
/4 = 1/29 is the weak coupling constant and we used g
W
= g
Z
cos
W
, where
W
=
28.7
o
, [sin
2

W
= 0.23]. Similar estimate is valid for the charge current reactions. Thus we obtain an
estimate

e


2
W
M
4
W,Z
T
2
. (3.87)
Using the scattering rate from kinetic theory with the cross-section (3.87) we obtain

(
e
n
e
..
T
3
)
1

2
W
M
4
W
T
5
. (3.88)
From the condition

t, the latter given by (3.76), we obtain the temperature at which the electron
neutrinos decouple M
T

e
=
2/3
W
M
4/3
W
1.5 MeV. (3.89)
The and neutrinos couple to the rest of the matter via their charge neutral scattering o the electrons
e +
,
e +
,
. (3.90)
The rates of these processes that proceed via Z
0
exchange are smaller than that for electron neutrinos,
therefore the and neutrinos decouple earlier, i.e., at higher temperature than the electron neutrinos.
After the decoupling neutrinos do not interact with matter and their spectrum remains Planckian
spectrum at T = 1.5 MeV. Their temperature changes because of the expansion of the universe as
T a
1
and can be calculated in any specic model of evolution of the universe.
The next event is the departure of the baryons from chemical equilibrium, which we will treat in
more detail later in the context of nucleosynthesis. For now we concentrate on the eect of electron-
positron annihilation as the temperature becomes of the order of electron mass. This process will rise the
temperature of the radiation. Since the neutrinos decouple before the annihilation, their temperature
will be unaected and will stay at the value derived above.
We can now make a robust estimate of the ratio of the neutrino to the radiation temperatures. Note
that after neutrino decoupling the entropy density stays constant. The net entropy of the remaining
matter also stays constant
s

= Const. s

+ s
e
= Const. (3.91)
Since s

T
3

and s

T
3

, then
s

+ s
e

=
s

+
s

s
e

=
T
3

T
3

_
1 +
s
e

_
= Const. (3.92)
Just before decoupling of neutrinos we have T

= T

and the ratio of the entropies can be computed


from Eq. (3.55), i.e., s
e
= (7/8)s

since the number of internal degrees of freedom of a photon (two


polarizations) and electron (two spin directions) are equal. Taking into account the positrons we have
s
e

=
7
4
. (3.93)
at the decoupling. Substituting this and T

= T

in (3.91) we obtain the value of the constant 11/4.


Asymptotically, when the electron-positron pairs are annihilated their entropy vanishes and we obtain
T

=
_
11
4
_
1/3
= 1.401. (3.94)
Since the measured microwave background temperature is 2.73 K, then the primordial neutrino tem-
perature should be equal 1.95 K. However, there is no detection of the background primordial neutrinos
to date.
3.2. NUCLEOSYNTHESIS 53
3.2 Nucleosynthesis
Current universe consist of 75% Hydrogen and about 25%
4
He. The remaining element contribute
only trace amount to the matter density.
The
4
H nucleus has mass number A = 4 and is a bound state of two neutrons and two protons, with
bound state energy 28 MeV. It is much more stable than the neighboring nuclei with mass number 3
and 6 (a stable nucleus with mass number 5 does not exist in Nature). The fact that helium is so stable
is the consequence of the closed shell structure of this nucleus. The key question we ask now is when
and where was such a large amount of Helium created?
We know that at least some elements were created in stars through nuclear fusion reactions. This
is however not the case for Helium. The reason is the following: The observed luminosity vs mass ratio
in the universe is
_
L
M
B
_
obs
0.05
L

. (3.95)
where L

is the solar luminosity and M

= 1.3310
33
g is the solar mass. Let us compute this number
assuming that it was created in stars. The energy released as radiation when a Helium nucleus is created
is E
H
= 28/4 = 7 MeV per nucleon; note that if we do the calculations in the cgs units common in
astrophysics we should convert the MeV to ergs. The conversion formula is 1 MeV= 1.610
6
erg. The
mass of a nucleon is m
B
= 1.67 10
24
g and the age of the universe is approximately 3.2 10
17
s. Then
L
M
B

E
H

1
4m
B
= 5
erg
gs
= 2.5
L

. (3.96)
The discrepancy between (3.95) and (3.96) can be attributed to the fact that most of the Helium was
created in the early universe and not in stars.
It is clear that when the temperature is much higher than the binding energy of the Helium nucleus
(28 MeV) the nucleus will be easily dissociated. Therefore the nucleus must have been created after
the temperature drop below 28 MeV, when the universe seconds old. The approximate analysis below
shows that the primordial elements were created seconds to minutes after the birth of the universe when
the temperatures are few MeV.
We start by the analysis of the evolution of neutron and proton numbers in the regime of temper-
atures relevant for the nucleosynthesis. Neutron and proton abundances are maintained in equilibrium
via the charge current weak interactions
n +
e
p + e, n +e
+
p +
e
. (3.97)
The momenta and energies involved in this process are much smaller than the mass scale of gauges
bosons. Therefore, we can approximate the gauge boson propagator by a contact term. In that case
the matrix elements of the processes (3.97) is given by
[/[
2
= 16(1 + 3g
2
A
)G
2
F
(p
n
p

)(p
p
p
e
), (3.98)
where G
F
= 1.17
5
GeV
2
. The rst term in the bracket is due to the vector coupling (i.e. the vertex
is just a Dirac spinor

) the second term corresponds to the axial-vector coupling, i.e., the vertex is of
the type

5
; the factor g
A
1.26 is the axial vector coupling constant.
Now we wish to calculate in some detail the cross-section for the process n+ p+e. The general
formula for the dierential cross-section is given by
d
e
d
=
1
(8)
2
[/[
2
(p

+ p
n
)
2
_
(p
p
p
e
)
2
m
2
p
m
2
e
(p
n
p

)
2
m
2
n
m
2

_
1/2
. (3.99)
54 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
Note that we deal here with four-vectors so that p
2
= E
2
p
2
and p
1
p
2
= E
1
E
2
p
1
p
2
. Furthermore,
since E
2
= p
2
+ m
2
we see that p
2
= m
2
. We will ignore the mass of electron neutrino, which is small
compared to the other scales in the problem (temperature, masses of other particles). Applying these
general relations to Eq. (3.99)
(p
n
+p

)
2
= p
2
n
+ p
2

+ 2p
n
p

= m
2
n
+ E
n
E

_
1
p
n
p

E
n
E

_
m
2
n
+ m
n
E

_
1
[p
n
[E

cos
E
n
E

_
m
2
n
,
(3.100)
where is the angle formed by the vectors p
n
and p

and in the last step we have taken into account


that for relevant temperatures of order few MeV neutrons (as well as protons) are non-relativistic, i.e.,
E
n
m
n
[p
n
[. Furthermore,
p
n
p

= E
n
E

[p
n
[[p

[ cos = E
n
E

[p
n
[E

cos = E
n
E

_
1
[p
n
[
E
n
cos
_
m
n
E

. (3.101)
Note that the velocity of an electron is dened as
v
e
=
E
[p[
=

[p[
_
[p[
2
+ m
2
=
[p[
E
=
_
E
2
m
2
E
2
=
_
1
m
2
E
2
. (3.102)
Therefore, we obtain for the following factor from the cross-section
(p
p
p
e
)
2
m
2
p
m
2
e
= (E
p
E
e
[p
p
[[p
e
[ cos
pe
)
2
m
2
p
m
2
e
E
2
p
E
2
e
_
1
m
2
p
E
2
p
m
2
e
E
2
e
_
m
2
p
E
2
e
_
1
m
2
e
E
2
e
_
. .
v
2
e
.
(3.103)
With the approximations above we obtain for the matrix element
[/[
2
= 16(1 + 3g
2
A
)G
2
F
m
n
m
p
E

E
e
, (3.104)
and the cross-section (m

= 0)
d
e
d
=
1
4
2
(1 + 3g
2
A
)G
2
F
E
2
e
v
e
_
m
p
m
n
_
2
. (3.105)
We would like now to nd the change in the number of neutrons in a unit volume in a comoving
coordinates due to the reactions above. Let us write down the kinetic equation for neutron distribution
function
f
n
t
=

i=e,,p
_
d
3
p
i
(2)
3
2E
i
. .
phase space element
f

[1 f
e
][1 f
p
][/[
2
(2)
4
(p
n
+ p

p
e
p
p
). (3.106)
We can obtain a rate equation for neutron density by summing up over the neutron phase space on the
left and right hand sides of the kinetic equation:
n
n
t
=
_
d
3
p
n
(2)
3
2E
n
f
n
t
=

i=e,,p,n
_
d
3
p
i
(2)
3
2E
i
f

[1f
e
][1f
p
][/[
2
(2)
4
(p
n
+p

p
e
p
p
). (3.107)
3.2. NUCLEOSYNTHESIS 55
Note that now the product on the right-hand-side includes also neutrons. The energy conservation
in the non-relativistic approximation can be approximated as (
n
+


e

p
) = (

+ Q
e
).
We now substitute the cross section and use the non-relativistic approximation E
i
= m
i
for i = p, n,
which cancel similar factor in the matrix element. The integration over the phase space of neutrons and
protons on the right-hand-side gives n
n
and 1 n
p
and we assume n
p
1. Furthermore for electron
and neutrinos we have
_
d
3
p
i
(2)
3
2E
i
=
_
4[p
i
[
2
dE
i
(2)
3
2E
i
=
_
E
i
dE
i
4
2
[p
i
[
E
i
..
v
i
(3.108)
We thus obtain for the rate equation (v

= 1)
n
n
t
= 8n
n
(1 + 3g
2
A
)G
2
F
_
E
2
e
dE
e
4
2
E
2

dE

4
2
f

[1 f
e
](2)(

+ Q
e
)
= n
n
(1 + 3g
2
A
)G
2
F
_

Q
E
2
e
dE
e
2
3
(E Q)
2

v
e
f(E Q)[1 f(E
e
)] =
n
n
n
. (3.109)
The one-dimensional integral
n
can be computed numerically. A good approximation is provided by
the formula

n
1.63
_
T

Q
_
3
_
T

Q
+ 0.25
_
2
s
1
. (3.110)
Where T

is the neutrino temperature and before electron-positron annihilation it is equal to the tem-
perature of the ambient matter (electrons).
The rate of the reaction ne
+
p is calculated in the similar fashion and diers only in the
integration limits
n
n
t
= n
n
(1 + 3g
2
A
)G
2
F
_
m
e

E
2
e
dE
e
2
3
(E Q)
2

v
e
f(E Q)[1 f(E
e
)] =
ne
n
n
. (3.111)
For T

= T and T > m
e
the both results are numerically close
ne

n
. The rates of the inverse
reactions
p + e

n + , p + n +e
+
, (3.112)
can be related to the direct reaction through the so-called detailed balance equations which state that

pe
= exp
_

Q
T
_

n
,
p
= exp
_

Q
T
_

ne
. (3.113)
Now let us introduce the concentrations of neutrons and protons as
X
n
=
n
n
n
, X
p
=
n
p
n
, (3.114)
56 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
where n = n
n
+n
p
is the net density of baryons. The rate equation for neutron concentration becomes
d
dt
X
n
= (
n
+
ne
)X
n
+ (
pe
+
p
)X
p
= (
n
+
ne
)
_

_
X
n
exp
_

Q
T
_
(1 X
n
)
. .
X
p
_

_
= (
n
+
ne
)
_
X
n
_
1 + exp
_

Q
T
__
exp
_

Q
T
__
= (
n
+
ne
)
_
1 + exp
_

Q
T
__
_

_
X
n

1
_
1 + exp
_
Q
T
_
. .

X
n
_

_
= A(X
n


X
n
), (3.115)
where

X
n
must be the asymptotic (equilibrium) value of the neutron concentration, since when dX
n
/dt =
0 we get automatically X
n
=

X
n
. The formal solution of this dierential equation is
X
n
(t) =

X
n
(t)
_
t
0
I(t, )
d

X
n
d
d, (3.116)
where the integral is given by
I(t, ) =
_
t

A(y)dy. (3.117)
We now wish to estimate the freeze-out concentration. We have

X
n
0 when T 0. (3.118)
Therefore the concentration is given by the second term in (3.116) when the limit t is taken. Thus
the freeze-out density is given by
X
f
=
_

0
I(t, )
d

X
n
d
d, (3.119)
where we need the relation between time and temperature [see Eq. (3.76)]
t
sec
= 1.39
1

T
2
MeV
, =

2
30
_
g
b
+
7
8
g
f
_
3.537, (3.120)
where we inserted g
b
= 2 for bosons, g
f
= 10 for fermions present in the lepton era. The numerical
calculation shows that
X
f
= 0.158, 3.537. (3.121)
If there is an additional neutrino (with its anti-particle) then kappa + 0.58 and then X
f
= 0.163.
We can generalize this to
X
f
0.158 + 0.005(N

3), (3.122)
where N

is the number of neutrino species (equal to 3 in the standard model).


3.2. NUCLEOSYNTHESIS 57
Neutron decay
On time-scales that are much shorter than the life-time of a neutron
n
886 s we can neglect the
process of neutron decay
n p + e +
e
. (3.123)
After the freeze-out this is the only reaction that changes the number of neutrons. The decrease of the
neutron concentration is then given
X
n
(t) = X
f
exp
_

n
_
. (3.124)
The nucleosynthesis occurs at about t 250 sec, which is non-negligible compared to the neutron
life-time; therefore we need to take into account the neutron decay to obtain the number of neutrons
at the start of the nucleosynthesis.
3.2.1 Deuterium
Now we turn to the question on how the light nuclei are created. This is a key test of Big Bang
cosmological model. Our goal will be to show how to compute the elemental abundances and compare
them with their observed abundances in the current universe. Most of them are created in a two-step
process. The rst process is
p + n D + . (3.125)
where D stand for deuterium. We need now to consider a mixture of neutrons, protons and deuterium.
Let us start by giving some general formulae. Suppose we have only one type of nuclei i with mass
number A
i
and Z
i
. Their density is then given by
n
i
= g
i
(2m
i
T)
3/2
e

i
T
, (3.126)
where g
i
is the number of the spin state of the nucleus, m
i
is the mass of the nucleus and the chemical
potential is given by

i
= Z
i

p
+ (A
i
Z
i
)
n
. (3.127)
We can eliminate the nucleon chemical potentials by dening
n
i
n
Z
i
p
n
A
i
Z
i
n
=
g
i
2
A
i
A
3/2
i
(2m
N
T)
3(1A
i
)/2
e
B
i
/T
, (3.128)
where the binding energies B
i
are dened as
m
i
= Z
i
m
p
+ (A
i
Z
i
)m
n
B
i
. (3.129)
Dening the fractions
X
i
=
n
i
n
B
, X
n
=
n
n
n
B
, X
p
=
n
p
n
B
, (3.130)
where n
B
is the total number of baryons we can rewrite the last equation as
X
i
=
g
i
2
X
Z
i
p
X
A
i
Z
i
n
A
3/2
i

A
i
1
e
B
i
/T
. (3.131)
58 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
Here we have dened a dimensionless quantity
=
1
2
n
B
(2m
n
T)
3/2
= 2.96 10
11
_
a
10
10
a
0
_
3
_
T
10
10
K
_
3/2
, (3.132)
where we have used for the baryon number
n
B
=
3
B
H
2
0
8Gm
N
_
a
0
a
_
3
. (3.133)
Sometimes the baryon density is parameterized in terms of the baryon to photon ratio, which dened

10
10
10

n
N
n

(3.134)
and is related to the parameter
B
via the relation

B
h
2
6.53 10
3

10
. (3.135)
Using the key formula (3.131) with g
D
= 3 the deuterium abundance can be written as
X
D
= 5.7 10
14

10
T
3/2
MeV
exp
_
B
D
T
_
X
p
X
n
. (3.136)
where from Eq. (3.129) follows that B
D
= m
p
+m
n
m
D
2.23 MeV.
When the build-up of substantial amounts of deuterium possible? We can not simply apply the
formula for X
D
, because high energy photons with energies

> B
D
will destroy deuterium nuclei. An
estimate of the number of such photons follows from
n

> B
D
)
n
D

B
2
D
Te
B
D
/T
n
N
X
D
10
10
1

10
X
D
_
B
D
T
_
2
e
B
D
/T
. (3.137)
This number becomes of the order of one at T 0.06 MeV.
The reates of the other reactions that convert deuterium into somewhat heavier elements scales with
the density of deuterons. Therefore, up to the point where substantial number of deuterons has been
built up heavier elements can not be created. This picture persist until large amounts of deuterons
opens up the way for built up of heavier elements. As pointed out above this occurs at temperatures
below 0.06 MeV. This picture is called deuterium bottleneck.
3.3 Cosmological nucleosynthesis
formation of light elements (D,
3
He,
H
e,
7
Li) in the early universe
goal is to compute abundances of these elements and to compare this with observations
key-test for Big-Bang model
the main cosmological parameter one can obtain from this analysis is baryon to photon ratio
=
n
B
n

. (3.138)
From one can obtain
B
, the current density in baryons over the critical energy density
3.3. COSMOLOGICAL NUCLEOSYNTHESIS 59

B
=
n
B
m
P

cr
=
n
B
m
P
(3H
2
0
)/(8G)
= n

m
P
(3H
2
0
)/(8G)
, (3.139)
H
0
= h 100
km
s Mpc
, (3.140)

cr
=
3 (100
km
sMpc
) h
2
8 6.67 10
11
m
3
kgs
2
= 1.0 10
26
h
2
kg
m
3
, (3.141)
h 0.70 , (3.142)
n

= 2
1
(c)
3
_
d
3
p
(2)
3
1
e
p
1
= 2
1

2
(3)
(k
B
T)
3
(c)
3
= 4.08 10
8
m
3
, (3.143)
T = 2.7358 K, (3.144)
(3) =

k=1
1
k
3

= 1.202 , (3.145)

B
= n

m
P

cr
= 4.08 10
8

1.67 10
27
1.9 10
26
h
2
, (3.146)

B
h
2
= 3.6 10
7
. (3.147)
We know (at universe!):
B
1 . Estimate:
0.7
2
3.610
7

= 1.4 10
8
.
(Attention: Zeile in Vorlage nicht lesbar!!!)
Properties light elements:
A
Z Binding Energy g(2I+1) observed abundance
m
A
n
A
m
H
n
H
2
H = D 2.22 MeV 3 2.78
+0.44
0.38
10
5
3
He 2.22 MeV 2 (1.1 0.2) 10
5
4
He 2.22 MeV 1 0.2477 0.0029
6
Li 2.22 MeV 3 (Attention:??????)
7
Li 2.22 MeV 4 2.07
+0.15
0.04
10
10
(
3
H decays to
3
He via decay, no stable nuclei with mass 5 and 8,
7
Be EC
7
Li) .
To cook nuclei, T 0.01 10 MeV. At this point all nuclei are non-relativistic (m
N
T) . The
universe then consisted of , and e
+
, e

. Since
n
B
n

10
8
, and from exercise 4.1 we know radiation
dominates for t < 2 10
4
y, nuclei are just a tiny perturbation to the energy density of the universe at
time they were formed.
60 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
We can get T(t) in the era where nuclei were cooked from the , , e
+
, e

plasma.
Radiation dominates: = g

2
30
T
4
a

t
a
a
=
1
2t
_
a
a
_
2
=
1
4t
2
=
8G
3
(3.148)
(t) =
3
8G
1
4t
2
= g

2
30
t
4
(3.149)
T =
_
g
0
32
3
gG
_
1/4
1

t
=
_
_
_
_
g
0
32
3
2
..
only
6.67 10
11
_
_
_
_
(Attention :???) (3.150)
t 0.2 s neutrino decoupling
T 1 2 MeV proton neutron ratio freezes out
t 1 s e
+
e

annihilation
T 0.5 MeV
t 200 300 s nucleosynthesis
T 0.05 MeV
t 10
11
s matter domination starts
T eV
t 10
12
10
13
s 3 10
5
y recombination
So how to form nuclei. We start building them from protons and neutrons. From freeze-out we
know (see previous lecture) that X
n
=
N
n
N
n
+N
p
0.158.
First one has to make Deuterium (p +p +n +n
4
He has very low cross section (rare process), we
have to go via D!)
D can be formed in the following reaction:
p + n D + (3.151)
for t < 10
3
s, H. (no problem this reaction).
Let us try to compare the Deuterium abundance per weight:
X
D

2n
D
n
N
(3.152)
where n
N
is the total number of nucleons.
g
D
= 3 , B
D
= 2.23 MeV, g
n
= g
p
= 2 . (3.153)
3.3. COSMOLOGICAL NUCLEOSYNTHESIS 61
For
p + n D + (3.154)
to be in chemical equilibrium we have

p
+
n
=
D
. (3.155)
All particles are non-relativistic,
n
A
= g
A
_
d
3
p
(2)
3
1
e
(

p
2
+m
2
)
+ 1

= g
A
_
m
A
T
2
_
3/2
exp
_

A
m
A
T
_
(3.156)
where T is temperature of the universe.
X
D
=
2n
D
n
N
, (3.157)
X
D
=
n
p
n
N
, (3.158)
X
D
=
n
n
n
N
, (3.159)
B
D
= m
p
+ m
n
m
D
= 233 MeV, (3.160)
X
D
=
2n
D
n
p
n
n
X
P
X
n
n
n
= 2
g
0
g
p
g
n
_
m
D
m
p
m
n
_
3/2
_
2
T
_
3/2
exp (B
D
/T) n
N
. (3.161)
Now we can use that n
N
/n

is constant.
=
n
N
n

, (3.162)
n

=
2

3
(3) T
3
, (3.163)
n
N
= n

, (3.164)
X
D
= 2
3
4

_
m
D
m
p
m
n
_
3/2
_
2
T
_
3/2
exp B
D
/T n

(3.165)
m
N
= 938 MeV m
D
2 m
N
. (3.166)
X
D
= 2
3
4
2
3/2
(2)
3/2

2
(3)
1
m
3/2
N
T
3/2
exp(B
D
/T)
= 5.67 10
4
T
3/2
MeV
exp(B
D
/T
MeV
)X
p
X
n
= 5.67 10
4

10
T
3/2
MeV
exp(B
D
/T
MeV
)X
p
X
n
, (3.167)
where
10
= 10
10
.
At T = 1 MeV X
D
= 5.27 10
13

10
X
p
X
n
extremely small,
10
7
T = 0.1 MeV X
D
= 2.7 10
4

10
X
p
X
n
T = 0.08 MeV X
D
= 7.2 10
2

10
X
p
X
n
T = 0.07 MeV X
D
= 3.9
10
X
p
X
n
T = 0.06 MeV X
D
= 7.8 10
2

10
X
p
X
n
62 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
Hence between T = 0.06 MeV and T = 0.08 MeV D concentration should raise enormously. Why only
signicant production at these rather low temperatures, T B?
Many photons although majority has energy B, the tail has photons with energy B, tail only
suppressed at small T.
The rates of converting D into heavier elements are proportional to the D concentration and negligible
after T < 0.1 MeV. This is called the Deuterium battleneck. We have to wait for D to be formed
before heavier elements which have much higher B can be formed.
Now how to make the earlier elements:
D +D
3
He + n, (3.168)
D +D T + p , (3.169)
where T =
3
H (Tritium) . It turns out that these reactions become ecient when X
D
10
4
. Bottle-
neck opens up. Complicated set of reactions possible.
It is not easy to go further.
4
He +
4
He
8
?, no stable 8, also no stable 5 other battleneck.
Show abundances as a function of time with computer.
Show picture of nal abundances.
Some words about
4
He.
It turns out that when X
D
10
2
in a short time nearly all available neutrons end up in
4
He.
X
D
10
2
at t
N
= 269 (1 0.07(N

3) 0.06 ln(???)
10
) s = 238 s (3.170)
X4
He
= 2X
n
(t) = 2X

n
e
t
N
/
= 0.23 0.012(N

3) + 0.005 log
10
= 0.24 , (3.171)
where = 886 s and X

n
= 0.158 .
Not so sensitive to
10
.
Conclusions:
3.4. RECOMBINATION 63
Light element abundances can be explained.
Remarkable since abundances span many orders of magnitude. One can extract the value of :
4 10
10
7 10
10
(old values from (19??)), then 0.014
B
h
2
0.025,
h = 0.7 0.029
B
0.051 .
D/H ratio:
= 5.9 0.5 10
10
(???) .
B
h
2
= 0.0214 0.020 .
Independent measurement CMB uctuations, WMAP:

B
h
2
= 0.0224 0.0009,
B
= 0.044 0.004, = (6.1 0.3) 10
10
(Perfect agreement!).
Note that
B
0.044 <
M
0.3. So most of matter is not baryonic. What can it be? Dark
matter.
Slight problem with
7
Li in nucleosynthesis: predicted value 3 10
10
, observed value 2 10
10
.
3.4 Recombination
After the nucleosythesis the matter composition of the universe is the following
Thermal radiation (photons)
Electrons
Protons (hydrogen nuclei)
Fully ionized Helium nuclei (i.e. n + n + p + p).
Next the recombination occurs. The process is very fast and we call it recombination moment.
Physically this means that the electrons are captured on protons to form Hydrogen nuclei
e + p H (3.172)
and two electron are captures on alpha particle to form
4
He
e + e + p +p + n + n
4
He. (3.173)
The ionization energy of Helium is much larger than that of the Hydrogen, therefore these two events
dierent somewhat in time. First comes the Helium recombination then the Hydrogen. When all the
ionized atoms have capture the available electrons and have become electrically neutral the universe
becomes transparent to the radiation. This radiation carries the imprint of the early universe and exists
up to our days - it is called Cosmic Microwave Background (CMB). We will return to CMB and study
its features in future lectures.
64 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
3.4.1 Helium recombination
Let us consider the recombination of Helium. The recombination of Helium occurs in two steps.
Step one: capture of one electron on He nucleus leads to a singly ionized Helium atom
B
+
= m
e
+ m
2+
m
+
= 54.4 eV (3.174)
where m
2+
and m
+
are the masses of He
2+
and He
+
. This energy corresponds to 6.32 10
5
K. Let us
estimate the temperature at which this process occurs. We assume a chemical equilibrium
He
2+
+ e

He
+
+ . (3.175)
The chemical equilibrium implies then that the chemical potentials satisfy the following relation

2+
+
e
=
+
. (3.176)
Now we consider the ratio
n
2+
n
e
n
+
=
g
2+
g
e
g
+
_
Tm
e
2
_
3/2
exp
_

B
+
T
_
(3.177)
where the densities are given by the usual equation for non-relativistic particles (3.57). The ratio of
statistical weights is equal 1. The complete recombination of Helium reduces the number of electrons
only by 12%. Therefore, the number density of electrons before the Hydrogen recombination is
n
e
0.8n
N
2 10
11

10
T
3
. (3.178)
We substitute Eq. (3.178) into Eq. (3.177) to obtain
n
2+
n
+
exp
_
35.6 +
3
2
ln
_
B
+
T
_

_
B
+
T
_
ln
10
_
. (3.179)
This is a transcendental equation for temperature which can be used by interactions. It is clear that
when the exponent is positive then the number of complete ionized He nuclei is much larger than that
of singly ionized ones. Thus, for the temperature at which this ratio is of order of unity:
T
+
=
B
+
42 ln
10
15000K(1 + 0.023 ln
10
)K. (3.180)
This half of the nuclei and doubly ionized whereas the other half is singly ionized. Now consider the
case when the temperature falls slightly, i.e., T = T
+
T T
+
is a small number. Then we can
substitute in Eq. (3.179) T = T
+
T and expand to leading order in T, which gives
n
2+
n
+
exp
_

B
+
T
T
2
+
_

_
42
T
T
+
_
(3.181)
This equation tells us that, for example, if the temperature falls by 1/5 of T
+
(typically going from 15
000 K to 12 000 K) we get n
2+
/n
+
0.0001.
A further observation is that the temperature T
+
varies as a logarithm of the baryon density .
Therefore the larger the baryon density the earlier the recombination of He takes place (at a higher
temperature).
3.4. RECOMBINATION 65
Second step. Another electron is captured on a singly ionized He atom. In this case we have for the
binding energy
B = m
e
+ m
1+
m
He
= 24.62 eV (3.182)
Note that both the rst and second electron end up in the rst orbit. Because the binding ionization
energy (3.182) is lower than for the rst electron this second stage occurs at lower temperature, which
is typically 5000 K. At this temperature Hydrogen is still ionized and the universe is still opaque to
photons.
3.4.2 Hydrogen recombination
After the Helium recombination the key reaction that maintains the equilibrium between the radiation
and matter is this
p + e

H + . (3.183)
Here H is the neutral Hydrogen atom. In the ground state the binding energy is
B
H
= m
p
+ m
e
m
H
= 13.6 eV (3.184)
which translates into temperature T
H
= 158000 K. We can again derive the Saha equation, which now
reads
n
p
n
e
n
H
=
_
Tm
e
2
_
3/2
exp
_

B
H
T
_
(3.185)
where n
H
is the number density of the hydrogen atoms in the ground state; the ratio of the statistical
weights is again equal 1. In principle, higher states of atoms can be populated (these are denoted as
2
S,
2
P after the scheme
(2S+1)
L
J
, where L is the orbital momentum, J is the total angular momentum, and
S is the total spin. For the temperatures of interest T 5000 K, the relative concentration is given by
n
2P
n
H
=
g
2P
g
H
= exp
_

3
4
B
H
T
_
< 10
10
, (3.186)
and therefore is negligible. It is common to introduce the ionization factor
X
e
=
n
e
n
e
+ n
H
. (3.187)
Since the charge neutrality implies n
p
= n
e
and
n
e
+ n
H
0.75 10
10

10
n

= 3.1 10
8

10
_
T
T
0
_
3
cm
3
, (3.188)
we can rewrite (3.185) as
X
2
e
1 X
e
= exp
_
37.7 +
3
2
ln
_
B
H
T
_

B
H
T
ln
10
_
. (3.189)
The expression in the exponent vanishes at the temperature
T
rec

B
H
43.4 ln
10
3650(1 + 0.023 ln
10
)K. (3.190)
At this point the recombination of hydrogen begins; the fractional value of the electrons is X
e
= 0.6.
The recombination proceeds very rapidly. We obtain from Eq. (3.189) that a fractional change in the
temperature of the order 0.1 reduce the fraction X
e
10
4
. The description above gives only very crude
picture of the recombination. In fact, the ionization fraction does not decrease exponentially, rather it
freezes-out. Furthermore, the equilibrium description fails shortly after the recombination starts.
66 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
Figure 3.4: The leakage scheme leading to formation of neutral hydrogen.
Nonequilibrium hydrogen recombination
The simple picture where an electron is captured to the ground state
p + e

H + (3.191)
does not describe properly the recombination because the photon emitted in this process is very energetic
and ionizes immediately another atom. More eective are the cascade processes where
p + e

H(excited) + H(2P) + H(1S) + . (3.192)


During the last process emitted photon is called Lyman- photon It has energy and cross-section
3B
H
/4 = 117000 K,

10
17
10
16
. (3.193)
The Lyman- photons are reabsorbed on short-timescales
(

n
H
)
1
10
3
10
4
s. (3.194)
This timescale should be compared with the cosmological timescale. At the recombination we have
matter dominated universe. We take the energy of cold particles, whose critical density is

cr
= (6t
2
)
1
. (3.195)
3.4. RECOMBINATION 67
Using for the temperature T = T
0
(1 + z) we obtain
t
sec
= 2.75 10
17
(
m
h
2
75
)
1/2
_
T
0
T
_
3/2
10
13
s. (3.196)
Since
t
sec
(3.197)
the photons does not see the cosmological expansion and therefore are not redshifted. We can neglect
the redshift eects.
To give a qualitative kinetic approach to the problem the following assumptions are made
Keep only 1S, 2S, and 2P states of neutral hydrogen and neglect all the excited states.
electrons, protons and thermal photons are present
non-thermal L

photons are present


The rate of ionization obeys the following equation
dX
e
dt
=
dX
1S
dt
= W
2S
X
2S
, (3.198)
where X
e
= n
e
/n
T
, X
2S
= n
2S
/n
T
, where n
T
is the total number of neutral atoms and electrons.
Eq. (3.198) is a good approximation once about 50% of neutral hydrogen is formed until the end of
recombination.
We consider now the 2S reservoir and the quasi-equilibrium condition for it becomes
v
ep2S
n
e
n
p

2Sep
n
eq

n
2S
W
2S1S
n
2S
= 0, (3.199)
where n
eq

is the number density of thermal photons. The relation between the cross-section for the
direct and inverse reactions ep 2S is given by

2Sep
n
eq

v
ep2S
=
n
eq
e
n
eq
p
n
eq
2S
=
_
Tm
e
2
_
3/2
exp
_

B
H
4T
_
, (3.200)
where we used the Saha equation and the fact that the binding energy for the 2S state is B
H
/4. Then
we have
X
2S
=
_
W
2S
v
ep2S
+
_
Tm
e
2
_
3/2
exp
_

B
H
4T
_
_
1
n
T
X
2
e
(3.201)
and then we become from (3.198)
dX
e
dt
= W
2S
_
W
2S
v
ep2S
+
_
Tm
e
2
_
3/2
exp
_

B
H
4T
_
_
n
T
X
2
e
. (3.202)
We have now two cases
The rst term in the bracket is small, then we have a Saha equation for excited state of Hydro-
gen and electrons. At T > 2450 K the reaction 2S ep is ecient in maintaining chemical
equilibrium between the electrons, protons and hydrogen 2S states.
68 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
for temperatures T < 2450 K the thermal radiation is unimportant; the rate is determined by the
equilibrium between recombination and to the 2S level and two-photon decay.
In the latter case we obtain
dX
e
dt
v
ep2S
n
T
X
2
e
. (3.203)
The rate is entirely determined by the recombination to the 2S level and does not depend on the W
2S
(two photon decay from the 2S state). The recombination cross-section can be tted by
v
rec
8.7 10
14
_
B
H
T
_
0.8
cm
3
s
1
. (3.204)
It is convenient to rewrite the rate equation in terms of the redshift parameter
z =
T
T
0
1 (3.205)
instead of cosmological time. One nds that
dX
e
dz
0.1

10
_

m
h
2
75
_
0.72
_
z
14400
_
0.3
+ 10
4
z exp
_

14400
z
__
1
X
2
e
(3.206)
which can be integrated to nd
X
e
(z) 6.9 10
4
_

m
h
2
75

10
__
z/14400
dy
0.72y
0.3
+ 1.44 10
8
y exp(1/y)
_
1
. (3.207)
At T > 2450 K the rst term in the denominator can be neglected; then we obtain
X
e
(z) 1.4 10
9
_

m
h
2
75

10
z
1
exp
_

14400
z
_
. (3.208)
At low temperatures, asymptotically the integral goes to 0.27, therefore
X
f
e
2.5 10
3
_

m
h
2
75

10
1.6 10
5

b
h
75
(3.209)
Finally, let us nd out when the universe becomes transparent to radiation. This occurs when the photon
scattering time exceeds the cosmological time. The photon opacity is due to Thomson scattering (elastic
scattering of radiation (photon) by free charged particle). The scattering cross-section
T
6.6510
25
cm
2
and the cosmological time
t
sec
= 2.75 10
17
(
m
h
2
75
)
1/2
_
T
0
T
_
3/2
10
13
s. (3.210)
and the total number density
n
e
+ n
H
0.75 10
10

10
n

= 3.1 10
8

10
_
T
T
0
_
3
cm
3
. (3.211)
3.4. RECOMBINATION 69
Substituting all this in the standard equation
t
1

T
n
t
X
e
(3.212)
we obtain
X
dec
e
6 10
3
_

m
h
2
75

10
_
T
0
T
dec
_
3/2
. (3.213)
For T
dec
2500 K z 900 independent of the cosmological parameters. For
m
h
2
75
0.3 and
10
5
the ionization fraction is 2 10
2
.
70 CHAPTER 3. THERMAL EVOLUTION OF THE UNIVERSE
Chapter 4
Gravitational instabilities
The universe was homogeneous at the time of decoupling of radiation, as is evidenced by the isotropy
of the microwave background radiation. Afterwards structures developed. The simple explanation of
how structures developed in a homogeneous universe is based on the notion of gravitational instability.
The build up of instabilities is natural for the gravity because matter is attracted to high density
regions which amplies the existing inhomogeneities. It is convenient to start with the Newtonian
theory.
4.1 Newtonian theory
In this section we treat matter as a perfect uid, which can be characterized by the energy density
(x), where x = (x, t), entropy S(x, t) and three-velocity v(x, t). The prefect uid is characterized by
conservation laws of energy density
(x, t)
t
+((x, t)v) = 0, (4.1)
Euler equation
v
t
+ (v

)v +
p

+ = 0, (4.2)
and entropy conservation
S(x, t)
t
+(S(x, t)v) = 0. (4.3)
This latter equation neglects the dissipation in matter. In the Newtonian theory these equations should
be supplemented by the Poisson equation
= 4G, (4.4)
and nally the equation of state
p = p(, S). (4.5)
These equations are sucient to determine the unknowns, which are
(x, t), S(x, t), v(x, t), (x, t) p. (4.6)
71
72 CHAPTER 4. GRAVITATIONAL INSTABILITIES
4.1.1 Instabilities in the static universe (Jeans instability)
We consider rst a static universe assuming that it is homogeneous and the energy density is constant
in time
(t, x) = const. (4.7)
We have seen in our exercise that this assumption is inconsistent with the Newtonian gravity and perfect
hydrodynamics. This problem is avoided however in the Einstein universe where the gravitational force
is compensated by oppositely directed cosmological constant. Let us look at Small perturbations
(x, t) =
0
+(x, t), (4.8)
v(x, t) = v
0
+v(x, t), (4.9)
(x, t) =
0
+ (x, t), (4.10)
S(x, t) = S
0
+ S(x, t). (4.11)
where all perturbations are much small than the background values, i.e.,
0
, etc. The pressure
can be written as
p(x, t) = p(
0
+ , S
0
+S) = p
0
+ p(x, t). (4.12)
where we can write
p = c
2
S
+ s (4.13)
where c
2
S
= (p/)
S
, with c
S
being the speed of sound, and = (p/S)

.
Perturbation equations are

t
+
0
(v) = 0, (4.14)
v
t
+
c
2
S

0
() +

0
(S) + = 0, (4.15)
S
t
= 0, (4.16)
= 4G (4.17)
The solution of Eq. () is independent of time
S(x, t) = S(x). (4.18)
We apply the operation to Eq. (4.15) and use the Poisson equation to eliminate the function
to obtain a second order dierential equation for the energy density

t
2
c
2
S
4G
0
= S(x). (4.19)
Adiabatic solutions
Adiabatic processes proceed at constant entropy S = 0. Then the solutions of Eq. () can be obtained
by inserting the Fourier transform
(x, t) =
_

k
(t)e
i

kx
d
3
k
(2)
3/2
. (4.20)
4.1. NEWTONIAN THEORY 73
Then we obtain

k
+ (k
2
c
2
S
4G
0
)

k
= 0. (4.21)
This is the equation of harmonic oscillator and has the standard solution

k
exp(i(k)t). (4.22)
where
(k) =
_
k
2
c
2
S
4G
0
. (4.23)
Dene the Jeans length from the condition (k) = 0

J
=
2
k
J
= c
S
_

G
0
_
1/2
(4.24)
We have now two cases: when <
J
the contribution from gravity (the term 4G
0
) only changes
the eective frequency of the sound modes. The solutions in this case are thus sound modes with the
eigenvector
sin(t +

k x + ) (4.25)
which propagate with the phase velocity
c

=

k
= c
S
_
1
k
2
J
k
2
. (4.26)
In the limit k k
J
the gravity is much smaller than the pressure, therefore, c

c
S
. In the opposite
limit >
J
the solutions are not oscillatory, i.e., we have

k
exp ([[t) . (4.27)
The plus sign describes exponentially fast growing mode while the minus sign describes exponential
decaying perturbations. If we look at very large scales, i.e., very small wave-vectors, we can write
[[t t/t
GR
, where t
GR
= 1/

4G
0
can be interpreted as a characteristic collapse time for a region
with initial density
0
. The Jeans length characterizes the scales at which the gravity and pressure
perturbations are coupled via the sound mode propagation.
4.1.2 Vector perturbations
Eq. (4.19) has trivial solutions = 0 and S = 0, but the complete system of hydrodynamical equations
has still a solution which give
v = 0,
v
t
= 0. (4.28)
The second equation requires only that the function be independent of time and could in principle be
an arbitrary function of coordinate. The rst equation tells us that the plane wave perturbations
v =

A

k
exp(i

k x). (4.29)
are transverse (i.e. are perpendicular to the wave vector k)

k = 0. (4.30)
(Recall that this is the case for the electromagnetic waves). The vector perturbations describe the shear
motions of the media which do not disturb the energy density. As in the case of electromagnetism we
can have two independent elds that satisfy the condition (4.30).
74 CHAPTER 4. GRAVITATIONAL INSTABILITIES
4.1.3 Entropy perturbations
Now consider the case where S ,= 0, in which case

k
+ (k
2
c
2
S
4G
0
)

k
= k
2
S

k
. (4.31)
This is an inhomogeneous equation whose solution is the sum of the general (homogeneous) solution
and some particular solution. The particular solutions is given by

k
=
k
2
S

k
k
2
c
2
S
4G
0
. (4.32)
It is easy to check that it satises (4.32) by direct substitution. The perturbations (4.32) are called
entropy perturbations. It is interesting that in the limit when k is large we have
lim
k

k
=
S

k
c
2
S
, (4.33)
i.e., the gravitational term does not contribute. That is the pressure perturbations are exactly compen-
sated by entropy perturbations so that p
k
= c
2
S

k
+ S
k
= 0.
The complete set of modes thus includes two adiabatic perturbations, two vector perturbations and
one entropy perturbation. The phenomenologically most important mode is the exponentially growing
one, which is responsible for the structure formation in the Universe.
4.1.4 Instabilities in the expanding universe
So far we have considered a universe which is static. Now we reconsider the same problem of gravitational
instability by taking into account the expansion of the Universe. The background energy density and
background velocities are given by
=
0
(t),

V =

V
0
= H(t)x. (4.34)
In the Newtonian theory we have the following equations

0
+ 3H
0
= 0, (4.35)

H + H
2
=
4G
3

0
. (4.36)
The rst one the energy conservation equation, the second one is the Friedman equation. Consider
again small perturbations that are given by
(x, t) =
0
+ (x, t), (4.37)
v(x, t) =

V
0
+ v(x, t), (4.38)
(x, t) =
0
+ (x, t), (4.39)
p(x, t) = p
0
+ p(x, t) = p
0
+ c
2
S
(x, t). (4.40)
Here we neglect the entropy perturbations. The perturbation equations are

t
+
0
(v) +(

V
0
) = 0, (4.41)
v
t
+ (

V
0


)v + (v

)V
0
c
2
S

0
() + = 0, (4.42)
= 4G. (4.43)
4.1. NEWTONIAN THEORY 75
Obviously we cannot proceed as before because

V
0
depends on the coordinates. So far we considered
only Eulerian perturbations (denoted by ) which are perturbations with respect to a x reference
frame. The method of Fourier transform does not help here to solve the equations. It is much more
convenient to work with Lagrangian perturbations which are dened with respect to the moving uid.
Let us see how this works. Denote by q the Lagrangian coordinate which we relate to the Eulerian
coordinate by
x = a(t)q, (4.44)
where a(t) is the scale factor. Now we need to be careful in computing the partial derivatives with
respect to time in the Eulerian and Lagrangian coordinate frames. Consider the time derivative of a
general function f(x = aq, t)
f
t

q
=
f
t

x
+ a q
f
x

t
=
f
t

x
+
a
a
(aq)
f
x

t
(4.45)

x
=

t

q
(V
0

x
). (4.46)
The coordinate derivatives are related by

x
= a
1

q
(4.47)
Now we rewrite the equations (4.41) -(4.43) in the Lagrangian coordinates and introduce the normalized
energy density /
0

t
+ a
1

q
v = 0, (4.48)

t
v +Hv +
c
2
S
a
+
1
a

q
= 0 (4.49)
= 4Ga
2

0
. (4.50)
Now we carry out the same steps as in the case of static universe; take the divergence of the equation
(4.49) and use the Poisson equation to obtain
+ 2H
c
2
S
a
2
4G
0
= 0. (4.51)
The latter equation describes the gravitational instability in the expanding universe.
Adiabatic perturbations
Next take a Fourier transform with respect to the comoving coordinate q

k
+ 2H
k
+
_
a
2
c
2
S
k
2
4G
0
_

k
= 0. (4.52)
where we have written
(t, q) =
_

k
(t) exp(ik q)
d
3
k
(2)
3/2
(4.53)
76 CHAPTER 4. GRAVITATIONAL INSTABILITIES
Now we introduce the Jeans length as

J
=
2a
k
J
= c
S
_

G
0
(4.54)
where

J
is the physical scale, which is related to the comoving wavelength via = 2/k and

= a.
Assuming at, mater-dominated universe
0
= (6Gt
2
)
1
, therefore

J
c
S
t. (4.55)
Jeans length is of the order of sound horizon. Now consider the two cases of being much smaller and
much larger than the Jeans length scale.
For the case
J
the solution is given by

k

1

c
S
a
exp
_
k
_
c
S
dt
a
_
. (4.56)
For the case
J
gravity dominates and one can neglect the k-dependent pressure term (c
2
s
k
2
= 0).
The solution is then simply = H(t), which corresponds to decaying solution of the perturbation in
matter dominated universe of arbitrary curvature.
The most general solution in the long-wave limit is then given by (we accept this result without
derivation)
= C
1
H
_
dt
a
2
H
2
+ C
2
H. (4.57)
If we assume matter-dominated universe, then a t
2/3
and H t
1
. Then
= C
1
t
2/3
+ C
2
t
1
. (4.58)
Thus in an expanding universe the gravitational instability is much less ecient, i.e., it increases as a
power law (as opposed to the static universe, where it was exponential). This implies that the initial
perturbation at the very early stages of the universe must have been quite large. This condition is
satised by the inationary scenario.
Vector perturbations
If we set = 0 in Eqs. (4.48)-(4.50) then we nd that
v = 0,
v
t
+ Hv = 0. (4.59)
If we assume that v
k
(t) exp(ik q) then from the rst Eq. (4.59) it follows that k v
k
(t) = 0. Upon
Fourier transformation, the second equation becomes
v
k
+
a
a
v
k
= 0. (4.60)
The solution of this equation is of the form
v
k

1
a
. (4.61)
We conclude that vector perturbations decay as the universe expands. Their current amplitudes must
be very small; in fact, if we assume that their amplitudes are rather large in the universe, this would
destruct the initial homogeneity of the universe. Vector perturbations can be important in explaining
the rotation of galaxies, i.e., after the stage when inhomogeneities are formed in the universe.
4.1. NEWTONIAN THEORY 77
4.1.5 Self-similar solutions
For large scale perturbations we can neglect the pressure, therefore the spatial derivatives can be dropped
in Eq. (4.51). The solutions can be written directly in the coordinate space as
(q, t) = A(q)
i
(t) +B(q)
d
(t). (4.62)
where the indices i and d refer to the increasing and decaying modes. The initial conditions are chosen
such that at the initial time t
0
we have
i
(t
0
) =
d
(t
0
) = 1. We assume that the energy distribution
at the initial moment is described by (q, t
0
) and the matter is at rest with respect to the Hubble ow
v = 0) initially. Then, using the initial conditions, we obtain
(q, t) = (q, t
0
)
_

i
(t)
1 (
i
/
d
)
t
0
+

d
(t)
1 (
d
/
u
)
t
0
_
, (4.63)
where we have eliminated the constants A and B in favor of the initial value of perturbation (q, t
0
) and
its rst derivative. Such perturbations preserve their initial shape and are called self-similar. This is
the case at late times (t t
0
), where we can omit the mode corresponding to the decay (i.e., B(q)).
4.1.6 Eects of dark matter on the instabilities
Here we wish to study the eect of dark energy component on the development of instabilities in the
universe within the Newtonian theory. We are interested in the case where the we have a homogeneous
relativistic energy component. This comes in two possible versions: radiation, with the equation of
state w = 1/3 and dark matter with w < 1/3.
The dynamical equations for instabilities
+ 2H
c
2
S
a
2
4G
0
= 0 (4.64)
contain the same perturbations for the cold component, but the Hubble constant now is determined
by energy density which contains components listed above as well as cold matter; the energy density is
given by

tot
=

eq
2
_
_
a
eq
a
_
3
+
_
a
eq
a
_
3(1+w)
_
, (4.65)
where a
eq
is the scale factor at the point where the energy densities of the relativistic and cold matter
components are equal. The value of the Hubble constant is given by that for the at universe
H
2
=
8G
3

tot
. (4.66)
We wish now to solve Eq. (4.64). Instead of time we introduce a variable x = a/a
eq
. We replace
0
by
the energy density of the cold matter

d
=

eq
2
_
a
eq
a
_
3
(4.67)
We also use Eq. (4.66) to write the Hubble parameter in terms of the variable x. Then Eq. (4.64)
becomes
x
2
(1 +x
3w
)
d
2

dx
2
+
3
2
x
_
1 + (1 w)x
3w
_
d
dx

3
2
= 0. (4.68)
where we dropped the term c
S
is negligible (it is determined by the cold component alone).
78 CHAPTER 4. GRAVITATIONAL INSTABILITIES
There are several special cases when we can solve (4.68) analytically The rst case is that of cosmo-
logical constant: w = 1. It can be veried that a special solution is this

1
(x) =

1 + x
3
(4.69)
by a direct substitution. The most general solution for w = 1 is given by
(x) = C
1

1 + x
3
+ C
2

1 + x
3
_
x
0
(
y
1 + y
3
)
3/2
dy. (4.70)
where Cs are constants of integration. At early times when cold matter dominates (x 1) the
perturbations grow as
(x) = C
1
x
3/2
+
2
5
C
2
x + O(x
3/2
), (4.71)
which is in agreement with the previous result when the cold matter dominates. In the later times when
the cosmological constant dominates and in the limit x 1 we obtain
(x) = (C
1
+ IC
2
)
1
2
C
2
x
2
+ O(x
3
) (4.72)
where
I =
_

0
_
y
1 + y
3
_
3/2
0.57. (4.73)
Thus we conclude that in the presence of cosmological constant the perturbation stop to grow and
the amplitude of the perturbations is frozen at the late times. This is not surprising because the
cosmological constant acts as anti-gravity.
Now we consider the case where the homogeneous relativistic background is given by the radiation.
This corresponds to setting w = 1/3 in previous equations. In this case the solutions are as follows. A
special solution is

1
(x) = 1 +
3
2
x. (4.74)
The most general solution can be checked to be
(x) = C
1
_
1 +
3
2
x
_
+ C
2
__
1 +
3
2
x
_
ln

1 +x + 1

1 + x 1
3

1 + x
_
. (4.75)
Now consider the limiting behaviour. In the limit x 1 (early times) we have the asymptotics
(x) = (C
1
3C
2
) C
2
ln
x
4
+O(x), (4.76)
i.e., the perturbations grow very slowly (logarithmically). Thus the radiation suppresses the growth of
instabilities in cold matter. In the limit x 1, i.e., when matter takes over the radiation we have the
asymptotics
(x) = C
1
(1 +
3
2
x) +
4
15
C
2
x
3/2
+ O(x
5/2
) (4.77)
Thus the perturbations grow proportional to the scale factor. It is clear that in the initial radiation
era the structures can not be formed, therefore one needs quite early transition to the cold matter
dominated epoch to get the perturbations grow. This implies that there should be a minimal amount
of cold matter so that it can dominate over the radiation.
4.1. NEWTONIAN THEORY 79
4.1.7 Nonlinear perturbations
Linear perturbations are stretched by the Hubble ow. Their energy density in the linear regime is
given by
=
0
(1 + + O(
2
)) (4.78)
and it does not dier much from the background value
0
. However, when the perturbations are of
order of unity (non-linear regime) the perturbation expansion does obviously break down.
What happens to the density perturbations when they reach the non-linear regime? The answer
is that the gravitational perturbation wins over the Hubble expansion, so that the dynamics is now
determined by the local gravitational forces; the increase in the energy density leads nally to a collapse
of the region to a stable non-linear structure.
It is a dicult numerical problem to nd exact solutions in the non-linear regime. It is useful to
consider two limiting cases
spherically symmetric perturbations; the solutions are called Tolman solutions
one-dimensional perturbations: the solutions are called Zeldovich solution
Obviously, the general case interpolated between these two limit and idealized cases. Nevertheless it is
useful to understand how the non-linear regime develops in these cases.
Hydrodynamics equations in the non-linear regime
We rewrite the continuity equation in the following form
_

t
[
x
+ V
i

i
_
+
i
V
i
= 0. (4.79)
where
i
/x
i
. Let us write the Euler equation in new notations and assume that the pressure is
equal zero
_

t
+ (V
j

j
)
_
V
i
+
i
= 0. (4.80)
Acting on this equation by a operator and using in the last term the Poisson equation we obtain
_

t
+ (V
j

j
)
_

i
V
i
+ (
i
V
j
)(
j
V
i
) + 4G = 0. (4.81)
We go over from the Eulerian to the Lagrangian coordinates to take into account the expansion of the
uid. As before we can write
x = x(q, t). (4.82)
These comoving coordinates can be used as long as the uid trajectories do not cross. By denition
V
i
=
dx
i
dt
=
dx
i
(q, t)
dt
(4.83)
Now we need to compute in these coordinates the quantity

j
V
i
=
q
k
x
j

q
k
_
x
i
(q, t)
t
_
=
q
k
x
j

t
_
x
i
(q, t)
q
k
_
. .
J
i
k
. (4.84)
80 CHAPTER 4. GRAVITATIONAL INSTABILITIES
The quantity J
i
k
is called the strain tensor. The time derivatives are related in the Eulerian and
Lagrangian description by the following relation
_

t
+ (V
j

j
)
_
=

t

x
+
x
i
(q, t)
t

x
i
=

t

q
. (4.85)
We have established the time and spatial derivatives entering the hydrodynamics equations; substituting
these into Eqs. (4.79) and (4.81) we obtain

t
+
q
k
x
i
J
k
i
t
= 0, (4.86)
and

t
_
q
k
x
i
J
i
k
t
_
+
_
q
k
x
i
J
i
k
t
__
q
l
x
j
J
i
l
t
_
+ 4G = 0. (4.87)
It is useful to dene a 3 3 matrix
J = [[J
i
k
[[. (4.88)
Now we note that
q
k
x
j
x
i
q
k
=
x
i
x
j
=
i
j
. (4.89)
This identity implies that the q
k
/x
j
J
1
, i.e., those coecients are elements of the inverse matrix.
In the matrix notations we can now write
+ Tr(

J J
1
) = 0, (4.90)
and
_
Tr(

J J
1
)
_

+ Tr
_
(

J J
1
)
2
_
+ 4G = 0. (4.91)
where the prime (as well as dot) denotes here the time-derivative for Lagrangian coordinates. Use the
identity
Tr(

J J
1
) = (ln Det J)

. (4.92)
which we substitute in Eq. (4.90) to obtain
+ (ln Det J)

= 0,

+ (ln Det J)

= 0. (4.93)
The variables separate, and we can trivially integrate it to obtain
ln + (ln Det J) C = 0, ln(Det J) = C, =
C
Det J
. (4.94)
Substitute the last two equations in (4.91) we rewrite it
[(ln Det J)

+ Tr
_
(

J J
1
)
2
_
+
4GC
Det J
= 0. (4.95)
This is our new perturbation equation. The general solution of this equation is a complicated numerical
problem. We will nd the solution of this equation in some special cases.
4.1. NEWTONIAN THEORY 81
4.1.8 Tolman solution
The Tolman solution refers to the spherically symmetrical solutions of Eq. (4.95). In the spherically
symmetrical case the matrix J
i
k
is diagonal, i.e., we can write
J
i
k
= a(R, t)
i
k
. (4.96)
Recall that the Eulerian and Lagrangian coordinates are related by the relation x = a(t)q and that is
why the scale factor enters this equation. Now we can compute
Det J = Det
_
_
a 0 0
0 a 0
0 0 a
_
_
= a
3
. (4.97)
and

J J
1
=
_
_
a 0 0
0 a 0
0 0 a
_
_
_
_
1/a 0 0
0 1/a 0
0 0 1/a
_
_
, (4.98)
and, therefore,
Tr(J J
1
)
2
= 3
_
a
a
_
2
. (4.99)
Thus Eq. (4.95) reduces to
a(R, t) =
4G
0
(R)
3a
2
(R, t)
, (4.100)
which after integration (multiply this equation by a to complete the integrand)
a
2
(R, t)
8G
0
(R)
3a(R, t)
= F(R). (4.101)
where F(R) is an integration constant. This would be simply the Friedman equation, if there was no
dependence on R of the functions a,
0
and F.
The solutions to Eq. (4.101) depend on the sign of the function F. If we assume negative F then
a(R, ) =
4G
0
3[F[
(1 cos ), (4.102)
t(R, ) =
4G
0
3[F[
3/2
( sin ) + t
0
, (4.103)
and when F is positive
a(R, ) =
4G
0
3[F[
(cosh 1), (4.104)
t(R, ) =
4G
0
3[F[
3/2
(sinh ) +t
0
. (4.105)
Here t
0
is an integration constant. We can put it to zero, if we assume that the initial singularity occurs
at the same moment of physical time t = 0 everywhere in space (at that point the scale factor a 0.)
Now we specify the general discussion to the most realistic case of at, matter-dominated universe.
Consider a perturbation of density which has as spherical form. On the spatial innite the matter
82 CHAPTER 4. GRAVITATIONAL INSTABILITIES
density is constant, i.e.,
0
(R ) =

= const. Because we assume at universe, F 0 as R


(this you can see by comparing Eq. (4.101) with the corresponding Friedman equation, which tells us
that the factor F plays the role curvature term.) We take the limit F 0 in Eq. (4.101); since the
variables separate, it can be easily integrated, and we nd
a(R , t) = (6G

)
1/3
t
2/3
. (4.106)
The energy density is then given by
(R , t) =

a
3
=
1
6Gt
2
. (4.107)
So far we have obtained the solutions at innity. Let us now characterize the evolution of the density
perturbation.
The sign of F is negative in the region where we have density perturbation. So we can use the
solutions given by Eq. (4.102). We see that because of the cos term the scale factor is oscillating; it is
maximal when = , i.e., cos = 1 and has the value
a
max
=
8G
0
3[F[
. (4.108)
The value of the physical time we then obtain on substituting = in Eq. (4.103):
t
max
=
4
2
G
0
3[F[
3/2
. (4.109)
The energy density at this point is

max
=

0
(R)
a
3
m
(R)
=
27[F[
3
(8G)
3

2
0
=
3
32Gt
2
max
. (4.110)
This should be compared with background value, which is given by Eq. (4.107) with t = t
max

max
(R )
=
9
2
16
5.6. (4.111)
What happens when this ratio exceeds the maximal value given above? The matter then decouples
from the Hubble ow and collapses. Note that since the Tolman solution admits a = 0 solutions, which
appear when = 2, physically speaking t = 2t
max
. Indeed
t

= 2t
max
=
8
2
G
0
3[F[
3/2
, = 2. (4.112)
In this case the energy density a
3
diverges (i.e., becomes innite). This unphysical behavior of the
Tolman solution is related to the idealized assumption that the perturbation is spherically symmetric.
We will now study another extreme limit of one-dimensional perturbation. The physical reality
corresponds to perturbations that are neither spherical nor one-dimensional, but these two extreme are
useful to get feeling of the behavior of more complicated solutions.
4.1. NEWTONIAN THEORY 83
4.1.9 Zeldovich solution
Zeldovochs solution describes the non-linear evolution of one-dimensional perturbation on the back-
ground of three-dimensional Hubble ow. The Eulerian and Lagrangian coordinates are related by
x
i
= a(t)
_
q
i
f
i
(q
j
, t)

. (4.113)
We implement this is the perturbation equation; for that let us rst dene the matrix J
i
k
= x
i
/q
k
.
Assume that the one-dimensional perturbation is along the x
1
-axis. The matrix J becomes
J = a(t)
_
_
1 (q
1
, t) 0 0
0 1 0
0 0 1
_
_
. (4.114)
where (q
1
, t) f
1
/q
1
. If one ignores the vector perturbations, then f
i
= /q
i
, where is the
potential that induces the peculiar velocities. The other elements in the perturbation equation are given
by
Det J = a
3
(1 ), Tr
_
_

J J
1
_
2
_
=
_
H

1
_
2
+ 2H
2
. (4.115)
We substitute these elements in the general perturbation equation
[(ln Det J)

+ Tr
_
(

J J
1
)
2
_
+
4GC
Det J
= 0. (4.116)
and assume that C(q) = const; then we obtain

H +H
2
=
4G
3

0
, (4.117)

+ 2H

4G
0
= 0, (4.118)
where
0
(t) = C/a
3
. The rst equation is the Friedman equation for the background. According to
Eq. (4.94)
(q, t) =
C
Det J
=

0
(t)
1 (q
1
, t)
. (4.119)
We can decompose then perturbations into decaying and increasing modes
(q
1
, t) = (q
1
)
i
(t) +(q
1
)
d
(t), (4.120)
where
i
(t) and
d
(t) are the growing an decaying modes of linearized theory. For at matter dominated
universe

i
t
2/3

d
t
1
. (4.121)
If 1 we can expand the exact solution (4.120). The decaying mode become quickly negligible, so
that we obtain
(q, t) =

0
(t)
1 (q
1
)
i
(t)
. (4.122)
For positive the energy density is larger than the average one
0
. However the total energy density is
decaying because of the Hubble expansion. The perturbation drops out the expansion in the non-linear
regime, where (q
1
)
i
(t) 1. This happens when (q, t) = 0, therefore from (4.122)
(q, t) = 0
(q, t)

0
(t)
= 1 + 3
H
[ln
i
]

. (4.123)
84 CHAPTER 4. GRAVITATIONAL INSTABILITIES
In a at, matter-dominated universe
i
a(t) then
[ln
i
]

= [ln a(t)]

=
a(t)
a(t)
= H(t). (4.124)
The second term in Eq. (4.123) is then of the order of 3; once the ratio (q, t)/
0
(t) > 4 the region starts
to collapse. This can be compared with Eq. (4.111), where this number was 5.6. The one-dimensional
collapse then leads to a two dimensional structures, which are sometimes called Zeldovich pancakes.
Inversely, when (q
1
) is negative the energy density decreases and the region becomes empty.
In general the perturbations are neither spherically symmetrical nor one-dimensional. A possible
generalization of the one-dimensional solution is given by the following function
(q, t) =

0
(t)
(1
i
(t))(1
i
(t))(1
i
(t))
. (4.125)
where , and are the deformation parameters along the three principle axis of the strain tensor. In
this case the strain tensor can be written as
J = aI a
i
_
_
0 0
0 0
0 0
_
_
, (4.126)
where I is a unit tensor, where I satises Eq. (4.95). If we substitute this solution into (4.95) we obtain
(q, t) =

0
(t)1 [( + + )
2
i
2
3
i
]
(1
i
(t))(1
i
(t))(1
i
(t))
. (4.127)
This is close to the postulated solution Eq. (4.125) when the second term in the numerator is small.
The Zeldovich approximation is unreliable in the non-linear regime and can not reproduce the very
basic properties of non-linear collapse.
Large-scale structure of the universe
We have seen that the strain tensor completely characterizes the inhomogeneities in the universe. It
can be parameterized by three scalar numbers , and . In any local regions the following possibilities
are possible (based on the results obtained above)
, the collapse is one dimensional and produces pancakes
the collapse is two dimensional and produces one dimensional laments
nearly spherical collapse
We can use the analogy with the mountains and planes to envision the uctuations in the density
superimposed on the homogeneous density background. The simulations and observations show that
the cosmic web consists of nearly spherical dense regions, with spherical and/or elliptical objects (case
3 in the list above).
The peaks are connected by side-ridges where one curvature scale is much smaller than the others
(case 2). The collapse is is two-dimensional and produces laments.
Thus, we have the picture where 3-dimensional spherically symmetrical dense regions are connected
by laments.
4.1. NEWTONIAN THEORY 85
In the regions when only one scale is large (case 1) (here we do not have a two-dimensional analog
in terms of mountains and valleys) the collapse leads to formation of domain walls. The walls are
connected to the laments.
The regions where the density is lower than the average will be expanding forever, and will be emptied
from matter. Smaller scales become parts of larger inhomogeneities and eventually upon equilibration
(virialization) disappear. Thus, the formation of structures is a transition from one homogeneous state
to another equilibrated state.
Presently we observe laments and pancakes on scales which recently entered the non-linear regime.
The density in these regions is about only factor 4 larger than the average density of the universe. The
spherical inhomogeneities are connected by laments.
If we make a distinction between the luminous and dark matter one can say that the baryons fall in
the gravitational wells formed by the cold dark matter to form galaxies. Such spherically wells contain
galaxies or clusters of galaxies of the size of several Mpc. Some galaxies could be found in the laments
with scales of the order of 10-30 Mpc, while others in the pancakes.
86 CHAPTER 4. GRAVITATIONAL INSTABILITIES
Figure 4.1: Numerical simulations of the cosmic web - structure formation in the universe
Figure 4.2: Photograph taken by the Hubble space telescope in the infrared. Each blob is a galaxy.
Chapter 5
Instabilities in General Theory of
Relativity
In this chapter we extend the discussion of the gravitational instabilities studied in the Newtonian
theory in the previous chapter to General Relativistic description. The key dierence in the physical
description is that the metric of space-time is not static, rather the matter motions and gravity are
coupled. So, apart from the usual hydrodynamics perturbations we need to look for metric perturbations
in a self-consistent manner.
5.1 Types of perturbations
Consider Friedman universe which is at. The metric of such universe, which includes small perturba-
tions, can be written as
ds
2
=
_
(0)
g

+ g

(x

dx

dx

. (5.1)
where the perturbation is much smaller than the background metric [g

[ [
(0)
g

[. The background
metric can be written as
(0)
g

dx

dx

= a
2
()(d
2

ij
dx
i
dx
j
). (5.2)
The metric perturbations can be of three types scalar, vector, and tensor. This reects the group
properties of the background metric. The scalar perturbations can be written
g
00
= 2a
2
. (5.3)
The spatial component is then written as a gradient of a scalar plus a vector (which should be divergence
free)
g
0i
= a
2
(B
,i
+ S
i
). (5.4)
In the following we will use B
,i
to denote the derivative B/x
i
. As pointed out the vector S
i
satises
S
i
,i
= 0 (i.e. it is transverse and has only two independent components). The most general form of
tensor perturbation is given by
g
ij
= a
2
(2
ij
+ 2E
,ij
+ F
i,j
+ F
j,i
+h
ij
), (5.5)
where and E are scalar functions, the vector F
i
satises the condition F
i
,i
= 0 and the tensor h
ij
satises the conditions
h
i
i
= 0, h
i
j,i
= 0, (5.6)
87
88 CHAPTER 5. INSTABILITIES IN GENERAL THEORY OF RELATIVITY
i.e. this tensor is traceless and transverse.
Let us now classify the perturbations
Scalar perturbations are described by fours scalar functions , , B and E.
Vector perturbations are described by two vectors S
i
and F
i
. These are related to rotational
motions, decay quickly and are not important for cosmology.
Tensor perturbations are purely general relativistic and describe gravitational waves. In the linear
order gravitational waves do not induce any perturbations in the perfect uid.
These perturbations are decoupled from each other and can be studied separately.
5.2 Gauge transformations and gauge invariance
We consider coordinate transformation
x

= x

. (5.7)
where

is innitisemally small change. Under such a transformation the metric changes according to
g

( x

) =
x

(x

) =
(0)
g

(x

) + g

(0)
g

(0)
g

,
. (5.8)
The rst relation follows directly from the fact that the interval ds
2
= g

dx

dx

is an invariant
quantity. This expansion keeps only terms that are linear in and g. At the new point x

we can
also split the metric into background and perturbation part
g

( x

) =
(0)
g

( x

) + g

(5.9)
where the unperturbed part is given by the Friedman metric (5.2), with the coordinate being now x
instead of x. Next we note that we can also expand the equilibrium part of the metric
(0)
g

(x

)
(0)
g

( x

) +
(0)
g
,

. (5.10)
Now compare Eqs. (5.8) and (5.9). If we take into account (5.10) we can deduce the following gauge
transformation law
g

= g

(0)
g
,

(0)
g
,

(0)
g

,
. (5.11)
Next we consider a 4-scalar function which is slightly perturbed from its background value
q(x

) =
(0)
q(x

) +q. (5.12)
We wish to establish the transformation of the quantity q. In principle this could be done in analogy
to the metric function. Let us write this scalar at the new point x

= x

as
q( x

) =
(0)
q( x

) + q =
(0)
q(x

) +
(0)
q(x

)
,

+ q. (5.13)
where in the last step we expanded the equilibrium background value of the function. Comparing (5.12)
and (5.13) we see that the scalar is invariant against transformations (5.7) if
q q = q
(0)
q
,

. (5.14)
5.2. GAUGE TRANSFORMATIONS AND GAUGE INVARIANCE 89
In the similar fashion we derive the gauge transformation for the covariant 4-vector
u

= u

(0)
u
,

(0)
u

,
. (5.15)
Let us decompose the spatial part of the innitisemal perturbation

= (
0
,
i
) in two components

i
=
i

+
,i
,
i
,i
= 0, (5.16)
where
i

is a three-vector with zero divergence and is a scalar function. In the Friedman universe
(0)
g
00
= a
2
() and
(0)
g
ij
= a
2
()
ij
[compare with (5.2)]. Now we can use the rule for the metric
transformations
g

= g

(0)
g
,

(0)
g
,

(0)
g

,
. (5.17)
to obtain for each component the following relations:
g
00
= g
00
2a(a
0
)

, (5.18)
g
0i
= g
0i
+ a
2
_

i
+ (

0
)
,i

, (5.19)
g
ij
= g
ij
+ a
2
_
2
a

ij

0
+ 2
,ij
+ (
i,j
+
j,i
)
_
, (5.20)
where the prime denotes derivative with respect to the conformal time .
5.2.1 Scalar perturbations
Suppose we have only scalar perturbations. We wish to write down the metric which includes the
background metric plus the scalar perturbations. In this case the metric takes the form:
ds
2
=
(0)
g

+ g
00
+ g
0j
+ g
i0
+ g
ij
= a
2
_
(1 + 2)d
2
+ 2B
,i
dx
i
d [(1 2)
ij
2E
,ij
]dx
i
dx
j
_
. (5.21)
First note that all the terms that are proportional to 1 refer to the background metric. There are
two such terms, one for the time-like coordinate, the other for the space-coordinates. Furthermore, we
substituted the scalar perturbations from Eqs. (5.3)-(5.5). Now we can use the denitions Eqs. (5.3)-
(5.5) and transformations (5.18) and (5.20) to establish the transformation properties of the scalar
functions


=
1
a
(a
0
)

, B

B = B +

0
, (5.22)


= +
a

0
, E

E = E +. (5.23)
We may also consider the combination

B

E

= B E

0
, (5.24)
which tells us that the combination BE

has purely time-like shift under transformations. Now we see


that only the perturbations
0
and contribute to the transformations of the scalars. They can be chosen
such that two of the four scalar functions vanish. The simplest gauge-invariant linear combination of
the scalar functions is given by

1
a
[a(B E

)]

, +
a

a
(B E

). (5.25)
90 CHAPTER 5. INSTABILITIES IN GENERAL THEORY OF RELATIVITY
This means that they do not change under coordinate transformations; if they vanish in one particular
coordinate system, then they should vanish in any other coordinate system. Note that one can construct
innite set of gauge invariant variables and it is a matter of choice and convenience which coordinate
system will be used.
To give a concrete physical example consider energy perturbations. We have seen that the four-
scalars are transformed according to (5.13). For the energy density this equation translates into
=

0
=

0
(B E

). (5.26)
In the homogeneous universe the uid four-velocity us given by
(0)
u

= (a, 0, 0, 0) and the velocity


perturbation are transformed according to
u
0
= u
0
[a(B E

)] , u
i
= u
i
a(B E

). (5.27)
5.2.2 Vector perturbations
Now consider the case where there are only vector perturbations
ds
2
= a
2
_
d
2
+ 2S
i
dx
i
d (
ij
F
i,j
F
j,i
)dx
i
dx
j
.

(5.28)
The vectors transform as
S
i


S
i
= S
i
+

i
, F
i


F
i
= F
i
+
i
. (5.29)
and we can construct a gauge-invariant combination as

V
i
= S
i
F

i
. (5.30)
Out of four independent function S
i
and F
i
only two describe real physical perturbations; the other two
reect the freedom to choose the coordinate system. The variables (5.30) span a two-dimensional space
of physical perturbations and describe rotational motions.
5.2.3 Tensor perturbations
For tensor perturbations
ds
2
= a
2
_
d
2
(
ij
h
ij
)dx
i
dx
j

. (5.31)
and h
ij
does not change under coordinate transformations and describes the gravitational waves in a
gauge-invariant manner.
5.3 Gauge xing
The gauge xing means that we can choose arbitrarily the functions and
0
which will provide
connections on the functions , B, E and velocity perturbations u
,i
=
,i
.
Gauge xing is equivalent to the choosing a special coordinate system. We will consider below two
examples
5.3. GAUGE FIXING 91
Longitudinal gauge
This gauge is dene by the condition
B = E = 0. (5.32)
This implies that the perturbations =
0
= 0 [to verify this substitute this condition in Eq. (5.25)].
The metric then takes the form
ds
2
= a
2
_
(1 + 2
l
)d
2
(1 2
l
)
ij
dx
i
dx
j

. (5.33)
The metric can be even further more simplied if the spatial perturbations of the energy-momentum
tensor are diagonal T
i
j

i
j
. We will see below that in that case
= , (5.34)
i.e., the metric perturbations are characterized just by one scalar function, say, . This is a generaliza-
tion of the Newtonian potential therefore this gauge and the corresponding coordinate system is often
called conformal Newtonian. In this gauge all the variables are the simple meaning
= = = (5.35)
are the amplitudes of the metric, is the amplitude of the energy density and u

are the amplitudes


of the velocity perturbations.
Synchronous gauge
In the synchronous the following conditions are imposed
= 0, B = 0. (5.36)
This choice does not x the gauge completely, since the following transformation dene an equivalent
systems of coordinates corresponding to x

, where
= +
C
1
a
, x
i
= x
i
+C
1,i
(x
j
)
_
d
a
+ C
2,i
(x
j
). (5.37)
The connection to the longitudinal gauge is easily established if we note the relations (5.25) in the
synchronous gauge can be inverted
=
1
a
[aE

E =
_
1
a
__

ad
_
d, (5.38)
=
a

a
E

s
= +
a

a
2
_
ad, (5.39)
so that they are expressed only the unknown function (which can be computed by some means, e.g.,
in the Newtonian conformal gauge). The energy density perturbation can be likewise written as

s
=


0
a
_
ad. (5.40)
92 CHAPTER 5. INSTABILITIES IN GENERAL THEORY OF RELATIVITY
5.4 Cosmological perturbations
In this section we consider the perturbation of the Einsteins equations on the background of Friedman
universe. The unperturbed equations read
G

= R


1
2

R = 8GT

. (5.41)
The background metric is given by Eq. (5.2)
(0)
g

dx

dx

= a
2
()(d
2

ij
dx
i
dx
j
). (5.42)
The components of the Einstein tensor are given by
(0)
G
0
0
=
3H
2

a
2
,
(0)
G
0
i
= 0,
(0)
G
i
j
=
2H

+ H
2

a
2

ij
, (5.43)
where H

= a

/a. Comparing (5.41) and (5.43) we conclude that


(0)
T
0
0
= 0,
(0)
T
i
j

ij
. (5.44)
For small perturbation from the background we can write the perturbed equations as
G

= 8GT

. (5.45)
The perturbed quantities are non-invariant. The transformation laws for the perturbations of the
energy-momentum tensor are
T
0
0
= T
0
0
(
(0)
T
0
0
)

(B E

), (5.46)
T
0
i
= T
0
i

_
(0)
T
0
0

T
k
k
3
_
(B E

)
,i
, (5.47)
T
i
j
= T
i
j

_
(0)
T
i
j
_

(B E

). (5.48)
Here T
k
k
is the trace of the spatial components, which is gauge invariant. Analogously, we have
G
0
0
= G
0
0
(
(0)
G
0
0
)

(B E

), (5.49)
G
0
i
= G
0
i

_
(0)
G
0
0

G
k
k
3
_
(B E

)
,i
, (5.50)
G
i
j
= G
i
j

_
(0)
G
i
j
_

(B E

). (5.51)
The perturbation equations can be written as
G

= 8GT

. (5.52)
Note that the both sides of this equation can be decomposed into scalar, vector and tensor perturbations.
Let us consider these separately.
5.5. METRIC PERTURBATIONS 93
5.5 Metric perturbations
Scalar perturbation
We need to compute the perturbations for the metric (5.53) containing only scalar perturbations
ds
2
= a
2
(1 + 2)d
2
+ 2B
,i
dx
i
d [(1 2)
ij
2E
,ij
]dx
i
dx
j
. (5.53)
A direct calculation of the perturbation G

gives the following set of equations


3H

+ H

) = 4Ga
2
T
0
0
, (5.54)
(

+ H

)
,i
= 4Ga
2
T
0
i
, (5.55)
_

+H

(2 + )

+ (2H

+ H
2

) +

2
( )
_

ij

1
2
( )
,ij
= 4Ga
2
T
i
j
. (5.56)
The equations for the metric perturbations in the conformal Newtonian and in terms of the gauge
invariant potentials are identical.
Vector perturbation
The perturbation equations for the vector perturbations take the form
V
i
= 16Ga
2
T
0
i
, (5.57)
(V
i,j
+ V
j,i
)

+ 2H

(V
i,j
+ V
j,i
) = 16Ga
2
T
i
j
. (5.58)
where T

is the vector part of the energy-momentum tensor and V


i
= S
i
F

i
[see Eq. (5.30)] and the
discussion there.
Tensor perturbation
As we already said the tensor perturbations correspond to the gravitational waves. The tensor pertur-
bations obey the equations
h

ij
+ 2H

ij
h
ij
= 16Ga
2
T
i
j
. (5.59)
where T
i
j
is the part of the energy-momentum tensor that have the tensor structure of h
ij
.
5.6 Perturbations of the matter
Now we consider the matter perturbations which arise from perturbations of the energy-momentum
tensor. For simplicity we assume that the matter can be described as a perfect uid, therefore, its
energy and momentum tensor can be written as
T

= ( + p)u

. (5.60)
We have already computed the relevant perturbations in Eq. (5.46). They can be written as
T
0
0
= , (5.61)
T
0
i
=
1
a
(
0
+ p
0
)(u
i
..
vector
+ u
i
..
scalar
), (5.62)
T
i
j
= p
i
j
. (5.63)
94 CHAPTER 5. INSTABILITIES IN GENERAL THEORY OF RELATIVITY
[see (5.26) and (5.27) for denitions]. All the perturbations contribute to the scalar perturbation except
the term that is marked by vector.
Now we specify the equations for the case of scalar perturbations. Consider the elements of Eq. (5.56)
with i ,= j. In this case it term
i
j
will not contribute and we obtain
( )
,ij
= 0 i ,= j. (5.64)
This equation has the solution
= . (5.65)
Next we substitute Eqs. (5.61)-(5.63) into the equations for scalar perturbations (5.54)-(5.56) and obtain
3H

+H

) = 4Ga
2
, (5.66)
(a)

,i
= 4Ga
2
(
0
+ p
0
)u
i
, (5.67)
a

+ 3H

+ (2H

+ H
2

) = 4Ga
2
p. (5.68)
It is instructive to consider the case of non-expanding universe, i.e., the limit H = 0. Then the
rst equation is just the Poisson equation for the gravitational potential. Thus Eq. (5.66) is just a
generalization of the Poisson equation to the case of Hubble expansion. We can use the relation
p = c
2
S
+S, (5.69)
to eliminate the pressure perturbations from Eq. (5.68). Combining the Eqs. (5.66)-(5.68) we nally
obtain one perturbation equation

+ 3(1 + c
2
S
)H

c
2
S
+
_
2H

+ (1 + 3c
2
S
)H
2

= 4Ga
2
S. (5.70)
Our aim in the following will be to nd the solutions to this equation in several interesting cases.
Nonrelativistic matter
We consider now at, matter-dominated universe, such that a
2
and H

= 2/. Then only the rst


two terms in Eq. (5.68) survive and one nds

+
6

= 0. (5.71)
Integration of this equation gives
= C
1
(x) +
C
2
(x)

5
(5.72)
where C
1
(x) and C
2
(x) are arbitrary functions of the comoving spatial coordinates. With this solution
at hand we can go back to Eq. (5.66) and nd the energy perturbations as

0
=
1
6
_
(C
1

2
12C
1
) + (C
2

2
+ 18C
1
)
1

5
_
. (5.73)
This equation can be analysed numerically. However, simple solutions can be obtained in the limiting
cases. Consider perturbations such that C
1,2
exp(ik x). Dene a physical scale a/k which is
related to the perturbation with wave number k; this can be compared with the Hubble scale 1/H a.
The ratio of the Hubble scale over the physical scale is given by the parameter k. The interesting
limiting cases are
5.6. PERTURBATIONS OF THE MATTER 95
k 1, which is the long-wave length limit, the energy perturbation is given by

0
2C
1
+
3C
2

5
. (5.74)
If we neglect the decaying mode (the second term) then we obtain simply /
0
= 2.
k 1 - the short-wave-length limit leads to

0

k
2
6
(C
1

2
+C
2

3
) =

B
1
t
2/3
+ B
2
/t, (5.75)
which is in agreement with (4.58). In this limit the Newtonian and GR treatments lead to the
same result.
Ultrarelativistic matter
Consider now the universe, which is dominated by relativistic matter, whose equation of state is p = w,
where w is a positive constant. The scale factor behaves as
a
2/(1+3w)
. (5.76)
For such equation of state c
2
S
= w. Assuming plane-wave perturbations =
k
exp(ik x) we have

k
+
6(1 +w)
(1 + 3w)

k
+ wk
2

k
= 0. (5.77)
The solution of this equation is

k
=
1

_
C
1
J

wk) + c
2
Y

wk)

,
1
2
_
5 + 3w
1 + 3w
_
. (5.78)
Here J

and Y

are the Bessel functions of order . Consider again the cases


wk 1, which is the long-wave length limit. In this case we can use the expansion of Bessel
functions for small arguments to obtain

0
2 (5.79)
If we neglect the decaying mode (the second term) then we obtain simply /
0
= 2.


wk 1 - the short-wave-length limit leads to

k

1/2
exp
_
i

wk
_
(5.80)
which is in agreement with (4.58). In this limit the Newtonian and GR treatments lead to the
same result.
96 CHAPTER 5. INSTABILITIES IN GENERAL THEORY OF RELATIVITY
Special case: w = 1/3 (radiation dominated universe)
In this special case the order of Bessel functions is xed = 3/2. And these could be expressed in
terms of elementary functions (x = k/

3)

k
=
1
x
2
_
C
1
_
sin x
x
cos x
_
+ C
2
_
cos x
x
+ sin x
_
_
, (5.81)
and the energy-density perturbation is given by

0
= 2C
1
__
2 x
2
x
2
__
sin x
x
cos x
_
+ 4C
2
_
1 x
2
x
2
_
_
cos x
x
+ sin x
_
_
. (5.82)
Entropy perturbations
If a medium is multi-component, then we can have in principle both adiabatic perturbation and entropy
perturbations. Quite generally the number of dynamical variables doubles in this case and the analysis
is rather complicated.
A simplied model arises when one has cold baryonic component coupled to radiation plasma. These
two components are coupled tightly to each other, i.e., there is no relative motion between them. Then
we can use a one-component perfect uid approximation. The main dierence is that the pressure
depends not only on energy density but also the entropy per baryon S T
3

/n
b
, where n
b
is the density
of baryons. This one can have entropy perturbations. Then our master equation (5.70)

+ 3(1 + c
2
S
)H

c
2
S
+
_
2H

+ (1 + 3c
2
S
)H
2

= 4Ga
2
S. (5.83)
can be used with the understanding that S is a constant because the entropy per baryon is conserved.
First we determine the parameter
=
_
p
S
_

. (5.84)
The contribution to the pressure arises from the radiation w = 1/3, so that we can write
p = p

=
1
3

. (5.85)
The total energy perturbation can be written as a sum of that of the components
=

+
b
. (5.86)
We need to nd the scaling of the entropy
S
T
3

n
b
=

3/4

S
S
=
3
4

b
(5.87)
Now we use Eqs. (5.85) and (5.86) to write
p =
1
3
_
1 +
3
b
4

_
1
_
+
b
S
S
_
. (5.88)
By matching this equation to [see Eq. (5.69)]
p = c
2
S
+S, (5.89)
5.6. PERTURBATIONS OF THE MATTER 97
we can identify the parameters
c
2
S
=
1
3
_
1 +
3
b
4

_
1
, =
c
2
S

b
S
. (5.90)
Now let us come back to the solution of Eq. (5.83). When there non-zero entropy perturbation the
solution of this equation is a sum of the homogeneous equation with S = 0 and some special solution.
We have found the solutions to the homogeneous equation in some interesting cases. Let us look for a
solution of the special case. We rst slightly modify the term as follows
2H

+ (1 + 3c
2
S
)H
2

= 8Ga
2
(c
2
S
p) = 2Gc
2
S
a
2

b
. (5.91)
Furthermore, for long-wavelength perturbations the term with spatial gradient can be dropped. The
new equation then reads

+ 3(1 +c
2
S
)H

+ 2Ga
2
c
2
S

b
= 4Ga
2
c
2
S

b
S
S
. (5.92)
where we substituted the value for the parameter. Thus we see that

p
=
2S
S
= const. (5.93)
is a solution to Eq. (5.92). However, in the very early universe we should have a homogeneous background
radiation plus uctuations in the baryon eld, so that one would expect that
0 ( 0). (5.94)
The solution that satises this condition is the following
=

5
(
2
+ 6 + 10)
( + 2)
3
S
S
. (5.95)
where = /

[for denition see Eq. (1.95)].


Vector and tensor perturbations
Substituting the expressions for the perturbations of energy-momentum tensor T
0
i
= a
1
(
0
+p
0
)u
i
we obtain the perturbation equations for the vector elds
V
i
= 16Ga(
0
+ p
0
)u
i
(5.96)
and
(V
i,j
+ V
j,i
)

+ 2H

(V
i,j
+ V
j,i
) = 0. (5.97)
The solution of the second equation is
V
i
=
C
i
a
2
(5.98)
where C is a constant of integration. For the physical velocity
v
i
= a
_
dx
i
ds
_
= a
1
u
i
(5.99)
98 CHAPTER 5. INSTABILITIES IN GENERAL THEORY OF RELATIVITY
we obtain
v
i

1
a
4
(
0
+ p
0
)
. (5.100)
In agreement with the Newtonian theory the vector perturbations decay inversely with the scale factor.
Therefore, the expansion of the universe leads to a damping of the vector perturbations.
The tensor perturbations; assuming a perfect uid T
i
j
= 0 and the perturbation equation simplies
to
h

ij
+ 2H

ij
h
ij
= 0. (5.101)
This equation can be brought to a compact form by dening
h
ij
=
v
a
e
ij
. (5.102)
where e
ij
is polarization tensor. Now we consider plane-wave perturbations to obtain
v

+
_
k
2

a
_
v = 0. (5.103)
In the radiation dominated era a , a

= 0 and v exp(ik) then there exists an exact solution


h
ij
=
1

[C
1
sin(k) + C
2
cos(k)]e
ij
. (5.104)
For long-wavelength perturbations with k 1 we can neglect the k
2
term in Eq. (5.103) and the
solution reads
v C
1
a +C
2
a
_
d
a
2
, (5.105)
which can be translated into
h
ij
=
_
C
1
+ C
2
_
d
a
2
_
e
ij
. (5.106)
For short-wavelength perturbations (k 1) we obtain k
2
a

/a and h exp(ik)/a.
Chapter 6
Phase transitions in the early universe
In this chapter we will study the physics of early universe, i.e., the evolution of the universe at high
energies. There are two important energy scales that mark the high energy landscape. One is the Planck
scale 10
19
GeV where we expect that the classical gravity will break down and quantum eects will
become important. Below the TeV = 10
3
GeV scale experimental studies of matter are feasible. This
energy domain is well described by the Standard Model of particle physics and we start our discussion
by an overview of the standard model.
6.1 Elements of the Standard Model
6.1.1 Local gauge invariance and quantum electrodynamics
There are four forces known in Nature: the strong, weak, electromagnetic and gravitational. The
Standard Model describes the physics of particles and elds interacting by the rst three forces. The
gravity is described by the theory of General Relativity. The electromagnetism and weak force are
unied in an electroweak theory, while the strong force is described by quantum chromodynamics.
These theories are based on the ideas of quantum elds theories or QFT, where particles are elementary
excitations of elds.
Let us start with an example. The fermions are described by the Dirac equation
(i

m) = 0, (6.1)
where is a four-component Dirac spinor

are 44 Dirac matrices. The Dirac equation follows from


the Euler-Lagrange equation of the Lagrangian
L = i

(6.2)
where

0
.
An important feature of this Lagrangian is that it is invariant with respect to global gauge trans-
formations. This means that it will not change is the elds are transformed as
e
i
(6.3)
where the phase is a constant in space and time. The transformed Lagrangian is
L

= i

e
i

e
i
m

e
i
e
i
= L. (6.4)
99
100 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
We see that the Lagrangian remained invariant because we could commute the phase factor with the
remaining terms.
The next question is what if e(x

), i.e., the phase depends on the local space-time coordinate; the


appearance of the constant charge e will be clear shortly. Such transformation is called local gauge
transformation. Now, obviously will not commute with the operator of dierentiation:
L

= i

e
i

e
i
m

= i

ie

) m

. (6.5)
The gauge invariance of the Lagrangian is now spoiled by the gradient term (which is a four-vector).
However, we can restore the gauge invariance by introducing a new vector called gauge eld A

. The
Lagrangian is now redened to contain a covariant derivative
D

+ ieA

. (6.6)
Physically, the new term in the Lagrangian describes the interaction of the gauge eld with the Dirac
fermion. What are the transformation properties of this new gauge eld? In fact, we use the freedom of
this transformation to postulate a transformation law such that the Dirac Lagrangian is locally gauge
invariant. The following transformation does the job:
A

= A

. (6.7)
The gauge invariant Lagrangian then reads
L = i

(6.8)
Simultaneously transforming the fermionic and gauge elds

e
ie(x

)
, A

= A

. (6.9)
we see that the changes are in the term
D

e
ie(x

)
. .

= (

+ ieA

)e
ie(x

)
= e
ie(x

+ ieA

e
ie(x

)
iee
ie(x

)
(x

)
= e
ie(x

+ieA

+ ie

e
ie(x

)
iee
ie(x

(x

) = e
ie(x

)
D

.
(6.10)
Now the kinetic term in the Lagrangian transforms as

= (

e
ie
)

e
ie(x

)
= (

e
ie
)

e
ie(x

)
D

, (6.11)
i.e., it is invariant with respect local gauge transformations. The eld A

contributes its energy density


to the Lagrangian (this contribution should be scalar that is invariant under Lorentz transformations).
It is given by the contraction of the eld strength tensor with itself F

, where the eld-strength


tensor is dened via
F

. (6.12)
The full Lagrangian can be now written as the sum of the fermionic part plus the gauge eld part
L = i


1
4
F

. (6.13)
This Lagrangian describes the Quantum Electrodynamics. The meaning of e constant is then clear
- it is the charge of the fermion. The coupling constant of this theory is the ne-structure constant
= e
2
/4 1/137. Since 1 the interaction term is small and one can develop systematic
perturbation theory to compute the observables. The Quantum Electrodynamics is in an excellent
agreement with the experimental data.
6.1. ELEMENTS OF THE STANDARD MODEL 101
6.1.2 Non-abelian quantum eld theories
The idea of gauge invariance can be extended to groups larger than the U(1) group of quantum elec-
trodynamics. Indeed, the gauge transformation we have discussed corresponds to a multiplication of
the wave function with a unitary 1 1 matrix with complex elements U = exp(ie(x

)), such that


U

U = 1. Next we follow the historical path and extend the idea of local gauge invariance to the so-
called SU(1) group. This program was carried out by Yang and Mills and was later used to construct
the electroweak theory.
We generalize our gauge transformation by going from a 1 1 matrix to an N N matrix. The
group of all N N matrices is called U(N) group and the gauge transformation generated by such
matrices is called U(N) gauge transformation. To see how this works let us consider a Lagrangian of
N free Dirac elds with equal masses. Let us introduce vectors of these elds
=
_
_

1
. . .

N
_
_
,

=
_

1
. . .
N
_
. (6.14)
Each of the elements is by itself a Dirac spinor (i.e. has four components). We have assumed that the
massed of the particles are the same but we will not imply that they all posses the same charges. The
new Lagrangian is now
L = i

m

. (6.15)
The property U

U = 1 guarantees that this Lagrangian is invariant under global gauge transformations,


that is it does not change under replacements
U. (6.16)
Now we repeat the steps that lead us to construction of local gauge invariant Lagrangian. First we
check that our Lagrangian is not invariant under local gauge transformations. Indeed

(U) = U

U = U[

+ U
1

U]. (6.17)
The second term spoils the gauge invariance of the Lagrangian. The solution of the problem is the same
as above; we introduce gauge elds via a new matrix
D

+ igA

(6.18)
and postulate their gauge transformations as
A

= UA

U
1
+
i
g
(

U)U
1
. (6.19)
Now we compute
D

= (

+ igA

U + igA

U = U[

+ U
1

U] + ig(UA

U
1
+
i
g
(

U)U
1
)U
= U

+ igUA

= UD

. (6.20)
Now noting that

=

U

and again that U

U = 1 we see that the Lagrangian is invariant under local


gauge transformations. The eld strength tensor for the gauge eld is constructed as
F

= D

= (

+igA

)A

+igA

)A

+ig(A

). (6.21)
102 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
The eld strength tensor transforms as F

= UF

U
1
, which follows from the commutation rules
D

U = UD

. The full Lagrangian is then given by


L = i

m

g


1
2
Tr(F

). (6.22)
The key dierence between the (6.12) and (6.21) is the appearance of the additional term g in the last
equation, which is a manifestation of the non-abelian nature of the gauge elds. Mathematically this
comes from the fact that the N N matrices do not commute as opposed to the numbers. Physically,
this means that self-interactions between the elds is possible. The eld theories with this property are
called non-abelian gauge theories.
A natural question is what is the number of independent gauge elds that are need to insure the
gauge invariance? For that purpose we turn to some properties of complex matrices and state these
properties without proof.
1. Any unitary matrix can be written as U = exp(iH), where H is a Hermitian matrix.
2. The number of independent real number in a N N matrix is equal N
2
.
3. Any Hermitian matrix can be decomposed into linear superposition of N
2
1 basis matrices and
the unit matrix
H = 1 +
N
2
1

i=1

i
T
i
(6.23)
where T
i
are traceless matrices and
i
are real numbers.
4. The unitary matrix can be written then as
U = exp(i)
. .
U(1)
exp
_
i

i
T
i
_
. .
SU(N)
. (6.24)
where U(1) and SU(N) are the subgroups of the group U(N).
Mathematically the lats statement is written as U(N) = U(1) SU(N). (The notation SU stands for
special unitary). The SU(N) group has N
2
1 independent generators. We can now consider separately
the gauge invariance within the SU(N) group by introducing N
2
1 independent gauge elds A
i

. The
Hermitian matrix of the gauge elds now can be written in the basis of the generators as
A

i
A
i

T
i
(6.25)
In the case of the SU(2) group the basis matrices are the given by /2, where are the Pauli matrices.
In the case of SU(3) group the 22 Pauli matrices are replaced by the 33 Gell-Mann matrices. Note
that the basis matrices are traceless.
6.1. ELEMENTS OF THE STANDARD MODEL 103
6.1.3 Quark-hadron phase transition in the early universe
The quantum chromodynamics is the theory of strong interactions. It is built on the idea of SU(3) local
gauge invariance of quantum eld theory of quarks (fermions of the theory) and gluons (gauge elds).
The phenomenology of hadrons - particles interacting via the strong force - is described in terms quarks
which come in six avors and three colors. The quark avors are up, down, strange, charm, bottom
and top and the colors are red, blue, and green. (Note that avor and color are just internal quantum
numbers characterizing the elementary particles, such as e.g. spin). The up, charm and top quarks have
+2/3 electric charge, while the others (down, strange and bottom) have -1/3 electric charge. The bare
(or so-called current) masses of the quarks are strongly avor dependent, i.e., for the up quark 1.5 4.5
MeV, down quark 5 8.5 MeV, strange quark 80 155 MeV, charm quark 1.3 GeV, bottom quark 4.3
GeV, top quark 170 GeV.
The quarks allow us to classify all the (several hundreds of) hadrons known in Nature. The baryons
(e.g. the neutron and proton) appear as the bound states of three quarks. The bosons (e.g. pion)
appear as bound states of quark and anti-quark pairs.
L =

f,i
_
i

f
m
f

f

1
2
g
s
(i

f
)A
i

1
2
Tr(F

), (6.26)
where f sums up the quark avors, the index i = 1, 2, . . . 8 labels the Gell-Mann matrices. The
eight gauge elds A
i

correspond to the gluons which are transmitters of the strong interactions (like
the photons are transmitters of the electromagnetic interactions). In writing the Lagrangian we have
chosen to organise the dierent colors into a color-triplet of given avor as

f
=
_
_
r
f
b
f
g
f
_
_
, (6.27)
where r stands for red, b for blue, and g for green. Each of the eight gluons (because there are
8 gluons we speak of gluon octet) carries one charge of color and one charge of anti-color. There
are nine combinations in which we can pair gluons and anti-gluons: r r, r g, r

b, g r, g g, g

b, b r, b g, b

b.
The colorless combinations are now allowed in the Lagrangian; such combination is r r + b

b + g g.
(Such combination would appear in a theory with U(3) symmetry where the unit matrix is among the
generators of the group, however the corresponding charge neutral boson would induce a long range
interaction between the hadrons, which would contradict the experimental data). The last term in
the Lagrangian (6.26) correspond to the self-interaction among gluons and includes three-leg and four-
leg processes. (This is in contrast to the quantum electrodynamics, where there is no self-interaction
between the photons).
There are a number of conservation laws associated with the processes involving quarks and gluons
The avor is unchanged in the interaction term of the Lagrangian; thus strong interactions do not
change the quark avors.
The total number of quarks minus the number of anti-quarks remains unchanged, i.e. the baryon
number is conserved (quarks carry the baryon number 1/3, antiquarks -1/3, so that a baryon has
baryon number 1, and an anti-baryon a baryon number -1).
The color is conserved in any process.
104 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
An important feature of the QCD is the property of connement, which states that all the particles
in nature must occur in color-neutral states. Familiar examples are neutron, which is built of ddu quarks
and proton, which built of duu quarks. Mesons are built of quark and anti-quark pairs, for example the

is a bound state of d u quarks.


Another important feature of the QCD is the asymptotic freedom, which is the consequence of the
running coupling constant in QCD. The latter means that the coupling constant is a function of the
momentum transfer: it is large at small momentum transfers and becomes large as it increases. The
dependence of the strong coupling constant on the magnitude of the momentum transfer is given by

s
(q
2
) =
12
11n 2f ln(q
2
/
QCD
)
, (6.28)
where n is the number of colors, f is the number of avors: f = 5 for energies below the top quark
mass 170 GeV and f = 6 above this energy.
QCD
= 220 MeV is a characteristic scale of QCD. For
example, for energies of the order 10 GeV the strong coupling constant is of the order of 0.2 (f = 5).
Thus, for energies GeV and larger QCD is a perturbative theory, i.e., one can carry out systematic
calculations where the processes are arranged in increasing powers of the coupling constant. For lower
energies one need non-perturbative methods (e.g. lattice calculations).
Now let us come back to the early universe. At temperatures T
QCD
the matter is most likely
to be in the deconned state in which the quarks are not localized in hadrons (as in our world) but form
a plasma of free particle, which in view of smallness of
s
at these energies interact weekly with each
other. As the temperature decreases the quarks become conned in hadrons for T
QCD
. Thus we
may expect that there occurs a phase transition from quark-gluon plasma to hadronic matter as the
universe expands and cools. This phase transition will be referred to as the cosmological quark-hadron
phase transition.
To understand this phase transition let us discuss the thermodynamic parameters of the quark-gluon
plasma. The total pressure of the plasma can be written as
p
q
=
k
q
3
T
4
B(T), (6.29)
where

q
=

2
30
[2 8
. .
gluons
+
7
8
3 2 2 N
f
. .
quarks
]. (6.30)
The counting of the degrees of freedom is as follows: rst term contains the contribution from 8 gluons
with two polarizations, the second term is the contribution from quarks with 3 colors, two spins, and
N
f
avors, where an extra factor of 2 arises from the anti-particles.
The terms B(T) is called the bag constant (which in general may depend on temperature and density)
and can be phenomenologically adjusted with simplied models, such as the MIT bag model. It seem
that the bag constant produced a negative pressure on quarks so that remain conned in a certain
volume corresponding to hadrons. The quark and gluon elds vanish outside the hadrons. This mimics
the connement of quarks with the hadrons. The quark-gluon plasma state corresponds to the one
where the bags overlap and therefore these can move free from one to another volume with feeling
the action of the bag constant over the space.
From the pressure (6.29) we can derive the energy density and the entropy of quark gluon plasma
6.1. ELEMENTS OF THE STANDARD MODEL 105
as
s
qg
=
4
3

qg
T
3

B
T
, (6.31)

qg
=
qg
T
4
+ B T
B
T
. (6.32)
When the temperature drops below the critical temperature of phase transition from the quark to
hadronic phase
T
c

QCD
200 MeV, (6.33)
the quarks become conned to hadrons - mostly the lightest hadrons, i.e., pions

,
0
. Since the pion
masses are about 134 140 MeV one can treat them at T
c
as massless bosons. Assuming that we deal
with an ideal gas of pions the pressure and entropy are given by
p
h
=

h
3
T
4
, s
h
=
4
h
3
T
3
, (6.34)
where
h
=
2
/10. We see that the number of degrees of freedom dramatically drops from 16 +12 N
f
to only 3. An important question to answer is the order of the phase transition. The answer depends
on the masses of the light quarks, and since the calculations are dicult it is hard to say which order
of phase transition we have or whether there is a phase transition at all (the alternative is a crossover).
Here are the possibilities:
Pure gluonic matter (no quarks) - the phase transition is rst order;
Two massless quarks and gluons - the phase transition is second order;
Three massless quarks and gluons - the phase transition is rst order;
Massive quarks and gluons - the phase transition becomes a crossover.
As mentioned above there is a sharp change in the thermodynamical parameters such as the entropy and
energy density around the critical temperature T
c
. But even in the case of rst order phase transition
the pressure of the two phases must be continuous, so that equating the pressure of these two phases
we obtain an estimate of the critical temperature as a function of the gap constant B
0
T
c
=
_
3B
0

qg

h
_
1/4

_
B
0
1.42 + N
f
_
1/4
= 0.7B
1/4
0
(6.35)
We see that to obtain T
c
= 200 MeV we need B
1/4
0
= 286 MeV.
A rst order phase transition occurs via formation of seeds of the new phase (in our case hadronic
phase) which grow and form larger bubble until the new phase does not ll the whole space (a good
example from daily life is the boiling of water as a special case of liquid-gas phase transition). Further-
more, during such phase transition there is a latent heat released, which is equal to the TS, where
S = (4/3)(
qg

h
)T
3
c
is the jump in the entropy at the phase transition. At the phase transition
T = T
c
, i.e., it stays constant throughout the phase transition. Thus, the released heat during the
cosmological phase transition can be absorbed by the other components of the uid (the radiation -
photons, and the leptons) and their temperature may stay constant despite of the fact that the universe
expands.
The entropy is continuous through the phase transition if it is second order or a crossover. In the
case of the rst order phase transition the formation of inhomogeneities may lead to some violent eect
that may be observables. Examples are formation of black holes, generation of gravitational waves and
magnetic elds.
106 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
6.1.4 Electroweak phase transition in the early universe
The electroweak theory unies the electromagnetic and the weak forces within a theory which is based
on the SU(2) U(1) gauge symmetry. We can not combine them into a U(2) theory, since the coupling
constants in each group should be adjusted to the experiment independently.
The particle content of the electroweak theory includes the following elementary particles:
d, u, s, c, b, t,
. .
quarks
e, , ,
neutrinos
..

e
,

. .
leptons
. (6.36)
All the fermions carry two spin degrees of freedom, except of neutrinos, which appear in nature only
in the left-handed states (anti-neutrinos only in the right handed states). Each quark avor carries the
color degrees of freedom. Furthermore, the neutrinos are special because they have very small masses
compared to the other fermions and can be treated, in a rst approximation, as massless. This implies
that helicity of neutrinos is a good quantum number (by convention the helicity of neutrinos is 1, that
of anti-neutrinos +1). The leptons can be organized into three generations
_

e
e
_
L
, e
R
;
_

_
L
,
R
;
_

_
L
,
R
. (6.37)
Since neutrinos appear only as left handed states we assign a SU(2) duplet to the left-handed states,
whereas the right-handed states are singlets with respect to the SU(2) group. The organisation in
generations reect the fact that the weak interactions convert particles only within the same generation
and the lepton numbers are conserved in each generation separately. The quarks can be arranged in
three generations as well
_
u
d

_
L
, u
R
, d

R
;
_
c
s

_
L
, c
R
, s

R
;
_
t
b

_
L
, t
R
, b

R
. (6.38)
The primed avors u

, s

, b

are linear superpositions of the avors u, s, d that are conserved in the


strong interactions. Thus the weak interactions violate the avor conservation laws.
There are three gauge bosons of the electroweak theory
W

, Z
0
. (6.39)
where the rst two are electrically charged, while the third is electrically neutral. The rst two are
responsible for charge current interactions and the third one for the charge neutral current interactions.
The masses of fermions can not be introduced in the Lagrangian in a ad hoc manner, since this is
known to spoil the gauge invariance of any chiral theory. We shell assume for the time being that the
fermions are massless and will return later to the problem of generation of fermion masses from their
interactions with a classical scalar eld (known as the Higgs eld). Now we are ready to write down
the Lagrangian of the electroweak theory
L
f
= i

+ igA

+ ig

Y
L
B

L
) + i

+ ig

Y
R
B

R
). (6.40)
Note that this Lagrangian is written for one particular avor and can be applied to each avor gen-
eration. Furthermore, the left and right handed sectors are separated - this is characteristic for the
so-called chiral theories (chiral means handedness). Note that the Lagrangian is not chirally symmetric.
The Lagrangian (6.40) is invariant under the SU(2) transformations

L
= U
L
,

R
=
R
(6.41)
6.1. ELEMENTS OF THE STANDARD MODEL 107
where U is a unitary 2 2 matrix and U(1) transformation

L
= e
ig

Y
L
(x)

L
,

R
= e
ig

Y
L
(x)

R
. (6.42)
The gauge transformation by which the gauge elds A

and B

have been dened in Eqs. (6.19) and


(6.7) respectively. Not only the fermions, but also three out of four gauge bosons should acquire masses
in the electroweak theory to comply with the experimental data. And as in the case of fermions, the
masses of bosons are generated via the Higgs mechanism. Let us see how this works on a simple example
Spontaneous symmetry breaking in scalar eld theories
To illustrate the mechanism of mass generation let us consider a theory described by the Lagrangian
L =
1
2
[(

+ ieA

)]

[(

+ ieA

)] V (

)
1
4
F

. (6.43)
This Lagrangian is invariant under U(1) gauge transformation

= e
ie
, A

= A

. (6.44)
If we write the complex eld in terms of an amplitude and a phase
= e
ie
(6.45)
we see that the transformation (6.44) leaves the amplitude unchanged, where the phase changes as

= . Then, the second relation in (6.44) suggest that we can dene a new vector
G

= A

. (6.46)
If we rewrite the Lagrangian in terms of the new variables and G

the kinetic term can be written


entirely in terms of gauge invariant quantities
L =
1
2

+
e
2
2

2
G

1
4
F

V (
2
). (6.47)
Suppose the amplitude has a minimum at
0
and consider small perturbations around this value
=
0
+. This new Lagrangian describes a real scalar eld , which posses mass m
Higgs
=
_

V (
0
)
and interacts with a massive vector eld of mass M
G
= e
0
. Given that the starting Lagrangian is
renormalizable we expect that rewriting the Lagrangian in new variables does not spoil the renormal-
izability of the Lagrangian. Thus the new elds representing the observable particles of the theory is a
real scalar eld (Higgs eld) and a massive vector eld.
We have seen how a massive eld arises from initial Lagrangian featuring only massless elds. This
mechanism of mass generation is known as Higgs mechanism.
Generating the masses of gauge bosons
We now consider step by step the generation of masses of gauge bosons, masses of fermions and their
interactions. Our discussion will concentrate on the essential ideas; please consult any textbook on
quantum eld theory for more details. We start by considering the SU(2) U(1) extension of the U(1)
model discussed above; the gauge part is left for later; keeping the kinetic and potential terms only we
obtain the Lagrangian
L
1
=
1
2
(D

(D

) V (

), D =

+ igA

2
B

. (6.48)
108 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
Note that the hypercharge is set to 1/2. The scalar eld can be written in terms of a real amplitude
and complex SU(2) matrix U as
= U
_
0
1
_
. .

0
. (6.49)
Upon substituting (6.49) in (6.48) one nds
L
1
=
1
2
(

) V (
2
) +

2
2

0
_
gG

2
B

__
gG

2
B

0
. (6.50)
where
G

= U
1
A

U
i
g
U
1

U. (6.51)
Conventionally one writes the matrix G

as
G

=
_
G
3

/2 W
+
/

2
W

2 G
3

/2
_
, (6.52)
and uses instead of the variables B

and G

the variables
_
A

_
=
_
cos
W
sin
W
sin
W
cos
W
__
B

G
3

_
, (6.53)
where
W
, which is dened via cos
W
= g/
_
g
2
+ g

2
is the Weinberg angle. In the new variables the
Lagrangian reads
L
1
=
1
2
(

) V (
2
) +
(g
2
+ g

2
)
2
8
Z

+
g
2

2
4
W
+

. (6.54)
Now assume that the eld has a non-zero expectation value
0
and consider as in the case of the
U(1) model small uctuations around this value. Expanding the eld potential V (
2
) we see that the
Lagrangian describes the Higgs scalar eld with mass
M
Higgs
=
_

V [

0
=

2
0
(6.55)
and massive vector mesons elds Z

and W

with masses
M
Z
=
_
g
2
+g

0
2
M
W
=
g
0
2
= M
Z
cos
W
. (6.56)
In addition it contains one massless gauge boson related to the eld A

, which is identied with the


photon. The relation between the coupling constants is given by
e = g

cos
W
= g sin
W
. (6.57)
The gauge bosons W

and Z
0
are the physical elds that appear in the weak interactions. We see that
the charged boson couple to matter at about the same strength as the electromagnetism. The weak
interactions appear to be weak because the eective coupling constant contains inverse second power
of the mass of the gauge boson, which is rather large M
W
= 80.4 GeV and M
Z
= 91.2 GeV. One can
dene the weak ne structure constant

W
=
g
2
4
=
e
2
4 sin
2

W
, (6.58)
which is in fact larger than the ordinary ne structure constant e
2
/4 = 1/137.
6.1. ELEMENTS OF THE STANDARD MODEL 109
Generating the masses of fermions
The masses of the fermions are generated via the interactions with the scalar eld . The Lagrangian
of this interaction can be written (for simplicity we take only the rst generation) as
L
int
= f
e
(

L
e
R
+ e
R

L
), (6.59)
where f
e
is the coupling constant. (In order to be U(1) invariant the hypercharges of the scalar eld
and the left/right handed fermions should satisfy the relation Y
L
= Y
R
+ Y

). As before we substitute
for the scalar eld = U
0
, which then leads to
L
int
= f
e

0
e
R
+ h.c. = f
e
( e
L
e
R
+ e
R
e
L
) = f
e
ee, (6.60)
where

L
= U
1

L
=
_

L
e
L
_
. (6.61)
Again, if the scalar eld has non-zero expectation value, we can write =
0
+ , and read-o the
electron mass as
m
e
= f
e

0
. (6.62)
Note that f
e
is a free parameter in the electroweak theory and can be adjusted to reproduce the electron
mass given an estimate for
0
. The uctuation term that arises from contribution is
f
e
ee (6.63)
and describes the interaction of Higgs boson with electrons. Note that only the electron, i. e., one
component of the doublet becomes a mass, while neutrino remains massless. The magnitude of the
Higgs eld can be xed from the measured masses of gauge bosons
M
W
= 80.4 GeV M
Z
= 91.2 GeV, (6.64)
and the weak coupling constant
G
F
=
1
4

2
g
2
M
2
W
= 1.166 10
5
GeV
2
. (6.65)
and the relations (6.56)

0
=
2M
W
sin
W
e
250 GeV. (6.66)
The electron mass is then reproduced if we choose f
e
= 2 10
6
.
Now we turn to the generation of quark masses. Consider SU(2) gauge invariant quark doublets
u
i
= (u, c, t) d

i
= (d

, s

, b

), i = 1, 2, 3. (6.67)
(recall that the three quark states d

, s

, b

acting in the weak interactions are linear superpositions


of mass eigenstates d, s, b and this leads to avor non-conservation in the weak interactions). In the
three-avor cases these vectors are related by the unitary 3 3 Kobayashi-Maskawa matrix V
i
j
d

i
= V
i
j
d
j
. (6.68)
110 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
The quark interaction term can be written as
L
Q
int
= f
d
ij


Q
i
L

0
d

j
R
f
u
ij


Q
i
L

1
u
j
R
+ h.c., Q
i
L
=
_
u
i
d

i
_
,
1
=
_
1
0
_
. (6.69)
The second term involving the upper components of the doublet allows one to generate also the masses
of the quarks belonging to that doublet. If we substitute Eq. (6.68) in (6.69) we and keeping only the
non-uctuating part of the Higgs eld
0
we obtain the part of the Lagrangian describing the masses
of quarks
L
Q
m
=
_
V
i
m
f
d
ij
V
j
k
_

d
m
L
d
k
R
f
u
ij

d
i
L
u
j
R
+ h.c. (6.70)
The quark masses are introduced as follows
_
V
i
m
f
d
ij
V
j
k
_

0
= m
d
k

mk
, f
u
ij

0
= m
u
i

ij
. (6.71)
To obtain for example to mass of the top quark m
t
= 170 GeV one has to choose f
t
= 0.7.
6.2 Phase transitions and symmetry restoration at high T
We now consider how the Higgs eld behaves as the temperature is varied. This is of obvious interests to
us in the context of expanding and cooling universe. For simplicity we start by consider an interacting
scalar eld
g

+ V

() = 0, V

() =
V

. (6.72)
The eld is inhomogeneous in general and can be written as the background homogeneous component
plus inhomogeneous component
(t, x) = (t) + (t, x). (6.73)
We substitute Eq. (6.73) into Eq. (6.72) and expand the potential in powers of (small) to obtain
g

+ V

( ) +
1
2
V

( )
2
= 0. (6.74)
to second order in the average of the uctuating eld (note that = 0). At the classical level the
inhomogeneous eld obeys the equation
g

+ V

( ) = 0, V

( ) m
2

( ). (6.75)
We will use here a result from quantum eld theory of scalar elds (for a derivation see any textbook
on QFT)

2
(x) =
_
d
3
k
(2)
3
)E
k
_
n
k
+
1
2
_
=
1
2
2
_
dk k
2
E
k
_
n
k
+
1
2
_
, (6.76)
where E
k
=
_
k
2
+m
2

( ). One contribution arises from vacuum uctuations and is 1/2 and the
other from thermal uctuations and is n
k
. The vacuum uctuation part is divergent as k
linearly in k. To regularized the integral we simply cut-o it at some frequency k
c
. It is easy to check
that the vacuum contribution to the third term in Eq. (6.74) can be written as a derivative of some
integral
1
2
V

( )
2
=
V


, V

=
1
4
2
_
k
c
0
dk k
2
E
k
. (6.77)
6.2. PHASE TRANSITIONS AND SYMMETRY RESTORATION AT HIGH T 111
The derivatives in Eq. (6.75) thus can be combined
g

+ V

e
( ) = 0. (6.78)
where V
e
= V +V

. The calculation of the integral in Eq. (6.77) and subsequent expansion in powers
of 1/k
c
gives
V

=
1
32
2
_
k
c
(2k
2
c
+ m
2

)E
k
c
+ m
4

ln
m

k
c
+ E
k
c
_
. (6.79)
In the limit k
c
we have E
k
c
k
c
and
V

=
1
32
2
_
k
2
c
(2k
2
c
+ m
2

) + m
4

ln
m

2k
c
_
=
1
64
2
_

_
2k
2
c
(2k
2
c
+ m
2

) m
4

ln
4k
2
c

2
. .
v

+m
4

ln
m
2

2
_

_
(6.80)
where is a characteristic scale, v

is a divergent term, which can be absorbed through a redenition


of constants. Thus the eective potential can be written as
V
e
= V +
m
4

64
2
ln
m
2

2
. (6.81)
The changes in the scale induces a term m
4

which can be reabsorbed in the denition of the


potential and therefore leads to nite renormalization of constants.
The thermal contribution can be written in terms of the integral dened by Eq. (3.31)

2
=
1
2
2
_
dkk
2
E
k
(e
E
k
/T
1)
=
T
2
4
2
I
(1)

(m

/T, 0). (6.82)


Further we have
1
2
V

( )
2
=
V
T


, V
T

=
T
4
4
2
_
m

/T
0
dxxI
(1)

(x, 0) =
T
4
4
2
I
T
(m

/T). (6.83)
Thus the total potential is the sum of the vacuum and thermal parts
V
e
= V +
m
4

64
2
ln
m
2

2
+
T
4
4
2
I
T
(m

/T). (6.84)
where the mass term is dened as m
2

= V

( ).
Now we consider the same problem, but instead of the scalar uctuations described by the eld,
we consider the vector particles described by the eld G

in the Lagrangian (6.47)


g

+V

e
( ) e
2
G

= 0. (6.85)
Here we neglect the eect of the scalar particles. Physically this corresponds to the limit where the
mass of the gauge bosons is much larger than the mass of the Higgs particles. As in the case of the
scalars there two contributions to the eective potential - one from vacuum uctuations, another from
thermal uctuations. The vacuum part can be written as
e
2
G

=
V
G

, V
G
= 3V

, (6.86)
112 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
where the factor 3 arises from the vector nature of the eld G

. The mass of the vector eld is


m
G
( ) = e . (6.87)
The thermal contribution can be computed as in the case of the scalar eld and the nal result is
V
e
= V +
3m
4
G
64
2
ln
m
2
G

2
+
3T
4
4
2
I
T
(m
G
/T). (6.88)
We can now specify some form of the potential V . As an example, we take the quartic potential and
consider the zero temperature limit
V
e
=

R
4

4
+
m
2
R
2

2
+
R
+
3e
4
32
4

4
ln

0
. (6.89)
where we substituted the value of m
G
. The renormalized coecients
R
, m
R
and
R
can be expressed
completely through the observable quantities e, m
H
, and
0
from the set of equations
V
e
(
0
) = 0, V

e
(
0
) = 0, V

e
(
0
) = m
2
H
, (6.90)
where m
H
is the Higgs mass. These are given by (for the computation see the exercise 11.1)

R
=
m
2
H
2
2
0
3, m
2
R
=
m
2
H
2
+ 2
2
0
,
R
=

2
0
4
_
m
2
H
2

2
0
_
,
3e
4
32
2
. (6.91)
It is seen that when m
2
H
< 4m
2
G
(
0
), then m
R
> 0 and the potential has a second minimum at the
origin. Furthermore if m
2
H
< 2m
2
G
(
0
), then
R
< 0 and the minimum at the origin is deeper than at
the nite
0
. Therefore we conclude that if
if m
2
H
> 2m
2
G
then symmetry will remain unbroken and the particle are massless;
if m
2
H
< 2m
2
G
the symmetry is broken and the particles obtain a mass through the Higgs
mechanism.
We do expect that the symmetry was indeed broken in the early universe, therefore this fact sets a
lower bound on the mass of the Higgs boson (within the model we discusses). This bound is known as
Linde-Weinberg bound.
For further applications it is useful to consider the case m
2
R
= 0, i.e. V

e
( ) = 0. In that case we
obtain
m
2
H
2
= 2
2
0
,
R
= ,
R
=

4
0
4
, (6.92)
and the eective potential becomes
V
e
=

4
4
+

4
0
4
+
4
ln

0
, (6.93)
which is known as the Coleman-Weinberg potential and can be used to construct inationary scenarios.
Now we include the nite temperature contribution to the eective potential. For that purpose we
need to calculate
I
T
(m
G
/T) =
_
m
G
/T
0
dxxI
(1)

(x, 0) . (6.94)
6.2. PHASE TRANSITIONS AND SYMMETRY RESTORATION AT HIGH T 113
Using the high temperature expansion
I
(1)

(x, 0) =

2
3
x
x
2
2
_
ln
_
x
4
_
+ C
1
2
_
, (6.95)
we obtain
V
e
( , T) =

T
4

4

e
3
4
T
3
+
m
2
T
2

2
+
R
(6.96)
where

T
=
m
2
H
2
2
0
+ 2ln
bT
2
e
2

2
0
, m
T
=
e
2
_
T
2
T
2
0
, ln b = 2 ln 4 2C 3.5. (6.97)
This analysis is valid in the limit m
G
= e T, i.e., in the regime where the expansion for the integral
(6.95) is valid. One may also use the low-temperature expansion of the integral to obtain the behavior
in that regime, or simply carry out the integration numerically.
6.2.1 Symmetry restoration at high temperatures
The potential (6.96) is illustrated in Fig. 6.1. It has the following features. For largest temperatures in
Figure 6.1: The eective potential dened in Eq. (6.96).
has a minimum only at the origin = 0. In this case the gauge bosons have no mass and the symmetry
is restored. Below the temperature
T
1
= T
0
_
1
6
(T
1
)
_
1/2
,
1
=
3e
3
T
1
8
T
(6.98)
there appears a second minimum at the non-zero value of the eld
1
. For smaller temperatures the
value of the minimum is larger than
1
. At the critical temperature and critical value of the eld
c
T
c
= T
0
_
1
16
3(T
0
)
_
1/2

c
=
e
3
T
c
2
T
(6.99)
114 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
the two minima are equal. The potential barrier separating these two minima has a height
V
c
=
e
12
T
4
c
4(4)
4
(T
c
)
3
, max( ) =
c
/2. (6.100)
with the maximum at =
c
/2. As the temperature is further lowered and reaches T
0
, we nd that
m
2
T
is negative and there is no minimum at the origin. Asymptotically the potential reaches its zero
temperature limit.
If the situation is realized where there is a double-well structure of the potential with two minima
the system will undergo a rst order phase transition from the state with = 0 to the state ,= 0. The
rst order phase transitions proceed by bubble nucleation of the new phase with subsequent coalescence
of the bubbles which nally occupy all the space. Such a phase transition in the early universe would
be a very violent event with strong deviations from the thermal equilibrium.
The other possibility is a second order phase transition where there is a smooth transition between
the phases. This would be the case where the potential does not have a double well structure, i.e., there
is no potential barrier separating the two minima. Because the Higgs symmetry is not broken in the
Figure 6.2: The eective potential in the case of second order phase transition; the double-well structure
is absent.
phase transition (practically there is no symmetry to be broken) one expects that there is a cross-over,
rather than a second order phase transition. For cosmological purposes one distinguishes basically the
violent phase transitions from smooth phase transitions.
One can give a qualitative answer to the question, when the phase transition is rst or second order.
For that purpose we note that the uctuations of the eld are given by
=
_

T

T
24
,
2

T
=
T
2
4
2
J
(1)

_
m

( )
T
, 0
_
. (6.101)
If the value of is smaller that the there is no mass generation and the vector bosons should be
treated as massless particles. In this case the
3
term that is responsible for the appearance of a barrier
will not be present in the eective potential (recall that m
G
= e ) and the phase transition will be
second order. In the opposite case there is a barrier and the phase transition is frist order.
6.2. PHASE TRANSITIONS AND SYMMETRY RESTORATION AT HIGH T 115
6.2.2 Electroweak phase transition in the early universe
Now we consider the electroweak phase transition in the early universe. The relevant pieces of the
Lagrangian are given by Eqs. (6.54), (6.60) and (6.69). The equation of motion of the eld that follows
upon assumption that the scalar particles can be neglected is

+ V

()
g
2
+ g
2
4
Z


g
2
2
W
+

+ f
t
t

t = 0. (6.102)
Here the coupling of the t-quarks is given by
f
t
t

t =
V
t

, V
t
= 3 4
_

I
T
(m
t
)
4
2
+
T
4
4
2
F
+
_
m
t
T
_
_
. (6.103)
where m
t
= f
t
and
F
+
_
m
t
T
_
=
_
m
t
/T
0
xI
(1)
+
(x, 0)dx. (6.104)
We have kept only the contribution from the t quarks since their coupling is largest and they dominate
the contribution from other quarks. The integral in Eq. (6.103) above is dened in Eq. (6.83) which
we reproduce for completeness
I
T
(m
t
) =
_
m
t
/T
0
dxxI
(1)

(x, 0). (6.105)


The extra factor 3 in Eq. (6.103) accounts for three colors of t-quark, factor 4 stands for the spin plus
antiparticle degrees of freedom. The calculation of the corresponding eective potential proceeds in full
analogy to the scalar potential and is then given
V
e
= V +
3
64
2
_
m
4
Z
ln
m
2
Z

2
+ 2m
4
W
ln
m
2
W

2
4m
4
t
ln
m
2
t

2
_
+
3T
4
4
2
_
F

_
m
Z
T
_
+ 2F

_
m
W
T
_
+ 4F

_
m
t
T
__
, (6.106)
where
m
Z
=
_
g
2
+ g

2

2
, m
W
=
g
2
, m
t
= f
t
. (6.107)
For most of the application it is sucient to consider the high-temperature expansion of the potential
V
e
, which is given by
V
e
( , T)

T
4

4


3
T
3
+
(T
2
T
2
0
)
2

2
+
R
. (6.108)
where the temperature-dependent coupling constant is given by

T
=
m
2
H
2
2
0
+
3
16
2

4
0
_
M
4
Z
ln
bT
2
M
2
Z
+ 2M
4
W
ln
bT
2
M
2
W
4M
4
t
ln
bT
2
M
2
t
.
_
(6.109)
where the values of the masses are taken at the minimum of the eld, i.e., M
Z
= m
Z
(
0
). The
constant are dened as ln b = 2 ln 4 2C 3.5, b
f
= 2 ln 2C = 1.4. The dimensionless constants
are given by
=
3(M
3
Z
+ 2M
3
W
)
4
3
0
2.7 10
2
(6.110)
=
M
2
Z
+ 2M
2
W
+ 2M
2
t
4
2
0
0.3. (6.111)
116 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
where the temperature of the phase transition is given by
T
2
0
=
1
2
_
m
2
H

3(M
4
Z
+ 2M
4
W
4M
4
t
)
8
2

2
0
_
= 1.7[m
2
H
+ (44 GeV)
2
]. (6.112)
This temperature explicitly depend on the unknown Higgs mass. As a next step we would like to study
the temperature dependence of the eective potential. First, in analogy to the scalar eld the masses
can be taken as real is the value of , which corresponds to the breaking of symmetry is outside of
the uctuation region of the eld at any given temperature. The terms corresponding to the t quarks
should be kept only if the temperature is much higher than the their masses m
t
( ) = M
t
/
0
T.
Within the range
T
M
Z,W

0
> >
T
M
t

0
(6.113)
the t quarks should be omitted as they are non-relativistic, whereas the W, Z-bosons are relativistic.
The analysis of the eective potential shows the following features. At very high temperatures
T T
1
=
T
0
[1 (
2
/4(T
c
))]
1/2
(6.114)
the potential has one minimum at the origin ( = 0) the symmetry is intact and the gauge bosons and
fermions are massless. Below the temperature T
1
the second minimum appears and at T = T
c
, dened
by
T
c
=
T
0
_
1 2
2
/9
T
c
, (6.115)
the depth of the minimum, which is dened by

c
=
2T
3(T
c
)
(6.116)
becomes of the same as that at = 0. At this point the transition to the phase with the broken
symmetry becomes possible. We have seen that the barrier at ,= 0 is real if the location of the
minimum is larger than the uctuation region of the eld, i.e.,
c
/2 > T/5. Using Eq. (6.116) we obtain
the condition on the coupling constant (T
c
) <
_
8/3 for the strong rst order phase transition to
occur. This condition can be translated upon a bound on the Higgs mass via Eq. (6.109), which implies
m
H
< 75 GeV. Thus, if the Higgs particle is light the electroweak phase transition will be rst order (for
m
H
50 GeV the critical temperature is of the order T
c
90 GeV. However, the experimental lower
bound on the Higgs mass if m
H
> 114 GeV. Therefore, the analysis above implies that the breaking of
the electroweak symmetry will be smooth. For large Higgs mass one can neglect the barrier inducing

3
-term in Eq. (6.108). In this case there is only one minimum of the potential, which at T < T
0
is
located at

2
c
=

T
(T
2
0
T
2
), (6.117)
and the symmetry is broken. Thus, we conclude that the data indicates that the Higgs mass is large
and therefore the cosmological electroweak transition is smooth. The value of the critical temperature
depends on the mass of the Higgs boson, e.g., if m
H
120 GeV then T
0
170 GeV or if m
H
180
GeV then T
0
240 GeV.
6.3. PHYSICS BEYOND THE STANDARD MODEL 117
6.3 Physics beyond the Standard Model
The experimentally veried Standard Model is expected to work up to energies of the order of 100 GeV.
Beyond this energy we expect the Standard Model to be replaced by a more general theory, whose low
energy limit is the standard model. Indeed the standard model contains many ne-tuned constants
whose specic values have no explanation within the standard model.
Cosmology needs to answer a number of questions, whose roots lie in the physics of the early universe.
The most prominent ones are the nature of the dark energy, the baryon asymmetry and those problems
that are solved by the inationary theory.
6.3.1 Grand-unication theories
Grand unied theories combine the electroweak and strong interactions into a single force at energy
much larger than those encountered in the Standard Model. The idea is based on the running of the
coupling constats of the two theories. The strong coupling constant runs according to Eq. (6.28) On
the other hand the weak coupling constant varies with the momentum transfer as

w
(q
2
) =
g
2
4


0
w
1 + 0.265
0
w
ln(q
2
/q
2
0
)
. (6.118)
where the low-energy value at q
0
100 GeV is
0
w
=
w
(q
0
) = 1/29. It can be see that the condition
that the both coupling constant of the same of order of magnitude denes the unication energy as
follows

s
(q
2
)
w
(q
2
) q 10
17
GeV. (6.119)
In the group theory terminology the idea of grand unication is the combination of the groups governing
Figure 6.3: The inverse running coupling constants in the Standard Model meet at about the Planck
scale
the strong and weak interactions into a larger group. The group of the strong interactions is SU(3),
118 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
whereas that of the weak interactions is SU(2) U(1). These can be combined minimally into a larger
group SU(5). To allow for a unication one needs the ne structure constant to meet the other coupling
constants at the same point. Within the standard model these three couplings do not meet at the same
point. The discovery of the neutrino masses rules out the models based on the SU(5) symmetry, because
it can incorporate the right-handed neutrinos in the fermionic content. Indeed the SU(5) group has
5
2
1 = 24 generators. The corresponding gauge bosons are 8 gluons, 1 photon, and 3 gauge bosons of
the electroweak theory. There remain 12 bosons form two charged colored triplets (whose elements are
labeled by the index i = r, b, g)
X
4/3
i
, Y
1/3
i
. (6.120)
The upper index corresponds to the charge of the bosons. After the SU(5) symmetry breaks down to
SU(3) SU(2) (1) these bosons acquire mass of the order of the breaking energy scale 10
17
and the
transitions from the dierent sectors are strongly suppressed. At larger energies one expects processes
which connect these two groups, i.e., transitions from quarks to leptons and visa versa. In the SU(5)
the baryon number is not conserved (an example of such a process is X ql (end baryon number is
2/3 2). On the other hand the number of baryons minus the number of leptons B L is conserved
and models which generate baryon asymmetry from non-conservations of BL number can not operate
with the SU(5) theory.
Thus, we conclude that we need to look for larger symmetry groups. We expect to be able to solve
the following issues
including the right-handed neutrinos in the theory, as there are much mode fermions present in
the model
there theories violate BL number and thus allow one to generate baryon asymmetry from this.
the coecients of the renormalization group equation for the running of coupling constant depend
on the fermion content and this can be tuned so that all the coupling constant meet at the same
point.
the new particles can serve as candidates of dark matter in the universe.
The candidate theories that include larger that SU(5) group are based on the following groups
SO(10), SO(14), SO(22), E
6
, E
7
E
8
, etc. (6.121)
It is not possible to distinguish between the competing theories as they operate at very high energies
of order of 10
17
GeV and in the low-energy limit always reduce to the standard model.
6.3.2 Supersymmetric theories
Supersymmetry states that every fermion has its superpartner which is a boson and, visa versa, every
boson have its fermionic supersymmetric partner. The superparticles are paired in so-called supermu-
tiplets, which contain both bosons and fermions. (This is in contrast to the standard model, where the
multiplets contain either fermions or bosons.)
Supersymmetry transformation cannot be decoupled from space transformation (the bosons and
fermions transform dierently under the Poincare group). The algebra of supersymmetry generators Q
closes only if one includes the generators of space and time translations. Supersymmetry transformations
at the local level require inclusion of curvature of the space and time. Local supersymmetry is called
supergravity.
6.3. PHYSICS BEYOND THE STANDARD MODEL 119
The extension of the standard model to the supersymmetry is not unique. First, supersymmetry
can be local or global. Supermutiplets can include a many pairs of fermions. For example N = 8
supergravity contains in the supermultiplet left and right-handed particles with spins 0, 1/2, 3/2, and
2.
Unfortunately all known fermions are pairs of unknown bosons and all known bosons are pairs to
unknown fermions. Therefore in supersymmetric theories there at least twice as many particles as in
the standard model. The absence of the observed superpartners implies that supersymmetry must be
broken at scales larger than that reached by current day accelerators, otherwise we had observed bosons
and fermions of the same mass.
Figure 6.4: The running coupling constants in the Minimal Supersymmetric Standard Model meet at
the same point.
The minimal supersymmetric extension of the Standard Model (MSSM) every lepton or quark has
a supersymmetric boson partner of spin 0 called slepton (squark). Every gauge boson have a fermionic
superpartner with spin 1/2, called gaugino (gluino is the superpartner of gluon, winos and zinos are
superpartners of gauge bosons of the electroweak theory). The Higgs particle is accompanied by a
higgsino. The lightest neutral combination s of all inos, called the neutralino, must be stable. If the
scale of breaking of supersymmetry was at the electroweak scale the neutralino would interact weakly
with the standard model particles. Thus, neutralino is an ideal candidate for cold dark matter. Finally
one should mention the fermionic counterpart of the graviton gravitino, which is a fermion of spin
3/2; it could also serve as candidate for dark matter.
One interesting property of supersymmetric theories is that the degrees of freedom of any particle
and ist supersymmetric partner is equal. For example, a spin-1/2 fermion has a bosonic superpartner
described by a complex scalar eld. The energy of vacuum uctuations of fermions and boson then
have the same magnitude but opposite sign, and therefore exactly cancel each other. In cosmology this
means that cosmological constant is exactly zero if the supersymmetry is unbroken (this is of course
not the case). If the supersymmetry is broken at the TeV scale the mismatch in the vacuum energies of
the superpartners can lead to a cosmological constant which is 10
64
in Planck units, which however
is much larger than the observational limit.
120 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
Finally, in supersymmetric theories the degrees of freedom combine in such a way that the run-
ning coupling constants meet at a single point. This impressive results hints to the correctness of
supersymmetric theories.
6.3.3 Dark matter
The origin of dark matter is not understood yet and, therefore, there are many potential candidates.
One class of dark matter candidates are thermal relics which originate via decoupling from matter.
Furthermore, if these relics where relativistic at the moment of decoupling these belong to the class
of hot dark matter (an example is neutrino). Those relics which at the decoupling are non-relativistic
belong to the class of cold dark matter (an example are neutralinos).
The non-thermal relics are, for example, condensates of weakly interacting massive scalar elds.
An example of such particle is axion, which despite of its small mass has a zero momentum in the
condensate and therefore belongs to the cold non-thermal relics.
Hot relics
Hot relics decouple from matter when their relativistic. Their number density decreases as the volume
expands, therefore n

a
3
. On the other hand the entropy density of the rest matter scales also as
a
3
so that the ratio n

/s = const throughout the evolution of the universe. We use the fact that the
entropy to density ratio is conserved. Suppose the relics have mass m

and zero chemical potential.


Their contribution to the cosmological parameter is given by

h
2
75
=
g

_
m

19 eV
_
, (6.122)
where g

is the total number of degrees o freedom of the hot relics and g

= g
b
+ (7/8)g
f
is the
eective number of bosonic and fermionic degrees of freedom at the freeze-out of all particles which
later annihilate and convert their energy to photons.
Let us give an example. Suppose the left-handed neutrinos constitute the hot relics. Then, g

= 6
(three families of neutrinos plus their anti-neutrinos). The decoupling temperature of neutrinos is
about O(1) MeV. Those particles that will annihilate later and therefore contribute to g

are electrons,
positrons and photons. Thus g

= 2(+(7/8) 4) = 5.5. Therefore, if each of the neutrinos would


contribute 17 eV, then we would obtain

=
total
, that is the neutrinos would close the universe.
However, observations tell us that the dark matter contributes about 30% of the density of the universe.
Therefore the sum of the masses of neutrinos should be smaller than 15 eV, or if they are equal then
m

< 5 eV.
However, the models which simulate the evolution of the universe with dominant contribution from
hot relics cannot explain the inhomogeneities of the universe; these are washed out by the free streaming
of this matter on all scales up to the horizon scale. The large scale of the universe can not be explained
if the dominant part of the dark matter are hot relics.
Thermal cold relics
The cold relics decouple from the matter at temperatures that are much less than their mass m

Therefore their number density is exponentially suppressed compared to the number of photons. An
estimate of the number of thermal cold relics one can simply equate the annihilation rate of relics to
6.3. PHYSICS BEYOND THE STANDARD MODEL 121
the Hubble expansion rate
n

=
_
8
3
90
_
1/2
g
1/2

T
2

(6.123)
where the left hand side is the standard classical rate equation. Alternatively
n

_
8
3
90
_
1/2
g
1/2

T
2

. (6.124)
The eective number of degrees of freedom g

counts all the particles which are relativistic at T

. The
number density of relics can be obtained from their equilibrium distributions by setting = 0, so that
n

T
3

(2)
3/2
x
3/2

e
x

, (6.125)
where x

= m

/T

. From the ratio of the last two equations we obtain


x

ln(0.038g

g
1/2

) 16 + ln
_
g

g
1/2

10
38
cm
2
_
_
m

GeV
_
_
. (6.126)
From the condition that the relic should be cold x

> 3 one obtains a bound on the product

>
10
3

/g

. The present day energy density of the relics can be estimated as

s
0
s

=
2
g

_
T
0
T

_
3
(6.127)
where T
0
= 2.73 K is the temperature of the background radiation, s
0
/s

is the ratio of he present day


entropy density of radiation to the total entropy density at freeze-out of all the components of matter
(eective degree of freedom g

) which later transfer their entropy to radiation. Finally substituting the


density of relics from Eq. (6.123) one nally nds

h
2
75

x
3/2

_
3 10
38

_
. (6.128)
Since x

depends only logarithmically on the mass of the relics, the main uncertainty in Eq. (6.127)
arises from the annihilation cross-section at the decoupling.
The leading candidates for cold relics are weakly interacting massive particles with masses in the
range 10 GeV m

1 TeV with the cross-section of the same of order of magnitude as in Eq. (6.128).
For these particles x

20 and therefore they can easily contribute the needed 30% to the total density
of the universe and thus can constitute the main component of dark matter. One example of such
particles is the lightest supersymmetric particle (most likely this the neutralino with large admixture
of bino and photino.)
Nonthermal relics
The nonthermal relics are never in equilibrium, since they are out of the equilibrium at all times.
Therefore, their contribution to the thermodynamical quantities depends on the dynamics of the system.
We consider here a specic example of a homogeneous condensate of massive scalar particles and neglect
their interactions with other particles in the environment. The dynamics of the scalar eld is described
by

+ 3H

+ m
2
= 0. (6.129)
122 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
We now introduce conformal time dt = ad and rescaled variable u = a. The rescaled equation of
motion can be written now as
u

+a
2
_
m
2

a
3
_
u = 0, (6.130)
where as before the prime denotes derivative with respect to the conformal time. In the case where
a

/a
3
H
2
m
2
the rst term in the bracket can be neglected and the equation motion has a solution
u a
_
C
1
+ C
2
_
d
a
2
_
. (6.131)
where C
1
and C
2
are integration constants. The second term describes a decaying mode, where as the
rst term is the dominant constant mode

1
= C
1
. (6.132)
We can conclude that when the mass of the particle is much smaller than the Hubble constant the
scalar eld is static and its energy density can be read-o from the Lagrangian as

=
1
2
_

2
+ m
2

2
_
. (6.133)
The energy density thus depends only on the mass, and if this is constant, then it can be interpreted
as a cosmological constant.
In the opposite limit H
2
m
2
we can neglect the second term in Eq. (6.130). The solution of the
equation of motion is then

1

ma
3
sin
__
mdt
_
. (6.134)
This we see that once the Hubble constant drops below the mass of the particle the condensate starts
to oscillate. The energy density which was constant now decreases as ma
3
, so that the condensate
essentially behaves as matter with zero pressure. The current energy density of scalar eld is

m
0
_

m
_

s
0
s

g
3/4

m
0

2
in
T
3
0
, (6.135)
where all the quantities label with 0 are present day values, whereas the quantities label with * refer
to the moment of the transition from the constant to oscillating solution (H m). The initial value of
the eld is
in
. To make an estimate of the contribution from the scalar eld to the energy density of
the universe we will assume that the mass stayed roughly constant since the oscillating solution sets in
(m
0
= m

). Then, the energy density is given by

h
2
75

g
3/4

_
m
0
100GeV
_
1/2
_

in
3 10
9
GeV
_
2
. (6.136)
The two parameters can be adjusted to t the required amount of the energy density of the dark matter.
Axions
Axions are an example of relic particles. The idea of axions arises from the conjecture that the vacuum
of the QCD is a so-called vacuum which is dened as a superposition of several vacua with dierent
winding numbers
[ =

e
in
[n, (6.137)
6.3. PHYSICS BEYOND THE STANDARD MODEL 123
where is an arbitrary constant to be obtained from experiment. Such a assumption induces an addi-
tional contribution to the eective Lagrangian
s
tr(F

F). Such term generally violates the following
symmetry transformations CP, P, and T. Experimental constraint coming from the measurements of
the neutron dipole moment constrain the magnitude of the parameter < 10
10
.
In the model by Peccei and Quinn the -parameter is replaced by dynamical axion eld. This is
generally described by a complex scalar eld = exp(i

), which has chiral U(1) symmetry. Below


a certain scale f this symmetry is broken and the eld acquires the expectation value f. The axion
eld is dened as the product a = f

. It is important to note that the symmetry is global and can


not be absorbed in gauge elds. Through interaction of the axions with gluons these acquire a mass of
order of
m
a

m
u
m
d
m
u
+ m
d
m

f

6 10
6
f
GeV
eV, (6.138)
where m
u
and m
d
are the masses of the light quarks, m

= 140 meV is the pion mass and f

= 93 MeV
is the pion decay constant. This is the key fact that we need to estimate the contribution of axions to
the energy density of the universe. Quite generally the axion mass depends on the temperature as
m
a
(T) m
a
_

QCD
T
_
4
. (6.139)
From the condition m
a
(T

) = H

we obtain the temperature at which the axions become cold relics


T

g
12

m
1/6
a

2/3
QCD
, m

g
1/3

m
1/3
a

4/3
QCD
. (6.140)
This value of m

can be used in (6.133) with a


in
= f
in
to obtain the estimate

a
h
2
75

_
6 10
6
m
a
_
7/6

2
in
, (6.141)
where
in
1. This axions of mass m
a
10
5
eV are sucient to provide a substantial contribution
to mass density of the universe. In some models (such as the inationary model) it is possible to have

in
1 in which case similar contributions can be achieved with much smaller masses of axions.
The theories beyond the Standard Model provide one with more candidates for dark matter; these
include for example the gravitino, axino etc.
6.3.4 Baryogensis
The observed universe around us is asymmetric with respect the baryon number, i.e., there are many
more baryons than anti-baryons. Although we observe anti-baryons in experiments on colliders, there
are no observations of anti-matter on larger scales. There is a relative excess of baryons over anti-
baryons B (n
b
n
b
)/s 10
10
. It is needed to explain the observed abundances of light elements
and the observed spectrum of CMB uctuations (to be discussed later). In principle one can put a
baryon asymmetry by hand as an initial condition. However, in the inationary scenario, any initial
asymmetries are erased by the inationary expansion and one needs a generation mechanism. The key
features of any dynamical mechanism of generation of baryogenesis must have the following features
baryon number violation
C and CP violation
124 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
departure from equilibrium
These criteria are known as Sakharov criteria. The rst condition is obvious as one needs to violate
the baryon conservation at some point to generate the present day observed asymmetry. If the baryon
number is not conserved in must vanish in thermal equilibrium, i.e., we need a non-equilibrium situation
to maintain some baryon number. The second condition insures that the rates of decays of particles
and anti-particles are dierent. Indeed, if this condition is not fullled, the baryons and anti-baryons
will be produced in equal amounts.
These conditions can not be fullled in the Standard model; for example, although weak interactions
violate CP the electroweak phase transition is most likely a crossover without any violent events and
deviations from equilibrium so that the third condition is not fullled.
Currently accepted scenarios of baryogenesis are all based on the ideas that lie beyond the standard
model. We review below some of the most popular ons.
Grand Unied Theories
As we have already discussed on the example of SU(5) theory there are gauge bosons that mediate the
interactions between the quark and lepton sectors. The possible channels of decay for such a gauge
boson are
X qq, B = 2 2/3, (6.142)
X q

l, B = 1/3. (6.143)
Similarly,

X q q, B = 2 (2/3), (6.144)

X ql, B = 1/3. (6.145)


The total decay rates for both bosons are equal by CPT invariance. However, in each channel they
could be dierent. If C and CP are violated then the transition rates
(X qq) r ,= (

X q q) r (6.146)
are not the same. The scenario of baryogenesis should work as follows. At very high temperature the
X and

X bosons are in thermal equilibrium and their abundances are the same n
X
= n
X
n

. Now
suppose that the temperature drops below the mass of the X boson m
X
and these bosons drop out
of equilibrium at concentration = n
X
/s. As given by the third criterion of Sakharov only out of
equilibrium decays are important. The mean baryon number in the decays can be estimated as
B
X
= (2/3)r + (1/3)(1 r), B
X
= (2/3) r + (1/3)(1 r) (6.147)
Thus the resulting baryon asymmetry is
B = (B
X
+B
X
) = (r r) = . (6.148)
The parameter characterizes the amount of CP violation. The term arises from higher order
perturbation and in the SU(5) this occurs at 10th order. Consequently, the order of magnitude of such
asymmetry turns out to be much smaller than the required value 10
10
. As mentioned above one non-
attractive feature of the SU(5) model is that B L is conserved and the baryon asymmetry generated
under this conditions will be washed out in any inationary stage. The SU(10) model does not have
this feature and the required magnitude of asymmetry can be obtained. There are problems for grand
unied theories related to the ination; one believes that ination ends below the electroweak energy
scale, which in turn implies that X bosons were never in equilibrium.
6.3. PHYSICS BEYOND THE STANDARD MODEL 125
Baryogenesis via leptogenesis
The key idea here is the following. In topological phase transitions in the early universe B + aL 0
whereas BL = Const, where a is a constant which is equal to 28/51 in the Standard Model. Suppose
initially we have B
i
= 0 and L
i
,= 0 (where i refers to the initial state). Using the conditions above we
obtain for the nal baryon number
B
f
=
a
1 + a
L
i
. (6.149)
This formula shows that if by some process we can generate nite lepton number L
i
the topological
phase transitions will do the work of converting this into nite baryon number - hence the name -
baryogenesis via leptogenesis.
Thus we need to concentrate below on the problem how the lepton number can be generated. One
possibility is that it is generated by lepton number non-conserving out-of-equilibrium decays of heavy
neutrinos.
We will now assume that there exist right-handed neutrinos and the neutrino masses are described
by the Lagrangian
L
()
=
1
2
(
L

c
R
)
_
0 m
D
m
D
M
__

c
L

R
_
+ h.c. (6.150)
The mass matrix in this Lagrangian term is not diagonal. We can turn to a new basis, where the elds
are dened as

L
+
c
L
N
R
+
c
R
. (6.151)
In this new basis the Lagrangian has the form
L
()
=
1
2
_


N
c
_
_
m

0
0 m
N
__

c
N
_
+ h.c. (6.152)
where the mass matrix is diagonal with elements dened as
m

=
m
2
D
M
, m
N
M. (6.153)
The fermions entering the mass matrix are so-called Majorana fermions ( =
c
, N = N
c
), i.e., the anti-
particles and particles are equivalent. If m
D
m
t
170 GeV and M 3 10
15
GeV then m

= 10
2
eV. This value of neutrino mass is in agreement with those found in experiments on neutrino oscillations.
This way of obtaining very small masses is called the seesaw mechanism.
The heavy neutrinos can decay into a lepton plus Higgs particle pair:
N

l +

, N l + , (6.154)
which violates the lepton number. In addition in the case of three generations of neutrinos the mass and
avor eigenstates are connected by the complex Kobayashi-Maskawa matrix. This has the consequence
that the couplings in the decay rates are complex and therefore violate the CP symmetry. As a result
the rates of these reactions can be written in general as
(N l) =
1
2
(1 + )
total
(N

) =
1
2
(1 )
total
, (6.155)
where the parameter 1 measures the amount of CP violation. We assume at some early point
of the evolution of the universe the neutrino concentration freezes out the decays above occur under
conditions of non-equilibrium (i.e. the third Sakharov condition is satised).
126 CHAPTER 6. PHASE TRANSITIONS IN THE EARLY UNIVERSE
There exist other ways to solve the problem of baryogenesis. One is the so called Aeck-Dine scenario
which is based on the supersymmetric theories where there could appear B, L and CP violating terms
in the Lagrangian related to the scalar superpartners of quarks and leptons. There are further scenarios
in the literature which we will not discuss.
Chapter 7
Ination
7.1 Motivating inationary theories
In classical mechanics the trajectories of particles are completely determined if we know the initial
positions of all particles and their velocities. The state of a mechanical system is completely determined
thereafter on the basis of the laws of mechanics. In cosmology we are interested in the initial energy
density distribution and velocities. When we attempt to determine these quantities from observations
we are stumbled over several problems. Let us address them.
Horizon problem (homogeneity and isotropy problem)
We have said that the universe is homogeneous and isotropic over scales larger than that about 100
Mpc. The maximal scale for which this statement is valid is the present horizon scale which is the age
of the universe times the speed of light; this scale is about l
0
= ct
0
10
28
cm. If the current value of
the scale factor is a
0
and at some moment its value (which we will call initial) was a
i
then the size of
the horizon was then a
i
/a
0
times smaller.
Assume that the expansion process does not aect the homogeneity and isotropy features of the
universe. The size of the homogenous/isotropic patch at the moment t
i
then should be
l
i
l
0

a
i
a
0
, l
i
ct
0
a
i
a
0
. (7.1)
The size of the causal region on the other hand is l
c
ct
i
. Comparing these two scales we obtain
l
i
l
c

t
0
a
i
t
i
a
0
. (7.2)
Let us now specify some initial value; we will assume that the initial time could be identied with the
Planck time t
i
t
PL
10
43
s, when the temperature is T
PL
10
32
K. Since the current temperature
is given by the Cosmic Microwave Background temperature and is O(1) we obtain an estimate
a
i
a
0

T
0
T
PL
10
32
. (7.3)
Then taking for t
0
10
17
we can estimate the ratio (7.2) as
l
i
l
c
10
32
10
17
10
43
10
28
. (7.4)
127
128 CHAPTER 7. INFLATION
We see that the elementary extrapolation back implies a size of the universe that was 10
28
times
larger than the size of the causal region. This translates into (10
28
)
3
causally disconnected volume
elements in each of which the energy density was distributed homogeneously with fractional uctuations
not exceeding 10
4
. Therefore, we come to an unnatural conclusion that the matter was distributed
homogeneously in a lagre number domains which had no causal contact.
If we assue that the scale factor grows as some power of time a t
n
, then a/t a
n1
and at the
same time a = na
n1
, so that a/t a. Thus, we immediately nd that
l
i
l
c

a
i
a
0
. (7.5)
Thus we conclude that the mismatch between the causal region and back-extrapolated size of the
universe is given by the ratios of the expansion rates of the universe at the initial time and at present.
Since the gravity always acts to contract the expansion one may conclude that homogeneity scale l
i
was
always larger than the causality scale l
c
. This problem is thus essentially a horizon problem.
Flatness problem (initial velocities problem)
Apart from the positioning the matter in a smooth way, we also need to specify the velocities so that
the Universe can be propagated in time. The initial velocities should obey Hubble law, otherwise their
relaxation to the Hubble law could spoil the homogeneity of the Universe.
Consider a spherical region of matter whose energy contains the kinetic energy of Hubble expansion
E
k
and gravitational energy of attraction E
g
. The total energy is conserved so that we can equate it
at some initial time and the present time
E = E
k
0
+ E
g
0
= E
k
i
+ E
g
i
= Const. (7.6)
Obviously E
k
v
2
a
2
, where v is the velocity, and we nd that
E
k
i
= E
k
0
_
a
i
a
0
_
2
. (7.7)
Now we can estimate
E
E
k
i
=
E
k
0
+ E
g
0
E
k
i
=
E
k
0
+ E
g
0
E
k
0
. .
O(1)
_
a
0
a
i
_
2
10
56
, (7.8)
where in the last step we used the estimate (7.4) and the relation (7.5). Thus the velocities need to
be tuned to this high accuracy to have the present evolution of the universe; otherwise the universe
will either collapse or become empty too fast. This problem is known also as the problem of initial
velocities.
These statements can be put somewhat dierently by using the Friedman equation in the following
form
(t) 1 =
k
(Ha)
2
. (7.9)
or

i
(t) 1 = (
0
1)
(Ha)
2
0
(Ha)
2
i
= (
0
1)
a
2
0
a
2
i
10
56
. (7.10)
Thus we see that the cosmological parameter needs to be very close to the unity, i.e., correspond to at
universe to very high accuracy. This problem therefore is also referred as the atness problem.
7.1. MOTIVATING INFLATIONARY THEORIES 129
Initial perturbations problem
A further problem is the origin of primordial inhomogeneities needed for the large scale of the universe
to develop. These should be of the order of / 10
5
on galactic scales. This problem will be
addressed later.
7.1.1 What is ination?
If we look back at the problems we have encountered, we will see that they all originate in the large
value of a
i
/ a
0
. Because gravity is attractive this ratio will be always larger than unity. That is, gravity
tends to decelerate the expansion of the universe, i.e., the acceleration is always larger at any earlier
time than at a given time of interest.
It turns out that the problem can be avoided if at some stage of its evolution the gravity acted as an
eective repulsive force, which would accelerate the expansion of the universe. Ination arises exactly
in this context: it can be view as a stage of accelerated expansion where the gravity acts as a repulsive
force.
Thus, the old picture of constantly decelerating Friedman universe is replaced with a new picture
where during an inationary phase a > 0, i.e., the expansion is accelerated. However, some of the
predictions of the Friedman universe agree well with observations (such as the nucleosynthesis) therefore
the ination must start early enough and also end early enough in order not to spoil this feature of
Friedman universe. The theory of primordial uctuations also restricts the stage of ination to times
earlier than 10
34
s. A part of successful inationary theory is the graceful exit which requires a smooth
transition to the Friedman stage of evolution. The ination thus produces a whole observable universe
from a small homogenous patch even if the universe was highly inhomogeneous outside of this patch.
Ination solves the problem of homogeneity of the universe by assigning an accelerated expansion to a
Figure 7.1: The big-bang is followed by a stage of ination before the universe enters via a graceful exit
into a stage of Friedman evolution.
small ball of the universe which blows to the size of a horizon at the end of ination. This processes is
independent of the conditions that are prevailing outside of the ball.
130 CHAPTER 7. INFLATION
Can the ination also damp the initial inhomogeneities? The answer is yes. Let us assume we have
an inhomogeneity / on the scale of the order of H
1
i
and it is of order of unity
_

_
t
i

(H
i
a
i
)
1
O(1), (7.11)
where is the derivative with respect to comoving coordinates. At some later time t t
i
the contri-
bution of this same inhomogeneity would be
_

_
t
i

[H(t)a(t)]
1
O(1)
a
i
a(t)
. (7.12)
where it is assumed that / does not change substantially over the period of the time. We see that if
a(t) > a
i
for t > t
i
, i.e., for an accelerated universe the perturbations are damped.
Thus, the ination also destroy large initial inhomogeneities and produces a homogeneous and
isotropic domain which blows to the size of the universe. The observations of the CMB require that the
variations of the energy density on the present horizon scale do not exceed 10
5
. This requires that the
ratio of the initial to present date accelerations be a
i
/ a
0
1. We rewrite Eq. (7.10)

i
(t) 1 = (
0
1)
(Ha)
2
0
(Ha)
2
i

0
= 1 + (
i
1)
_
a
i
a
0
_
2
. (7.13)
We see that if [
i
1[ 1, then

0
= 1, (7.14)
to a very high accuracy. This is a robust prediction of the inationary theory, namely, that the total
energy density of all components of matter irrespective of their origin should be equal to the critical
density today.
7.1.2 Equation of state for ination
To understand how the universe undergoes a stage of accelerated expansion let us look at the Friedman
equation
a =
4
3
G( + 3p)a. (7.15)
We see that the answer is in the sign of the factor + 3p. If it is positive then a < 0 and gravity
decelerates the expansion. The universe can undergo a stage of accelerated expansion with a > 0 if
we could arrange of an equation of state that leads to + 3p < 0. One special case that we already
encountered in that of the cosmological constant for which p

and therefore +3p = 2

< 0.
Then solutions of the Einsteins equations in this case correspond to the de Sitter universe which we
discussed in Chapter 1. In particular we found that the de Sitter universe expands exponentially,
a exp(H

t) for t H
1

. However, the proper de Sitter solution fails to produce a graceful exit to


a Friedman stage. To obtain the latter feature one needs to allow for time-dependent Hubble constant.
We now turn to some general conditions that must be satised by inationary models. We start
with the equation for the acceleration
a
a
= H
2
+

H. (7.16)
7.2. FIELD THEORY MODELS OF INFLATION 131
The requirements are a > 0 during the ination and a < 0 at the transition to the Friedman universe
(graceful exit). Obviously H
2
is always positive, so that

H needs to become negative at the graceful
exit. Also we need that

H H
2
. A crude estimate for these two terms to be equal is
t
f

H
i
[

H
i
[
, (7.17)
where i refers to the initial time.
The ination should last long enough so that the small seed expands to the size of the universe. We
now rewrite the condition a
i
/ a
0
< 10
5
as
a
i
a
f
a
f
a
0
=
a
i
a
f
H
i
H
f
a
f
a
0
10
5
. (7.18)
We have obtained that a
f
/ a
0
10
28
(the size of the causal domain). Therefore using this in the above
we obtain
a
f
a
i
> 10
33
H
i
H
f
. (7.19)
Let us assume that the Hubble time variations can be neglected. Then
a
f
a
i
> 10
33
. (7.20)
On the other hand, taking into account that the inationary phase is similar to the de Sitter expansion
we obtain
a
f
a
i
exp(H
i
t
f
), t
f
H
1
i
log
a
f
a
i
75H
1
i
. (7.21)
Using the Friedman equations with k = 0 we can obtain a rough estimate on the equation of state
( + p)
i

i
10
2
. (7.22)
This brings us to the conclusion that before the ination the equation state must be very close to the
vacuum one to better than 1 per cent.
7.2 Field theory models of ination
We now turn to a discussion of a eld theory model in which the requirements above can be realized. The
key idea is the introduction of a scalar eld inaton. A homogeneous classical eld is characterized
by energy density and pressure
=
1
2

2
+ V (), p =
1
2

2
V (). (7.23)
The spatial derivative are neglected as the matter is homogenous through the ination. From the sum
of Eqs. (7.23) we see that
p + =

2
p = +

2
. (7.24)
Thus, we would have a vacuum type equation of state if the kinetic term vanishes

0. This implies
that during the inationary stage the kinetic term should be small. We now study the equations of
132 CHAPTER 7. INFLATION
motion of the scalar massive eld and the conditions under which the solutions admit an inationary
stage. The equation of motion of the eld is given by

+ 3H

+ V
,
= 0, (7.25)
where V
,
= V/. To nd the evolution of the eld in the expanding universe we need to supplement
it with the Friedman equation
H
2
=
8
3
_
1
2

2
+ V ()
_
(7.26)
where G = 1 and k = 0 (at universe). Eq. (7.25) and (7.26) form a couple set of equations that need
to be solved self-consistently.
We now consider the special case of the eld
V () =
1
2
m
2

2
. (7.27)
Substituting Eq. (7.26) into (7.25) we obtain a closed equation for the eld

+
_
12(

2
+ m
2

2
)
_
1/2

+m
2
= 0. (7.28)
This equation is second order and non-linear. It does not contain any time-dependence. It can be
reduced to a rst order equation for

(). We note rst that trivially

d
. (7.29)
Then we can rewrite Eq. (7.28) as

d
=
_
12(

2
+ m
2

2
)
_
1/2

+ m
2
. (7.30)
The numerical solutions of this equation are shown in Fig. 7.2. One can see that there are attractor
Figure 7.2: Numerical solution to Eq. (7.30).
7.2. FIELD THEORY MODELS OF INFLATION 133
solutions in this gure, i.e. all the solutions tend to the same point. Of special interest is the solution
corresponding of ultra-hard equation of state, which corresponds to the limit

V . In this case we
nd immediately from (7.23) that the equation of state is ultra-hard p +. In Eq. (7.30) we can
neglect m compared to

we obtain
d

12



= C exp
_

12
_
(7.31)
where C < 0 is a constant of integration. Solving the last equation for (t) we obtain
= C
1
(12)
1/2
ln t. (7.32)
Substituting this result in (7.26) we obtain
H
2

_
a
a
_
2

1
9t
2
. (7.33)
Thus we see that a t
1/3
and a
6
. This solutions are exact for massless elds. From the solutions
we see that the initial large value of

is quickly damped within time that is much shorter than that
for the eld itself.
We see that the diagram has an attractor solution. To nd that solution we put d

/d = 0, which
gives us the solution

atr

m

12
. (7.34)
which can be integrated to obtain

atr

i

12
(t t
i
)
m

12
(t
f
t), (7.35)
where t
i
is the time when the trajectory joins the attractor, t
f
is the moment when the = 0. Thus
the scalar eld decreases linearly with time during the inationary stage. The equation of state has the
form
p +
m
2
12
. (7.36)
Thus when the additional term in Eq. (7.36) becomes of the order of the potential term the ination is
over, i.e.,
m
2
12
m
2

2
. (7.37)
This occurs when 1 (in Planckian units). An important quantity is the value of the scale factor
during the ination. We start with Eq. (7.26) where we neglect the kinetic term
H(t)
_
4
3
m(t). (7.38)
Upon substituting (7.35) herein and integrating we obtain
a(t) a
f
exp
_

m
2
6
(t
f
t)
2
_
a
i
exp
_
H
i
+ H(t)
2
(t t
i
)
_
, (7.39)
134 CHAPTER 7. INFLATION
where the i index refers to the initial values of the scale factor and Hubble constant. From the solution
(7.35) we also obtain the duration of the ination
t = t
f
t
i

12

i
m
. (7.40)
The increase in the scale factor is
a
f
= a
i
exp
_
2
2
i
_
. (7.41)
Now we wish to obtain the largest possible increase of the scale factor during the ination. Consider a
eld of mass 10
13
GeV. Since the Planckian energy domain starts at 10
19
GeV, the maximal value of
the eld which will allows us to stay at sub-Planckian energies is 10
6
. Therefore, we nd that
_
a
f
a
i
_
max
= exp
_
10
12
_
. (7.42)
Thus the ination oers far more increase in the scale factor than the 75 e-fold increase we have deduced
in the our hydrodynamical study. Another important feature of ination is that the Hubble constant
decreases only by a factor 10
6
, i. e., we conclude that
H
i
H
f

a
f
a
i
. (7.43)
Now we turn to the discussion of the graceful exit and the following evolution. In this regime the eld
is described by the equation of motion

2
+ m
2

2
=
3
4
H
2
. (7.44)
We introduce two new independent variables, the Hubble constant H and an angle which obviously
obey the following equations

=
_
3
4
H sin m =
_
3
4
H cos . (7.45)
Then we can rewrite (7.30) as a set of two rst order dierential equations

H = 3H
2
sin
2
(7.46)

= m
3
2
H sin 2. (7.47)
We see that the second equation contains an oscillating term, whose amplitude is decaying (the decaying
feature is seen from the rst equation). If we neglect this decaying term we obtain the solution
mt +
0
, (7.48)
where the constant phase
0
can be set to zero. Thus we conclude that the scalar eld oscillates at a
frequency m. Substituting this in Eq. (7.47) we can integrate it to obtain
H(t) =
2
3t
_
1
sin(2mt)
2mt
_
1
(7.49)
7.2. FIELD THEORY MODELS OF INFLATION 135
where the constant of integration is removed by a shift in time. The solution is applicable for mt 1
(as we have ignored the decaying oscillations with decaying amplitude above). Therefore, the oscillating
term is small compared to unity and we can expand the left-hand side in powers of (mt)
1
.
Substituting Eq. (7.48) and (7.49) in the second equation in (7.45) we obtain the solution for the
eld
(t) =
cos(mt)

3
2
mt
_
1 +
sin(2mt)
2mt
+ O((mt)
3
)
_
. (7.50)
Furthermore Eq. (??) can be integrated to obtain the evolution of the scale factor
a t
2/3
_
1
cos(2mt)
6m
2
t
2

1
24m
2
t
2
+ O((mt)
3
)
_
. (7.51)
We see that the scale factor to leading order 1 behaves as in a universe which is matter dominated
with zero pressure.
To summarize, the ination can be modeled in terms of heavy scalar particles. The theory has a
smooth exist into a Friedman universe. The inationary stage last long enough to solve the horizon
problem. Eventually we want the cold matter to be converted into radiation, leptons and baryons.
7.2.1 Slow-roll approximation
So far we have studied a specic form of the quadratic potential. Let us go back to Eq. (7.25) and see
what can be said in the case of a general potential. The equation of motion of the eld in the general
case reads

+ 3H

+ V
,
= 0, (7.52)
This equation has the form of an oscillator whereby the term proportional to the Hubble constant acts
as a friction. As is well know if we wait enough we can neglect the acceleration term compared to the
frictional damping term (this is called slow-roll approximation). Thus we will omit the

term we nd
3H

+ V
,
0 H =
d ln a
dt

_
(8/3)V (). (7.53)
Now we use the identity
d ln a
dt
=

d ln a
d
=
V
,
3H
d ln a
d
(7.54)
to rewrite Eq. (7.54) in the form
V
,
d ln a
d
= 8V. (7.55)
Integration gives
a() = a
i
exp
_
8
_

i

V
V
,
d
_
. (7.56)
This solution is valid in the limit of slow role approximation
[

2
[ [V [ [

[ 3H

[V
,
[. (7.57)
This conditions can be written entirely in terms of the properties of the potential
_
V
,
V
_
2
1 [
V
,
V
[ 1. (7.58)
136 CHAPTER 7. INFLATION
If now we assume a general power law potential
V =
1
n

n
, (7.59)
then both conditions are satised for [[ 1. The solution for the scale factor is then
a((t)) a
i
exp
_
4
n
(
2
i

2
(t))
_
. (7.60)
One can use the power law potential above to compute the behavior of the scale factor and the duration
of the ination.
7.3 Reheating after ination
The theory of reheating deals with creation of thermal Friedman universe after the inationary stage
is over. The details of the theory obviously depends on the details of the particle theory. Nevertheless,
some robust futures of the reheating theory can be studied in general terms. Consider the inaton eld
interacting with a scalar eld and a spinor eld . The tree-level interaction Lagrangian is given
by
L
int
= g
2
h

. (7.61)
The rates of the processes leading to decay of inaton are given by
( ) =
g
2
8m
( ) =
h
2
m
8
. (7.62)
We model the time dependence of the inaton eld as
(t) = (t) cos(mt), (7.63)
where (t) is the slowly decaying amplitude. The number density of inatons is given by
n

m
=
1
2m
(

2
+ m
2

2
)
1
2
m
2
. (7.64)
Since 1 after the end of the ination and m 10
13
GeV we have n

10
92
cm
3
. Out of the
two processes above the second one is much more ecient. The reason is that g < m and h <

m in
order the loop corrections to be negligible, them ( ) m and ( ) m
2
. Thus while
the rst process will die out quickly, the second process will be the one that will eectively entre the
rate equations. Another important eect is the phase space occupation of the nal states. In the case
of fermions the nal states must be non-occupied in order the process to work. On the other hand for
scalar particle, there is no such restrictions since the nal states are always open and even further the
eects of Bose condensation can enhance the decay rate.
Keeping only the scalar particles we obtain the following rate equations
1
a
3
d
dt
(a
3
n

) =
e
n

;
1
a
3
d
dt
(a
3
n

) = 2
e
n

; (7.65)
(the factor 2 is due to the fact that one particle decays into two particles.)
7.3. REHEATING AFTER INFLATION 137
An estimate can be made of the temperature of the matter after the ination and reheating. Suppose
the number of inaton oscillations are N
T
and N is the number of degrees of freedom of light elds at
these temperatures. The temperature is given by the formula
T
R

N
T
N
1/4
. (7.66)
For typical values N
T
10
6
, N 10
2
and m
13
GeV one nds T
R
10
12
GeV.
We have seen that one can produce thermal Friedman universe through the reheating process. The
details were not given here and depend on the underlying particle theory. It should be noted that apart
from the reheating the particle theories need to answer the problem of baryogensis.
The precise nature of the inaton depends on the particle theory that describes physics and sub-
Planckian scales. This could be a scalar eld or fermionic condensate. However, there is a further
possibility which is related to modied theory of gravity. If one accepts that Einstein gravity is a low-
curvature limit of more complicated theory whose action contains higher powers of curvature then the
inaton can be purely general-relativistic eect. Consider for example the action
S =
1
16
_
(R + R
2
R

+ R
3
+. . . )

gd
4
x. (7.67)
As opposed to the Einstein theory this theory contains higher that the second order derivatives. There-
fore, in addition to the gravitational waves, which are related to the existence of the graviton, there
could exist spin 0 elds.
Another example are the eld theories described by the action
S =
1
16
_
f(R)

gd
4
x, (7.68)
where f(R) is an arbitrary function of the scalar curvature. The equation of motion for this action are
given by
f
R
R

1
2

f +
_
f
R
_
;
;

f +
_
f
R
_
;
;
. (7.69)
Consider conformal transformation
g

= Fg

. (7.70)
The Ricci tensor and the scalar curvature transform as
R

= F
1
R

F
2
F
;
;

1
2
F
2
F
;
;

+
3
2
F
3
F
;
F
;
. (7.71)
R

= F
1
R 3F
2
F
;
;
+
3
2
F
3
F
;
F
;
. (7.72)
For a scalar eld dened as
=
_
3
16
ln F(R) (7.73)
the eld equations
R


1
2
R

= 8T

() (7.74)
coincide with (7.69) if we assue that F = f/R and the potential for the scalar is
V () =
1
16
f RF/R
(f/R)
2
. (7.75)
138 CHAPTER 7. INFLATION
We conclude that the higher derivative gravity theory is conformally equivalent to the Einstein theory
with an extra scalar eld.
Inationary solution can be realized even without any potential in the so-called Born-Infeld type
theories where the action depends nonlinearly on the kinetic energy of the scalar eld. The action of
such theory can be written as
S =
_
p(X, )

gd
4
x, (7.76)
where p is an arbitrary function of and X = (1/2)

. The energy-momentum tensor of this eld


can be written as
T

= ( + p)u

, (7.77)
where the Lagrangian p play the role of the pressure and
= 2X
p
X
p, u

2X
. (7.78)
Figure 7.3: The behavior of the potential in the old, new and chaotic inationary scenarios.
If the Lagrangian p satises in some range of parameters the condition
2X
p
X
p (7.79)
we obtain the desired equation state p needed for the inationary solution. The inationary
scenarios that are based on the non-trivial dependence of the Lagrangian on the kinetic terms are called
k-ination.
7.4 Scenarios for ination
There are several inationary scenarios: they can be roughly divided into
scalar elds with a potential
higher-derivative gravity theories
7.4. SCENARIOS FOR INFLATION 139
k-ination theories
The physical predictions of these theories are almost identical. We shall discuss a bit more the case
of the scalar eld with canonical energy. The potential can have three dierent shapes, each of which
corresponds to a scenario called old ination, new ination, and chaotic ination.
In the old ination the scalar eld reaches the minimum of the potential at = 0 as a result of
supercooling of initially hot universe. After the inationary expansion the universe undergoes graceful
exit via bubble nucleation. However, the energy here is concentrated at the walls of the bubbles and
the graceful exit is not guaranteed. In the new inationary scenario the potential is at and has a
maximum at = 0. It reaches its minimum via quantum uctuations and slow roll, where the energy
of the phase transition is released homogeneously. The chaotic ination correspond to a borad class of
potentials that satisfy the slow-roll conditions. The name chaotic refers to the initial conditions, which
can be almost arbitrary.
140 CHAPTER 7. INFLATION
Chapter 8
Cosmic microwave background
The cosmic microwave radiation we observe today are those photons which have been decoupled from the
rest of the universe at the recombination, when universe became transparent to photons. To summarize
the physical picture of the CMB:
no interaction after recombination - photons have been freely streaming
last interaction of photons with rest of matter was at z 1000
the CMB provides a snapshot of the universe when it was 10
5
times younger than now, i. e.,
thousand times smaller
CMB is extremely isotropic - variations across sky on average less 0.01%. Therefore we conclude
that the universe was very isotropic at the time of recombination.
The purpose of this chapter is to study the properties of the CMB and discuss what can be learned
from such studies.
Figure 8.1: The microwave background uctuations as measured by the COBE satellite.
141
142 CHAPTER 8. COSMIC MICROWAVE BACKGROUND
8.1 Basic equations
Before recombination the matter is a perfect uid. After recombination, when sucient hydrogen is
formed, photons are out of equilibrium and should be described by a kinetic equation. Let us start by
introducing the notion of the phase-space. A photon at conformal time is completely characterized by
its position in space x
i
() and its three-momentum p
i
(), i = 1, 2, 3 (spatial index). The phase-space
element is given by
d
3
xd
3
p = dx
1
dx
2
dx
3
dp
1
dp
2
dp
3
. .
covariant components
of momentum
. (8.1)
The phase-space is invariant under general coordinate transformations. To prove this statement let us
dene a new system, which we denoted by tilde
= (, x
i
) , x
i
= x
i
(, x
i
) . (8.2)
Take a constant =const
d
3
xd
3
p = J d
3
xd
3
p , (8.3)
where J is the Jacobian of the transformation:
J =
( x
1
, x
2
, x
3
, p
1
, p
2
, p
3
)
(x
1
, x
2
, x
3
, p
1
, p
2
, p
3
)
(8.4)
for transition
x
i
x
i
(x
j
, ) , p
i
p
i
=
_
x

x
i
_
x
p

. (8.5)
Since x/p = 0 we obtain
J = det
_
x
i
x
j

=const
_
det
_
x
i
x
k

=const
_
= det(
i
j
) = 1 . (8.6)
Therefore, we conclude that the phase volume is a invariant quantity.
Boltzmann equation
Consider ensemble of non-interacting identical particles. The number of particles dN is related to the
phase-space volume element d
3
xd
3
p via
dN = f(x
i
, p
j
, t) d
3
xd
3
p . (8.7)
where f(x
i
, p
j
, t) is the distribution function. Since d
3
xd
3
p is invariant it follows that f(x
i
, p
j
, t) is a
spacetime scalar. If there are no scattering processes the distribution function obeys the collisionless
Boltzmann equation
Df(x
i
(), p
i
(), )
D
=
f

+
dx
i
d
f
x
i
+
dp
i
d
f
p
i
= 0. (8.8)
The derivatives dx
i
/d and dp
i
/d are calculated along geodesics.
8.1. BASIC EQUATIONS 143
Temperature
Consider universe which is homogeneous and isotropic and is lled by slightly perturbed thermal ra-
diation. Suppose an observer measures the frequency (or energy) of a photon. If p

is the photon
four-momentum, then the observer measures the zero component in comoving local inertial frame. In
an arbitrary frame, where the observer has the velocity u

, this frequency is
= p

. (8.9)
Dene a direction of radiation by the vector
l
i

p
i

p
2
i
. (8.10)
The spectrum of photons is then Planckian
f = f
_

T
_
=
2
exp
_

T(x

,l
i
)
_
1
, (8.11)
where factor 2 stands for the two polarizations. The temperature T(x

, l
i
) depends on the direction l
i
,
observers location x
i
and time-moment . For nearly isotropic universe
T(x

, l
i
) = T
0
() +T(x

, l
i
) , (8.12)
where T T
0
. How do the uctuation T depend on the coordinate system? Consider two observers O
and

O in coordinate systems and , which are related by the coordinate transformation x

= x

.
in the rest frame of each observer:
g

= g
00
(u
0
)
2
= 1 u
0
=
1

g
00
. (8.13)
Then the photon frequencies:
= p

= p
0
u
0
p
i
u
i
..
0
=
p
0

g
00
, (8.14)
=
p
0

g
00
. (8.15)
p
0
and p
0
as well as g
00
and g
00
are related by coordinate transformations. Since the photon is massless,
p

= 0, the transformation law is


=
_
1 +

i

l
i
_
,

1 =

i

l
i
(8.16)
to rst order in
i
. Now we wish to relate the temperature uctuations in two reference frames. First
note that the distribution function is a scalar, so that

T(x

)
=

T( x

)
. (8.17)
Furthermore:

T
0
( x

) = T
0
(x

) = T
0
(x

) + T

0
. (8.18)
144 CHAPTER 8. COSMIC MICROWAVE BACKGROUND
By denition:


T =

T(x

, l
i
)

T
0
() =

T(x

) T
0
(x

) T

0
. (8.19)
Now we eliminate T
0
(x

):


T =

T(x

) T(x

) + T(x

, l
i
) T

0
=
_

1
_
T(x

) + T(x

, l
i
) T

0
=

i

l
i
T
0
. .
dipole
T

0
+ T(x

)
. .
monopole
. (8.20)
Both components depend on the coordinate system. If we observe only from one point (which we do)
then the monopole term can be removed by redenition of the background temperature. The dipole
component depends on the motion of observer with respect to the preferred frame dened by the
background radiation. Therefore, we have to look at higher order multipoles (quadrupole and higher),
which do not depend on coordinate system and motion of observer.
8.2 Sachs-Wolfe eect
In this section we will be working in the conformal Newtonian coordinate system where the metric is
given by ( 1 is the gravitational potential of scalar metric perturbations)
ds
2
= a
2
_
(1 + 2)d
2
(1 2)
ik
dx
i
dx
k

. (8.21)
The equation for geodesics quite generally is given by
du

ds
+

= 0 (8.22)
and it can be rewritten as
du

ds

1
2
g

= 0 . (8.23)
For the case of photons we use somewhat dierent notation:
dp

d
=
1
2
g

, p

=
dx

d
, (8.24)
where is called ane parameter. Since photon is massless p

= 0, p
0
p
0
p
i
p
i
= 0
p
0
=
1
a
2
_

p
2
i
_
1/2
=
p
a
2
, p
0
= (1 + 2)p . (8.25)
We can use equation Eq. (8.24) to write
dx
i
d
=
p
i
p
0
=

1
a
2
(1 + 2)p
i
p
0
= l
i
(1 + 2) , (8.26)
8.2. SACHS-WOLFE EFFECT 145
where l
i
= p
i
/p. Now express p
0
and p
i
in terms of p
i
and substitute the metric (8.21) in the frist
equation of (8.24) to obtain
dp

d
=
1
2
g

p
0
= 2p

x

. (8.27)
Now we substitute dx
i
/d and dp

/d in the Boltzmann equation (8.8) to obtain


f

+ l
i
(1 + 2)
f
x
i
+ 2p

x
j
f
p
j
= 0 . (8.28)
To leading order in and T
y =

T
=
p
0
T

g
00

=
p
T
0
a
_
1 +
T
T
0
_
. (8.29)
Since f = f(y) to leading order:
(T
0
a)

= 0 (8.30)
and to next-to-leading order:
_

+ l
i

x
i
__
T
T
+
_
= 2

. (8.31)
Solutions
From the Boltzmann equation we conclude that
The zeroth order equation implies T
0
a = Const and T
0
a
1
.
The next-to-leading order allows to solve for T. After recombination the universe is matter
dominated and

= const. Therefore / = 0. Then we obtain


T
T
+ = const. (8.32)
along the null geodesics. The eect of gravitational potential on T is called Sachs-Wolfe eect.
Because at recombination there is a small contribution to the energy density from radiation, is
actually varying in time. Then T/T + varies as an integral of / along a geodesics of a
photon. This is called early integrated Sachs-Wolfe eect.
Currently the vacuum energy overtakes the matter energy and is varying again. This causes
contributions to T. It is called late integrated Sachs-Wolfe eect.
Both eects contribute less than 10 20% to the amplitude of T and will be neglected below.
Initial conditions
Assume a photon from direction l
i
arrives at present space-time point
0
, x
i
0
. Its geodesics is given as
x
i
()

= x
i
0
+ l
i
(
0
) . (8.33)
Using the fact that T/T + = Const we can nd the quantity T/T in the direction of l
i
on the sky
today to be
T
T
(
0
, x
i
0
, l
i
) =
T
T
(
r
, x
i
(
r
), l
i
) + (
r
, x
i
(
r
)) (
0
, x
i
0
) . (8.34)
146 CHAPTER 8. COSMIC MICROWAVE BACKGROUND
Here
r
is the conformal time at recombination and x
i
(
r
) is dened via Eq. (8.33). The last term
contributes to the monopole and can be ignored. We see that present day temperature uctuations are
set by two factors:
Initial temperature uctuations at recombination (i.e. the last scattering surface);
Value of the gravitational potential at the same location point (x
r
,
r
) .
The rst term in Eq. (8.34) can be expressed via the gravitational potential and the uctuations of the
photon energy density on the last scattering surface. To do this we use matching conditions between the
hydrodynamic energy-momentum tensor, which describes the radiation before decoupling, and kinetic
energy-momentum tensor, which describes the photon gas after decoupling . The latter is given by
T

=
1

g
_
f
p

p
0
d
3
p , (8.35)
Substituting the metic (8.21) and assuming the Planckian distribution we obtain for the 00 component
of the energy-momentum tensor
T
0
0
=
1
a
4
(1 2)
_

f
_

T
_
p
0
d
3
p

= T
4
0
_ _
1 + 4
T
T
0
_

f(y)y
3
dyd
2
l . (8.36)
where y = /T and p
0
and p are expressed in terms of using (8.25) Before recombination
T
0
0
=

(1 +

) . (8.37)
By matching

= 4
_
T
T
d
2
l
4
. (8.38)
For other components of energy momentum tensor
T
i
0
= 4

_
T
T
l
i
d
2
l
4
. (8.39)
Upon taking the divergence of this expression and comparing with
T
i
0i
=
4
3

u
0
u
i
,i
= (4

(8.40)
the matching gives

= 4
_
l
i

i
_
T
T
_
d
2
l
4
, (8.41)
where we have neglected the radiation contribution to the gravitational potential and we have set

(
r
) = 0. These equation are satised if
_
T
T
_

k
(

l,
r
) =
1
4
_

k
+

i
k
2
(k
m
l
m
)
_

k
. (8.42)
Substituting back

l (l
1
, l
2
, l
3
), and the observed location x
0
= (x
1
, x
2
, x
3
) for the Fourier expansion
of (8.34) we obtain
T
T
(
0
, x
0
,

l) =
_ __
+

4
_

k
4k
2

0
_
exp i

k( x
0
+

l(
r

0
))
d
3
k
(2)
3/2
. (8.43)
8.3. CORRELATION FUNCTIONS 147
Since
r
/
0
< 1/30 we can put
r

0
and neglect it. First term in the square brackets comes from
initial inhomogeneities in the radiation energy density and Sachs-Wolfe eect; second term is related
to the velocities of the baryon-radiation plasma at recombination. The latter term is referred to as
Doppler term contribution to the uctuations.
8.3 Correlation functions
The sky map of the cosmic microwave background temperature uctuations can be completely charac-
terized by the two-point correlation function (assuming that the spectrum of uctuations is Gaussian)
C()
_
T
T
(

l
1
)
T
T
(

l
2
)
_
. (8.44)
denotes averaging over all directions of

l
1
and

l
2
, such that

l
1

l
2
= cos . The squared temperature
dierence between two directions separated by angle , averaged over the sky is related to the correlation
function C() as
_
_
T
T
0
()
_
2
_
=
__
T(

l
1
) T(

l
2
)
T
0
_
2
_
= 2
_
C(0) C()

. (8.45)
The correlation function C() can be used to discriminate between the cosmological models and within
a given model to determine the fundamental parameters of the model.
Apart from the temperature correlation function one can also dene polarization correlation function,
since the CMB shows some degree of polarization. But we will not discuss this correlation function any
further.
Now we dene the cosmic mean of temperature correlation function: it is obtained by keeping the
direction

l
1
and

l
2
constant and averaging over all the possible observers at positions x
0
. We substitute
the expression (8.43) for temperature uctuations into the denition of correlation function and integrate
over the angular part of

k integration. We then obtain
C() =
_ _

k
+

4

k
4k
2

1
__

k
+

4

k
4k
2

2
_

sin
_
k

l
1

l
2

_
k

l
1

l
2

k
2
dk
2
2
, (8.46)
where after the dierentiation with respect of
1
and
2
one should set
1
=
2
=
0
. We can make use
of the expansion
sin
_
k

l
1

l
2

_
k

l
1

l
2

l=0
(2l + 1) j
l
(k
1
)j
l
(k
2
) P
l
(cos ) , (8.47)
where P
l
(cos ) and j
l
(k) are Legendre polynomials and spherical Bessel functions of corder l, respec-
tively. Quite generally we can also expand the correlation function in the Legendre polynomials
C() =
1
4

l=2
(2l + 1) c
l
P
l
(cos ). (8.48)
148 CHAPTER 8. COSMIC MICROWAVE BACKGROUND
Here the monopole and dipole components corresponding to l = 0, 1 are excluded, while the remaining
coecients can be identied by comparing (8.48) with (8.46)
c
l
=
2

k
(
r
) +

k
(
r
)
4
_
j
l
(k
0
)
3

k
4k
2
j
l
(k
0
)
(k
0
)

2
k
2
dk. (8.49)
Similarly we can expand the temperature uctuations in spherical harmonics
T(, )
T
0
=

l,m
a
l,m
Y
l,m
(, ) (8.50)
and the coecients (in a homogeneous and isotropic universe) satisfy the condition
a

m
, a
lm
=
ll

mm
c
l
, (8.51)
where the brackets refer to taking the cosmic mean. The multipole moments c
l
[a
lm
[
2
receive
their main contribution from angular scale /l and l(l + 1)c
l
is the typical squared temperature
uctuation on this scale.
8.3.1 Large scale anisotropies
It is convenient to discuss the anisotropies on the large and small scales separately. With start with
the former. We assume that the angular scales characterizing the CMB uctuation are 1. For
such large angular scales the inhomogeneities have wave lengths that are larger than the Hubble radius
at recombination. Therefore they had no chance to evolve substantially after the ination.
The relative energy density uctuations in the radiation component can be expressed in terms of
the gravitational potential (see ....) for adiabatic perturbations with wave-vectors k
r
1.

k
(
r
)
8
3

k
(
r
),

k
0. (8.52)
We substitute this result in (8.43) to obtain
T
T
(
0
, x
0
,

l)
1
3
(
r
, x
0

l
0
). (8.53)
Now we can substitute (8.52) into (8.49) and carry out the integral with the help of the formula
_

0
y
n1
j
2
l
(y)dy = 2
m3

(2 n)(l + m/2)

2
((3 m)/2)(l + 2 m/2)
. (8.54)
The result is
l(l + 1)c
l
=
9B
100
= const. B = [(
0
k
)
2
k
3
[. (8.55)
This result is valid for
l 200 (8.56)
but is a good approximation for only l < 20. For larger l the eects of subhorizon modes can not be
neglected in the computation of c
l
which includes k integration over all the modes (sub and supra-
horizon).
8.3. CORRELATION FUNCTIONS 149
Finite thickness eects
In the previous section we assumed that the recombination is an instantaneous eects. This is not
true and can lead to modications on the modes characterized with small angular scales. In fact the
recombination time is nite, and this leads to a suppression of the temperature uctuations on small
scales.
Suppose a photons arrives at the detector from the direction

l. The redshift spanned by the re-
combination lies in the range 1200 > z > 900. If the last scattering was at the conformal time
L
the
photon position is given by
x(
L
) = x
0
+

l(
L

0
). (8.57)
The photons emitted during the duration
r
(roughly the duration of recombination) will be registered
from the direction

l. The temperature inhomogeneities that have scales smaller than


r
wil be smeared
out.
The probability that a photon scattered within the time interval t
L
and a time t
L
(conformal time

L
) and propagated further without interactions to the detector is given by (prime is derivative with
respect to the conformal time)
dP(
L
) =

(
L
) exp[(
L
)]
. .
visibility function
d
L
(8.58)
and (
L
) is the optical depth
(
L
) =
_
t
0
t
L
dt
(t)
=
_

0

T
n
t
X
e
a()d
. .
dt
(8.59)
where is the mean-free time for Thomson scattering, n
t
is the number density of all electrons and X
is the ionization fraction. The expression for temperature uctuations now need to be weighted with
the probability above, i.e., instead of (8.43) we obtain
T
T
(
0
, x
0
,

l) =
_ __
+

4
_

k
4k
2

0
_
exp i

k( x
0
+

l(
L

0
))

d
L
. .
P
L
d
3
k
(2)
3/2
. (8.60)
The visibility function vanishes in two limits:
L
0 then , and
L
when

0. The
visibility function has it maximum in the range 1200 > z > 900. And this condition follows from
d

d
= 0, (

2
)e

= 0,

2
. (8.61)
The main quantity that changes in this time interval is the ionisation fraction X. Substituting (8.59)
into the last condition in Eq. (8.61) we obtain
X

(
r
) (
T
n
t
a)
r
X
2
r
(8.62)
where r refers to the time
r
. The time variation of electron fraction is well described by
X
e
(z) 1.4 10
9
_

m
h
2
75

10
z
1
exp
_

14400
z
_
. (8.63)
150 CHAPTER 8. COSMIC MICROWAVE BACKGROUND
Using this equation we can compute
X


1.44 10
4
z
1X, 1 =
a

a
, (8.64)
substituting this in Eq. (8.62) we obtain
X
r
= 1
r
(
T
n
t
a)
1
r
, =
14400
z
r
. (8.65)
We can then deduce that the visibility function reaches its maximum at z
r
1050.
We can approximate the visibility function near the maximum as a Gaussian

exp() exp
_

1
2
( ln

r
(
L

r
)
2
_
. (8.66)
We can calculate the derivatives with the help of (8.64) and (8.65) to obtain

exp()
(1)
r

2
r
exp
_

1
2
(1)
2
r
_

r
1
_
2
_
. (8.67)
where the pre-factor is chosen to satisfy the normalization
_

exp () d
L
= 1. (8.68)
To obtain an new expression for (8.60) we note that the factor in front of the exponent does not change
over the time-scales we are interested, so that it can be taken at
L
=
r
. Substituting (8.67) in (8.60)
and performing the integration over
L
we get
T
T
=
_ _
+

4

3

4k
2

0
_

r
exp
_
(k
r
)
2

exp
_
i

k(x
0
+

l(
r

0
)
_
d
3
k
(2)
3/2
, (8.69)
where

k

l = k
3
/3 due to isotropy, one can neglect
r
compared to
0
and
=
1

6(1)
r
. (8.70)
Thus the nite time-scale of recombination introduces an extra exponential factor in the temperature
uctuations. The dependence of the function on cosmological parameters can be obtained if we ignore
the dark energy contribution at the recombination. Then,
(1)
r
= 2
1 + (
r
/

)
2 + (
r
/

)
. (8.71)
and
_

_
2
+ 2
_

z
eq
z
r
. (8.72)
Then we nally obtain
0.015
_
1 +
_
1 +
z
eq
z
r
_
1/2
_
. (8.73)
8.3. CORRELATION FUNCTIONS 151
The value of z
eq
depends on the matter contribution to the total energy density and the number of
ultra-relativistic species in the early universe. For three types of neutrinos we have
z
eq
z
r
= 12.8(
m
h
2
75
). (8.74)
To summarize the nite duration of the recombination leads to the extra exponential factor in the
temperature uctuations and Eq. (8.49) now reads
c
l
=
2

k
(
r
) +

k
(
r
)
4
_
j
l
(k
0
)
3

k
4k
2
j
l
(k
0
)
(k
0
)

2
exp[2(k
r
)
2
]k
2
dk. (8.75)
i.e., there is an extra exponential factor in the integrand.
8.3.2 Small-scale anisotropies
So far we have discussed results valid for l 200. The multipole moment l 200 corresponds to
the sound horizon scale at recombination. Therefore (i) for perturbations responsible for uctuations
with l > 200 entered the horizon before the recombination. These perturbations undergo substantial
evolution. Before computing the perturbations for these l of interest we need to improved source
functions in the integrands, i.e., we need a better approximation to functions
k
+
k
/4.
The source function appropriate for the conditions at recombination can be written for an input
gravitational potential
0
k
as

k
+

k
4

0
k
_
T
p
_
1
1
3c
2
s
_
+ T
0

c
s
cos
_
k
_

r
0
c
s
d
_
e
k
2
/k
2
D
_
(8.76)
and
(

k
)
r
4T
0
lc
3/2
s
sin
_
k
_

r
0
c
s
d
_
e
k
2
/k
2
D

0
k
. (8.77)
The transfer functions T
p
and T
0
can be calculated analytically in two limiting cases. When k
eq
1
(where
eq
corresponds to the equipartition between the matter and radiation) we have
T
p

9
10
, T
0

9
10
3
3/4
0.4. (8.78)
For perturbations with k
eq
1 we have the limits
T
p

ln(0.15k
eq
)
(0.27k
eq
)
2
0, T
0
=
3
5/4
2
= 1.97. (8.79)
However one generally needs to complete transfer functions since the perturbations with k
eq
O(1)
determine the amplitude of the temperature uctuations in the region of the rst peaks in the CMB
spectrum. Figure 8.2 shows the dependence of the transfer functions with the dimensionless wave-
vectors k
eq
. Now we can compute the multipoles of the CMB by substituting (8.76) and (8.77) into
(8.75). The result of lengthly calculation is [we put l(l + 1) = l
2
for large l]
c
l
=
1
16
_

l
1
0
_
[4
k
+
k
[
2
k
2
(k
0
)
_
(k
0
)
2
l
2
+
9
_
(k
0
)
2
l
2
(k
0
)
3
[

[
2
_
exp
_
2(k
r
)
2

dk. (8.80)
152 CHAPTER 8. COSMIC MICROWAVE BACKGROUND
Figure 8.2: The behavior of the transfer functions as a function of k
eq
.
Here we need to substitute (8.76) and (8.77) with the transfer functions dened above and carry out
integrals to obtain the multipoles.
We consider now the case of scale invariant spectrum of initial density perturbations [(
0
k
)
2
k
3
[ = B,
where B is a constant.
Then Eq. (8.79) can be written as
l
2
C
l
=
B

(O + N) (8.81)
where O and N refer to oscillating and non-oscillating parts of the integral. The oscillating part can
be written as a sum of two parts O = O
1
+ O
2
where
O
1
= 2

c
s
_
1
1
3c
2
s
__

1
T
p
T
0
exp[
1
2
l
2
x
2
(l
2
f
+l
2
s
)
2
] cos(lx)
x
2

x
2
1
dx. (8.82)
O
2
=
c
s
2
_

1
T
2
0
(1 9c
2
s
)x
2
+ 9c
2
s
x
4

x
2
1
exp(l
2
x
2
/l
2
s
) cos(2lx), (8.83)
where x = k
0
/l. These two terms give rise to constructive and deconstructive interference and to
associated peaks in the spectrum. The parameter
=
1

0
_

r
0
c
s
()d (8.84)
gives the locations of the peaks.
The non-oscillating part contains three integrals N = N
1
+ N
2
+N
3
, which are
N
1
=
_
1
1
3c
2
s
_
2
_

1
T
2
p
e
(l/l
f
)
2
x
2
x
2

x
2
1
dx (8.85)
(note that this integral vanishes as c
2
s
1/3, i.e., when baryon density vanishes),
N
2
=
c
s
2
_

1
T
2
0
e
(l/l
f
)
2
x
2 1
x
2

x
2
1
, (8.86)
N
3
=
9c
3
s
2
_

1
T
2
0
e
(l/l
f
)
2
x
2

x
2
1
x
4
. (8.87)
8.3. CORRELATION FUNCTIONS 153
The integrals which determine the multipoles depend on the following parameters
c
s
, l
f
, l
S
, , T
0
, T
p
, (8.88)
which can be related to the cosmological parameters

B

m
,

, w, h
75
. (8.89)
For the range of interest 1000 > l > 200 the dominant contribution to integrals comes from the vicinity
of the point x 1 and around this point the transfer functions can be approximated as
T
p
(x) 0.74 0.25(P + ln x) (8.90)
T
0
(x) 0.5 + 0.36(P + ln x) (8.91)
(8.92)
where
P = ln
_
I

l
200
_

m
h
2
75
_
. (8.93)
where
I

= 3
_

m
_
1/6
__
y
0
dx
(sinh x)
2/3
_
1
(8.94)
As pointed out above the main contribution to the integrals O
1
and O
2
arises from the vicinity of the
lower bound x = 1. One can use this fact to obtain approximate expressions for the total oscillating
part
O
_

l
_
A
1
cos(l + /4) + A
2
cos(2l + /4) exp (l/l
S
)
2

(8.95)
where the amplitudes which are weakly dependent on l are dened as
A
1
= 0.1
(P 0.78)
2
4.3
(1 +)
1/4
exp
_
l
2
2
(l
2
S
l
2
f
)
_
, (8.96)
A
2
= 0.14
(0.36P + 0.5)
2
(1 + )
1/2
. (8.97)
Here the parameters are dened as
=
1
3c
2
s
1 17(
b
h
2
75
). (8.98)
where c
s
is the speed of sound. For the non-oscillating parts one can use the following analytical ts
N
1
0.063
2
P 0.22(l/l
f
)
0.3
2.6
1 + 0.65(l/l
f
)
1.4
exp(l
2
/l
2
f
), (8.99)
N
2

0.037

1 +
P 0.22(l/l
f
)
0.3
+ 1.7
1 + 0.65(l/l
s
)
1.4
exp(l
2
/l
2
s
), (8.100)
N
3

0.037
(1 + )
3/2
P 0.5(l/l
s
)
0.55
+ 2.2
1 + 2(l/l
s
)
2
exp(l
2
/l
2
s
). (8.101)
154 CHAPTER 8. COSMIC MICROWAVE BACKGROUND
Figure 8.3: Upper panel gives multipoles on the linear scales which allows to see the details over the
large l. The lower panel is logarithmic and allows us to see the better the eect of small l
The ratio of small-scale to large scale uctuations is given by
l(l + 1)C
l
l(l + 1)C
l
[
low l
=
100
9
(O + N
1
+ N
2
+N
3
). (8.102)
The results for the concordance model which assumes the following values of the parameters

m
= 0.3

= 0.7,
b
= 0.04,
tot
= 1, H = 70 km/s/Mpc (8.103)
are in agreement with the experimental data, see gure.

Das könnte Ihnen auch gefallen