1

Cosmology: a brief refresher course

Fundamental assumptions
• Standard model of cosmology is based on two fundamental assumptions:
– On sufficiently large scales, the Universe is isotropic – i.e. there is no preferred
direction
– Our position in the Universe is not special – the Copernican principle
• Together, these two assumptions imply that on large scales, the Universe is homoge-
neous
• We have good observational evidence for isotropy. For example, consider the cosmic
microwave background. Once we subtract o↵ the dipole due to our motion through
space, we find that the remaining background is very close to flat, with inhomogeneities
only at the 10 5 level.
• Direct observation evidence for homogeneity is harder to come by, but large-scale galaxy
surveys find results consistent with a transition to homogeneity on scales ⇠ 100 Mpc;
see e.g. Scrimgeour et al. (2012, MNRAS, 425, 116).
• In addition to these two key assumptions, we also typically assume that general rela-
tivity (GR) is the correct theory of gravity on large scales.

The Friedmann-Robertson-Walker metric

• A key lesson from relativity: space and time are not distinct things; instead, we live
in a four-dimensional space-time.
• The geometry of space-time is described by a metric tensor gµ⌫ . Locally, we can
choose a coordinate system such that space-time appears flat (i.e. locally, space-time
is described by the Minkowski metric of SR).
• On large scales, we cannot a priori assume that space-time is flat. However, the
large-scale homogeneity and isotropy of the Universe allow us to write the metric in a
relatively simple form:
⇥ ⇤
ds2 = c2 dt2 a2 (t) dr2 + fK2 (r) d✓2 + sin2 ✓d 2 . (1)
Here, (r, , ✓) are polar coordinates, a(t) is a (time-dependent) scale factor and fK is
a function describing the curvature of the Universe:
8 1/2
< K
> sin K 1/2 r K>0
fK (r) = r ⇣ ⌘ K=0 (2)
>
: |K| 1/2 1/2
sinh |K| r K<0

Note that the constant K here is the Gaussian curvature of space-time.

 2

• We can write this in a slightly simpler form by choosing our radial coordinate so that
the three cases correspond to K = 1, K = 0, and K = 1. This yields
8
< sin (r) K=1
fK (r) = r K=0 (3)
:
sinh (r) K= 1

• In the most general case, the curvature K could vary as a function of our location within
that space. For example, consider the example of a 2D surface that is flat in most
locations but that has a small peak in one region. Depending on where we measure,
we might conclude either that this surface is flat (K = 0) or that it is positively curved
(K > 0). However, if the Universe is homogeneous and isotropic, then observers at
di↵erent points in space time must agree on the sign of the curvature; i.e. K must be a
constant. This leaves only three possibilities for the spatial curvature of the Universe
on large scales. The case K > 0 corresponds to positive curvature, the case K < 0
to negative curvature, and the special case K = 0 to no curvature, i.e. a spatially
flat Universe. As we will see later, current observations suggest that the K = 0 case
best describes the Universe we inhabit.

• The scaling factor a in the FRW metric cannot depend on our location in the Uni-
verse, but the assumptions of homogeneity and isotropy do not prevent it from being
a function of time. The case ȧ > 0 corresponds to a Universe which is expanding; the
case ȧ < 0 to one which is contracting.

Redshift
• If a is not constant, then photons propagating to us from distant sources will be red-
shifted (if the Universe is expanding) or blueshifted (if the Universe is contracting).
The fact that the light that we observe from distant galaxies is redshifted tells us that
we live in an expanding Universe.

• Consider light emitted from some source which is comoving with the expansion of the
Universe at a time te , and which is observed by a comoving observer at r = 0 at time
to . For light, we know from relativity that ds = 0, and if the direction of propagation
is purely radial then the angular terms also vanish. We therefore have

c |dt| = a(t)dr (4)

The coordinate distance between the source and the observer is simply:
Z to
reo = dr = constant. (5)
te

Alternatively, we can write this as:

Z to
cdt
reo = = constant. (6)
te a(t)
 3

• If reo is constant, then ṙeo must vanish. Therefore:

dreo c dto c
= = 0, (7)
dt a(to ) dte a(te )
which means that
dto ao
= . (8)
dte ae
In an expanding Universe, a time interval dte at the source is lengthened by a factor
ao /ae by the time it arrives at the observer.
• If we now take dt to be the time elapsed during the propagation of a single period of
our light wave, i.e. dt = ⌫ 1 , then it is easy to show that
⌫e o ao
= =1+z = . (9)
⌫o e ae
• Note that if we use the common convention that a = 1 at the present day, this implies
that the redshifts we measure now are related to the scale factor at the time the light
was emitted by
1
1+z = . (10)
a
It also follows that the redshift of any particular astrophysical object that we observe
is a function of time (since at later times, we observe the light emitted when a was
slightly larger than at earlier times). From the above, we have
dz ȧ ȧ
= 2 = (1 + z). (11)
dt a a
The fractional change in the redshift per unit time is therefore z/z / ȧ/a (for z 1).
In practice, as we will see shortly, ȧ/a is very small and so this e↵ect is not currently
observable.

The Friedmann equations

• So far, we have not used any results from GR. Redshift is a consequence of the FRW
metric, and the FRW metric is a consequence of our assumptions of isotropy and
homogeneity. GR only enters the picture when we want to determine how the scale
factor a(t) evolves with time.
• If we assume that the Universe is filled with a perfect fluid with energy density ⇢(t)
and pressure p(t), and is described by the FRW metric, then we can use GR to derive
the following equations describing the behaviour of the scale factor a(t):
✓ ◆2
ȧ 8⇡G Kc2 ⇤
= ⇢ + , (12)
a 3 a2 3
✓ ◆
ä 4⇡G 3p ⇤
= ⇢+ 2 + . (13)
a 3 c 3
Here, ⇤ is a dimensionless quantity known as the cosmological constant.
 4

• We can combine these two equations to yield a third equation

d 3 2 d 3
a ⇢c + p a = 0. (14)
dt dt
These three equations are known as the Friedmann equations.

Cosmological parameters
• We can divide the matter content of the Universe into two forms: relativistic and
non-relativistic (aka “radiation” and “dust”).

• For relativistic particles, the pressure is related to the energy density via
1
p = ⇢c2 , (15)
3
while for non-relativistic particles, p is much smaller than ⇢ and hence it is a good
approximation to set p = 0.

• From equation 14, we see that for non-relativistic matter we have

d 3 2
a ⇢c = 0 (16)
dt
and hence
⇢(t) = ⇢0 a 3 , (17)
where ⇢0 is the present-day energy density and we have chosen distance units such that
a0 = 1.

• For relativistic matter we instead have

d 3 2 ⇢c2 d 3
a ⇢c + a = 0, (18)
dt 3 dt
d⇢ da
a3 + 4⇢a2 = 0, (19)
dt dt
and hence
⇢(t) = ⇢r,0 a 4 . (20)

• The energy density of non-relativistic particles is dominated by their mass, and hence
varies with redshift only because the spatial density of the particles varies. On the other
hand, relativistic particles lose energy at a faster rate due to the e↵ects of redshift.

• At this point, it is convenient to write the Friedmann equations in a simpler form with
the help of a number of dimensionless parameters that we will now introduce.
 5

• First of all, we have the Hubble parameter:

ȧ
H(t) ⌘ . (21)
a
The value of H(t) at the present time is given by
km 18
H0 = 100h = 3.22 ⇥ 10 h s 1, (22)
s Mpc
with h ' 0.70, and is known as the Hubble constant.
• Next, we have the critical density
3H 2
⇢crit ⌘ . (23)
8⇡G
The present day value is written as ⇢crit,0 and is given by
29 2
⇢crit,0 = 1.88 ⇥ 10 h g cm 3 . (24)

• By dividing the densities of non-relativistic and relativistic matter by the critical den-
sity, we obtain the dimensionless density parameters ⌦m and ⌦r :
⇢m ⇢r
⌦m ⌘ , ⌦r ⌘ . (25)
⇢crit ⇢crit

• We can rewrite the first of the Friedmann equations in terms of H, ⌦m and ⌦r as:

⇤ Kc2
H = H0 ⌦r,0 a 4 + ⌦m,0 a 3 +
2 2
. (26)
3H02 a2 H02

• If we define two further dimensionless density parameters,

⇤
⌦⇤ ⌘ , (27)
3H02
and
Kc2
⌦K ⌘ = 1 ⌦r,0 ⌦m,0 ⌦⇤ , (28)
H02
then we can rewrite Equation 26 as
⇥ ⇤
H 2 = H02 ⌦r,0 a 4 + ⌦m,0 a 3 + ⌦⇤ + ⌦K a 2 . (29)

• A hugely important consequence of Equation 29 is the fact that the sign of ⌦K depends
on the sum of the other density parameters. If ⌦m,0 +⌦r,0 +⌦⇤ < 1, then ⌦K > 1. This
in turn implies that K < 0 and hence that the Universe has negative spatial curvature.
Similarly, if ⌦m,0 + ⌦r,0 + ⌦⇤ > 1, then ⌦K < 1, K > 0 and the Universe has positive
spatial curvature. Finally, if ⌦m,0 + ⌦r,0 + ⌦⇤ = 1, then K = 0 and the Universe is
spatially flat. Furthermore, since K is a constant, if this result holds at the present
time, it must hold at all times.
 6

• Therefore: the contents of the Universe determine its curvature.

• Another important point to note here is that the terms on the right-hand side of Equa-
tion 29 have di↵erent dependencies on a and hence evolve at di↵erent rates. At the
present-day, ⌦r,0 is composed of contributions from the cosmic microwave background
and the cosmic neutrino background. The energy densities of both of these are well
constrained, and we know that ⌦r,0 ⌧ ⌦m,0 . Moreover, it also appears that in our Uni-
verse, ⌦K = 0. Therefore, at the present time, the right-hand side of the Friedmann
equation is dominated by the contributions from non-relativistic matter and the cos-
mological constant; in practice, ⌦m,0 ' 0.315, ⌦⇤ ' 0.685, and hence the cosmological
constant term dominates.
• An important consequence of this follows from the second of the Friedmann equations.
We can use the dimensionless density parameters to rewrite this as an expression for
ä: 
2 1 H2
ä = aH0 ⌦⇤ (⌦m + ⌦r ) (30)
2 H02
If the cosmological constant dominates, then H 2 ' H02 ⌦⇤ , and so the term in square
brackets is positive now and at all times in the future, i.e. the expansion of the Universe
is accelerating.
• As we move to higher redshift, however, the size of the cosmological constant term does
not change, while the non-relativistic matter term evolves as a 3 . Therefore, the latter
dominates for a < (⌦m,0 /⌦⇤ )1/3 , corresponding to redshifts z > 0.3. In the regime
where matter dominates, it is simple to use Equation 30 to show that ä < 0, i.e. that
the expansion in this regime decelerates.
• In addition, since the radiation and matter terms also evolve at di↵erent rates, there
will come a time when both are equal. This occurs at a redshift
⌦m,0
zeq = 1, (31)
⌦r,0
known as the redshift of matter-radiation equality. Evaluating this, we find that zeq ⇠
3500.
• In this course, we are mostly concerned with the range of redshifts between z ⇠ 1000
and z ⇠ 20, the so-called Dark Ages. In this range of redshifts, the matter term
dominates, and we can write Equation 29 in the form
H 2 ' H02 ⌦m,0 a 3 . (32)
However, it is important to remember that this is an approximation that breaks down
at lower and at higher redshifts.
• In the regime governed by this approximation – known as the Einstein-de Sitter
limit – there is a simple relationship between the redshift z and the time t:
2 1
t= . (33)
3 H0 ⌦1/2
m,0 (1 + z) 3/2
 7

Thermal history
• How does the temperature change as the Universe expands? The answer to this di↵ers,
depending on whether we are considering radiation or non-relativistic matter.
• We know that for a black-body radiation field, the energy density urad scales with
temperature as urad / T 4 ; this is the just the Stefan-Boltzmann law. Since the
energy density of a radiation field scales as urad / (1+z)4 , this implies that T / (1+z),
provided that the radiation field retains its black-body shape.
• Suppose that at redshift z1 , the Universe is filled with black-body radiation with a
temperature T1 . A volume V1 then contains
8⇡⌫12 d⌫1 /c3
dN1 = V1 ⇣ ⌘ (34)
h⌫1
exp kT 1
1
photons in the frequency range ⌫1 ! ⌫1 + d⌫1
• Provided that photons are not created or destroyed, but merely redshifted, the same
set of photons at redshift z2 occupy the frequency range ⌫2 ! ⌫2 + d⌫2 , where ⌫2 = f ⌫1
and d⌫2 = f d⌫1 , where f = (1 + z2 )/(1 + z1 ).
• The volume V2 at z2 corresponding to our original volume is given by V2 = V1 f 3 .
Therefore, the number of photons dN2 in the frequency range ⌫2 ! ⌫2 + d⌫2 in a
volume V2 is simply dN1 , and is given by
V1 8⇡f 3 ⌫12 d⌫1 /c3
dN2 = dN1 = 3 ⇣ ⌘ (35)
f exp hf ⌫1 1
kf T1

However, we can rewrite this as

8⇡⌫22 d⌫2 /c3
dN2 = V2 ⇣ ⌘ , (36)
h⌫2
exp kT 2
1
where T2 = f T1 . Since we can apply the same argument to any frequency interval in
our original spectrum, we see that an initial black-body spectrum retains its black-body
shape and simply changes its temperature as the Universe expands.
• For a non-relativistic ideal gas, we can derive the evolution of the temperature from the
relationship p = K⇢ , where is the adiabatic index. The expansion of the Universe
is an adiabatic process, and hence K does not change as the Universe expands. Since
we can write the pressure of an ideal gas in the form
⇢kT
p= , (37)
m̄
where m̄ is the mean particle mass, this means that the relationship
⇢kT
= K⇢ (38)
m̄
must continue to hold as the Universe expands.
 8

• From this, we see that the temperature of the gas must scale with the density as
T / ⇢ 1 , and since the density evolves with redshift as ⇢ / (1 + z)3 , this means that
T / (1 + z)3 3 . For an atomic (or ionized) gas with = 5/3, we therefore arrive at
the result:
T / (1 + z)2 . (39)
• The temperature of a non-relativistic gas therefore falls o↵ more rapidly than the
radiation temperature, in the absence of any energy transfer between gas and radiation
(or vice versa).
• In practice, the gas and radiation temperatures are strongly coupled at high redshift
by Compton scattering. When photons Compton scatter o↵ electrons, they may either
lose or gain energy, depending on the details of the collision. However, we know from
simple thermodynamics that in the limit of a large number of scatterings, energy will
flow from the gas to the radiation field if the gas temperature Tgas is greater than the
radiation temperature Trad , and from the radiation field to the gas if Trad > Tgas .
• The energy transfer rate per unit volume can be written as
4
4 T asb T kne
⇤Comp = (T Tgas ) , (40)
me c
where T is the Thompson scattering cross-section and asb is the Stefan-Boltzmann
constant.
• If we compare this with the cooling rate due to the adiabatic expansion of the Universe,
we find that Compton scattering dominates when
⇤Comp
> 1. (41)
3nkTgas H(z)
Evaluating this, we find that Compton scattering dominates at redshifts greater than
a few hundred.
• At high redshift, therefore, both gas and radiation temperatures evolve as T / (1 + z).
• In this course, we will mostly be concerned with the evolution of the Universe between
redshifts z ⇠ 1000 and z ⇠ 10. However, it is useful at this point to remind you that
there is considerable physics occurring at higher redshifts. In particular, primordial
nucleosynthesis has already occurred.
• Protons and neutrons first form at the point where the temperature of the gas and
radiation corresponds to kT ' 1 GeV.1 At this point, the ratio of neutrons to protons
is maintained in equilibrium by the conversion reactions
p+e ! n+⌫ (42)
p + ⌫¯ ! n + e+ , (43)
where p represents a proton, n a neutron, e an electron, e+ a positron, and ⌫ and ⌫¯
are a neutrino and an anti-neutrino, respectively.
1
Immediately prior to this, the Universe was filled with a so-called quark-gluon plasma.
 9

• Once the temperature of the Universe drops to kT ' 800 keV, these reactions “freeze-
out” – the timescale associated with them becomes longer than the expansion timescale
of the Universe. At the time that this happens, the neutron-to-proton number density
is given by
nn 2 1
= e mc /kT ' , (44)
np 6
where mc2 = 1.4 MeV is the mass di↵erence between neutrons and protons.

• Although it is energetically favourable for the neutrons and protons to fuse together
to form heavier nuclei, they cannot immediately do so, as immediately after freeze-out
there are still too many extremely high energy photons around, and these photo-
disintegrate any heavy nuclei that form. As the Universe expands and cools, however,
the number density of these photons falls o↵ exponentially, and once the temperature
is kT ' 80 keV, heavier nuclei start to form in abundance. This occurs roughly three
minutes after t = 0.

• The ratio of neutrons to protons at this point is around 1/7. It is smaller than the
value at freeze-out because some of the neutrons have undergone beta decay.

• Almost all of these neutrons wind up in 4 He, while the vast majority of the remaining
protons remain free. Small fractions of deuterium (i.e. 2 D), 3 He and lithium are
also formed, but elements heavier than lithium form only in truly negligible amounts,
owing to the lack of any stable nuclei with weights A = 5 or A = 8; the process of
nucleosynthesis becomes “stuck” at helium, and cannot progress further.

• The precise abundances of the various nuclei depend on the details of the cosmological
model, and in particular on the photon-to-baryon ratio. For the currently-favoured
⇤CDM model, we have a helium mass fraction of around 24.7%, a deuterium to hy-
drogen ratio of around 2.6 ⇥ 10 5 and a lithium to hydrogen ratio of around 4 ⇥ 10 10
(Planck collaboration, 2013).

• At the point where we first enter the cosmological “Dark Ages” – the epoch of re-
combination – the chemical composition of the gas is therefore roughly 75% ionized
hydrogen, 25% ionized helium, and tiny traces of D, 3 He and Li.