Chapter 2

Expansion of the Universe

In this chapter we will consider the smooth expansion of the Universe by studying the zero-th order
Einstein Equations in the Friedman-Robertson-Walker metric.

2.1 Introduction
2.1.1 Symmetries and the Cosmological Principle
Think of the Universe as a 3-d spatial surface evolving in time. It is typical to assume a Universe
with lots of symmetry on its spatial hypersurfaces. In the early days, assumptions were motivated
mainly by the search for simplest possible solutions to the Einstein Equations, but now are backed
up by recent observations.

Cosmological Principle: Assumption that the Universe is homogeneous (same at every point,
therefore symmetric under translations) and isotropic (same in all directions, therefore symmetric
under rotations).

Can have Universe that is homogeneous but not

isotropic. Example: Associate an arrow (e.g. a ve- B E

locity field) at every point in the Universe, pointing

in a certain fixed direction. Such universe is homo- D
geneous because all points are equivalent. But it is C
not isotropic because there is a preferred direction. A

Can have Universe that is isotropic (with re-

spect to a point) but not homogeneous. Example:
Spherically Symmetric Universe with radial depen-
dent comoving density. It is isotropic because it
looks the same in all directions (but only with re-
spect to a point, e.g. us). But it is not homoge- Figure 2.1: Universe isotropic with respect to A
neous. and B implies same density at C = D = E. By
changing radius of B sphere, can show that all
points in the shaded region must have same den-
Not trivial to discard completely inhomoge- sity. Since argument can be extended to larger
neous and/or anisotropic models, because typical spheres, whole Universe must be homogeneous.
(Peacock).
59
60 CHAPTER 2. EXPANSION OF THE UNIVERSE

observations only access our past light-cone. In principle must check if homogeneneity and isotropy
holds at every spatial hypersurface, not only today.

A Universe that is isotropic with respect to any point is homogeneous (see Fig. 2.1)

Observations from galaxies and CMB radiation indicate an isotropic universe on large scales.

Copernican Principle: We do not occupy a special place in the Universe.

Isotropy + Copernican Principle imply that the Universe must be isotropic with respect to any
point. Therefore it must be homogeneous on large enough scales.

For isotropy, the velocity field must be radial at any given time t

v = H(t)r (2.1)

Any component in the tangential direction would indicate a rotation around a (prefered) axis,
breaking isotropy. If isotropy is guaranteed at a point and homogeneity is assumed, then isotropy
holds for any other point since

v2 − v1 = H(t)(r2 − r1 ) (2.2)

Notice ∇ · v = 3H(t) =constant at a given time.

2.1.2 Expansion and Comoving Coordinates

We have a number of good evidences that the Universe is expanding.

Hubble observed in 1929 that the velocity of galaxies was roughly proportional to their distance,
as would be expected for a homogeneous and isotropic expansion.

It is convenient to define comoving coordinates x, which do not change with the expansion, and
parametrize the expansion in terms of the scale factor a(t) (see Fig. 6.1).

Then physical coordinates r change with the expansion such that

physical distance = a(t) × comoving distance. (2.3)

or

r(t) = a(t)x (2.4)

2.2. THE FRIEDMANN-ROBERTSON-WALKER METRIC 61

Figure 2.2: Scale factor and expansion. Comoving coordinates do not change, but physical coordianates
expand with the scale factor a(t). (Dodelson).

We can think of the comoving coordinates as a grid that follows the expansion by definition.
Thinking about the expansion in this way provides exactly the expected relation between velocity
and distance for an isotropic Universe, because since r = a(t)x
ȧ
v = ṙ = ȧx = r (2.5)
a
and we obtain the same relation if we define the Hubble parameter H(t):
ȧ
H(t) = (Hubble Parameter) (2.6)
a

2.2 The Friedmann-Robertson-Walker metric

The Friedmann-Robertson-Walker metric generalizes the Minkowski metric to include expansion
on the spatial hypersurfaces, maintaining spatial isotropy and homogeneity. For a flat Universe it
is given by

ds2 = −dt2 + a2 (t)dl2 (2.7)

where

dl2 = dx2 + dy 2 + dz 2 (2.8)

is the 3D spatial metric without expansion. However, there are more general spaces (with curvature)
that also have homogeneous and isotropic 3D spatial hypersurfaces, and in principle could describe
our Universe expansion.

2.2.1 Curvature
2 dimensions
For pedagogical purposes, let us star with the case of 2 spatial dimensions. Flat space in 2 dimen-
sions is described by an euclidian metric, either in cartesian coordinates (x,y) as

dl2 = dx2 + dy 2 (2d flat space) (2.9)

62 CHAPTER 2. EXPANSION OF THE UNIVERSE

or in polar coordinates (D, φ) defined as (see left panel in Fig. 2.3.)

x = D cos φ , (2.10)
y = D sin φ (2.11)

such that the metric is

dl2 = dD2 + D2 dφ2 (2d flat space) (2.12)

Now, curved spaces in RN that satisfy the cosmological principle (homegeneity and isotropy)
are necessarily N-spheres, i.e. S N = {x ∈ RN +1 : |x| = R} or N-hyperboloids. Let us focus
on N-spheres for now. Thus, one way to obtain a homogeneous and isotropic curved space in 2
dimensions, e.g. a 2-sphere or spherical shell, is to think of it as flat space in 3 dimensions confined
to a fixed radius R.
In this case, we start with 3 dimensional spatial flat line element as

dl2 = dx2 + dy 2 + dz 2 (3d flat space) (2.13)

with the constraint that

x2 + y 2 + z 2 = R 2 (2.14)

We can express the metric in spherical coordinates (θ,φ) for fixed R

x = R sin θ cos φ , (2.15)

y = R sin θ sin φ , (2.16)
z = R cos θ (2.17)

and we have

dl2 = R2 (dθ2 + sin2 θdφ2 ) (2d curved space) (2.18)

Suppose you are an observer at the North Pole and you don’t know you are in this curved
geometry in 2 dimensions. You will be tempted to use the coordinates that seem more natural to
you, namely D and φ, as if you were in flat space (see right panel in Fig. 2.3.). Let us express the
metric in terms of these coordinates. Note that

D = Rθ → dD2 = R2 dθ2 (2.19)

Moreover, we define an interesting coordinate DA called angular diameter distance as

 
D
DA = R sin θ = R sin (Angular Diameter Distance) (2.20)
R
DA is called this way because it is the correct distance to give you the angular piece of the metric.
For instance, in terms of DA , we have

x = DA cos φ , (2.21)
y = DA sin φ , (2.22)
q
z = R 1 − DA 2 /R2 (2.23)
2.2. THE FRIEDMANN-ROBERTSON-WALKER METRIC 63

which is similar to the flat case with DA instead of D.

From these we have
dDA
dθ = q (2.24)
2 /R2
R 1 − DA
dD
dθ = (2.25)
R
Therefore, we can express the metric either in (D, φ) coordinates as
 
2 2 2 2 D
dl = dD + R sin dφ2 (2d curved space) (2.26)
R

or in (DA , φ) coordinates as

dDA2
dl2 = 2
2
+ DA dφ2 (2d curved space) (2.27)
1 − DA /R2

or even in mixed coordinates

dl2 = dD2 + DA
2
dφ2 (2d curved space) (2.28)

A comparison between 2d flat and curved spaces is shown in Fig. 2.3.

3 dimensions
Flat space in 3 dimensions is described by an euclidian metric, either in cartesian coordinates (x,y)

dl2 = dx2 + dy 2 + dz 2 ( 3d flat space) (2.29)

or in spherical coordinates (D, θ, φ) defined as

x = D sin θ cos φ , (2.30)

y = D sin θ sin φ , (2.31)
z = D cos θ , (2.32)

such that the metric is

dl2 = dD2 + D2 (dθ2 + sin2 θdφ2 ) = dD2 + D2 dα2 (3d flat space) (2.33)

Similarly to the previous section, we can obtain a curved space in 3 dimensions as a flat space
in 4 dimensions confined to a fixed curvature radius R.
In this case, we start with 4 dimensional spatial line element as

dl2 = dx2 + dy 2 + dz 2 + dw2 ( 4d flat space) (2.34)

with the constraint that

x2 + y 2 + z 2 + w 2 = R 2 (2.35)
64 CHAPTER 2. EXPANSION OF THE UNIVERSE

Observer

ds z
D
DA
Observer
dφ D ds
θ
R
φ
x
φ dφ
R sin θ

x R sin θdφ y

Figure 2.3: Left: Observer in 2 dimensional flat space (2d plane). Here (x, y) axes are shown in magenta.
Using polar coordinates (D, φ), this observer finds that the line element ds is given by ds = Ddφ, as expected
for Euclidian geometry. Right: Another observer in 2 dimensional space with constant curvature (2d sphere
or spherical shell with curvature radius R). Similarly to his friend in flat space, this observer sets up (x, y)
axes (magenta lines) and polar coordinates (D, φ) (notice the radial distance D = Rθ). He is somewhat
surprised to find that ds 6= Ddφ. Instead, he finds ds = DA dφ, where the angular diameter distance
DA = R sin(D/R). This allows him to infer the value of R, even though he is confined to the spherical shell
surface. Only when R → ∞, the spherical shell degenerates into a plane, curvature effects become negligible
and DA → D. Notice that both observers live in homogeneos and isotropic spaces.
2.2. THE FRIEDMANN-ROBERTSON-WALKER METRIC 65

We express the metric in generalized spherical coordinates (D, θ,φ) for fixed R
x = R sin(D/R) sin θ cos φ , (2.36)
y = R sin(D/R) sin θ sin φ , (2.37)
z = R sin(D/R) cos θ (2.38)
w = R cos(D/R) (2.39)
and we have:
 
2 2 2 2 D
dl = dD + R sin (dθ2 + sin2 θdφ2 ) (3d curved space) (2.40)
R
and gain, in terms of DA = R sin(D/R), this looks quite similar to the 3d flat case:
dl2 = dD2 + DA
2
(dθ2 + sin2 θdφ2 ) (3d curved space) (2.41)
or in (DA , θ, φ) coordinates as
dDA2
dl2 = 2
2 /R2 + DA dα
2
(3d curved space) (2.42)
1 − DA
Notice that for R → ∞ (infinite curvature radius), we have
DA = R sin(D/R) ≈ RD/R = D (2.43)
i.e. we recover a flat universe with comoving angular diameter distance equal to comoving distance.

Now, in terms of scaled (comoving) coordinates D → RD, and DA → RDA , we have

 
dl2 = R2 dD2 + sin2 (D) (dθ2 + sin2 θdφ2 ) (2.44)
 2 
dDA
= R2 2
2 + DA dα
2
(2.45)
1 − DA

Repeating the computation for N-hyperboloids H N = x ∈ RN +1 : |x| = iR , these equations
would be
 
dl2 = R2 dD2 + sinh2 (D) (dθ2 + sin2 θdφ2 ) (2.46)
 2 
dDA
= R2 2
2 + DA dα
2
(2.47)
1 + DA
Therefore, all cases can be put together as
 
dl2 = R2 dD2 + fk2 (D)dα2 (2.48)
 2 
2 dDA 2 2
= R 2 + DA dα (3d curved space) (2.49)
1 − kDA
such that [because −i sin(ix) = sinh(x)]:

√ sinh(D), k = −1, Negative Curvature, Open Universe
sin( kD) 
DA = fk (D) = √ = D, k = 0, Zero Curvature, Flat Universe (2.50)
k 
sin(D), k = +1, Positive Curvature, Closed Universe
Since a flat universe is favored by observations, we will take a flat geometry, unless stated
otherwise. Including the space expansion and the time piece, we have
ds2 = gµν dxµ dxν = −dt2 + a2 (t)dl2 (2.51)
66 CHAPTER 2. EXPANSION OF THE UNIVERSE

2.3 The Friedmann Equations

Let us consider the case of flat space, in which the metric is given by (c = 1):

ds2 = gµν dxµ dxν = −dt2 + a2 (t)(dx2 + dy 2 + dz 2 ) (2.52)

so that
   
−1 0 0 0 −1 0 0 0
 0 a2 0 0   0 1/a2 0 0 
gµν =  and g µν =  (2.53)
 0 0 a2 0   0 0 1/a 2 0 
0 0 0 a2 0 0 0 1/a2

Notice that gij = a2 δij . The affine connection or Christoffel symbols are given by:
1 σρ
Γσµν = g (∂µ gνρ − ∂ρ gµν + ∂ν gρµ ) (2.54)
2
and its componentes are

Γ000 = Γ00i = Γ0i0 = 0 (2.55)

and

Γ0ij = δij aȧ (2.56)

and

Γi00 = 0 (2.57)

and
ȧ
Γi0j = Γij0 = δij (2.58)
a

and

Γijk = 0 (2.59)

Bottom line: Γαµν is non-zero only when two indices are spatial and one is zero.

Note that the non-zero components above can be expressed in terms of the Hubble parameter
H = ȧ/a (Eq. 2.6) and the spatial metric gij = a2 δij , as Γ0ij = Hgij and Γi0j = Γij0 = Hδij .

The Ricci tensor is given by

Rµν = Γαµν,α + Γαµν Γβαβ − Γαµα,ν − Γαµβ Γβαν (2.60)

Therefore, its componentes are

ä
R00 = −3 (2.61)
a
2.3. THE FRIEDMANN EQUATIONS 67

and

R0i = Ri0 = 0 (2.62)

and

Rij = δij (aä + 2ȧ2 ) (2.63)

and the Ricci scalar is given by

R = g µν Rµν = g 00 R00 + g ij Rij (2.64)

or
 
ä ȧ2
R=6 + (2.65)
a a2

Finally, the Einstein Tensor

1
Gµν = Rµν − gµν R (2.66)
2
We have for the 00 component
1
G00 = R00 − g00 R
2  
ä 1 ä ȧ2
= −3 + 6 +
a 2 a a2
(2.65)

so
 2
ȧ
G00 = 3 (2.66)
a

and for the ij component

1
Gij = Rij − gij R
2   
δij a2 ä ȧ2
= δij (aä + 2ȧ2 ) − 6 + 2
2 a a
= δij [aä + 2ȧ2 − 3aä − 3ȧ2 ]
(2.64)

or

Gij = −[2aä + ȧ2 ]δij (2.65)

Therefore homogeneity and isotropy in the metric automatically imply a diagonal Gµν and this
guarantees a diagonal stress-energy tensor, i.e. as that of a perfect fluid. To get Einstein Equations,
we express the diagonal energy-momentum tensor as:
68 CHAPTER 2. EXPANSION OF THE UNIVERSE

 
−ρ 0 0 0
 0 P 0 0 
T µν =
 0 0
 (2.66)
P 0 
0 0 0 P
with components that can also be written as
T00 = g0α T α0 = g00 T 0 0 = ρ (2.67)
α i 2
Tij = giα T i = gij T j = a δij P (2.68)
The 00 Einstein Equation becomes
G00 = 8πT00
 2
ȧ
3 = 8πGρ
a
and we obtain the first Friedmann equation
 2
ȧ 8πG
= ρ (2.67)
a 3
The ij equation becomes
Gij = 8πTij
−2aä − ȧ = 8πGa2 P
2
 
ä 1 ȧ 2
− − = 4πGP
a 2 a
or
 2
ä 1 ȧ
+ = −4πGP (2.65)
a 2 a
Using the first Friedmann equation into this, we get
ä 4πG
+ ρ = −4πGP (2.66)
a 3
or the second Friedmann equation:
ä 4πG
=− (ρ + 3P ) (2.67)
a 3
Notice that for cosmic acceleration, we need ä > 0 and therefore (ρ + 3P ) < 0, i.e. P < −ρ/3.
Therefore negative pressure is required for cosmic acceleration within General Relativity.
Above ρ and P are sums over all components of the universe (and their possible interactions):
X
ρ = ρi (2.68)
i
X
P = Pi (2.69)
i
where i can be total matter (cold dark matter + baryons + massive neutrinos), radiation (photons
+ massless neutrinos), dark energy (cosmological constant, dynamical dark energy, scalar fields,
interacting fields) and even curvature (which we are neglecting for considering only a flat space-
time).
2.3. THE FRIEDMANN EQUATIONS 69

2.3.1 Hubble and Deceleration Parameter

It is convenient to define the Hubble parameter H(t) as the relative rate of change of the scale
factor (recall Eq. 2.6)

ȧ
H(t) = (2.70)
a

whose present value denoted H0 = H(t0 ) is the Hubble constant. In terms of H(t) the first
Friedmann equation becomes

8πG
H 2 (t) = ρ (2.71)
3

Historically, we defined the Deceleration parameter q(t) as a uniteless quantity constructed from
the second derivative of the scale factor as

ä/a ä
q(t) = − 2 2
=− (2.72)
(ȧ /a ) aH 2

Using the ratio of the second and first Friedmann equation we find

1
q(t) = (1 + 3w) (2.73)
2

where w = P/ρ is the equation of state. For w < −1/3 the deceleration parameter is negative,
indicating that the Univese is actually accelerating.

2.3.2 Curvature
If we consider metrics with curvature, still consistent with homogeneity and isotropy as in Eqs. 6.89,
one can show that the first Friedmann equation becomes
 2
ȧ 8πG k
= ρ− 2 (2.74)
a 3 a

This motivates us to interpret the curvature term k as an additional term in the total density:

8πG
k=− ρk,0 (2.75)
3

On the other hand, the second Friedmann equation is unmodified:

ä 4πG
=− (ρ + 3P ) (2.76)
a 3

2.3.3 Newtonian Derivation

Even though we derived these results within General Relativity, they can be obtained more easily
within the Newtonian context, providing some intuition to the effects of the expansion.
70 CHAPTER 2. EXPANSION OF THE UNIVERSE

Energy Conservation
Consider a test particle (or shell) of mass m at a radius R from the center of a uniform mass
distribution with mass density ρ̄m . Suppose the particle is moving only radially with velocity
v = Ṙ. We may replace the mass distribution by a point mass with mass M = 4πR3 ρ̄m /3 at the
center. The total energy E for this test particle is conserved and as
1 GM m 1 G(4πR3 ρ̄m )m
E = K + U = mv 2 − = m(Ṙ)2 − (2.77)
2 R 2 3R
so we may write
2E 8πG 2
= (Ṙ)2 + R ρ̄m (2.78)
m 3
Assuming that the radius is expanding with scale factor from a comoving value R(t) = a(t)Rc we
find
 2
ȧ 8πG 2E
= ρ̄m + (2.79)
a 3 mRc2 a2
This is exactly the Friedmann equation, where the last term appears if we consider a universe with
curvature. So the curvature in an expanding universe is related to its total energy. A flat universe,
with no curvature, has no total energy, i.e. its ”escape velocity” is just enough to keep expanding
forever.

Force Equation
Considering the same test particle (or shell), we may compute the gravitational force acting on it:
d2 R GM m
F =m = −
dt2 R2
d2 R G(4πR3 ρ̄m )
= − (2.79)
dt2 3R2
So in this case we would have
ä 4πG
=− ρm (2.80)
a 3
This is only consistent with our general relativity result for a Universe with only dust matter, for
which the pressure Pm ≪ ρm and can be neglected. The fact that the correct equation contains an
extra pressure P is an effect of General Relativity and does not have a Newtonian analog. So in
general relativity ρ + 3P sometimes plays the role of an effective density.

2.3.4 Density Parameter

We define the critical density ρcrit as the density for a flat universe, i.e.
8πG
H 2 (t) = ρcrit (t) (2.81)
3
or, similarly
3H 2 (t)
ρcrit (t) = (2.82)
8πG
2.3. THE FRIEDMANN EQUATIONS 71

The critical density today ρcrit,0 is then

3H02
ρcrit,0 = = 1.879 × 10−29 h2 g cm−3 = 2.775 × 1011 h2 M⊙ Mpc−3 (2.83)
8πG
It is convenient to normalize the energy density for all species in the Universe relative to the critical
density, defining the density parameter Ωi (t):
ρi (t)
Ωi (t) = (2.84)
ρcrit (t)
It is a standard notation to refer to Ωi as the present value of the density parameter
ρi,0
Ωi = (2.85)
ρcrit,0
Notice that for the curvature, from Eq. 2.75
ρk,0 −3k 8πG k
Ωk = = 2 =− 2 (2.86)
ρcrit,0 8πG 3H0 H0
so we have k = −Ωk H02 . In terms of the density parameters then, the Friedmann equation can be
written as
8πG X
H 2 (t) = ρi (t) (use 8πG/3 = H02 /ρcrit,0 )
3
i
X ρi (t)
= H02 (2.86)
ρcrit,0
i
We will see later that the density evolution of each species is determined by its equation of state
w. For constant (time-independent) equation of states wi for each species, we have
ρi (t) = ρi,0 a−3(1+wi ) (2.87)
Therefore
X ρi,0 a−3(1+wi )
H 2 (t) = H02 (2.88)
ρcrit,0
i
or finally, defining the scaled Hubble parameter E(t) = H(t)/H0
H 2 (t) X
E 2 (t) = = Ωi a−3(1+wi ) (2.89)
H02 i
In terms of the most typical components, i.e. total matter (w = 0), radiation (w = 1/3), cosmo-
logical constant (w = −1) and curvature (effective w = −1/3), we may write:
H 2 (t)  
E 2 (t) = 2 = Ωk a−2 + Ωm a−3 + Ωr a−4 + ΩΛ (2.90)
H0
Evaluating this today (a = 1), we have
1 = Ωk + Ωm + Ωr + ΩΛ (2.91)
so we may solve for one of the Ω’s in terms of the others. Typically we choose the curvature, so
Ωk = 1 − (Ωm + Ωr + ΩΛ ) (2.92)
The Friedmann eq. can then be written as
H 2 (t)  
E 2 (t) = 2 = (1 − Ωm − Ωr − ΩΛ ) a−2 + Ωm a−3 + Ωr a−4 + ΩΛ (2.93)
H0
72 CHAPTER 2. EXPANSION OF THE UNIVERSE

2.4 Solutions to the Friedmann Equation

Matter
For a matter dominated Universe, we have
ȧ2
= H02 Ωm a−3
a2
 2
da
= H02 Ωm a−1
dt
p
a1/2 da = H0 Ωm dt
(2.91)

integrating and setting imposing a = 0 at t = 0, we obtain

2 3/2 p
a = H0 Ωm t (2.92)
3
or
 p 2/3
3
a(t) = Ωm H0 t (Matter Domination) (2.93)
2
For an Einstein-de Sitter (EdS) Universe, Ωm = 1 and the age of the Universe (t =today, with
a = 1) is .
2
t = H0−1 (2.94)
3

Radiation
For a radiation dominated Universe, we have
ȧ2
= H02 Ωr a−4
a2
 2
da
= H02 Ωr a−2
dt
p
a da = H0 Ωr dt
(2.92)

integrating, we obtain
1 2 p
a = H 0 Ωr t (2.93)
2
or
 p 1/2
a(t) = 2 Ωr H0 t (Radiation Domination) (2.94)

Notice the Universe grows slower than in matter domination. For Ωr = 1 we have the age is
1
t = H0−1 (2.95)
2
2.4. SOLUTIONS TO THE FRIEDMANN EQUATION 73

Cosmological Constant
For a Universe dominated by a cosmological constant, we have

ȧ2
= H02 ΩΛ
a2
 2
da
= H02 ΩΛ a2
dt
da p
= H0 ΩΛ dt
a
(2.93)

integrating, we obtain
p
ln(a) = H0 ΩΛ t + const (2.94)

and we find
p 
a(t) ∝ exp ΩΛ H0 t (Cosmological Constant Domination) (2.95)

Notice the Universe grows exponentially fast in this case. In this case we cannot set a = 0 initially.

Curvature Dominated
For a Universe with only curvature, we have

ȧ2
= H02 Ωk a−2
a2
 2
da
= H02 Ωk
dt
p
da = H0 Ωk dt
(2.93)

integrating, we obtain
p
a(t) = H0 Ωk t (Curvature Domination) (2.94)

With Ωk = 1 the age is

t = H0−1 (2.95)

Matter + Radiation
For a flat Universe with both matter and radiation, we have

ȧ2 2
 −3 −4

= H 0 Ω m a + Ω r a
a2
 2
da
= H02 a−2 [Ωm a + Ωr ]
dt
(2.94)
74 CHAPTER 2. EXPANSION OF THE UNIVERSE

Therefore
da
→ H0 dt = √
a−1Ωm a + Ω r
1 ada
= √ p (2.94)
Ωr 1 + (Ωm /Ωr )a
Integrating, we have
Z a
1 a′ da′
t = √ p (2.95)
H0 Ωr 1 + (Ωm /Ωr )a′
0
1 2 Ω2r 
n
′
p
′+1
oa
= √ (Ω m /Ω r ) a − 2 (Ω m /Ω r ) a (2.96)
H0 Ωr 3 Ω2m 0

We may express this in terms of the scale factor of equality (aeq = Ωr /Ωm , see section 2.11.6 ):
 
1 2a2eq q
t = √ 2 + [a/aeq − 2] a/aeq + 1 (2.97)
H0 Ωr 3
or finally,
 
1 4a2eq q
t(a) = √ 1 − [1 − a/2aeq ] 1 + a/aeq (Matter + Radiation) (2.98)
H0 Ωr 3
This gives t(a), so implicitly a(t). For a ≪ aeq we have, up to a2 /a2eq :
 

 

1 4a2eq   2 2
 
t ≈ √ 1 − [1 − a/2aeq ] 1 + a/2aeq − a /8aeq (2.99)
H 0 Ωr 3   | {z }
 
(1−a2 /4a2eq )−a2 /8a2eq = 1−3a2 /8a2eq
 
1 4a2eq 3a2 a2 p
≈ √ 2
= √ → a(t) = (2H0 Ωr t)1/2
H 0 Ωr 3 8aeq 2H0 Ωr
which agrees with the radiation dominated solution. For a ≫ aeq we have:
 

 

 
1 4a2eq  q 
t ≈ √ 1 − [−a/2aeq ] a/aeq (2.99)
H0 Ωr 3   | {z } 


 
3/2 3/2 
1+a3/2 /2aeq ≈a3/2 /2aeq
( )  2/3
1 4a2eq a3/2 2 1 1/2 3/2 2 1 3/2 3 p
≈ √ = √ a a = √ a → a(t) = H0 Ωm t
H0 Ωr 3 3/2
2aeq 3 H0 Ωr eq 3 H0 Ωm 2
which agrees with the matter dominated solution.

Matter + Curvature
For a Universe with both matter and non-zero curvature, we have
ȧ2 2
 −3 −2

= H 0 Ω m a + Ω k a
a2
 2
da  
= H02 Ωm a−1 + Ωk
dt
(2.97)
2.4. SOLUTIONS TO THE FRIEDMANN EQUATION 75

Therefore
da
→ H0 dt = p
Ωm a−1 + Ωk
1 a1/2 da
= √ p (2.97)
Ωm 1 + (Ωk /Ωm )a
It turns out that it is easier to first solve for the conformal time dη = dt/a. We have
Z Z Z
dt 1 a−1/2 da
η = dη = = √ p (2.98)
a H0 Ωm 1 + (Ωk /Ωm )a
Let us assume we have a closed
p universe, i.e. k > 0 and p therefore Ωk = −k/H02 < 0. Then let
u2 = −Ωk /Ωm a, so that u = −Ωk /Ωm a1/2 and du = 1/2 −Ωk /Ωm a−1/2 da. We have
r Z
1 Ωm du
η = √ 2 √
H 0 Ωm −Ωk 1 − u2
2
= √ sin−1 u (2.98)
H0 −Ωk
or inverting
u = sin(θ/2) (2.99)
p
θ = H0 −Ωk η (2.100)
Under the same change of variables (a → u), u2 du = 1/2(−Ωk /Ωm )3/2 a1/2 da, and the equation for
t becomes
Z
1 a1/2 da
t = √ p
H0 Ωm 1 + (Ωk /Ωm )a
  Z
1 Ωm 3/2 u2 du
= √ 2 √
H0 Ωm −Ωk 1 − u2
Z
2Ωm u2 du
= √
H0 (−Ωk )3/2 1 − u2
Now changing u = sin(θ/2), du = cos(θ/2)dθ/2, we have
Z
2Ωm sin2 (θ/2) cos(θ/2)dθ/2
t = p
H0 (−Ωk )3/2 1 − sin(θ/2)2
Z
Ωm
= 3/2
sin2 (θ/2)dθ
H0 (−Ωk )
(2.96)
Now using cos(θ) = cos2 (θ) − sin2 (θ) = 1 − 2 sin2 (θ/2), we find
Z
Ωm
t = 1 − cos(θ)dθ
2H0 (−Ωk )3/2
Ωm
= (θ − sin(θ))
2H0 (−Ωk )3/2
(2.95)
76 CHAPTER 2. EXPANSION OF THE UNIVERSE
3.5
Ωm=1.0, -Ωk=0.0
Ωm=1.1, -Ωk=0.1
3 Ωm=2.0, -Ωk=1.0
Ωm=4.0, -Ωk=3.0
2.5

2
a(t)
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4
H0 t
Figure 2.4: Scale factor a(t) as a function of H0 t (cosmic time normalized by the Hubble time H0−1 ) for a
universe with only matter and curvature, with different values of Ωm and Ωk = 1 − Ωm . Since Ωk = −k/H0 ,
Ωk < 0 corresponds to a closed Universe (k > 0), which reaches a maximum turn-around scale factor
3/2
ata = Ωm /(−Ωk ) at time H0 tta = (π/2)Ωm /(−Ωk ). As Ωk → 0, both √ ata , tta → ∞ and the solution
approaches that of a flat Universe without turn-around, i.e. a(t) = (3/2 Ωm H0 t)2/3 .

Finally, recall that a = (Ωm / − Ωk )u2 = (Ωm / − Ωk ) sin2 (θ/2) so that we have a parametric solution
for a cycloid
Ωm
a = [1 − cos(θ)] (2.96)
2(−Ωk )
Ωm
t = [θ − sin(θ)] (Matter + Curvature) (2.97)
2H0 (−Ωk )3/2
p
θ = H0 −Ωk η (2.98)
Notice that for small values of θ, we have sin θ ≈ θ − θ3 /6 and cos θ ≈ 1 − θ2 /2, so that:
 
Ωm 3 12H0 1/3
t≈ 3/2
θ →θ= (−Ωk )1/2 t1/3 (2.99)
12H0 (−Ωk ) Ω m
(2.100)
so that
 2/3  p 2/3
Ωm Ωm 12H0 3
a≈ θ2 = 2/3 (−Ωk )t 2/3
= Ωm H 0 t (2.101)
4(−Ωk ) 8 (−Ωk ) Ωm 2

Matter + Cosmological Constant

For a flat Universe with both matter and cosmological constant, we have
ȧ2 2
 −3

= H 0 Ω m a + Ω Λ
a2
 2
da  
= H02 Ωm a−1 + ΩΛ a2
dt
(2.100)
2.5. PHOTON GEODESICS AND ENERGY 77

Therefore
da
→ H0 dt = p
Ωm a−1 + ΩΛ a 2
1 a1/2 da
= √ p ( Let u2 = ΩΛ /Ωm a3 )
Ωm 1 + (ΩΛ /Ωm )a3
2/3 du
= √ √
ΩΛ 1 + u 2
(2.98)
integrating, we obtain
r !
2/3 2/3 ΩΛ 3/2
H0 t = √ sinh−1 (u) = √ sinh−1 a (2.99)
ΩΛ ΩΛ Ωm
and
 1/3  √ 
Ωm 2/3 3 ΩΛ H0
a(t) = sinh t (Matter + Cosmological Constant) (2.100)
ΩΛ 2
which reduces to the matter dominated solution for small t:
   √ 2/3  p 2/3
Ωm 1/3 3 ΩΛ H0 2/3 3
a(t) ≈ t = Ωm H0 t (2.101)
ΩΛ 2 2
and recovers the cosmological constant solution for large t [sinh(at) = (eat − e−at )/2 → eat /2 ]
  √ 2/3 p 
3 ΩΛ H 0
a(t) ∝ exp t = exp ΩΛ H0 t (2.102)
2

2.5 Photon Geodesics and Energy

Recall we defined the 4-momentum P α = (E, p) for a massive particle as
dxα
Pα = m (2.103)
dτ
But for a massless particle (e.g. a photon), both m = dτ = 0, so we need an alternative definition.
We define it then with respect to a general implicit parameter λ along the particle trajectory:
 
α dxα dP α dE dp
P = and = , (2.104)
dλ dλ dλ dλ
This definition is valid both for massive and massless particles. We can replace λ in favor of t as
d dx0 d d
= 0
=E (2.105)
dλ dλ dx dt
The zeroth component of the Geodesics equation for a photon then reads
dP 0
= −Γ0 αβ P α P β = − Γ0 ij P i P j
dλ |{z}
δij aȧ
dE
E = −δij aȧP i P j (2.105)
dt
78 CHAPTER 2. EXPANSION OF THE UNIVERSE

Now in the FRW metric we have

P µ Pµ = gµν P µ P ν = g00 (P 0 )2 + gij P i P j = −E 2 + δij a2 P i P j = −m2 = 0 (2.106)

Therefore δij P i P j = E 2 /a2 and

dE E2
E = −aȧ
dt a2
dE da
→ = − (2.106)
E a
Therefore ln E ∝ ln a−1 and
1
E∝ (2.107)
a
i.e. the energy of photons decreases as a grows in an expanding Universe. This is yet another way
of seeing the effect of the radiation redshift. This is why the energy density of matter decreases as
a−3 following the volume dilution, but that of photons decrease with a−4 : there is an extra factor
of 1/a due to the redshift of the photon energy itself.
Suppose a photon is emitted at a with energy Eem = α/a (with α =const.) and observed at
aobs with Eobs = α/aobs . We have that Eobs = Eem a/aobs . For observation today aobs = a0 = 1
and Eobs = aEem . Since emission happens at a < 1, we have Eobs < Eem .

2.6 Redshift
Photons travel through null geodesics in which ds2 = 0. Consider a light ray emitted from a galaxy
at time te and observed at time to . We have
Z Z to
2 2 2 2 2 cdt
ds = c dt − a dD = 0 → D = dD = (2.108)
te a

Suppose now that the same galaxy emits another light pulse at te + δte – where δte is the wave
period – which is then observed at to + δto . The radial comoving distance D is obviously the same,
so we have
Z to +δto
cdt
D= (2.109)
te +δte a

Comparing both expressions gives

Z to +δto Z to
cdt cdt
= (2.110)
te +δte a te a
Z te +δte Z to +δto
cdt cdt
→ = (2.111)
te a to a

For δt ≪ t, we can approximate a as constant in the integrals and write

δte δto
= (2.112)
a(te ) a(to )
2.6. REDSHIFT 79

Taking cδt = λ to be the wavelength, we have

λe cδte a(te )
= = (2.113)
λo cδto a(to )

Therefore, defining the redshift z at observation as the fractional change in wavelength

∆λ λo − λe
z(to ) = = (2.114)
λe λe

or equivalently

λo a(to )
1 + z(to ) = = (2.115)
λe a(te )

For to =today, we define a(to ) = 1 and

1
1+z = (2.116)
a
Therefore, the redshift defined in terms of change in wavelength of photons from galaxies is related
to the scale factor at emission. By measuring shifts in line emissions of galaxies, we are directly
measuring the scale factor at the time of emission and, as we shall see soon, we can obtain the
distance to the galaxy, given a cosmology.

2.6.1 Sandage-Loeb Test

References:
Sandage, ApJ 139, 319, (1962);
Loeb, ApJ 499, L111 (1998);
Corasaniti et al (2008), arxiv:0701433).

Consider now that, in addition to making an observation at time to , we make another observation
at time to + ∆to of the same galaxy’s redshift. For instance, ∆to = 10 years. Again, the invariance
of the comoving distance gives us

∆te ∆to
= (2.117)
a(te ) a(to )

and

a(to + ∆to )
1 + z(to + ∆t0 ) = (2.118)
a(te + ∆te )

The difference in the two redshift measurements is

∆z = z(to + ∆to ) − z(to )

a(to + ∆to ) a(to )
= − (2.118)
a(te + ∆te ) a(te )
80 CHAPTER 2. EXPANSION OF THE UNIVERSE

Now we still have ∆te ≪ te , so we expand to first order

a(to + ∆to ) a(to ) + ȧ(to )∆to a(to )(1 + ȧ(t o)

a(to ) ∆to )
= = (2.119)
a(te + ∆te ) a(te ) + ȧ(te )∆te a(te )(1 + ȧ(t e)
a(te ) ∆te )
  
a(to ) ȧ(to ) ȧ(te )
≈ 1+ ∆to 1 − ∆te (2.120)
a(te ) a(to ) a(te )
 
a(to ) ȧ(to ) ȧ(te )
= 1+ ∆to − ∆te (2.121)
a(te ) a(to ) a(te )
a(to ) ȧ(to ) a(to )ȧ(te )
= + ∆to − ∆te (2.122)
a(te ) a(te ) a2 (te )
a(te )
Using ∆te = a(to ) ∆to (see Eq. 2.112) , we have

a(to + ∆to ) a(to ) ȧ(to ) a(to )ȧ(te )

− = ∆to − ∆te
a(te + ∆te ) a(te ) a(te ) a2 (te )
 
ȧ(to ) a(to )ȧ(te ) a(te )
= ∆to − ∆to
a(te ) a2 (te ) a(to )
ȧ(to ) − ȧ(te )
= ∆to (2.121)
a(te )

Using H0 = H(to ) = ȧ(to )/a(to ) and a(to )/a(te ) = 1 + z(t0 ), we have

ȧ(to ) − ȧ(te )
∆z = ∆to
a(te )
 
ȧ(to ) a(to ) ȧ(te )
= − ∆to (2.121)
a(to ) a(te ) a(te )
= (H0 (1 + z(t0 )) − H(te )) ∆to
(2.121)

Setting to =today, a(to ) = 1, z(t0 ) = z, H(te ) = H[a(te )] = H(z) and E(z) = H(z)/H0 we have
 
H(z)
∆z = (1 + z) − H0 ∆to (2.122)
H0
 
∆z E(z)
→ = 1− H0 ∆to (2.123)
1+z 1+z
(2.124)

where, from the Friedmann equation, in a flat Universe

 1/2
E(z) = Ωm (1 + z)3 + ΩDE (2.125)

For Ωm = 0.3 and ΩDE = 0.7, at z = 1 we have E(z = 1) = 1.76, so

∆z
≈ 0.12H0 ∆to (2.126)
1+z
(2.127)
2.7. COSMOLOGICAL DISTANCES AND TIMES 81

For H0 = 70 km s−1 Mpc−1 = 0.72 × 10−10 year−1

∆z
≈ 0.86 × 10−11 year−1 ∆to (2.128)
1+z
(2.129)

So observing a galaxy today at z = 1 and again 10 years from now, we would only have a
fractional change of ∼ 10−10 in its redshift.

2.7 Cosmological Distances and Times

2.7.1 Comoving Radial Distance
The comoving radial distance D can be obtained by considering the a radial path of photons, in
which we have dα2 = 0 (radial) and ds2 = −dt2 + a2 (t)dD2 = 0 (photons), so that D can be
expressed as
Z Z age Z 1 Z 0
dt da dz
D = dD = = =−
t a(t) a ȧa z H(z)
Z z
dz
= (2.129)
0 H(z)

where we used da = −a2 dz and H(z) = ȧ/a. Notice that D depends on the curvature only via the
Hubble parameter from the Friedmann’s equations. We may also define a physical radial distance
dp = a(t)D.

2.7.2 Comoving Horizon

The comoving horizon DH is similar to D, but instead of integrating from z = 0 to a certain redshift
z, we integrate from z to z = ∞, effectivelly finding the comoving size of the universe at z:
Z t Z a Z ∞
dt da dz
DH = = = (2.130)
0 a(t) 0 ȧa z H(z)

We may also define a physical horizon dH = a(t)DH .

2.7.3 Comoving Angular Diameter Distance

The comoving angular diameter distance DA is defined such that it gives an object’s comoving size
dl when it is multiplied by the object angular size dα

dl = DA dα (2.131)

From the metric definition, with dD = 0 we can see that it is given in terms of D by

√ sinh(D), k = −1, Negative Curvature, Open Universe
sin( kD) 
DA = fk (D) = √ = D, k = 0, Zero Curvature, Flat Universe (2.132)
k 
sin(D), k = +1, Positive Curvature, Closed Universe
82 CHAPTER 2. EXPANSION OF THE UNIVERSE

or similarly, with k = −H02 Ωk :

 sinh[√Ω H D]
√  √ k 0 , Ωk > 0, Negative Curvature, Open Universe
sin[ −Ωk H0 D]  Ωk H 0
DA = fk (D) = √ = D, √ Ωk = 0, Zero Curvature, Flat Universe (2.133)
−Ωk H0 
 sin[√ −Ωk H0 D]
−Ω H
, Ωk < 0, Positive Curvature, Closed Universe
k 0

We may also define a physical angular diameter distance dA (t) = a(t)DA

2.7.4 Comoving Luminosity Distance

In Euclidean space, the flux F (energy/time/area) crossing a spherical surface at distance x (area
A = 4πx2 ) due to an object emitting radiation with luminosity L (energy/time) is defined as
L
F = (2.134)
4πx2
because the luminosity is constant through any sphere. In an expanding Universe, we imagine the
source is at emission scale factor a = a(te ) and at the center of a comoving grid, while the observer
is at a = 1, sitting in a shell at a distance D(a) from the source. In this case the comoving flux
crossing the observer is
Lo (D)
F = 2 (2.135)
4πDA

where DA is the comoving angular diameter distance and Lo (D) is the observed luminosity at D.
But Lo (D(a)) is the number of photons per unit time, multiplied by the energy of each photon
crossing the observer shell, so we have
dN
Lo (D) = Eo (2.136)
dto
where Eo = aEe is the energy of a photon at distance D(a) given its energy at emission was Ee
(redshift) and dto = dte /a is the time to cross shell D given that the time at emission was dte (time
delay). Therefore
dN dN
Lo (D) = Eo = a 2 Ee = a2 L (2.137)
dto dte
where L is the luminosity at emission. Therefore

La2
F = 2 (2.138)
4πDA
The physical luminosity distance dL is defined such that the Euclidean relation remains valid for
the comoving flux, i.e.
L
F = (2.139)
4πd2L
and comparing with the previous equation, we conclude that
DA dA
dL = = 2 (2.140)
a a
2.7. COSMOLOGICAL DISTANCES AND TIMES 83

In the case of a flat universe we have

D d
dL = = 2 (Flat) (2.141)
a a
In any case, the relation a2 dL = dA is always true for FRW cosmologies, independent of curvature
and/or cosmology. It provides a consistency check for the homogeneity and isotropy of the Universe.
Finally, the comoving luminosity distance is
dL DA fk (D)
DL = = 2 = (2.142)
a a a2

2.7.5 Comoving Volume

The comoving volume element in spherical coordinates is given by
2 (z)
DA
dV (z) = (DA dθ)(DA sin θdφ)dD = dzdΩ, (2.142)
H(z)
where DA (z) is the comoving Angular Diameter Distance at redshift z, H(z) is the Hubble param-
eter and dΩ = sin θdθdφ is the solid angle.

2.7.6 Comoving and Physical Distances

As mentioned above, all distance definitions have a physical and comoving version. The physical
distance d is always obtained by multiplying the comoving distance D by the scale factor a(t). This
holds also for the luminosity and angular-diameter distances such that:
dp = a(t)D (2.143)
dH = a(t)DH (2.144)
dA = a(t)DA (2.145)
dL = a(t)DL (2.146)
and the physical volume is
2 (z)
DA
dVphys = (dA dθ)(dA sin θdφ)d(dp ) = a3 (t) dzdΩ = a3 (t)dV (2.147)
H(z)

2.7.7 Conformal Time

The conformal time η is defined such that the FRW metric looks the same for time and space
coordinates:
 
 
ds2 = −dt2 + a2 (t)dl2 = a2 (t) − dt2 /a2 (t) +dl2  = a2 (t) −dη 2 + dl2
 
(2.148)
| {z }
dη 2

Therefore, the conformal time is related to the physical time t as

dt
dη = (2.149)
a(t)
just like comoving coordinates dx are related to physical coordinates dr via dx = dr/a(t). In fact,
in units where c = 1, the conformal time is equivalent to the comoving horizon:
Z Z t
dt
η = dη = = DH (2.150)
0 a(t)
84 CHAPTER 2. EXPANSION OF THE UNIVERSE

2.7.8 Examples
In general, cosmological distances must be computed by numerical integration of the inverse Hubble
parameter. In a some cases though we may find simple analytical solutions. Let us see a few cases
where we can find solutions.

Einstein de Sitter (Ωm = 1)

For Ωm = 1.0 (all others zero), we have from the matter dominated solution:
p
H(z) = H0 Ωm (1 + z)3/2 (2.151)

so the comoving radial distance is:

Z z Z z
dz 1 dz
D = = √
0 H(z) H 0 Ω m 0 (1 + z)3/2
 
2 1
= √ 1− √ (2.151)
Ωm H0 1+z
In this case, DA = D.

Matter + Curvature
We have:
p
H(a) = H0 Ωm a−3 + Ωk a−2 (2.152)

So
Z 1 Z 1 Z 1
da 1 da 1 a−1/2 da
D(a) = 2
= p = √ q (2.153)
a a H(a) H0 a a2 Ωm a−3 + Ωk a−2 H 0 Ωm a 1 + ΩΩmk a

This integral is the same integral found in the computation of the conformal time Eq. 2.98
for this cosmology, but with different integration limits. Therefore we can read immediately the
answer:

 h q i  q
2 −1 Ωk Ωk −1

H 0 Ωk
sinh √ − sinh Ωm a ,
Ωm k < 0,

2 √
D(a) = √ [1 − a] ,
Ωm H 0 h
k = 0, (2.154)

 q  q i
 √2 −1 −Ωk −1 −Ωk
H −Ω
sin Ωm − sin Ωm a , k > 0.
0 k

The angular-diameter distance DA is given by:

 sinh[√Ω H D]
k 0
 √Ωk H 0 D ,
 k < 0, Ωk > 0,
DA = D, √ k = 0, Ωk = 0, (2.155)

 sin[√ −Ωk H0 D]
−Ω H D
, k > 0, Ωk < 0.
k 0

We can see that, e.g. in the closed Universe case (k > 0), the sin will be applied to a combinations
of sin−1 , so we expect some simplification.
2.7. COSMOLOGICAL DISTANCES AND TIMES 85

Indeed, after a long and detailed algebra (see Appendix D), this simplification occurs and we
find a simple analytical expression that is valid for all values of k:
2 h p i
DA (z) = 2 − Ω m (1 − z) − (2 − Ω m ) Ω m z + 1 . (2.156)
H0 Ω2m (1 + z)
This result was first derived in 1958 by W. Mattig (see Mattig, Astronomische Nachrichten 284
pg. 109, 1958). Notice that in the case of a flat universe Ωm = 1 and Ωk = 0, we find
 
2  √  2 1
DA (z) = 2 − (1 − z) − (2 − 1) z + 1 = 1− √ , (2.157)
H0 (1 + z) H0 z+1
so we indeed recover the Einstein-de Sitter solution.
Unfortunately, the real universe seems to i) contain dark energy and ii) be very close to flat, so
Mattig’s formula is not very useful nowadays.

Matter + Radiation
For a flat universe with only matter and radiation, we have:
p
H(a) = H0 Ωm a−3 + Ωr a−4 (2.158)
and Ωk = 1 − Ωm − Ωr = 0. So
Z 1 Z 1 Z 1
da 1 da 1 da
D(a) = 2
= √ = √ q (2.159)
a a H(a) H0 a a2 Ωm a−3 + Ωr a−4 H0 Ωm a a + ΩΩmr
" r #1 "r r #
1 Ωr 2 Ωr Ωr
= √ 2 a+ = √ 1+ − a+ (2.160)
H0 Ωm Ωm H0 Ωm Ωm Ωm
a

Matter + Radiation + Curvature

For a universe with matter, radiation and curvature, we have:
p
H(a) = H0 Ωk a−2 + Ωm a−3 + Ωr a−4 (2.161)
and Ωk = 1 − Ωm − Ωr . So
Z 1 Z 1
da 1 da
D(a) = 2
= p (2.162)
a a H(a) H0 a a Ωk a + Ωm a−3 + Ωr a−4
2 −2
Z 1
1 da
= √ q (2.163)
H0 Ωk a a2 + ΩΩmk a + ΩΩr
k
r !1
1 Ω Ω Ω
= √ log 2 a2 + m a + r + 2a + m (2.164)
H0 Ωk Ωk Ωk Ωk
" r ! a r !#
1 Ωm Ωr Ωm Ωm Ωr Ωm
= √ log 2 1 + + +2+ − log 2 a2 + a+ + 2a +
H0 Ωk Ωk Ωk Ωk Ωk Ωk Ωk
(2.164)
p
Since Ωk + Ωm + Ωr = 1 and a2 H(a)/H0 = Ωk a2 + Ωm a + Ωr we have
    2 
1 2 Ωm 2a H(a) Ωm
D(a) = √ log √ +2+ − log √ + 2a + (2.165)
H 0 Ωk Ωk Ωk H0 Ωk Ωk
86 CHAPTER 2. EXPANSION OF THE UNIVERSE

2.7.9 Caution and Notation

The notation we use here is not standard in the literature. In most books, the angular diameter
distance is always physical. People do not even talk about comoving angular diameter distance.
When they refer to it, they call it comoving transverse distance and denote it by DM (see e.g. Hogg
2000). Likewise, the luminosity distance is always considered to be physical, not comoving.
I find this a bit confusing, because for any distance you can in principle define its comoving and
physical versions. Especially for the comoving angular diameter distance, which appears directly
in the metric, depending of how you write it.
Even though it is only a matter of notation, beware that most likely you will find the following
notation in the literature:

Comoving radial distance : Literature: χ

Comoving transverse distance : Literature: DM or r
Angular diameter distance : Literature: dA or DA (always physical)
Luminosity diameter distance : Literature: dL or DL (always physical)

2.8 Distance-Redshift and Acceleration

Let us consider the distance-redshift relation for small redshifts so that we can have a sense of the
effects of an accelerated expansion on distances. Recall the Deceleration parameter q(t) was

ä/a ä 1
q(t) = − 2 2
=− 2
= (1 + 3w) (2.162)
(ȧ /a ) aH 2

where w = P/ρ. For small enough redshifts, we may expand the scale factor around today t0 as

1
a(t) = a(t0 ) + ȧ(t0 )(t − t0 ) + ä(t0 )(t0 − t)2 (2.163)
2

Since a(t0 ) = 1, we have H0 = ȧ(t0 ) and q0 = −ä(t0 )/H02 and therefore

1
a(t) = 1 − H0 (t0 − t) − q0 H02 (t0 − t)2 (2.164)
2

Using (1 + x)α = 1 + αx + α(α − 1)x2 /2, with α = −1, we find a(t)−1 up to second order in (t0 − t):

1
a(t)−1 ≈ 1 + H0 (t0 − t) + q0 H02 (t0 − t)2 + H02 (t0 − t)2 (2.165)
2
 
1
≈ 1 + H0 (t0 − t) + 1 + q0 H02 (t0 − t)2 (2.166)
2

Therefore the redshift z becomes:

z(t) = (a(t)−1 − 1)
 
1
≈ H0 (t0 − t) + 1 + q0 H02 (t0 − t)2 (2.166)
2
2.9. ENERGY EVOLUTION 87

so that (t0 − t) becomes up to second order in z:

 
z 1
(t0 − t) ≈ − 1 + q0 H0 (t0 − t)2 (2.167)
H0 2 | {z }
≈z 2 /H02
 
z 1 z2
≈ − 1 + q0 (2.168)
H0 2 H0
and the comoving distance becomes, up to second order in (t0 − t):
Z t0 Z t0   t0
dt′ ′
 ′
 ′ 1 ′ 2
D = ≈ dt 1 − H0 (t − t0 ) = t − H0 (t − t0 ) (2.169)
t a(t′ ) t 2 t
1
≈ (t0 − t) + H0 (t − t0 )2 (2.170)
2
Now we use Eq. 2.168 and keep terms up to second order in z:
   2
z 1 z 1 z2
D ≈ − 1 + q0 + H0 2 (2.171)
H0 2 H0 2 H0
z z2 1 z2 1 z2
≈ − − q0 + (2.172)
H0 H0 2 H0 2 H0
z z2 1 z2
≈ − − q0 (2.173)
H0 2H0 2 H0
 
z 1
≈ 1 − (1 + q0 )z (2.174)
H0 2
or using q0 from the equation above,
 
z 3
D≈ 1 − (1 + w)z (2.175)
H0 4
Therefore, in a universe that is accelerating, q0 < 0 (e.g. w = −1) and distances at a fixed redshift
are larger than expected in a decelerating universe (e.g. w = 0). This effect on the luminosity
distances of supernovae allowed us in 1998 to conclude that the universe is accelerating.
Given the exact solution for a matter dominated (Ωm = 1) universe, we may expand it in z:
    
2 h −1/2
i 2 z 3 2 2 z 3 2
D = 1 − (1 + z) ≈ 1− 1− + z ≈ − z (2.176)
H0 H0 2 8 H0 2 8
 
z 3
≈ 1− z (2.177)
H0 4
from which, comparing to Eqs. 2.174 and 2.175 we find q0 = 1/2 and w = 0, as expected.

2.9 Energy Evolution

2.9.1 Covariant Conservation
The Bianchi identity says that the covariant derivative of the Einstein Tensor is zero:

∇µ Gµν = 0 (2.178)
88 CHAPTER 2. EXPANSION OF THE UNIVERSE

which, through the Einstein equations, automatically imply that the Energy-Momentum tensor is
covariantly conseved:

∇µ T µν = 0 (2.179)

Therefore energy conservation is automatically guaranteed in General Relativity.

The components of this equation give us the equations describing energy and momentum con-
servation in an expanding universe. For the energy-momentum tensor of a perfect fluid, the ν = 0
equation implies (T 00 = g 00 T 0 0 = ρ and T ij = g ik T i k = −δik /a2 (−δik P ) = δij P/a2 ):

∇µ T µ0 = ∂µ T µ0 + Γµµλ T λ0 + Γ0µλ T µλ
= ∂0 T 00 + Γ00λ T λ0 + Γiiλ T λ0 + Γ00λ T 0λ + Γ0iλ T iλ (Γ00λ = 0)
= ∂0 T 00 + Γiiλ T λ0 + Γ0iλ T iλ (Γiij = Γ0i0 = 0)
= ∂0 T 00 + Γii0 T 00 + Γ0ij T ij
 
ȧ δij P
= ∂0 ρ + δii ρ + (δij aȧ)
a a2
ȧ ȧ
= ρ̇ + 3 ρ + 3 P = 0 (2.175)
a a
or with P = wρ:

dρ
+ 3H(1 + w)ρ = 0 (2.176)
dt

Without expansion (H = 0) this is simply the usual energy conservation dρ/dt = 0.

2.9.2 Combined Einstein Equations

Since the conservation of Tµν is not independent of the Einstein equations themselves, Eq. 6.112
is not independent from the Friedmann equations. In fact, it can be obtained by combining them.
Differentiating the first Friedmann equation, we have

2ȧä 2ȧ3 8πG

2
− 3 = ρ̇
" a  a # 3
ȧ ä ȧ 2 4πG
− = ρ̇ (2.176)
a a a 3

Using both the first and second Friedmann equations above we have
 
4πG 8πG 4πG
H − (ρ + 3P ) − ρ = ρ̇
3 3 3
−H [(ρ + 3P ) + 2ρ] = ρ̇
(2.175)

or

ρ̇ = −3H(ρ + P ) (2.176)
2.9. ENERGY EVOLUTION 89

2.9.3 Thermodynamics
Yet another way to obtain the conservation equation is simply from energy conservation, encoded
in the first law of thermodynamics

dE = −pdV
d(ρV ) = −pdV (V ∝ a3 )
d(ρa3 ) = −pd(a3 )
a3 dρ + 3ρa2 da = −3a2 pda (2.174)

and:
3ρ 3P da
dρ = − da − da = −3(ρ + p)
a a a
from which we get

dρ da/dt
= −3(ρ + P ) (2.174)
dt a
or

dρ
= −3Hρ(1 + w) (2.175)
dt

2.9.4 Solution
We can obtain the general solution to this equation as

dρ da/dt
= −3 ρ[1 + w(t)]
dt a
dρ da
= −3[1 + w(t)]
ρ a
d ln ρ = −3[1 + w(t)]d ln a
Z
ln ρ = −3 [1 + w(a)]d ln a + const.
 Z a 
(1 + w(a))
ρ(a) = ρ(1) exp −3 da (2.172)
1 a

In terms of redshift z, a = (1 + z)−1 , da = −(1 + z)−2 dz, so that da/a = −dz/(1 + z) and:

 Z z 
[1 + w(z)]
ρ(z) = ρ(0) exp 3 dz (2.173)
0 1+z

Solutions for constant w

We can find solutions for cases when the universe content is dominated by different species with
constant w:
90 CHAPTER 2. EXPANSION OF THE UNIVERSE

 Z z 
dz
ρ(z) = ρ(0) exp 3(1 + w) = ρ(0) exp [3(1 + w) ln(1 + z)] (2.174)
0 1+z
or

ρ(z) = ρ(0)(1 + z)3(1+w) (2.175)

For matter w = 0, radiation w = 1/3 and for a cosmological constant w = −1. Therefore:

ρ(z) = ρ(0)(1 + z)3 Matter Domination (2.176)

ρ(z) = ρ(0)(1 + z)4 Radiation Domination (2.177)

ρ(z) = ρ(0) = const. Cosmological Constant Domination (2.178)

2.10 Equilibrium Thermodynamics

Kolb & Turner

For some of the components of the universe, all we will need to specify will be their energy
density and their pressure, or equation of state. In this case we will be able to use the fluid
approximation to get the relevant evolution equations.
More generally, we need to specify the full distribution function f (x, p, t) of a species in phase
space (x, p) and time t, defined such that

dN = f (x, p, t)d3 xd3 p (2.179)

is the number of particles in phase space element d3 xd3 p.

In thermodynamical equilibrium, the distribution function is independent of position and an-
gular direction, and given by

1 + Fermi-Dirac
f (x, p, t) = f (p, t) = (E−µ)/T (2.180)
e ± 1 − Bose-Einstein
p
where E = p2 + m2 , and both cases reduce to the Maxwell-Boltzmann distribution in the classical
limit (high temperatures and low densities):

f (p, t) ∝ e−(E−µ)/T Classical (2.181)

where µ is the chemical potential. From the distribution function, we may define quantities averaged
in momentum, such as the the number density, energy density and pressure, respectively:
Z
d3 p
n(x, t) = g f (x, p, t) (2.182)
(2π)3
Z
d3 p
ρ(x, t) = g Ef (x, p, t) (2.183)
(2π)3
Z
d 3 p p2
P (x, t) = g f (x, p, t) (2.184)
(2π)3 3E
2.10. EQUILIBRIUM THERMODYNAMICS 91

where g is the number of internal degrees of freedom. For the Equilibrium distributions, isotropy
allows us to use d3 p = 4πp2 dp and get
Z
g p2 dp
n(t) = √ (2.185)
2π 2 2 2
e( p +m −µ)/T ± 1
Z
g p2 dp p
ρ(t) = √ p 2 + m2 (2.186)
2π 2 e ( p 2 +m2 −µ)/T
±1
Z
g p2 dp p2
P (t) = √ p (2.187)
2π 2 e( p +m −µ)/T ± 1 3 p2 + m2
2 2

The chemical potential is small in the early Universe, such that µ ≪ T . For our purposes in
the chapter we may set it equal to zero µ = 0. We will restore it in the next chapter.

2.10.1 Relativistic
For relativistic particles (p ≫ m) or (T ≫ m), and assuming (T ≫ µ), we have
Z Z 
g p2 dp gT 3 x2 dx (ζ(3)/π 2 )gT 3 , − BE
n(t) = 2
= 2 x
= 2 )gT 3 , + FD (2.188)
2π e p/T ±1 2π e ±1 (3/4)(ζ(3)/π
| {z }
(3/4) × Γ(3) ζ(3)
| {z } |{z} |{z}
+sign =2 =1.202
Z Z 
g p3 dp gT 4 x3 dx (π 2 /30)gT 4 , − BE
ρ(t) = = = 2 4 (2.189)
2π 2 ep/T ± 1 2π 2 ex ± 1 (7/8)(π /30)gT , + FD
| {z }
(7/8) × Γ(4) ζ(4) ,±
| {z } |{z} |{z}
+sign =6 =π 4 /90
Z
1 g p3 dp ρ(t)
P (t) = = , (2.190)
3 2π 2 ep/T ±1 3

where we used properties of the integrals above (see Appendix G). The latter equation implies
w = 1/3 for radiation.
The second law of thermodynamics in an expanding Universe can be written as

0 = T dS = dE + P dV = d(ρV ) + P dV = d[(ρ + P )V ] − V dP (2.191)

where the zero above is because the energy evolution equation is equivalent to the first law dE =
−P dV , i.e. the expansion happens adiabatically. Since dS = 0, we will obtain a more useful
relation for the conserved entropy S. From T dS = d(ρV ) + P dV , we may similarly write

V (ρ + P )
dS = dρ + dV (2.192)
T T
and we have

∂S  ρ+P
= , (2.193)
∂V T T

∂S  V dρ
= . (2.194)
∂T V  T dT
92 CHAPTER 2. EXPANSION OF THE UNIVERSE

The equality of crossed partial derivatives gives

∂2S ∂2S
=
∂T ∂V ∂V ∂T
ρ+P 1 dρ 1 dP 1 dρ
− + + = (2.194)
T2 T dT T dT T dT
and we find
dP (ρ + P )
= (2.195)
dT T
or
ρ+P
dP = dT (2.196)
T
One can also obtain this by inspection of the definitions of P and ρ from the distribution function.
Plugging this into the second law, we have
 
d[(ρ + P )V ] V d[(ρ + P )V ] V (ρ + P ) (ρ + P )V
dS = − dP = − dT = d , (2.197)
T T T T2 T
| {z }
S

so the entropy S = (ρ + P )V /T is constant. Defining the entropy density s

S (ρ + P )
s= = (2.198)
V T
since V ∝ a(t)3 , we find that a(t)3 s is conserved:

a(t)3 s = const. (2.199)

For relativistic species, we have P = ρ/3 and

(ρ + P ) 4ρ
s= = (2.200)
T 3T
Using Eq. 2.189 in the equation above, we have the entropy density in species i is

(2π 2 /45)gi Ti3 BE
si = 2 3 (2.201)
(7/8)(2π /45)gi Ti FD
and the total entropy density can be written
"   #
2π 2 X
3 7 X
s= g i Ti + gi Ti3 (2.202)
45 8
i=bosons i=fermions

For a common temperature Ti = T to all species, we may write

2π 2
s = g∗ T 3 (2.203)
"45 #
 
X 7 X
g∗ = gi + gi (2.204)
8
i=bosons i=fermions

where g∗ is an effective number of (bosonic) degrees of freedom.

2.11. COSMIC INVENTORY 93

2.10.2 Non-relativistic
For non-relativistic particles (p ≪ m) or (T ≪ m), the energy is
p2
E ≈m+ (2.205)
2m
So we recover the Maxwell-Boltzmann distribution
2 /2mT
f (p, t) = e−E(p)/T ∼ e−m/T e−p (2.206)
Then we have
Z
g −m/T 2 /2mT
n(t) = e p2 e−p dp
2π 2
Z  3/2
g −m/T mT
= e (2mT )3/2 x2 e−x dx = g e−m/T , (2.206)
2π 2 2π
| {z √ }
Γ(3/2)
2
= 4π
Z
g −m/T 2 /2mT
ρ(t) = e p2 |{z}
E e−p dp ≈ mn(t) , (2.207)
2π 2
≈m
Z
g −m/T p4 −p2 /2mT
P (t) = e e dp
2π 2 3m
 
mT 3/2 −m/T
Z
g −m/T (2mT )5/2 4 −x
= e x e dx = gT e ≈ T n(t) , (2.207)
2π 2 3m 2π
| {z √ }
Γ(5/2)
2
= 32 4π

Since n = N/V , the latter is simply the ideal gas law: P V = N kB T . Also since T ≪ m, we have
P ≪ ρ. That is why we take w = 0 for non-relativistic matter.

2.11 Cosmic Inventory

2.11.1 Photons
The photons in the Cosmic Microwave Background (CMB) were first detected in the 1960’s by
Penzias and Wilson, for which they earned the Physics Nobel Prize in 1978. The temperature and
anisotropies of the CMB photons were more precisely measured by the COBE satellite, for which
Mather and Smoot earned the Physics Nobel prize in 2006. This temperature has since been further
measured with ever increasing precision in CMB surveys such as WMAP, SPT, ACT and Planck.
The CMB radiation is actually the most perfect Blackbody radiation known to date, with present
temperature of Tγ ∼ 2.725 K.
Photons satisfy the BE statistics with g = 2 (polarizations) and µ = 0 (Planck spectrum) and
we have
π2 4
ργ = T (2.208)
15 γ
Notice that we derived before that ργ ∝ 1/a4 , and therefore we must have Tγ ∝ 1/a. For Tγ =
2.725K, the radiation density parameter value is
ργ 2
Ω γ h2 = h = 2.47 × 10−5 (2.209)
ρcrit,0
94 CHAPTER 2. EXPANSION OF THE UNIVERSE

Later we will consider departures from the BE distribution, which will lead us to perturbations
in the temperature field and in the energy density of photons. This will then bring us to study the
CMB anisotropies later.

2.11.2 Neutrinos
Neutrinos are fermions, and even though we believe they have a non-zero mass, at the early Universe
they were relativistic and kept in equilibrium in the cosmic plasma via the weak interactions. At
T ∼ 1 MeV, the weak interactions become ineffective in an expanding Universe (the interaction
rate becomes smaller than the expansion rate). As a result, neutrinos decouple from the plasma.
Right after that, at T ∼ me = 500 keV= 0.5 MeV, the reaction
e− + e+ ↔ 2γ (2.210)
becomes ineffective to the left, because photons do not have enough energy on average to make
the reaction happen (Eγ must be at least me for pair production conserving both energy and
momentum). As the reaction proceeds only to the right, electrons and positrons annihilate each
other and dump extra energy into the photons. Because neutrinos had already decoupled from the
plasma, they do not participate of the energy injection from electron/positron annihilation, and
end up having a temperature smaller than photons later on.
Let a1 denote the scale factor at some time before electron/position annihilation, and a2 at some
point after. From entropy conservation we have a31 s(a1 ) = a32 s(a2 ), which allows us to compute the
neutrino temperature relative to the photon temperature.
At a1 we have as bosons: photons (g = 2), and as fermions: electrons (g = 2), positrons (g = 2),
neutrinos (g = 3), antineutrinos (g = 3) all with the same temperature T1
 
2π 2 3 2π 2 3 43
s(a1 ) = T [2 + (7/8)(2 + 2 + 3 + 3)] = T (2.211)
45 1 | {z } 45 1 4
43/4

At a2 , electrons and positrons disappeared. We have only photons and neutrinos, but they have
different temperatures, so
2π 2  3 
s(a2 ) = 2Tγ (a2 ) + (7/8)(3 + 3)Tν3 (a2 ) (2.212)
45 "   #
2π 2 3 Tγ 3 21
= T (a2 ) 2 + (2.213)
45 ν Tν 4

and
a31 s(a1 ) = a32 s(a2 ) (2.214)
  "   #
3 3 43 3 3 Tγ 3 21
a 1 T1 = a2 Tν (a2 ) 2 + (2.215)
4 | {z } Tν 4
(a1 T1 )3

so that we have
 3
43 Tγ 21
= 2 + (2.216)
4 Tν 4
 3
Tγ 22 11
→2 = = (2.217)
Tν 4 2
2.11. COSMIC INVENTORY 95

or, finally
 1/3
Tν 4
= (2.218)
Tγ 11
i.e. the neutrino temperature at any time after electron/position annihilation becomes smaller than
the photon temperature. For Tγ = 2.725 K today, we have Tν = 1.95 K.
In terms of energy density, we have from Eqs. 2.189 for massless (relativistic) neutrinos:
7 π2 π2
ρν = gν Tν4 and ργ = gγ Tγ4 (2.219)
8 30 30
so the ratio is given by
 4
ρν 7 gν Tν
= (2.220)
ργ 8 gγ Tγ
From Eq. 2.218 and for gν = 2Nν = 6 and gγ = 2, we find
 
7 4 4/3
ρν = Nν ργ (2.221)
8 11
and evaluating this today for Nν = 3 generations of neutrinos:
 
2 7 4 4/3
Ων h = 3 (Ωγ h2 ) = 1.68 × 10−5 , mν = 0, Nν = 3 (2.222)
8 11
| {z }
≈0.68

In terms of number density, we have from Eqs. 2.188 for one generation of massless neutrinos
(gν = 2Nν = 2), and evaluating today in the last step:
 3 
3 ζ(3) 3 3 (1.202) 3 k
nν = g ν Tν = 2 (1.95) = 1.12 × 108 m−3 = 112 cm−3 (2.223)
4 π2 4 π2 h3 c 3
| {z }
units

whereas for photons we have

 
ζ(3) (1.202) k3
nγ = 2 gγ Tγ3 = 2 (2.725)3 = 4.11 × 108 m−3 = 411 cm−3 (2.224)
π π2 h3 c 3
| {z }
units

so, using Eq. 2.218, the ratio is given by (for Nν = 1 generation of neutrinos gν = gγ = 2):
   
nν 3 gν Tν 3 3 4 3/3 3
= = = (2.225)
nγ 4 g γ Tγ 4 11 11
For massive neutrinos, at early times when they were relativistic, everything
p goes as if they
were massless. When they become non-relativistic however, we have Eν = mν + p2 ≈ mν and
2

for a single neutrino generation (gν = 2) we have

Z
ρν = gν d3 pf (p)Eν (2.226)
Z
≈ mν gν d3 pf (p) = mν nν (2.227)
| {z }
nν
96 CHAPTER 2. EXPANSION OF THE UNIVERSE

Neutrinos decouple while still relativistic, and after decoupling their number density simply con-
tinues to decrease as a−3 ∝ Tν3 , so the number density found above holds true even after neutrinos
become non-relativistic1 .
In particular, the relation nν = 3nγ /11 continues to be valid even after neutrinos become
non-relativistic. We then have:
3
ρν = mν nγ (2.228)
11
Finally, since

ρc = 2.775 × 1011 h2 M⊙ Mpc−3 = 1.054 × 104 h2 eV cm−3 (2.229)

we have, evaluating today

ρν h 2 n ν h2 112 h2 cm−3 mν
Ω ν h2 = = mν = mν 4 2 −3 ≈ 94 eV , mν 6= 0, Nν = 1 (2.230)
ρc ρc 1.054 × 10 h eV cm

2.11.3 Baryons
There are four ways to measure the baryon density, which seem to agree well with each other. We
can observe directly in galaxies, mostly in the gas in groups of galaxies, but also in the stars inside
galaxies. We can also observe spectra of distant quasars at high redshifts, as their light is absorbed
by baryons (mostly hydrogen). We can also observe the CMB anisotropies in the Universe, which, as
we will see later, depend on the amount of baryons. Finally, the observed light element abundances
in the early Universe is sensitive to the amount of baryons (see next chapter). Altogether these
methods place consistent constraints of

Ωb ∼ 0.05 (2.231)

All these methods rely on the interactions between matter and radiation, which is expected from
baryons. As a result, we infer this density based on the light that results from such processes.

2.11.4 Total Matter

There are methods to measure mass that do not rely on the interaction between light and matter.
They rely on measuring the gravitational field produced by the total matter. Measurements of the
mass-to-light ratio M/L (mass over luminosity) of various systems (mostly groups and clusters of
galaxies) as a function of scale indicates a plato to the total matter density parameter of these
systems above R ∼ 1 Mpc, indicating

Ωm ∼ 0.25 (2.232)

These methods include galaxy rotation curves, velocity dispersions of galaxies in clusters, grav-
itational lensing, cluster abundance, CMB anisotropies, galaxy clustering and velocity field and
1
In fact, after decoupling of a particle species from the primordial plasma, its distribution function shape freezes
in, both for relativistic or non-relativistic particles. Recall the momentum always decays as p ∝ a−1 . For relativistic
particles E = p ∝ a−1 and T ∝ a−1 to ensure that n ∝ a−3 , so that the ratio E/T freezes and the BE/FD distribution
function maintains its shape. For non-relativistic particles the kinetic energy E = p2 /2m ∝ a−2 and T ∝ a−2 to
ensure that n ∝ a−3 . As a result, the ratio p2 /(2mT ) freezes and the Maxwell-Boltzmann distribution maintains its
shape. In the latter case, the chemical potential must change as well in a specific manner to guarantee that n ∝ a−3
(for details, see Kolb & Turner, pags. 69-70).
2.11. COSMIC INVENTORY 97

supernovae. They are all basically sensitive to the total amount matter and indicate that most of
the matter is not made of baryons (otherwise it would shine), but is instead dark. We call this
component dark matter.
A typical method to infer the total mass is to use observations that are sensitive to the baryon
fraction fb = Ωb /Ωm and use the value of Ωb measured from other methods to infer Ωm . For
example, in galaxy clusters Ωb ∼ Ωgas . The Mgas /Mtotal can be measured by the emission of X-ray
radiation from clusters or by the CMB that is heated by the cluster gas electrons in the direction
of the cluster (the inverse Compton scattering of the SZ effect). These measurements indicate
fb ∼ 0.2. Since Ωb ∼ 0.05, we find Ωm ∼ 0.05/0.2 ∼ 0.25. Wiggles in the power spectrum are also
sensitive to the baryon fraction and a number of galaxy surveys have been made.
Therefore the amount of total matter is about 5 times that of shining matter. They can’t
be baryons, because they would interact electromagnetically and would emit light. The models
of dark matter that are consistent with multiple observations indicate that it is cold, i.e. it was
non-relativistic by the time the Universe became matter dominated.
There have been attempts to set the neutrino as warm dark matter, since due to their small
mass neutrinos would be relativistic during se onset of matter domination and structure formation.
However, this is incompatible with structure formation because they neutrinos wash out struc-
ture in the Universe, in a way that is inconsistent with observations. and dark matter must be
cold. Therefore, their mass cannot be too large, and they cannot be a significant energy density
component of the Universe, i.e. they represent a large fraction of dark matter.

2.11.5 Dark Energy

Measurements of the luminosity of supernovae (exploding stars with roughly fixed and known
intrinsic luminosity) as a function of redshift have indicated that the Universe expansion seems to
be accelerating recently, i.e. ä > 0 for a ∼ 1. This discovery led to the Nobel Prize in 2011. Within
GR, this observation can only be explained assuming an energy component with negative pressure,
(P < −ρ/3). Such exotic component has been known as dark energy.
The simplest model of dark energy is a cosmological constant Λ, for which ρΛ = constant, and
PΛ = −ρΛ . For consistency with recent observations, we must have

ΩΛ ∼ 0.7 (2.233)

In the simplest interpretation, this component is the result of vacuum energy. However, attempts
to compute this energy in the realm of Quantum Field Theory lead to a value that is incompatible
with observations by 120 orders of magnitude. This is one of the cosmological constant problems.
Another problem is that, despite the fact that Ωm ∼ 1/a3 whereas ΩΛ ∼ constant, only recently
these components have become comparable in size, and not orders of magnitude different as they
had been for most of the Universe history. One therefore poses the question of ”why now?”, known
as the coincidence problem.
Other models of dark energy have been proposed, in which dark energy can be dynamical,
cluster and interact. Yet other alternative explanations for the cosmic acceleration rely on the fact
that GR may break down at cosmological scales. Even though in this case there is no dark energy
but only a ficticious effect, such explanations are usually also put in the class of ”dark energy”
models.
98 CHAPTER 2. EXPANSION OF THE UNIVERSE

2.11.6 Matter-Radiation Equality

The time when matter starts to dominate over radiation is important cosmologically:

Ωm a−3 = Ωr a−4 (2.234)

Ωr (2.47 + 1.68) × 10−5
→ aeq = = (2.235)
Ωm Ω m h2

or in terms of redshift

1 + zeq = 2.4 × 104 Ωm h2 (2.236)

We will see that photons decouple from matter at zd ∼ 103 so after equality and well into the
matter-dominated era.

2.12 Boltzmann Equations

In the absence of collisions, the collisionless Boltzmann equation (Vlasov equation) describes the
evolution of the distribution function as a result of phase-space incompressibility:

df ∂f dxi ∂f dpi ∂f
= + + =0 (2.237)
dt ∂t dt ∂xi dt ∂pi

In the equation above the only interactions included are gravitational effects in the metric
connecting the various terms. In the presence of other interactions (e.g. scattering of photons off
electrons before recombination), the full Boltzmann equation describes the space-time evolution of
f (x, p, t) as
 
df ∂f dxi ∂f dp ∂f dp̂i ∂f ∂f
= + + + = (2.238)
dt ∂t dt ∂xi dt ∂p dt ∂ p̂i ∂t C

where the right-hand side represents the collision term and must be computed from some fun-
damental theory of interactions, e.g. quantum field theory. Here p = pi = pp̂i is the physical
momentum, where p = |p| and p̂i a unit directional vector, i.e. δij pi pj = p2 and δij p̂i p̂j = 1.
In equilibrium, the distribution f (x, p, t) = f0 (p, t) is either the BE or FD distribution, and the
collision term is zero (collisions/reactions in one direction cancelled by terms in opposite direction)
such that the collisionless Boltzmann is satisfied and

df0 ∂f0 dxi ∂f0 dp ∂f0 dp̂i ∂f

= + + + = 0 (2.239)
dt ∂t ∂xi dt ∂p
dt |{z} dt ∂ p̂i
|{z}
0 0
∂f0 dp ∂f0
→ + = 0 (2.240)
∂t dt ∂p

This simplified equation can be used in the background expansion. When studing perturbations
we must include all terms.
2.12. BOLTZMANN EQUATIONS 99

From the 0-index Geodesics equation, using Γ0ij = δij aȧ = δij a2 (ȧ/a) = Hgij , we have

dP 0
= −Γ0αβ P α P β (2.241)
dλ
dP 0 dt
= − Γ0ij P i P j (2.242)
dt |{z}
|{z} dλ |{z}
dE P 0 =E Hgij
dt

dE
E = −Hgij P i P j (2.243)
dt
Now recall that

P 2 = gµν P µ P ν = −(P 0 )2 + gij P i P j = −E 2 + p2 = −m2 → E 2 = p2 + m2 (2.244)

| {z } | {z }
−E 2 =p2

Therefore EdE = pdp and the Geodesics equation, Eq. 2.243, gives
dp
p = −Hp2 (2.245)
dt
or finally
dp
= −Hp (2.246)
dt
This implies dp/p = −da/a, so that p ∝ 1/a. Notice this is valid for both massive and massless
species. The Boltzmann equation, Eq. 2.240, then becomes
∂f0 ∂f0
− Hp = 0 (2.247)
∂t ∂p
Multiplying by g and integrating over momenta
Z Z
∂ d3 p d3 p ∂f0
g f 0 − Hg p = 0 (2.248)
∂t (2π)3 (2π)3 ∂p

But,
Z
∂ d3 p ∂n
g f0 = (2.249)
∂t (2π)3 ∂t

and
Z Z
d3 p ∂f0 4πp2 dp ∂f0
g p = g p (2.250)
(2π)3 ∂p (2π)3 ∂p
Z
4πdp 3 ∂f0
= g p (Integrate by parts) (2.251)
(2π)3 ∂p
Z
4πdp 2
= −3g p f0 (Surface term ∝ p3 f0 (p) = 0 at p = 0, ∞) (2.252)
(2π)3
Z
dp3
= −3g f0 = −3n (2.253)
(2π)3
100 CHAPTER 2. EXPANSION OF THE UNIVERSE

therefore
∂n
+ 3Hn = 0 (2.254)
∂t
or
d(na3 )
a−3 =0 (2.255)
dt
Therefore
1
n∝ (2.256)
a3
So the number density of a species in equilibrium simply scales as a−3 following the dilution from
the volume increase. Again, this is valid for all species (massive and massless) in equilibrium.

2.12.1 Photons
For photons m = 0 and P 0 = E = p. In equilibrium, we assume photons have a BE (Planck)
distribution f (x, p, t) = f0 (p, t):
1 1
f0 (p, t) = = (2.257)
eE/kB T −1 ep/T (t) −1
such that the collisionless Boltzmann, Eq. 2.247 is satisfied. But because

∂f0 ∂f0 dT p ∂f0 dT Ṫ ∂f0

= =− =− p (2.258)
∂t ∂T dt T ∂p dt T ∂p
we have from the Boltzmann equation, Eq. 2.247:
" #
Ṫ ȧ ∂f0
− − p =0 (2.259)
T a ∂p

so
dT da
=− → ln T = − ln a + const. (2.260)
T a
and
1
T ∝ (2.261)
a
Therefore the temperature scales as 1/a. Since both p and T scale as 1/a, we have p/T =const.,
and the Planck distribution is preserved during evolution.

2.12.2 Matter
For non-relativistic matter, we assume that

p2
E ≈m+ (2.262)
2m
2.13. ENERGY-MOMENTUM TENSOR 101

and a Maxwell-Boltzmann distribution:

2 /2mT
f0 (p, t) = e−E(p)/T ∼ e−m/T e−p (2.263)

such that again the collisionless Boltzmann, Eq. 2.247 is satisfied. Now we have

∂f0 ∂f0 dT p ∂f0 dT Ṫ ∂f0

= =− =− p (2.264)
∂t ∂T dt 2T ∂p dt 2T ∂p
so that from the Boltzmann equation
" #
Ṫ ȧ ∂f0
− − p =0 (2.265)
2T a ∂p

or
dT da
= −2 → ln T = −2 ln a + const. (2.266)
T a
and
1
T ∝ (2.267)
a2
Therefore the temperature scales as 1/a2 for non-relativistic particles. Since both p2 and T scales
as 1/a2 , we have p2 /2mT =const., and the Maxwell-Boltzmann distribution is also preserved dur-
ing evolution, as long as the chemical potential µ(t) evolves as well such that the term e−µ/T
compensates e−m/T .
Finally, since ρ = mn for non-relativistic particles, Eq. 2.256 tells us that
1
ρ∝ (2.268)
a3

2.13 Energy-Momentum Tensor

We defined previously the energy-momentum tensor for discrete particles and specialized to perfect
fluids. But, as we will see, we cannot treat all species as fluids. Therefore we would like to define
Tµν for each species i, at each space-time point in terms of the distribution function f (x, p, t) in
phase space. We do that by averaging over the components of the 4-momentum P α :
Z
µ dP1 dP2 dP3 1 P µ Pν
T ν,i (x, t) = gi √ fi (x, p, t) (2.269)
(2π)3 −g P 0

where g = det gµν . Computing specific components, one can check that this recovers the usual
components of the energy momentum tensor (e.g for a fluid) in terms of the distribution function.
For instance, let us check it for µ = ν = 0. We use

P0 = g0µ P 0 = g00 P 0 = −P 0 = −E (2.270)

and

Pi = giµ P µ = gij P j = δij a2 P j = a2 P i (2.271)

102 CHAPTER 2. EXPANSION OF THE UNIVERSE

We define pi via
Pi Pj
p2 ≡ g ij Pi Pj = δij a−2 Pi P j = δij (2.272)
a a
but since we must also have

p2 = δij pi pj (2.273)

we find that the components pi relate to Pi via

Pi
pi = = aP i (2.274)
a
Here P i is the comoving momentum and pi the physical (proper) momentum. Therefore

dP1 dP2 dP3 = (adp1 )(adp2 )(adp3 ) = a3 d3 p (2.275)

From the diagonal FRW metric gµν =diag(−1, a2 , a2 , a2 ), we have g = −a6 , so

√
−g = a3 (2.276)

Finally, we have

−m2 = gµν P µ P ν = −E 2 + gij P i P j (2.277)

But gij P i P j = a2 δij P i P j = a2 δij (pi /a)(pj /a) = δij pi pj = p2 . So we have as usual:
p
−m2 = −E 2 + p2 → E(p) = p2 + m2 (2.278)

Therefore, we find for µ = ν = 0:

−E
Z z}|{ Z
a3 d3 p 1 P0 P0 d3 p
T 0 0 (x, t) = g f (x, p, t) = −g E(p) f (x, p, t) = −ρ(x, t) (2.279)
(2π)3 a3 P0 (2π)3

Likewise, for µ = ν = i (summing over indices), and since P i Pi = gij P i P j = p2 , we have

Z Z
i 1 2 3 a 3 d 3 p 1 P i Pi d 3 p p2
T i =T 1+T 2+T 3 =g f (x, p, t) = g f (x, p, t) = 3P (x, t)
(2π)3 a3 P 0 (2π)3 E
(2.279)

Therefore

T11 = T22 = T33 = P (2.280)