University of Porto

Lecture course notes of Curricular Unit

Tópicos Avançados em Galáxias e Cosmologia (UC AST603)
Module: Cosmological Structure Formation and Evolution

Doctoral Program in Astronomy

Elements of Cosmology and

Structure Formation

António J. C. da Silva

2012 / 2013
2
Contents

1 The Standard Model of Cosmology 7

1.1 The basis of the standard model of Cosmology . . . . . . . . . . . . . . . . . 7
1.2 Fundamental equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.1 The dynamical equations . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.2 Epochs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2.3 The observed universe . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3 Exact solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3.1 The age of the universe . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3.2 Scale factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3.3 Distances, horizons and volumes . . . . . . . . . . . . . . . . . . . . . 15
1.4 Initial conditions and Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.4.1 Problems with the Big Bang . . . . . . . . . . . . . . . . . . . . . . . . 17
1.4.2 The theory of inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2 Cosmological Structure Formation 23

2.1 Density contrast and the power spectrum . . . . . . . . . . . . . . . . . . . . . 23
2.2 Linear perturbation theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3 Sub-horizon perturbations in non-relativistic fluids . . . . . . . . . . . . . . . . 27
2.4 Relativistic fluids and super-horizon perturbations . . . . . . . . . . . . . . . . 30
2.4.1 Relativistic multi-fluid case . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.2 Density transfer function and the power spectrum . . . . . . . . . . . 32
2.5 Non-linear evolution of perturbations . . . . . . . . . . . . . . . . . . . . . . . 34
2.5.1 Spherical collapse model . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.5.2 Zel’dovich approximation . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.5.3 Press–Schechter theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

1
 2.5.4 Numerical simulations of large–scale structure . . . . . . . . . . . . . . 37

3 Cosmic Microwave Background Radiation 47

3.1 The intensity spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 The angular power spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3 Fontes geradores de anisotropias CMB: “A primer” . . . . . . . . . . . . . . . 52
3.3.1 Anisotropias Primárias . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.2 Anisotropias Secundárias . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.3 Anisotropias Terciárias . . . . . . . . . . . . . . . . . . . . . . . . . . 61

A Cálculo das anisotropias do CMB a grandes escalas angulares 73

A.1 Integração da equação das geodésicas para fotões . . . . . . . . . . . . . . . . 73
A.2 O Efeito de Sachs-Wolfe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
A.2.1 Cálculo de Cl para grandes ϑ. Patamar de Sachs-Wolfe . . . . . . . . . 81

B Geodésicas e velocidades próprias 85

2 2
B.1 Geodésicas nulas e velocidades próprias em ds e ds̄ . . . . . . . . . . . . . . 85
B.2 Velocidades próprias de partı́culas materias em universos perturbados (1a ordem
de aprox.) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

C Temperature fluctuations in a small patch of sky 87

List of Figures

2.1 Gas cooling functions for collisional ionization equilibrium in the optically-thin
limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.2 Simulation Box at redshift zero . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3.1 CMBR spectral distortion caused by energy release in the early universe. . . . 48
3.2 General features of the CMBR angular power spectrum . . . . . . . . . . . . . 51
3.3 O efeito Sachs Wolfe integrado . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.4 Efeitos do lensing gravitacional, reionização global e efeito Vishniac sobre o
espectro CMBR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.5 Efeito da reonização na radiação de fundo . . . . . . . . . . . . . . . . . . . . 59
3.6 Distorções espectrais CMBR provovadas pelo efeito Sunyaev– Zel’dovich . . . 59
3.7 Funções de distorção espectral dos efeitos SZ térmico e cinético . . . . . . . . . 60
3.8 Sensibilidade angular e espectral de várias experiências CMB . . . . . . . . . . 63
3.9 Foregrounds e ruı́dos do fundo de micro-ondas . . . . . . . . . . . . . . . . . . 64

A.1 Parametrisação das geodésicas luz no espaço não perturbado de ds̄2 . . . . . . 76

A.2 Última superfı́cie de scattering. . . . . . . . . . . . . . . . . . . . . . . . . . . 82

3
List of Tables

3.1 Processos que geram anisotropias na CMBR . . . . . . . . . . . . . . . . . . . 52

5
Chapter 1

The Standard Model of Cosmology

In the following sections we review the basic elements of standard cosmology needed for the
remainder of the course. The majority of the results presented here can be found in many
textbooks of cosmology, e.g. in Refs. [18, 61, 65, 75, 76, 79, 80].

1.1 The basis of the standard model of Cosmology

The standard model (SM) of cosmology is presently based on the Hot Big Bang paradigm. The
universe is expanding and cooling from an initial ultra dense state, where matter and radiation
were prisoners in a gaseous hot plasma of fundamental particles. The basic mathematical
framework of the SM is set by the following assumptions:

• The universe is homogeneous and isotropic when observed on large scales.

• The dynamics of space-time is described by Einstein’s theory of general relativity.

The first of these assumptions derives from the cosmological principle, which states that the
universe on large scales should have no privileged positions or directions and therefore at a
given time should look the same to all observers. The most general homogeneous and isotropic
solution of the Einstein equations of GR is the FLRW line element,

dr 2
! $
2 2 2 2 2
" 2 2 2
#
ds = c dt − a (t) + r dθ + sin θdφ , (1.1)
1 − kr 2

where c is the speed of light, t is the universal time, r, θ, φ are the comoving spatial coordinates
and a(t) is the scale factor, which describes the overall expansion (or contraction) of the
three-dimensional space. The constant k gives the spatial curvature of the universe. It can be
negative, zero or positive depending on whether the universe is open, closed or flat. If k ≤ 0
the universe has an infinite extension and the r coordinate ranges √ from zero to infinity. If k
is positive the universe is finite in size and r ranges from 0 to 1/ k. With an appropriate
rescaling of coordinates it is always possible to make k to take the values: −1, 0, +1. Observers

7
with fixed coordinates in the comoving coordinate system experience no external forces and are
said to be fundamental comoving observers as they move with the expansion (or contraction)
of the cosmological fluid. The proper distance between two of these observers scales as:
a(t)
ℓ(t) = ℓ0 , (1.2)
a0
where a0 and ℓ0 are the scale factor and the proper distance at some initial time t0 . Taking
the derivative of this expression with respect to time we obtain the relative speed between
fundamental observers:
dℓ ȧ(t) ȧ(t)
v(t) = = ℓ0 = ℓ(t) ≡ H(t)ℓ(t). (1.3)
dt a0 a(t)
This expression shows that in an expanding universe (ȧ(t) > 0) the further away any two
observers are, the faster they recede from one another due to the cosmic expansion. This is
known as the Hubble law, first derived from observations by Edwin Hubble in 1929. The
proportionality factor between velocity of recession and distance,
ȧ(t)
H(t) = , (1.4)
a(t)
gives the expansion rate of the universe at a given time. Its present value is usually parametrized
as H0 = 100 h km s−1 Mpc where h is the so-called Hubble parameter. H0 is also known as
the Hubble constant and has units of an inverse of time.
The cosmological expansion produces a redshift in the spectrum of the light emitted from
distant objects. This redshift is defined as the change of energy of a photon measured by
fundamental observers between the epochs of emission of the radiation, at t, and the present
time, t0 ,
E − E0 ν λ0 a (t0 )
z= = −1= −1= − 1. (1.5)
E0 ν0 λ a (t)
Here E, ν and λ are the photon’s energy, frequency and wavelength at t and the subscript
‘0 ’ denotes the same quantities at the present. To obtain this expression we use the quantum
mechanics’ proportionality between the energy and frequency of a photon E = hν = hc/λ (h
is the Planck’s constant) and the fact that the wavelength of a free photon stretches as the
other lengths with a(t).

1.2 Fundamental equations

In GR the dynamics of space-time is set by Einstein’s field equations,
1 8πG
Gab = Rab − Rgab = 4 Tab + Λgab , (1.6)
2 c
where G and Λ are the gravitational and cosmological constants, gab is the metric of space-
time, Gab is the Einstein tensor, Rab is the Ricci tensor and R is the Ricci scalar (this last
two quantities result from successive contractions of the Riemann tensor and are functions of
the metric and its derivatives). The inclusion of the cosmological constant term in Eq. (1.6),
originally introduced by Einstein in order to describe a static universe, is presently supported
by observations of distant Type Ia supernovae [83, 84, 89]. Finally Tab is the stress–energy
tensor, which describes the gravitational contributions from all forms of energy in the universe.
To satisfy the requirements of homogeneity and isotropy implied by the cosmological principal,
the stress–energy tensor has to be that of a perfect fluid:
p p
Tab = (ρ + 2 )Ua Ub − 2 gab (1.7)
c c
where ρ = ρ(t) and p = p(t) are the fluid’s energy density and pressure and Ua is the four
velocity field of a fundamental observer. In this case the solution of Einstein’s field equations
is the metric (1.1). The energy–momentum conservation law is expressed by the condition
T ab ;b = 0 (1.8)
where the symbol ‘;’ denotes covariant derivative. This is a set of four equations giving the
conservation of the energy density and the 3-momentum. The ‘temporal component’ equation
gives,
ȧ % p&
⇒ d ρc2 a3 = −pd a3 .
" # " #
ρ̇ = −3 ρ + 2 (1.9)
a c
This expression has a simple physical interpretation. It translates the first law of thermo-
dynamics applied to a comoving volume element: in an adiabatical process the variation of
energy is given by the work produced by the pressure forces. It is often assumed that during
several periods of the history of the universe the energy density and pressure can be related
by an equation of state with a simple form:
p = wρc2 − 1 ≤ w ≤ 1, (1.10)
where w is a constant. In this case the integration of (1.9) gives
' (−3(1+w)
a(t)
ρ(t) = ρi , (1.11)
ai
where ρi = ρ (ti ), ai = a (ti ) are the energy density and the scale factor at some initial time
ti . The cases w = 0, w = 1/3 and w = −1 are classical examples which are appropriate to
describe the phases when the universe is dominated by a fluid of (non relativistic) matter,
radiation (and relativistic matter) and vacuum energy associated with the Λ term in (1.6).

1.2.1 The dynamical equations

Using Eqs. (1.1) and (1.7) in (1.6) one can derive two equations which give the expansion and
acceleration rates of the universe as a function of its matter contents and geometry:
' (2
ȧ 8πG Λc2 kc2
= ρ+ − 2 (1.12)
a 3 3 a
ä 4πG % p & Λc2
= − ρ+3 2 + . (1.13)
a 3 c 3
These are known as the Friedmann and Raychaudhuri (or acceleration) equations, respectively.
If we set Λ = 0 we find that ä is negative whenever p > −ρc2 /3 (a(t) is always positive). In
this case both radiation (w = 1/3) and matter (ω = 0) dominated epochs are periods of
decelerated expansion (if the universe has a flat or open geometry) or decelerated contraction
(only possible with closed geometries).
If we take as an observational fact that the universe is expanding (ȧ > 0) and assume that
in the past ä was always negative (ä < 0) then going backwards in time there’s a moment,
ti , where a(ti ) = 0. This is the Big Bang event. The instant ti is usually redefined as the
origin of time, ti = 0. The FLRW models are Big Bang universes for many combinations of
energy densities, cosmological constant and geometries. However one should note that there
are FLRW models (with Λ > 0, k = 1), which do not “start” from a Big Bang. This is the
case of the Eddington–Lemaitre and Einstein universes (see e.g. Ref. [22]). Because equations
(1.12), (1.13) and (1.9) are related by the Bianchi identities (see e.g. Ref. [97]) we only need to
consider two of these equations to describe the dynamics of FLRW models. These are usually
taken to be the Friedmann and energy conservation equations (the later usually in the form
Eq. (1.11)).
We can re-write the Friedmann equation as a “conservation law of densities”,

8πG Λc2 kc2

ρ+ − =1 ⇔ Ω + ΩΛ + Ωk = 1 (1.14)
3H 2 3H 2 a2 H 2
where,
ρ Λc2 kc2 3H 2
Ω= , ΩΛ = , Ωk = − , ρc = (1.15)
ρc 3H 2 a2 H 2 8πG
are the matter (Ω), vacuum (ΩΛ ) and curvature (Ωk ) density parameters and ρc is the crit-
ical energy density of the universe. All these quantities evolve in time satisfying Eq. (1.14).
Therefore only two of the above density parameters are independent. For Λ = 0 (ΩΛ = 0)
models, the universe is geometrically closed (k = 1), flat (k = 0) or open (k = −1), depending
on whether its total energy density, ρ, is greater, equal or smaller than the critical density, ρc .
The acceleration rate of the universe is often expressed in terms of the deceleration parameter
which is defined as:
äa ä 1 1 + 3w
q≡− 2 =− 2
= Ω − ΩΛ (1.16)
ȧ aH 2
where the last equality results from combining Eq. (1.15) with Eqs. (1.13) and (1.11).

1.2.2 Epochs
In general the cosmological fluid can be regarded as a mixture of ideal fluids describing different
matter components, each of these having a particular energy density, ρi , and pressure, pi .
Considering that we have only two components consisting of non-relativistic matter, ρm , and
radiation, ρr , (ρ = ρr + ρm ) we can write the Freedmann equation as:

8πG kc2 Λc2

H 2 (t) = (ρr + ρm ) − 2 +
3 a 3
 ! % a &4 % a &3 % a &2 $
0 0 0
= H02 Ωr0 + Ωm0 + Ωk0 + ΩΛ0 . (1.17)
a a a
The second equality results from using Eq. (1.11) for matter (w = 0) and radiation (w = 1/3),
with ti = t0 being the present time (a(ti ) = a(t0 ) = a0 ). The quantities H0 , Ωr0 , Ωm0 , Ωk0
and ΩΛ0 are respectively the Hubble constant, H0 = H(t0 ), the matter, radiation, vacuum
and curvature density parameters evaluated at the present time (Ωr0 = 8πGρr (t0 ) /3H0 2 ,
Ωm0 = 8πGρm (t0 ) /3H02 ).
Equation (1.17) shows the relative contribution from the different fluid components to the
expansion rate of the universe. In a Big Bang scenario a → 0 in the limit t → 0. The expansion
rate is therefore initially dominated by the radiation component, Ωr . As the universe expands,
the contribution from the other terms becomes progressively important and can dominate the
expansion. When matter becomes the dominant component, H(t) is driven by the Ωm term
in Eq. (1.17). If the universe keeps on expanding and has a non-zero cosmological constant
the dynamics of the expansion becomes dominated by the vacuum term, ΩΛ . By comparing
the first with second term inside the square brackets of Eq. (1.17) we obtain the redshift at
which matter and radiation contribute equally to the expansion rate. This is

1 + zeq = Ωm0 /Ωr0 ≃ 2.4 × 104 Ω0 h−2 ,

where the last equality derives from the observed energy density of radiation (see next section).
The redshift 1 + zeq gives the time when the universe changes from being radiation dominated
to became dominated by matter. It is usually referred to as the matter–radiation equality
redshift.
Going back enough in time we find the universe in a stage where photons, electrons and baryons
are tightly coupled in a collisional plasma. When the temperature dropped to about 3600
Kelvin (zrec ≃ 1300) [61], electrons and baryons recombined to form the first neutral Hydrogen
atoms. Soon after this short period, known as recombination, the number of free electrons
drops dramatically and the scattering between the remaining free electrons and photons is
no longer sufficient to keep matter and radiation in contact. At this point (zdec ≃ 1100) the
CMBR photons decoupled from the fluid. Present observations of Ω0 and h indicate that the
matter–radiation equality happens before recombination (zeq ≃ 14300, see next section).

1.2.3 The observed universe

Our understanding of the universe relies ultimately on our ability to make measurements
and to compare those measurements with theoretical models. As (1.17) indicates, the key
observational parameters in the FLRW Big Bang models are the Hubble parameter and the
present-day densities of the mass–energy contents of the universe. Despite great progress in
the past decades there is still a considerable amount of uncertainty regarding the majority of
these parameters.

• Hubble parameter: The quest for the measurement of the Hubble parameter, h, is
the oldest among the cosmological parameters. It started soon after the discovery of the
 universal expansion and today it is still not known to high accuracy (see Refs. [47, 54]
for reviews). There are two basic ways of measuring h. One involves the measurement
of distances to nearby galaxies, typically by observing the periods and luminosities of
Cepheid stars within them, and then use these determinations to calibrate other methods
of measuring distances to more distant galaxies. This strategy is known as the cosmic
distance ladder [92]. The second way consists of using fundamental physics methods,
which permit the direct measurement of distances to faraway objects without using
the cosmic distance ladder approach. This is the case of methods involving Type Ia
or Type II supernovae, gravitational lensing, and the Sunyaev–Zel’dovich effect (see
Section ??). Although different approaches can still lead to different results, the range
of h determinations has been shrinking with time. Presently different observational
techniques seem to start to converge inside the range h ∈ [0.5, 0.9] to a mid-value of
h ≃ 0.7. This is the case of the methods based on the observation of distant Cepheids
using the Hubble Space Telescope [36, 37] and the observation of Type Ia supernovae
[56, 90].

• Total matter/energy density: The total matter or energy density of the universe,
Ωtot,0 = Ωr0 +Ωm0 +ΩΛ0 +Ωk0 , is presently most accurately constrained from observations
of the angular scale (or multipole) of the first acoustic peak in the angular power spec-
trum of the CMB anisotropies. The position of the peak is highly sensitive to Ωtot,0 . Ac-
cording to recent determinations from different CMB experiments its position is located
at a multipole value of lp ∼ 210. Assuming Gaussian adiabatic initial perturbations,
the Boomerang, MAXIMA and DASI experiments provide the following constraints on
the total matter/energy density: Ωtot,0 = 1.02+0.06 +0.18
−0.03 [21, 74]; Ωtot,0 = 0.9−0.16 [103]; and
Ωtot,0 = 1.04±0.06 [88], respectively. These estimates are remarkably consistent with an
Ωtot,0 = 1 (flat) universe.

• Radiation density: The energy density in all forms of electromagnetic radiation, Ωγ0 ,
is dominated by the contribution of the CMB, ΩCM B,0 (see e.g. Ref. [96]). This can be
computed accurately from the observed CMB mean temperature, TCM B = 2.725 ± 0.001
K [33], by using the Stefan-Boltzmann law, Ωγ0 ≃ ΩCM B,0 = 2.48×10−5 h−2 . To estimate
Ωr0 we need also to consider the contribution from the other relativistic species. Of
particular importance is the neutrino background. If in addition to radiation we assume
the existence of three families of massless neutrinos we obtain [18], Ωr0 ≃ 4.17×10−5 h−2 .

• Matter density: Many different techniques have been used to infer constraints on the
present day value of the matter density parameter, Ωm0 . These include methods based
on observations from CMB anisotropies, Type Ia supernovae, gravitational lensing, the
evolution of the abundance of X-ray clusters with redshift, gas mass fraction in galaxy
clusters, observational fits to the matter power spectrum of extra galactic objects, and
measurements of large scale peculiar velocities of galaxies (for an overview on these and
other methods see e.g. Ref. [91] and references therein). Results indicate that we are still
far from having an accurate determination of Ωm0 . In some cases different techniques
 can even show some degree of inconsistency.1 However, combined data analysis using
results from many of these methods give evidence in favor of a matter density parameter
of about Ωm0 = 1/3. For example, the authors in Ref. [44] have performed a likelihood
analysis using results from six independent data sets and found Ωm0 = 0.31 ±0.04 ±0.04,
assuming a flat Universe. In this determination the first error is mainly statistical and
the second is systematical. The observational data included constraints from recent
CMB observations made with Boomerang and MAXIMA, Type Ia supernovae, double
radio galaxies, lensing and large scale structure formation data.

• Cosmological constant: The strongest evidence for a positive cosmological constant

derives from observations of high-redshift Type Ia supernovae, which indicate that the
universe is currently evolving in accelerated phase of expansion. Independently, a posi-
tive Λ is also supported from a combination of observations indicating that Ωm0 < 1 in
conjunction with the CMB results which show that the universe is approximately flat.
Under the assumption of a flat Universe, with matter density Ωm0 ∼ 0.3, the energy
density associated with a cosmological constant term is ΩΛ0 ∼ 0.7.

Although there’s still a considerable amount of uncertainty in the determination of many of

the above parameters, the combination of data from present observations seem to indicate
that the Universe is consistent with being flat, and the ratio between the vacuum and matter
densities, ΩΛ /Ωm0 , is of the order of 2. For discussions on the current status of measurements
of cosmological parameters see e.g. Refs. [35, 87].

1.3 Exact solutions

The time dependence of the scale factor and the age of the universe result from the integration
of Eq. (1.17). The usual way to proceed is first to multiply Eq. (1.17) by (a/a0 )2 and use
Ωk0 = 1 − Ω0 − ΩΛ0 ,
) ,
* ' (−1 ' (−2 ' (2 -
d a(t) * a a a
= H0 +1 − Ω0 + Ωm0 + Ωr0 − ΩΛ0 1 − . (1.18)
dt a0 a0 a0 a0

1.3.1 The age of the universe

The integration of this expression, with the condition a(t = 0) = 0, gives
a(t)
=(1+z)−1
1
.
a0
t= H0−1 / dx, (1.19)
0 1 − Ω0 + Ωm0 x + Ωr0 x−2 − ΩΛ (1 − x2 )
−1

1
For example, preliminary constraints from the Cosmic Lens All-Sky Survey (CLASS) [46] appear to be in
strong conflict with the results from the Type Ia supernovae data [83, 84, 89].
where we have put x = a(t)/a0 = 1/(1 + z). This gives the age of the universe as a function
of the scale factor and the present day density parameters, t = f (a, Ω0 , Ωm0 , Ωr0 , ΩΛ0 ). Ob-
servationally we know that Ωr0 ≪ Ωm0 ≃ Ω0 but at early times the radiation term dominates
Eq. (1.19). Using current estimations of the density parameters it is easy to see that the
radiation-dominated period is very short when compared to the present age of the universe,
t0 . This means that in practice t0 can be calculated to a very good approximation by setting
Ωr0 = 0 in Eq. (1.19). Analytical expressions for the age of the universe can be found in
many textbooks for a range of cosmologies. Three cosmological scenarios of historical interest
are the flat universe with cosmological constant (Ω0 + ΩΛ0 = 1), the critical density universe
(Ω0 = 1) and the open universe without Λ (Ω0 < 1; ΩΛ0 = 0). The integration of Eq. (1.19)
for these models is also analytical and can be found for example in Ref. [61].
In some situations of interest is useful to define time in terms of the so-called conformal time,
dη = dt/a. This gives,
. t . a(t)
dt′ da
η(t) = ′
= . (1.20)
0 a(t ) 0 a(t )ȧ(t′ )
′

1.3.2 Scale factor

The inversion of t = f (a, Ω0 , Ωm0 , Ωr0 , ΩΛ0 ) with respect to a gives the dependence of the scale
factor with time. However since the expansion rate of the universe is dominated at different
phases by different fluid components (see Eq. (1.17)) it’s quite useful to examine the solutions
of the Friedmann equation for each of these phases. Restricting ourselves to the case Λ = 0,
the Friedmann equation for a single component fluid reads,
, ' (−(1+3w) -
a
ȧ2 = a20 H02 1 − Ωw0 + Ωw0 , (1.21)
a0

where Ωw0 is the present-day density of the fluid and w is the equation of state parameter in
Eq. (1.10). The solution of Eq. (1.21) is straightforward in the case of Ωw0 = 1 (flat geometry)

' (2/(3(1+w))
a(t) 3(1 + w)
= H0 t (1.22)
a0 2
ȧ 2
H(t) = = (1.23)
a 3(w + 1)t
äa 1 + 3w
q(t) = − 2 = = const . (1.24)
ȧ 2

These expressions are particularly useful for fluids dominated by matter (w = 0) and radiation
(w = 1/3). General solutions of Eq. (1.21) for this type of fluids are also not difficult to derive
and can be found in many cosmology textbooks (see e.g. Ref. [61]).
1.3.3 Distances, horizons and volumes
Of particular importance to the study of physical processes acting on different cosmological
scales is to determine the size of the largest causally connected regions at a given time. In a
universe described by Eq. (1.1) the regions in causal contact with an observer of coordinates
O = (t, r0 , θ0 , φ0 ) are those for which light rays emitted at the instant te reach O before or at
the instant t. Light rays arriving at (r0 , θ0 , φ0) later than t are beyond the horizon of O. Since
light rays travel along null geodesics (ds2 = 0), the coordinate distance travelled by light
between te = 0 and t is easily obtained from Eq. (1.1)
. re . t
dr dt′
√ =c ′
= c η(t), (1.25)
0 1 − kr 2 0 a (t )

where re is the radial coordinate at emission. Without loss of generality we set r0 = 0 and
assumed radial dθ = dφ = 0 geodesics. The corresponding (physical) proper distance is
. re / . re
dr
dH (t) = |grr |dr = a(t) √ = c a(t)η(t). (1.26)
0 0 1 − kr 2
This forms a spherical surface centered at (r0 , θ0 , φ0 ) known as the particle horizon of
O. For any observer, dH (t) separates the regions which can establish causal contact with
the observer at t (regions within the horizon) from those which cannot (regions beyond the
horizon). Using Eqs. (1.20) and (1.21) in Eq. (1.26) one obtains the following expression, valid
for 1 + z ≫ (1/Ωw0 − 1)1/(1+3w) and w > −1/3 (see Ref. [18])
' (3(1+w)/2
2 c 1/2 a 1+w
dH (t) ≃ Ω =3 ct . (1.27)
3w + 1 H0 w0 a0 1 + 3w

For Ωw0 = 1 this is in fact an exact solution of Eq. (1.26), whenever w > −1/3 (see e.g.
Ref. [61]). For universes with w < −1/3, the distance to the particle horizon becomes infinite.
This is the case of the vacuum-dominated de Sitter universe (w = −1), for which there’s no
particle horizon.
Another important length scale is the Hubble length, RH (also referred to as Hubble radius
or speed of light sphere). This is defined as the distance to the spherical surface (centered in
O) made by all points that at the time t have cosmological recessional velocities equal to the
speed of light. Setting v = c in the Hubble expansion law (1.3) one finds,

c 3(w + 1)
RH (t) = = ct, (1.28)
H(t) 2

where the last equality results from Eq. (1.23) and therefore it is only true for Einstein–de Sitter
universes. The Hubble length can be thought as the proper distance travelled by light during
the characteristic time scale of expansion, H −1 . Comparing Eq. (1.28) with Eq. (1.27) we see
that RH and dH only differ by a factor of the order of the unity (in particular for w = 1/3,
RH = dH ). This explains why both of these quantities are often used interchangeably and
referred to as the horizon. In practice the largest distance one can observe with electromagnetic
radiation is limited by what is called the visual horizon. This is defined as the distance to
the surface where the Cosmic Microwave Background Radiation was last scattered. Beyond
this last scattering surface (LSS) the universe becomes opaque due to the strong interaction
(Compton scattering) between matter and radiation. The CMB photons we observe today
suffered their last scattering at z ∼ 1000 when they were at a distance of about 6h−1 Mpc.
This corresponds to a present distance to the LSS of ∼ 6000 h−1 Mpc.
Let us now briefly examine how angular sizes and distances to faraway objects are defined.
Light emitted from the edges of an object (e.g. a galaxy cluster) located at a coordinate
position r and time t, occupies an angular size in the sky given by
D D D(1 + z)
θ= = = , (1.29)
dA a(t)r a0 r

where D is the object’s proper diameter and dA (t) = a(t)r is the proper distance to the
object at the moment of light emission. The second equality results from the fact that the
object is observed presently with redshift z. The quantity dA is called the angular diameter
distance. The present distance to the object is,
. 0 . a0 . a0
dr da c da
a0 r = a0 √ = a0 c = . (1.30)
re 1 − kr 2 a(t) aȧ H0 a(t) aH (a)

where H = H/H0 = (1−Ω0 +Ωm0 x−1 +Ωr0 x−2 +ΩΛ0 (1−x2 ))0.5 and x = a/a0 = 1/(1+z) (see
Eq. (1.18)). The relation between the angular size of objects and their redshifts is therefore
dependent on the underlying cosmological model and can in principle be used to constrain
cosmological parameters, provided one can find “standard rulers” (see e.g. Ref. [96] for a review
on “the standard tests of classical cosmology”). The evaluation of Eq. (1.30) is of particular
interest for matter-dominated universes (the phase we believe structures like galaxies and
clusters form). For Λ = 0 the expression Eq. (1.30) can be computed analytically (it was first
derived by Mattig in 1958 [70]). One obtains,
"√ #
2c Ω0 z + (Ω0 − 2) 1 + Ω0 z − 1
a0 r = . (1.31)
H0 Ω20 (1 + z)

In the case Λ ̸= 0, the evaluation of Eq. (1.30) has to be done numerically. A widely used
fitting formula for flat cosmologies with Λ was derived by Ref. [82]:
c
a0 r = [η(0, Ω0 ) − η(z, Ω0 )] , (1.32)
H0
where
√
η(z, Ω0 ) = 2 s3 + 1 (1 + z)4 − 0.1540s(1 + z)3 + 0.4304s2(1 + z)2
0
1−1/8
+ 0.19097s3(1 + z) + 0.066941s4 (1.33)

and s3 = 1/Ω0 − Ω0 . This fitting formula shows an accuracy better then 0.4% for 0.2 < Ω0 < 1
and any z.
Also useful is the definition of the volume element in the FLRW universes. This is,
/ a dr
dV = |g| dr dθ dφ = (ar)2 √ dΩ (1.34)
1 − kr 2

where |g| is the determinant of the metric (1.1) and dΩ = sin θ dθ dφ is the solid angle element.
Using Eq. (1.1) and the definition of redshift, one obtains the (physical) volume element per
unit of solid angle and unit of redshift,

dV c (a0 r)2 c d2A

= = (1.35)
dΩ dz H(z) (1 + z)3 H0 H (z)(1 + z)

where H (z) = H(z)/H0 and a0 r is given by Eq. (1.30).

Combining Eqs. (1.27) and (1.31) we can write an expression for the angular size of the horizon
scale at a given time in matter-dominated universes with Λ = 0. Setting D = dH we find,
3/2 √
θH Ω0 1+z
θH ≃ 2 tan = "√ #. (1.36)
2 Ω0 z + (Ω0 − 2) 1 + Ω0 z − 1

Although Eq. (1.27) is an approximate expression for non-flat universes, it’s possible to show
that Eq. (1.36) is in fact exact for all geometries (see e.g. Refs. [18, 106, 111] for general
expressions of dH and θ). At high redshifts Eq. (1.36) reduces to
2
180 Ω0
θH ≃ deg. (1.37)
π z
This expression tells us that the size of the horizon at the time of last scattering (z ∼ 1000)
occupies today an angular area in the sky no larger then ∼ 2 degrees.

1.4 Initial conditions and Inflation

As a mathematical framework of the Big Bang theory, the FLRW models have the great
virtue of describing well the dynamical properties of the observable universe (in particular
its expansion rate and age). They also give the correct temperature dependence on redshift,
which allows Big Bang nucleosynthesis to reproduce the observed light element abundances
so well. However, as we will see next, these models alone show an extreme sensitivity to the
“initial conditions” required to explain why the universe is the way we observe today. This
extreme “fine tuning” of the initial conditions raises a number of important questions which
led to the development of the theory of inflation.

1.4.1 Problems with the Big Bang

Some of the main problems regarding the initial conditions of the Big Bang theory are:
 • Horizon problem: According to Eq. (1.37) there are about 14000 to 65000 causally
disconnected regions in the CMB sky, assuming Ω0 ∈ [0.2 , 1]. If this is true why is the
CMBR very isotropic (showing blackbody spectrums with so similar temperatures in all
directions)? This is very difficult to understand if these regions were never in thermal
contact. Without any other mechanism to explain why the whole sky presents such
similar properties one is forced to impose this as an initial condition.
• Flatness problem: At early times the Friedmann equation can be written as:
|k| |k|
|Ω(t) − 1| = = , (1.38)
a2 (t)H 2 (t) ȧ2 (t)
where the quantity a2 H 2 = ȧ2 (t) is a decreasing function of time in all matter or radiation
dominated Big Bang FLRW universes.2 This means that as we go back in time the energy
density of universe has to be very close to the critical density, Ω(t) → 1. Dividing
Eq. (1.38) by itself written at the present we find that in order to get Ω0 ∼ 1, the energy
density needs to be extremely fine tuned in the past. For instance, when the universe
was t = 1 second of age (nucleosynthesis period) |Ω − 1| as to be of the order of ∼ 10−18 .
This becomes even more drastic at earlier times. At the Planck epoch (t ∼ 10−43 – a
time scale beyond which the classical GR equations should not be used) Ω can deviate
from the unity only one part in 1060 ! This shows Ω = 1 as an unstable critical point,
from which any initial deviation larger then what’s allowed by Eq. (1.38) leads to a
universe much different from that we observe today. So why has the universe to start
with an energy density so close to one?
• Monopoles and other relics problem: According to particle physics, the standard
Big Bang model meets the necessary conditions for variety of “exotic” particles (such
as the magnetic monopole, a very stable and massive particle) to be produced during
the early radiation dominated phase of the universe. Since these particles are diluted
by the expansion as a−3 they can very easily become the dominant component of the
universe. However no such particles have yet been observed. This either implies that
the predictions from particle physics are wrong, or their densities are very small and
therefore there’s something missing from this evolutionary picture of the Big Bang.
• Origin of structure problem: Locally the universe is not homogeneous. It displays a
complex hierarchical pattern of galaxies, clusters and super clusters. The general view
is that structure forms via gravitational instability from very small “initial” density per-
turbations. But, what is the origin of these initial perturbations? Without a mechanism
to explain their existence one has to assume that they “were born” with the universe
already showing the correct amplitudes on all scales, so that gravitational instability can
correctly reproduce the present-day structures.
• Homogeneity and isotropy problem: Why is the universe homogeneous on large
scales? At early times this “homogeneity” had to be even more “perfect”. The homoge-
neous and isotropic FLRW universes form a very special subset of all types of solutions
2
This can be easily verified by noting that for both matter and radiation dominated universes the second
member of Eq. (1.13) is always negative.
 of the GR equations. So why would nature “prefer” homogeneity and isotropy from the
beginning as opposed to, e.g., having evolved into that stage?

1.4.2 The theory of inflation

The theory of inflation (or simply inflation) was originally proposed by Guth [40] in an attempt
to solve some of the above difficulties of the standard Big Bang model. This theory does not
replace the Big Bang scenario. It’s rather an additional mechanism attached to the early
phases of the universe (prior to the radiation-dominated period), which liberates the Big
Bang model from its extreme sensitivity to the initial conditions. The mechanism proposed
by Guth solves the horizon, flatness and monopole problems. With time it would turn out
that inflation is a powerful falsifiable theory for the origin of cosmic structure, which makes
predictions that can be confirmed or ruled out against observations.
Inflation is simply defined as any period of the universe’s history during which the scale factor
a(t) is accelerating,
d " −1 #
Inflation ⇔ ä > 0 ⇔ cH /a < 0. (1.39)
dt
These are all equivalent ways of defining inflation. From the last equality we see that during
an inflationary phase the comoving Hubble length always decreases. This is a necessary and
sufficient condition for inflation to happen. As we noted before, in the early phases of the
standard Big Bang scenario aH = ȧ is bound to decrease continuously. This leads to the
flatness problem. So if we reverse this situation (aH = ȧ > 0) i.e if the universe experiences a
period of accelerated expansion, the energy density parameter in Eq. (1.38) is forced to move
away from the critical value Ω = 1 instead of approaching it. This would solve the flatness
problem. But when and in what way can inflation occur within the Big Bang scenario?
In its simplest form, inflation is sourced by a homogeneous scalar field ϕ, known as the
inflaton, with a stress energy tensor given by Tab = ϕ;a ϕ;b − gab L (ϕ) and a Lagrangian
L (ϕ) = 21 ϕ;a ϕ;b g ab − V (ϕ). The shape of the potential V (ϕ) depends on the details of the
model of inflation under consideration. Scalar fields play a central role in describing spon-
taneous symmetry breaking phenomena in particle physics theories. They usually represent
particles with spin zero, such as the Higgs field which is the particle responsible for electro-
weak symmetry breaking. At this point and for the reminder of the chapter we adopt a unit
system where c = ! = 1.
When we include the contribution of the inflaton field into the second term of Eq. (1.6),
the FLRW metric is still the solution of the Einstein equations of GR. The Friedmann and
acceleration equations are now modified to:
' (2
ȧ 8πG k
= (ρ + ρϕ ) − 2 (1.40)
a 3 a
ä 4πG
= − (ρ + ρϕ + 3(p + pϕ )) . (1.41)
a 3
where ρϕ and pϕ are the density and pressure of the field, respectively given by ρϕ = ϕ̇2 /2 +
V (ϕ) and pϕ = ϕ̇2 /2−V (ϕ). One should note that although we are not assuming the existence
of the Λ term in Eq. (1.6), the above equations possess the same form of Eqs. (1.12) and (1.13)
if we have Λ = −8πGpϕ = 8πGρϕ ⇔ pϕ = −ρϕ .
By either applying the conservation law (1.6) to Tab (ϕ) or using the Euler–Lagrange equations
on L (ϕ), we can derive the following equation of motion for the inflaton field,

ȧ dV
ϕ̈ + 3 ϕ̇ + = 0. (1.42)
a dϕ

This tells us that whenever the first two terms are negligible (ϕ̈ , ϕ̇ ≃ 0) the field experiences a
slow-roll period, during which its pressure and density are related by pϕ = −ρϕ ≃ V = const.
Moreover since ρϕ is approximately constant and initially ρ decreases as ρ ∝ a−4 , after some
time ti from the Big Bang the energy density of the inflaton field dominates the dynamics of
the expansion. This is the beginning of the inflationary period, during which the solution of
the Friedmann equation (1.40) is,
⎧
cosh [H(t − ti )] if Ω > 1
a(t) ⎨
= exp [H(t − ti )] if Ω = 1 (1.43)
ai
sinh [H(t − ti )] if Ω < 1
⎩

where ai = a(ti ) = (3/8πGV )1/2 and the Hubble expansion rate H = (8πGV /3)1/2 ≃ const.
This is the same dynamical behaviour we find in the de Sitter universe. Regardless of the
geometry, after a time interval t − ti ≫ (3/8πGV )1/2 the scale factor begins to grow expo-
nentially and the universe rapidly behaves as flat (Ω = 1). This is sufficient to remove the
unpleasant fine-tuning condition on Ω required by the standard Hot Big Bang theory. How-
ever inflation cannot continue indefinitely. At some point the universe needs to re-enter in
a decelerating phase which allows primordial nucleosynthesis to reproduce the correct light
elements abundances. With an exponential expansion we would now be inhabitants of an
“empty” universe.
Inflation ends when the scalar field approaches the minimum of the potential, V . At this
point ϕ ends its slow-roll motion and starts to oscillate around the minimum, Vmin . As it
oscillates the particles described by the field annihilate and the resulting energy is transferred
to the cosmological fluid, which experiences a (very rapid) temperature increase. This process
is known as re-heating. After the re-heating phase the universe becomes radiation dominated
and returns to its standard evolution. In a typical inflationary scenario, inflation starts just
before the GUT (Grand Unified Theory) phase transition, for ti ∼ 10−34 s, and finishes soon
after this at tf ∼ 10−32 s. The “amount of inflation” generated during this very short period
is usually expressed in terms of number of e-foldings,
tf ϕf
a(tf )
. .
N = ln ≃ Hdt ≃ −8πG V /V ′ dϕ, (1.44)
a(ti ) ti ϕi

where the last equality results from the use of Eqs. (1.40) and (1.42) under the slow-roll
approximation and ϕi and ϕf are the values of the inflaton field at the beginning and the end
of the inflationary phase. The amount of inflation produced during this period is therefore
dependent on the inflationary potential. The distance travelled by light during the same time
interval is then
. tf
dt
= H −1 eH(tf −ti ) − 1 ≃ H −1 eN ,
" #
d(tf , ti ) = a(tf ) (1.45)
ti a(t)

which shows that a causally-connected region with size equal to the Hubble volume is expo-
nentially expanded by eN at the end of inflation, whereas the Hubble radius itself remains
approximately constant during the same period, RH = H −1 ≃ const. After the end of infla-
tion the Hubble radius starts to grow again and eventually may enclose at some point regions
which were beyond the Hubble volume before inflation started. Another way of restating this
is to think in terms of comoving coordinates. During inflation the comoving Hubble length
decreases proportionally to ∼ e−N . Comoving scales of the size of the Hubble radius and
smaller are therefore pushed outside the comoving Hubble sphere. After inflation these scales
re-enter progressively the Hubble volume as the comoving Hubble radius starts to increase. If
the number of e-foldings is sufficiently large, scales that didn’t have time to establish causal
contact before inflation are still today beyond our observable horizon. This would explain the
high degree of isotropy and homogeneity we observe today.
This can also explain why magnetic monopoles and other Big Bang relics are unobservable
today. If inflation happens before or during the phase when these particles are created, their
density at the end of the inflationary period will decrease by a maximum factor of e3N . Again,
if N is big enough the density of these particles can still be very small today and therefore,
in practice, unobservable. Of course, this only works if inflation has enough time to dilute
these relics (see Ref. [65] for a discussion on the conditions required for this argument to hold
for different kinds of relics). It can be shown that the minimum amount of inflation needed
to solve these and other problems of the standard Big Bang model is about N ∼ 60 (see e.g.
Refs. [64, 65]).
Despite inflation’s success to bring the primordial universe towards homogeneity and a flat
geometry (as required by the standard Big Bang model), inflation’s most remarkable feature is
that it provides a theory for the origin of the primordial inhomogeneities. In the inflationary
scenario these inhomogeneities arise from quantum fluctuations about the vacuum state of
the inflaton field, which are always present due to the Uncertainty Principle (see Refs. [4, 41,
45, 65, 101]). The resulting irregularities can be of scalar (density perturbations) or tensorial
(gravitational waves) nature. Their amplitudes on a given scale can be fully specified by
the value of inflationary potential, and its derivative with respect to the inflaton field, at
the time the scale crosses outside the Hubble radius during the inflationary phase (see e.g.
Refs. [63, 64]).
Nowadays the title “standard model of cosmology” usually refers to the Hot Big Bang scenario
plus the theory of inflation as the mechanism responsible for the origin of cosmic structure.
Inflation generated perturbations are solely produced during the inflationary period and their
subsequent evolution is governed by the effects of gravity and cosmic expansion alone. For
this reason inflation-generated perturbations are said to be passive. Non-gravitational effects,
such as gas cooling and heating, only become important at later times when highly non-
linear structures (e.g. galaxy clusters) form. An alternative paradigm to inflation are the
theories which consider topological defects as the main source of cosmic structure (see e.g.
Refs. [25, 49, 60, 110]. In this case, perturbations are said to be active as they can arise at any
time and evolve also under the effect of non-gravitational forces. This significantly complicates
their treatment. However topological defect theories have been recently excluded from being
the main source of cosmic structure as they fail to predict the observed features (namely a
pronounced first peak and the indication for the existence of secondary acoustic peaks) in the
CMB anisotropy spectrum (see Ref. [25]).
Chapter 2

Cosmological Structure Formation

The formation of structure in the universe is one of the most outstanding problems of astron-
omy. The key point is to understand how small inhomogeneities present in a highly uniform
primordial universe can give rise to the complex pattern of structures we observe today. In
1902 Jeans [55] was the first to demonstrate that small density perturbations can grow with
time in a homogeneous and isotropic self-gravitating fluid. Slightly overdense regions provide
extra gravitational attraction towards them, which causes the in fall of the surrounding ma-
terial and the increase of their densities. Pressure gradients can resist the effect of gravity
and stop the growth of perturbations below a critical minimum length. This process became
known as the gravitational Jeans instability, and it was first studied in an attempt to describe
the formation of planets and stars in a static universe. The theory of gravitational instability
was applied to an expanding universe by Lifshitz in 1946 [66], which presented the first general
analysis on the evolution of inhomogeneities in FLRW models using linear perturbation theory.
Since then many authors have worked on achieving a theory that successfully describes the
formation of structure in the universe (see e.g. Ref. [79]).
Presently, it is widely accepted that structures grew via gravitational instability from passive
density perturbations produced in the primordial universe. As long as the perturbations are
small they can be well described in linear perturbation theory. However the accuracy of this
perturbative method starts to decrease as perturbations grow larger and higher-order terms
would need to be considered at later times. Eventually the perturbative theory itself breaks
down at the point when complex non-linear structures form. At this stage other methods need
to be used to follow the evolution of structure.

2.1 Density contrast and the power spectrum

Density perturbations are conveniently described by the density contrast or excess function
field,
ρ(x, t) − ρ0 (t)
δ(x, t) ≡ , (2.1)
ρ0 (t)

23
which is a dimensionless quantity that measures, at each point of the space–time, the deviation
of the density field ρ(x, t) from the mean density of the universe, ρ0 (t) = ⟨ρ(x, t)⟩. If nothing
is said in contrary, from hereafter we will assume the spatial coordinates, x, to be comoving.
In perturbation theory ρ0 (t) is the density of the unperturbed background universe and δρ0 (t)
is the perturbation to that background. The time dependence of δ can be followed in linear
perturbation theory, which applies whenever δ is small. Predictions for the spatial properties
of the density contrast can only be inferred from a statistical approach. Specific predictions
for each region of the universe would require the knowledge of its initial conditions. This
is something we do not know a priori and even if we had information on the specific initial
conditions of each point, it would be practically impossible to follow the evolution of large
regions of space.
In cosmology, fields like the overdensity contrast, are assumed to be random fields, which
are fully specified by an infinite set of joint probability distribution functions of the form,
⟨A(x)⟩, ⟨A(x)A(x′ )⟩,... , A(x)A(x′ )...A(x(n) ) ,... . These are respectively the one-point,
6 7

two-point,..., n-point probability distribution or correlation functions of the field A. The

angle brackets denote averages over an ensemble of possible universes. Since we are limited
to our own universe, these distributions are inaccessible to us even in principle. In practice
it is always assumed that cosmological random fields are ergodic and therefore the true
probability distribution functions (pdf) can be estimated by replacing the ensemble averages
by spatial averages over very large volumes of our universe. Another common assumption is
that cosmological random fields are invariant under rotations and translations. This implies,
for example, that the one-point pdf is independent of x and the two-point correlation function
is only function of the distance r = |x − x′ |.
A direct consequence of the ergodic hypothesis is that the one-point correlation function
of the overdensity contrast is zero, ⟨δ(x)⟩ = 0. The two-point correlation function ξ(r) ≡
⟨δ(x)δ(x + r)⟩, can be computed by expanding the overdensity field in a Fourier series1 over
a large box of volume V = L3
8
δ(x, t) = δk (k, t)e−ik·x (2.2)
1
.
δk (k, t) = δ(x, t)eik·x d3 x (2.3)
V
with periodical boundary conditions, which restrict the allowed wave numbers to be k =
2π(n1 , n2 , n3 )/L. The numbers n1 , n2 and n3 are integers. If the scales of interest are small
compared to the box size the summation in Eq. (2.2) can be replaced by an integral multiplied
by a prefactor of V /(2π)3 . The correlation function ξ(r) then reads,
V
.
2
6 7 −ik·r 3
ξ(r, t) = |δ k (k, t)| e d k (2.4)
(2π)3
1
.
2
ξ(r, t) eik·r d3 r
6 7
P (k, t) = |δk (k, t)| = (2.5)
V
1
We should note that Fourier expansions are only appropriate for flat geometries. For volumes larger
then the curvature radius in non-flat universes it’s more appropriate to use the generalized solutions of the
Helmholtz equation as a basis, see Ref. [51].
where P (k, t), the Fourier transform of the correlation function ξ(r, t), is the power spectrum
of the density perturbations. The power spectrum is usually redefined in a dimensionless form,
as variance per (natural) logarithm interval (e.g. as in Ref. [65, 76]),

V 2 3 ∞ sin kr 2
.
3
P(k, t) ≡ 3
4πk P (k, t) = k ξ(r, t) r dr. (2.6)
(2π) π 0 kr

With this definition the variance of the overdensity field (the zero-lag correlation function
ξ(0, t)) is simply given by
. ∞
2
6 2
7 dk
σ (t) ≡ δ(x, t) = P(k, t) . (2.7)
0 k

Most models of inflation favour density perturbations of adiabatic nature, with a Gaussian
probability distribution of variance σ 2 and a primordial density power spectrum, Pδ (k), well
described by a power law,
Pδ (k) = A k n . (2.8)
Here n is called the spectral index of scalar perturbations and in most cases it’s expected to be
close to the unity. The value n = 1 is also known as the scale–invariant or Harrison–Zel’dovich
power spectrum, [43, 119]. From Eq. (2.1) we see that values of δ < −1 are unphysical. This
means that, in reality, the density perturbations can only be considered Gaussian as long as
σ(t) is small (σ(t) ≪ 1).
A very useful quantity in the study of the statistical properties of the density field is the
smoothed density contrast, δ(R, x, t). This is defined as the convolution of the overdensity
field with some normalized window function, W (R, r = |x − x′ |) (the smoothing kernel), which
falls rapidly to zero beyond r > R,
1
.
δk (R, x, t) = W (R, |x − x′ |) δ(x, t) d3 x′ . (2.9)
V

The normalization factor is such that V = W (R, |x − x′ |) d3 x′ . The concept of smoothing is

9

important not only because many observational procedures automatically introduce smoothing
(due to resolution effects), but also because it produces a cut-off on the large k modes of δ that
describe small-scale structure. In practice smoothing erases the contribution from modes with
k larger then ∼ 1/R. Smoothing is also useful as it reduces the variance of the overdensity
field and therefore it allows to use linear perturbation theory to follow the evolution of the
smoothed density field for longer times. Common smoothing kernels are the Gaussian and the
Top-hat filters (see e.g. Ref. [76]),

V 2 2 2 R2 /2
Gaussian : W = 3/2 3
e−r /2R ; Wk = e−k (2.10)
(2π) R
: ! $
3V 1 if r < R sin(kR) cos(kR)
Top − hat : W = ; Wk = 3 − , (2.11)
4πR 3 0 otherwise (kR)3 (kR)2

where the right-hand side expressions are the Fourier transforms of each of the filters.
Of more importance than the smoothed overdensity field itself (which is zero for large inte-
gration volumes) is its variance and higher moments of the power spectrum. These can be
defined as,
V
.
2
σN (R, t) ≡ Wk2 (R, k) P (k, t) k 2N d3 k , (2.12)
(2π)3
from which is easy to obtain the variance of the smoothed overdensity on the scale R,
. ∞
2
6 2 dk
Wk2 (R, k)P(k, t) .
7
σ (R, t) ≡ δ (R, x, t) = (2.13)
0 k

A similar expression is valid for the variance of the smoothed peculiar velocity field,
. ∞
2
6 2 dk
Wk2 (R, k)PV (k, t) .
7
σV (R, t) ≡ δV (R, x, t) = (2.14)
0 k

where PV is the power spectrum of the peculiar velocity field. At the present we have
PV (k, t0 ) ≃ H0 Ω0.6 2
0 P(k, t0 )/k , [65]. The smoothing scale R is often specified by the mass
within the volume defined by the window function at the present time. In the case of the Top-
hat filter this is M = ρ0 (t0 )4πR3 /3 = 1.16 × 1012 h−1 (R/(h−1 Mpc))3 M⊙ and the variance of
the smoothed overdensity field coincides with the mass fluctuation on the mass scale M [18].

2.2 Linear perturbation theory

The evolution of the density perturbations in an expanding FLRW universe is in general a
difficult problem to tackle as it implies the study of the behaviour of perturbations on a
large range of scales for different matter components of the cosmic fluid. At a given time,
the total energy density depends on the nature and relative abundance of the individual
fluid components, which can exist in the form of collisional or collisionless material and be
of relativistic or non-relativistic nature. Free-streaming effects and the relative size of the
perturbations with respect to both the Hubble and the curvature radius, defined as RK =
a|k|−1/2 , are other factors which affect the growth of perturbations and influence the shape of
the observable density power spectrum P(k, t).
An important simplification to the structure formation problem is that perturbations are
expected to be small during the first stages of the structure formation process and therefore
the equations governing their evolution can be written in a perturbative form where all non-
linear terms can be neglected. In general this linear regime proves to be a good approximation
whenever δ ≪ 1. Although the detailed analysis of the growth of perturbations in a multi-
component cosmological fluid requires a general relativistic treatment (see Ref. [65]), many
fundamental aspects of the evolution of perturbations on scales smaller then the Hubble radius
can be followed by using Newtonian mechanics.2 This is ultimately justified by Birkhoff’s
theorem [9] of GR which states that the space inside a spherical volume (smaller then the
2
Note that for “reasonable” values of Ω + ΩΛ the Hubble radius is always smaller then curvature radius,
RK . From Eq. (1.14) we have RK = a|k|−1/2 = |Ω + ΩΛ − 1|−1/2 cH −1 > RH in the range 0 < Ω + ΩΛ < 2.
curvature radius) embedded in a FLRW universe must be flat and unaffected by the matter
distribution outside the sphere. Perturbations on scales larger then the Hubble radius need to
be treated within the framework of GR. However one should note that as the universe expands
the Hubble radius gradually encompasses perturbations of larger wavelengths. This is true
for both matter and radiation-dominated eras where the Hubble radius, RH ∝ t, grows faster
then physical wavelength of any given scale, λphy = a(t)λcom ∝ tα with α < 1 in both cases. In
the reminder of this section we will use the terms horizon and Hubble length interchangeably.

2.3 Sub-horizon perturbations in non-relativistic fluids

Here we introduce the basic equations which govern the linear growth of perturbations on
scales smaller then the horizon, in non-relativistic fluids, for both collisional and collisionless
systems. In a collisional system particles interact and experience short-lived accelerations
due to “shocks” with other particles. Collisionless particles practically don’t interact and
their motions are usually a lot less susceptible to strong accelerations as they generally move
under the influence of a mean force field. Examples of collisional and collisionless systems are
respectively the baryon-photon fluid at matter–radiation equality and the cold dark matter
component of the universe.
Collisional matter: the classical equations describing the dynamical evolution of an ideal
fluid of collisional gas particles, evolving adiabatically, in an expanding universe are respec-
tively the continuity, Euler and Poisson equations:

∂ρ
+ ∇ · (ρv) = 0 (2.15)
∂t ' (
∂v p
+ (v · ∇)v = −∇ φ + (2.16)
∂t ρ
2
∇ φ = 4πGρ (2.17)

where ρ(r, t), p(r, t) and φ(r, t) are the density, pressure and gravitational potential at the
physical position r = a(t)x and time t. The velocity field v is simply v = ȧx + au where
u = ẋ is the peculiar velocity in a comoving reference frame, which measures the deviations of
the velocity field v to the Hubble flow v0 = ȧx = Hr. In these a(t) is the scale factor obeying
the Friedmann equation. Since we are assuming the existence of small inhomogeneities, the
peculiar velocity field u is generally non-zero and constitutes a perturbation to v0 . The usual
way to proceed to describe the perturbations is to expand the other fields ρ, φ and p in terms of
their unperturbed values, respectively, ρ0 (t), φ0 (t) and p0 (t), plus perturbations and substitute
them in Eqs. (2.15–2.17). Rewriting the resulting system of equations in comoving coordinates,
dropping all non-linear terms and Fourier transforming all quantities with respect to x, one
obtains (after some algebra) the following second-order differential equation, describing the
evolution of each Fourier mode of the density contrast:
' 2 2 (
ȧ cS k
δ̈ + 2 δ̇ + − 4πGρ0 δ = 0 . (2.18)
a a2
Note that k = |k| is a comoving wavenumber and c2S ≡ (∂p/∂ρ)S is the adiabatic sound speed.
This well-known result introduces an important length scale (and corresponding wavenumber),
known as Jeans length (and Jeans wavenumber),

2π 2a /
λJ = ; kJ = πGρ0 , (2.19)
kJ cS
which separates two different regimes for the evolution of δ:

1. For k < kJ (λ > λJ ): perturbations on these scales are gravitationally unstable. In

this case Eq. (2.18) admits two independent solutions, one with increasing and another
with decreasing amplitude. The decaying solution erases itself with time. The growing
solution is physically more interesting. Its growth depends on the evolution of the scale
factor.

2. For k > kJ (λ < λJ ): perturbations on these scales do not grow. In fact their ampli-
tudes oscillate like an acoustic wave and even may decrease depending on the evolution
of the sound speed, cS .

An important example of a collisional system is the baryon–photon fluid prior to recombina-

tion. Perturbations of size smaller then the horizon can be followed with Eq. (2.18) during the
period between matter radiation equality and recombination, zeq >√z > zrec . The adiabatic
sound speed is that of a plasma with matter and radiation, cS = c/ 3(1 + 3ρm /4ρr )−1/2 , and
varies with redshift. Using this result in Eq. (2.19) we find that λJ is always larger then the
Hubble radius, for z > zrec (see e.g. Ref. [18]). This implies that all perturbations re-entering
the horizon during this period are unable to grow as they start to oscillate soon after the
horizon re-entry. For wavemodes larger then the Hubble radius, Eq. (2.18) is no longer appli-
cable. At this point one has to resort to GR linear perturbation theory. Equation (2.18) is
also not accurate for redshifts earlier then zeq , as the energy density of the fluid is dominated
by relativistic particles.
Collisionless matter: An important characteristic of collisionless systems is that they can
develop multi-stream flows for which there may not exist a unique matter velocity at a given
position. In this case the continuity and Euler equations, (2.15), (2.16), are not applicable
and need to be replaced by the Liouville-Vlasov or collisionless Boltzmann equation. The
dynamical evolution of a collisionless system is then described by

∂f ∂f ∂f
+v· − ∇φ · =0 (2.20)
∂t ∂r ∂v
∇2 φ = 4πGρ , (2.21)

where f (r, v, t) is the phase space density or distribution function of the system. It gives the
number density of particles in the volume element d3 r centered at r, with velocities in the range
d3 v around v. The density field is then ρ = m f (r, v, t)d3v, where m is the mean mass of the
9

particles. To describe the linear evolution of density perturbations one proceeds in a similar
way to what was done in collisional case. The starting point is to write f , φ and ρ in terms
of their unperturbed values, f0 (v, t), ρ0 (t) = m f0 (v, t)d3 v and φ0 (t) plus perturbations and
9

then substitute these in the above equations rewritten in terms of a comoving coordinate
system. After dropping all non-linear terms and Fourier transforming all quantities, one finds
again two distinct regimes of evolution separated by the wavenumber,
9 k·∂f0 /∂v 3
2a / −2 d v
kFS = πGρ0 ; v∗ = 9 k·v , (2.22)
v∗ f0 d3 v

where v∗ is generally of the order the mean velocity dispersion of the particles. When f0 (v)
is assumed to be Maxwellian distributed, v∗ coincides with the value of σ of the distribution.
We have:

1. For k < kFS : perturbations on these scales are gravitationally unstable. The equation
describing the evolution of the density contrast for each of these modes is [79]:
ȧ
δ̈ + 2 δ̇ − 4πGρ0 δ = 0 . (2.23)
a
As in the collisional case, the growing solution of this equation depends on the expansion
rate of the universe.

2. For k > kFS : perturbations on these scales are damped due to the free-streaming of
particles from overdense regions into regions of low density. This dissipation process is
known as phase mixing or Landau damping. The typical time-scale for the dissipation
of a perturbation of size λ is τ ≃ λ/v∗ . In the case of CDM particles the “dissipation
speed”, v∗ , is very small and therefore perturbations on cosmological scales are practically
unaffected. However in a HDM scenario this effect can be very strong. Assuming HDM is
made of massive (light) neutrinos, perturbations on scales smaller then the free-streaming
scale λF S ≃ 20 (mν /30 eV)−1 Mpc, are strongly suppressed (see e.g. Ref. [61]). In fact
more accurate calculations indicate even larger λF S ’s for neutrino masses of the order of
mν ≃ 30 eV, [12].

Due to the similarity with Eq. (2.19), the wavenumber kFS is sometimes referred to as Jeans
wavenumber and denoted by kJ .
It’s particularly interesting to investigate the solutions of Eq. (2.23) for matter and curvature
(if it exists) dominated universes. Since pressure is negligible the results derived also apply to
collisional systems (Eqs. (2.18) and (2.23) coincide for cS ≃ 0). The main conclusions are that
perturbations grow as δk ∝ t2/3 during matter domination and remain practically constant as
soon as curvature begins to be the driving term of the expansion (if ever). Detailed solutions
following through a matter–curvature transition can be found in Ref. [79]. Another important
scenario for the late evolutionary stages of the universe is when the Friedman equation becomes
dominated by a cosmological constant term. In this case the growth of perturbations is also
strongly suppressed as soon as ΩΛ ≫ Ωm [79].
Equation (2.23) can also be used to study the evolution of perturbations in the CDM compo-
nent during the radiation-dominated phase of the universe (z > zeq ). During this period the
“effect” of radiation on any collisionless non-relativistic material is simply that of controlling
the overall rate of expansion. In this case the solution of Eq. (2.23) is δk ∝ 1 + 3y/2 with
y = Ωm /Ωr = a(t)/a(teq ) [79]. Therefore perturbations on the collisionless non-relativistic
component remain ‘frozen’ practically until 1 + zeq = a0 /a(teq ). The transition from radiation
to matter domination is smooth and matches the law δk ∝ t2/3 found above, soon after zeq .
This “stagnation” period is known as the Mészáros effect.

2.4 Relativistic fluids and super-horizon perturbations

As mentioned earlier, during the radiation-dominated period, z > zeq , photons and baryonic √
matter constitute a collisional system, with sound speed rapidly approaching cS = c/ 3.
During this period the classical equations which lead to Eq. (2.18) are no longer applicable.
On scales smaller then the horizon, perturbations can be well described by special relativistic
fluid mechanics. The resulting equation for the evolution of δk has the same form of Eq. (2.18)
with the factor ‘4’ in the last term inside the parenthesis replaced by ‘32/3’
/ (see e.g. Ref. [18]).
This ‘relativistic
√ correction’ decreases the Jeans length definition by 8/3. Accounting also
for cS = c/ 3, the baryonic Jeans length is λJ ≃ 3 c t, which is larger then the Hubble radius.
This extends what we said earlier regarding baryon-photon perturbations to redshifts higher
then z > zeq . Density perturbations in the baryon-photon fluid are forced to oscillate (without
growth) as soon as they re-enter the horizon.
There’s another important effect in which perturbations in the baryon-photon fluid can be
suppressed. Before recombination photons and baryons are tightly coupled due to Thomson
scattering. The mean free path of the photons is λγ ≡ 1/(ne σT ) ∝ x−1 −3 2 −1
e (1 + z) (ΩB h ) ,
where σT is the Thomson cross section and xe and ne are respectively the fraction and number
density of free electrons. The ideal fluid approximation holds as long λγ is small. As the
universe recombines and the fraction of free electrons decreases, the mean free path of the
radiation becomes larger and the perfect fluid approximation starts to break down. Photons
can then diffuse out of perturbations of size comparable to the mean free path, dragging
with them the electron-baryon plasma. This effect is known as Silk damping and can erase
perturbations on scales smaller than λS = 3.5(Ω0 /ΩB )1/2 (Ω0 h2 )−3/4 Mpc [61].
Density perturbations on scales larger than the Hubble radius (super-horizon modes) need
to be treated within the general relativistic theory of cosmological perturbations. The main
difficulty with these modes is that the density perturbation, δρ, is a gauge dependent quantity
and therefore different gauge choices can lead to distinct predictions for the behaviour of
δρ outside the horizon. A natural approach to the problem is to formulate the evolution of
the perturbations in terms of gauge-invariant quantities (see e.g. Refs. [2, 73]). Of particular
importance to δρ is the gauge-invariant variable ξ, which is related to the density perturbation
by ξ = δρ/(p0 + ρ0 ) at horizon crossing [4]. A very nice feature of ξ is that it is constant
outside the horizon. So the problem of finding the amplitude of a given mode of the density
perturbations at the horizon re-entry reduces to the problem of computing ξ at the moment
that modes crossed outside the Hubble radius during inflation. Evaluating ξ at both horizon
crossings we have:

V ′ (ϕ)δϕ
' ( ' ( ' (
1 δρ δρϕ
= = , (2.24)
(1 + w) ρ0 hor. in ρϕ0 + pϕ0 hor. out ϕ̇2 hor. out

where V is the inflationary potential, ϕ the inflaton field and, during inflation, ρϕ + pϕ = ϕ̇2
(see Section 1.4.2). The value of w depends on the instant a given scale re-enters the horizon
(w = 1/3 during radiation domination and w = 0 during the matter-dominated period). This
result directly connects the amplitude of the density perturbations at the horizon re-entry with
the primordial quantum mechanical fluctuations of the inflaton field, which can be computed
for a given model of inflation. The evolution of the perturbations after horizon re-entry can
then be followed by using the above equations of fluid mechanics, valid inside the horizon.

2.4.1 Relativistic multi-fluid case

Unlike the systems we have considered so far, the real universe does not behave like a simple
fluid. We know for a fact that the universe is made of different species of relativistic and
non-relativistic particles co-existing in a multi-component system. Detailed predictions for
evolution of the density perturbations therefore require a careful analysis of the evolution of
the perturbations in each of these components. In general each component can be considered
as a single fluid subjected to its own pressure (described by an individual equation of state)
and the gravity produced by all existing particles. In some cases collisions between particles
of different species give rise to energy exchanges between fluids, which then evolve as coupled
systems. The evolution of such a multi-component fluid is therefore described by a system of
coupled second-order equations (each representing a different component), which usually can
only be solved numerically.
Observations indicate that we need to consider at least four types of particles: Collisionless
dark matter, baryons (which in an abuse of language account also for the electron popula-
tion), collisional photons and collisionless (relativistic) neutrinos. We are mainly interested in
the evolution of the total density perturbations of this system and in particular we want to
determine how the amplitude of a perturbation mode at the horizon re-entry evolves to the
present. The most general treatment involves solving the appropriate Boltzmann transport
equations for both matter and radiation in a evolving universe. This is a non-trivial task as
diffusion and free-streaming effects on both collisional and collisionless particles need to be
considered and relativistic particles can become non-relativistic as the temperature decreases.
Particular difficulties arise due to the fact that photons change from being a tightly-coupled
fluid with the baryons to become a “collisionless” component after decoupling.3 The result of
solving the transport equations is usually expressed in terms of a transfer function, T (k, t),
which encodes all of the above effects and gives the amplitude of the total density perturbation
of the mode k at the time t.
3
Photons are believed to interact again with free electrons when the universe becomes re-ionized at later
times.
2.4.2 Density transfer function and the power spectrum
The transfer function is commonly defined as the ratio between the amplitude of a given mode
and the amplitude of a mode with an arbitrary large wavelength, so that in the limit k → 0 we
have T (k, t) → 1. In practice δk (k, t) = T (k, t) δk (khor , t) where khor is the largest mode inside
the horizon at t. Many codes have been developed to compute numerically the detailed shape
of the transfer function given a particular model of the universe (e.g. the recently popular code
cmbfast [1, 98]). Results depend on the type of mechanism responsible for the perturbations
and on the relative abundance and nature of the matter and radiation components assumed for
that model. Once these are specified the shape of T (k, t) reflects the effects of growth under
self-gravity, pressure and damping due to diffusion and free-streaming, discussed earlier. In
general, the overall effect is to reduce the amplitude of the short-wavelength modes relative
to the amplitude of the long-wavelength modes. In a typical CDM scenario, the transition
point is around the wavenumber corresponding to the size of the horizon at the redshift of
matter–radiation equality, keq ≃ 14Ω−1 0 h
−2
Mpc−1 . This is because the dark matter component
dominates the fluid and perturbations entering the horizon after zeq (k < keq ) can start to
grow more efficiently, whereas modes with k > keq are unable to grow essentially due to the
Mészáros effect and Silk damping.
The transfer function for cold dark matter models can be approximated by the following
analytical fitting formula derived in Ref. [3]:
ln(1 + 2.34q)
T (q) = [1 + 3.89q + (16.1q)2 + (5.46q)3 + (6.71q)4]−1/4 , (2.25)
2.34q
where q ≡ (kθ/h Mpc−1 )/Γ, θ = TCM B /2.73, and TCM B is the CMBR mean temperature.
The shape parameter, Γ, is given by [104]
√
Γ = Ω0 h exp[−ΩB0 (1 + 2h/Ω0 )] . (2.26)

The fitting formula (2.25) is known as the BBKS transfer function for CDM models and it was
first derived assuming a pure cold dark matter scenario (ΩB0 = 0). The inclusion of a baryonic
component preserves the same functional shape of Eq. (2.25) but has the effect of lowering the
“apparent” dark matter density parameter by the exponential factor in Eq. (2.26), [77, 104].
These expressions also assume a number of families of massless neutrinos equal to Nν = 3.
For different values of Nν a multiplicative factor of (g∗ /3.36)−1/2 is sometimes considered in
Eq. (2.26), where g∗ = 2 + 2(7/8)(4/11)4/3 Nν is the number of effective degrees of freedom of
the neutrinos [24]. Fitting formulae for other popular structure formation transfer functions
can be found in Refs. [3, 29].
According to Eq. (2.5), the density power spectrum P (k, t) differs from its primordial value
Pδ (k) simply by the square of the transfer function, P (k, t) ∝ T (k, t)2 Pδ (k). Since most of
the modulating processes shaping T (k, t) finish by the recombination epoch, trec , only times
after t > trec are of interest. After recombination the power spectrum P(k, t) can be written
as [64, 65, 76]
(4
g 2 (Ω, ΩΛ )
'
k
P(k, t) = 2 T 2 (k, t) δH
2
(k) , (2.27)
g (Ω0 , ΩΛ0 ) aH
 2
where g(Ω, ΩΛ ) is the linear growth suppression factor, δH (k) ≡ ⟨(δρ/ρ)2 ⟩aH=k specifies the
amplitude of the density power spectrum at the horizon re-entry and the quantities Ω, ΩΛ
and aH are to be computed at the time t. The suppression factor g(Ω, ΩΛ ) gives the rate of
growth of the density perturbations relative to the growth in the Einstein–de Sitter Universe,
δk ∝ (aH)−2 ∝ t2/3 . Note that with this definition all cosmological models give the same
density power spectrum at the present. Since perturbations grow less rapidly in low-density
models (with or without cosmological constant), structures need to form earlier so that the
same density power spectrum is obtained with these models at the present. The suppression
growth factor can be calculated analytically for open models without cosmological constant
(see Ref. [79]). For flat models with Λ the calculation involves an integral which cannot be
done analytically. In practice it’s common to use the following accurate approximation, valid
during the matter domination era, for both cases [16],
! ' (' ($−1
5 4/7 1 1
g[Ω(z), ΩΛ (z)] = Ω(z) Ω(z) − ΩΛ (z) + 1 + Ω(z) 1 + ΩΛ (z) , (2.28)
2 2 70
where the redshift dependence of the matter and cosmological constant density parameters
(Ω ≡ Ωm ) are respectively,
Ω0 (1+z)
⎧
⎨ Ω0 (1+z)+(1−Ω 0)
, if Ω < 1; Λ = 0
Ω(z) = (2.29)
Ω0 (1+z)3
, if Ω = 1 − Ω
⎩
Ω0 (1+z)3 +1−Ω0 Λ

and
1 − Ω0
ΩΛ (z) = , (2.30)
Ω0 (1 + z)3 + 1 − Ω0
which is only valid for flat cosmologies. If the primordial power spectrum is well described by
the power-law in Eq. (2.8) we have,
' (n−1
2 2 k
δH (k) = δH (k0 ) , (2.31)
k0
where δH (k0 ) is a normalization factor at some arbitrary comoving scale k0 . The value of
δH (k0 ) can be set so that (2.27) reproduces the correct amplitude of the CMBR anisotropies
on the scales observed by COBE. Another way of specifying the normalization of the density
power spectrum is by quoting the rms dispersion of the overdensity field smoothed on some
scale probed by large-scale structure observations. Traditionally, the smoothing chosen is the
Top-Hat filter on a scale of size 8h−1 Mpc−1 . The corresponding smoothed rms dispersion
is denoted by σ8 ≡ σ(R = 8h−1 Mpc−1 , t0 ). The normalization is then σ82 = P(keff ), where
keff is an effective wavenumber well approximated by keff = 0.172 + 0.011 [ln(Γ/0.34)]2 for a
CDM-like power spectrum [76]. One of the best methods to constrain the value of σ8 is by
confronting the observed number density of local galaxy clusters (selected in various ways) with
the predictions given by the Press–Schechter theory [86] or other mass-function models. This
method, originally proposed in Refs. [32, 48] and later popularized by the work in Ref. [116],
has been developed by many authors (see e.g. Refs. [81, 85, 109] and references therein) to
alow the normalization of the density power spectrum for a range of cosmologies. For example,
according to the work in Ref. [109]: σ8 = 0.56 Ω−α0 , with α = 0.47 and α = 0.34 respectively
for flat (Ωk0 = 0) and open (Ω0 < 1, ΩΛ0 = 0) cosmologies.
2.5 Non-linear evolution of perturbations
As mentioned earlier, linear perturbation theory applies when perturbations are small, δ ≪ 1.
This is guaranteed almost everywhere in space as long as the rms fluctuations of the density
contrast are much smaller then the unity, σ(t) ≪ 1. Regions of space with δ(x, t) ! 1 are
initially very rare, but as perturbations evolve under the effects of gravitational instability more
and more of such regions appear and the linear perturbation theory become inappropriate to
describe a progressively larger number of regions. A way of extending the use of linear theory
is to consider the linear evolution equations applied to the smoothed density contrast δ(R, x, t)
instead of δ(x, t). As before the linear regime applies as long as σ(R, t) ≪ 1, in which case
perturbation modes smaller then k " R−1 evolve linearly. Of course, this is only useful if the
scales of interest are larger then the smoothing scale R.
A natural way to proceed to describe perturbations during their first stages of non-linear
evolution (quasi-linear regime) is to consider second order terms in the evolution equations of
δ [79]. This method has however a limited range of applicability because higher-order terms
become also important very rapidly. In general is considered that the transition from the
linear regime to the stage of quasi-linear or non-linear evolution occurs around σ(R, t) ∼ 1.
The best way to follow the full evolution of perturbations is, undoubtly, by using numerical
N–body simulation techniques. Some of these will be described in detail in Chapter ??.
However there are some very useful analytical approximations which can give a good insight
on the general problem of describing perturbations during the first stages of their non-linear
evolution. Some of these will be reviewed next. The major advantage of such analytical or
semi-analytical methods is that they are generally more flexible and a lot less time consuming
then numerical simulations.

2.5.1 Spherical collapse model

The spherical collapse model (see e.g. Refs. [75, 79]) can be used to give an estimate of how
dense and when a spherically symmetric overdense distribution of matter abandons the general
expansion of the universe and starts to gravitationally collapse towards its centre.
When the density of the overdense region becomes slightly larger then the critical density it
starts to behave as a small closed universe, which evolves independently of the outside space.
Initially the sphere will keep on expanding with the universe until it reaches a maximum radius,
known as the turnaround point, and subsequently collapses after a time interval, tcoll , equal to
double the time it takes to get to maximum expansion, tmax . Assuming matter domination and
that the universe is globally Einstein–de Sitter, it’s possible to show that the linear evolution of
the density perturbation inside the sphere is given by δlin = 3/20(6πt/tmax)2/3 . This indicates
that the linear overdensity at turnaround and collapse are respectively:

turn
δlin = 3/20(6π)2/3 ≃ 1.06 ; coll
δlin = 3/20(12π)2/3 ≃ 1.686 . (2.32)

The spherical collapse model also allows one to compute the non-linear density contrast at
both of these instants. In the first case we have
turn
1 + δnonlin = (3π/4)2 ≃ 5.55 (2.33)

and at collapse the spherical model predicts an infinite density. This happens because pressure
forces resisting the gravitational infall are ignored. More realistically we expect that part of
the kinetic energy of collapse is converted into random motion of particles and the object
will eventually reach a state of virial equilibrium characterized by a temperature, Tv , and
radius, Rv . Using the virial theorem to equate the virial energies at the instants of maximum
expansion and compression one finds that Rv = Rmax /2, where Rmax is the radius of the
spherical region at maximum expansion. Numerical simulations confirm that Rv is in fact a
coll
good estimate of the radius of the collapsed sphere and that δlin is a good indicator of when
the object becomes virialized. If we assume that at the time of collapse, tcoll = 2tmax , the
object virializes with radius Rv , the non-linear overdensity is then
vir
1 + δnonlin = 18π 2 ≃ 178 (2.34)

which is in very good agreement with results from simulations, see Ref. [17].
For non Einstein–de Sitter universes the predictions from the spherical collapse model are
slightly changed. In general both linear and non-linear overdensities at collapse are functions
of the matter density, Ω0 . However in low-density models, it’s possible to show that the
coll
variations in δlin are less than ∼ 5% in the range Ω0 ∼ 0.1–1 and that the overdensity at
collapse needs to be corrected by Ω−0.70 for ΩΛ = 0 universes and by Ω−0.45
0 for flat universes
with Λ [30, 31, 116]. Sometimes the non-linear overdensity is expressed in terms of the critical
density instead of the background density, ∆ ≡ ρ(t)/ρcrit = (1 + δnonlin ) Ω. In this case it is
usually referred to as the non-linear density contrast (instead of non-linear overdensity, see
e.g. Refs. [30, 31]). For detailed calculations involving the evolution of the density contrast at
collapse, ∆c (z), there is the following fitting formula, from Ref. [13]:

18π 2 + 60x − 32x2 , if ΩΛ = 0

:
∆c (z) = (2.35)
18π 2 + 82x − 39x2 , if Ω = 1 − ΩΛ ,

where x = Ω(z) − 1. This expression is accurate to 1% level in the range Ω ∼ 0.1–1.

2.5.2 Zel’dovich approximation

The Zel’dovich approximation, [99, 119], is a very successful first–order Lagrangian perturba-
tion theory, which describes perturbations with respect to particle positions rather then to
the density field (as described in the Eulerian perturbative method). In this approach, par-
ticles move in straight lines after being displaced from their original (comoving) positions by
the action of a initial velocity field sourced by the density perturbations. According to this
approximation, the comoving position and peculiar velocity of a given particle at the time t
are:
x(t) = q + D(t) u(q) ; ẋ(t) = Ḋ(t) u(q) , (2.36)
where x and q are respectively the final and initial comoving coordinates, u(q) is the initial
velocity displacement field and D(t) ∝ g(Ω, ΩΛ ) t2/3 is the growing mode factor of the den-
sity perturbation. In fluid mechanics terminology, x and q are said to be the Eulerian and
Lagrangian coordinates, respectively. In these expressions u(q) is usually assumed to be an
irrotational field, which satisfies

δ(q) = −∇ · u(q) ; ∇ × u = 0. (2.37)

Due to its nature the Zel’dovich approximation is commonly used to set up the initial condi-
tions in numerical N–body simulations. The initial velocity field is computed from Eq. (2.37)
by generating a random Gaussian realization of the density power spectrum and then equa-
tions (2.36) are used to generate the initial matter and velocity distributions required for the
first integration step of the simulations.
The main limitation of the Zel’dovich approximation is that particle trajectories are uniquely
set by the initial displacements and as motions intercept, particles keep moving away from
the forming structures. At this point, usually referred to as shell crossing, the Zel’dovich
approximation breaks down. However, being a first-order Lagrangian perturbative method,
the Zel’dovich approximation usually achieves the same level of accuracy in describing the
perturbations for longer then in the (Eulerian) linear perturbative case.

2.5.3 Press–Schechter theory

Perhaps the most successful analytical method to describe the density perturbations during
the early stages of their non-linear evolution is the Press–Schechter (PS) formalism [86]. The
central quantity in the theory is the PS mass function (or multiplicity function), which gives
the comoving number density, n(M, z)dM, of virialized objects forming at the epoch of redshift
z, with masses in the infinitesimal range M to M + dM,
2
δc2
' (
2 ρ0 δc dσ(M, t)
n(M, z)dM = − exp − 2 dM . (2.38)
π M σ 2 dM 2σ (M, z)

Here ρ0 is the comoving matter density, δc is the linear overdensity threshold at virialization,
σ 2 (R(M), z) is the variance of the smoothed density contrast filtered on the comoving scale
R(M) and M is the mass enclosed in the volume occupied by the smoothing filter, M = ρ0 V 3 .
With this expression, the total number of virialized objects, per unit of solid angle, with mass
above Mmin is simply
. ∞ . ∞
dN dV dn
(> Mmin ) = dz dM (2.39)
dΩ 0 dzdΩ Mmin dM

where dV /dzdΩ is the volume element (1.35). In general, the above integral requires the
use of Eq. (2.13) to evaluate σ 2 (R(M), z), with the appropriate smoothing filter and density
power spectrum, at each step of the redshift and mass integrations. In practice this can
be much simplified by using accurate fitting formulae to approximate σ 2 (R(M), z) in the
scales of interest and then use those formulae in the integration process. A particular good
approximation on scales in the vicinity on 8h−1 Mpc is the following fitting formula from
Ref. [109] for a CDM power spectrum and a Top-Hat filter (R = (3M/4πΩρcrit )1/3 ):
' (−γ(R)
R
σ = σ8 (z) −1
, (2.40)
8h Mpc

where σ8 (z) is the normalization of the power spectrum at z and

! ' ($
R
γ(R) = (0.3Γ + 0.2) 2.92 + log . (2.41)
8h−1 Mpc

Using Eqs. (2.13) and (2.27) we obtain

σ8 (0) g[Ω(z), ΩΛ (z)]

σ8 (z) = , (2.42)
1 + z g[Ω0 , ΩΛ0 ]

where σ8 (0) is the power spectrum normalization at the present. The shape parameter Γ is well
constrained, in the vicinity of the 8h−1 Mpc scale, by fitting observations of the galaxy and
cluster correlation functions with a CDM power spectrum, Γ = 0.230+0.042
−0.034 at 95% confidence
level [77, 108].
Although triaxial perturbations collapse preferentially into flattened structures (known as pan-
cakes) it’s usually assumed that rare objects like galaxy clusters collapse in a fairly symmetric
coll
way [6]. Therefore the density threshold in Eq. (2.38) is usually assumed to be δc = δlin = 1.69,
which, as mentioned before, is practically independent on the cosmology for many situations
of interest.

2.5.4 Numerical simulations of large–scale structure

Over the past two decades, numerical N–body simulations have become a powerful theoretical
tool to investigate the formation of structures in the universe. Complex non-linear physics,
acting over a wide range of scales, makes the structure formation process hard or impossible
to describe without the use of numerical computation methods. These methods track the
growth of structure in time by integrating the equations of motion of the individual particles
and following their trajectories under the action of gravity and other physical forces involved.
Numerical simulations which only account for the effects of gravity are appropriate to describe
the evolution of the dark matter component of the universe. These are usually referred to as
N-body simulations. To describe the evolution of perturbations on both baryons (gas) and
dark matter, simulations need also to describe the dynamics of the gas and to include the
appropriate gas physics in the range of scales probed by the simulations. These are known as
hydrodynamical N–body simulations.
Cosmological simulations of structure formation require the appropriate modelling of the phys-
ical processes involved (e.g. gravity, gas dynamics, radiative phenomena), a clear definition
of the starting (initial) conditions (power spectrum of density perturbations at the simulation
start and background model) and a set of parameters intrinsic to the numerical simulation
scheme (e.g. number of particles, box size and softening). Follows a brief description of how
these aspects are usually set up in numerical codes of large-scale structure formation.

Initial conditions

All numerical simulations require a clear definition of the initial conditions, from which the
simulations start. For structure formation simulations these consist of specifying the back-
ground cosmology and the perturbations to this background.
The background model is specified by a set of cosmological parameters, which describe the
matter and radiation contents of the universe (e.g. the density parameters of dark matter,
baryons and cosmological constant), and the geometry of the universe (usually taken to be
flat or open Friedman–Robertson–Walker space-times). The initial perturbations to the back-
ground model are set by assuming that when simulations start, at a time ti , perturbations are
still small and well described in linear perturbation theory (see Section 2.2). The Zel’dovich
approximation (described in Section 2.2) is then used to assign initial positions and velocities
to the simulation particles.
The procedure consists as follows. Particles are originally placed on a cubic grid, with Nrow
cells per side, inside the simulation volume V = L3 (simulation box). Assuming periodic
boundary conditions the overdensity field, δ, can be expanded in a Fourier series of the form
(2.2), where wavemodes above the Nyquist frequency, kNy = (2π/L)(Nrow /2), are excluded to
prevent aliasing effects. The largest mode in the box corresponds to a minimum wavenumber
kmin = 2π/L. The next step is to generate a Gaussian random realization of the linear
theory matter power spectrum, P(k, ti ), which is normalized from observations of the local
abundance of rich clusters, as described at the end of section 2.4.2. The amplitude of σ8 at
z(ti ) is derived from σ8 (z = 0) by using the scaling relation (2.42). The initial positions and
velocities of the particles are then calculated by using the Zel’dovich approximation, Eq. (2.36),
see e.g. Ref. [59].
The matter power spectrum is usually an input of the initial conditions generator codes.
This can be giving by reading in a pre-computed table of the power spectrum generated by
Boltzmann codes, or by specifying the parameters of transfer function fitting formulas from the
literature. For CDM simulations with a cosmological constant, the BBKS transfer function
(given by equations (2.25) and (2.26)) is a common choice for setting up the simulations
initial conditions. Due to symmetry properties of the Fourier components, δk∗ = δ−k , there’s
only a finite number (Nrow /2) of possible modes that can be accommodated in the box. To
sample the matter power spectrum in a wide range of modes one needs to have both large
simulation volumes and number of particles (high resolution simulations). This also implies
large computational times and system resources.
Gravity

Gravity plays a central role in the structure formation problem. Even though pure N-body
codes are sometimes used to probe scales as big as the Hubble volume in non-flat geometries,
gravity is usually modelled by Newton’s laws of gravity:
dx 1
= v (2.43)
dt a
dv ∇φ
+ Hv = − (2.44)
dt a
2 2
∇ φ = 4πGa [ρ(x, z) − ρ̄(t)] . (2.45)

These equations are written in a spatial coordinate system comoving with the cosmological
expansion of the universe. Here, v is the peculiar velocity, a(t) is the cosmological scale factor,
φ is the gravitational potential, ∇ = ∂/∂x is the gradient in comoving coordinates, H = ȧ/a
is the Hubble constant and ρ and ρ̄ are the density and spatial mean density (density of the
background model). The quantity g = −∇φ/a is the peculiar acceleration field. For the
numerical computations, periodic boundary conditions are usually adopted, so that a finite
expanding volume probed in a simulation is properly embedded in the perturbed background
space-time.
In order to obtain the evolution of the particle trajectories in time one needs to integrate
equations (2.43) to (2.45). This always has to start with the evaluation of the gravitational
potential φ. Since gravity is a long-range force, the potential seen by a particle depends, in
theory, on the mass and location of all other particles. The art of N-body simulations lies
exactly in how numerical codes compute the gravitational potential. In the next subsections
we will review some of today’s most common algorithms developed to perform this task.

• The Particle–Particle algorithm: The usual way to obtain the gravitational po-
tential φ is to integrate Eq. (2.45). For a population of point particles, the peculiar
acceleration field, g(xi ), acting upon a particle at the position xi is then given by a
summation over all the other mass particles, mj (see e.g. Ref. [79]):

∇φ (xi ) G8 rij
g (xi ) = − = 2 mj (2.46)
a a j |rij |3

where G is the gravitational constant and rij = xj −xi is the displacement vector between
the particles i and j. The most straightforward way to obtain the gravitational force act-
ing upon each particle is to perform the direct summation in Eq. (2.46). The algorithms
which implement this summation are called Particle–Particle or PP algorithms. Usually
they also include a smoothing length procedure to soften the gravitational force between
particles (gravitational softening). This is necessary to prevent two-body effects from
dominating, which can lead to the formation of unphysical binaries (see e.g. Ref. [27]).
The main problem with the PP technique is that the number of calculations needed to
follow the evolution of a population of particles increases with the square of the number
 of particles O(N 2 ). This makes the whole enterprise very expensive in computational
time and even with the largest parallel supercomputers the task becomes prohibitive for
a number of particles greater than N > 107 .
• The Particle–Mesh algorithm Particle–Particle interactions are more important on
small scales. On larger scales, such as those probed in large cosmological simulations,
the long-range force seen by a particle can be approximated by an average potential
generated by a new set of surrounding particles. This requires assigning average masses
to points on a Cartesian grid and then evaluate the potential seen by one particle from
the interaction of that particle with its surrounding grid neighbours. This is the heart
of the so-called Particle–Mesh or PM algorithm, and its operation can be summarised
in tree basics steps:
– Mass assignments: first, the mass density field is computed over Ng grid points
using a certain assignment scheme. One of the most commonly used schemes is the
Cloud–in–Cell (CIC) algorithm, which uses multi-linear interpolation to the eight
grid points on the corners of the cubic mesh cell containing the particle.
– Potential estimation on the grid: the Poisson equation is then used on the grid
to evaluate the gravitational potential. This is usually accomplished by the use
of a fast Poisson solver, which is a Fast Fourier Transform (FFT) algorithm that
typically requires O(Ng log Ng ) operations. The gravity field is then computed on
the Ng mesh points by differentiation of φ.
– Potential estimation on the particles: the last step is to interpolate the gravity
field back to the particles. According to Ref. [50] the interpolation scheme should
be the same used in the mass assignment procedure (first step), to ensure that
self-forces on the particles vanish.
The main advantage of the PM algorithms is that they require fewer operations, O(N) +
O(Ng log Ng ), to evaluate the forces acting on all particles. This means less memory and
higher speed per time step than the PP algorithm. However, for separations between
particles less then several grid spacings, the forces between particles become poorly
approximated and the method starts to give worse results then the PP algorithm.
• The P3 M and Adaptive P3 M P3 M stands for the concatenation of four words
Particle–Particle–Particle–Mesh. It’s an hybrid algorithm, first applied to cosmology
in Ref. [28], that makes use of the two previous (PP and PM) methods. At short
range distances the Particle–Particle method is employed, while the long range forces
are computed using the Particle–Mesh algorithm. The gravitational force, Fij , acting
on the particles is then given by a sum of two terms,
Fij = Fsr lr
ij + Fij (2.47)
where Fsr lr
ij and Fij represent the short and long range forces acting on the particles. The
short range term, evaluated using the PP method, vanishes after a few inter-particle
distances, while Flrij is computed using the PM algorithm and is set to zero at short
range distances.
 The Adaptive P3 M algorithm [19] is a mesh refinement of the P3 M method, which uses
sub-grids in the regions of higher density. The P3 M scales as O(N log N), where N is
the number of particles. This method is presently one of the best algorithms to compute
the gravitational force for large cosmological N-body simulations.

Gas dynamics

A description of structure formation exclusively based on the behaviour of matter under its
self–gravity is obviously insufficient. Gas dynamics, shocks and pressure gradients become
more and more important as structures evolve. Hydrodynamical effects are clearly important
to determine the behaviour of baryonic matter in galaxies and clusters of galaxies (even if
baryons are only a small fraction of the total mass), and play a central role in determining
how collapsing structures can reach virial equilibrium.
In comoving coordinates, gas dynamics is described by the following cosmological fluid equa-
tions (see e.g. Ref. [8]),
' (
∂ ρb 1
+ ∇ · vb = 0 (2.48)
∂t ρ̄b a
∂vb 1 1
+ (vb · ∇)vb + Hvb = − ∇p + g , (2.49)
∂t a aρb
where g = −∇φ/a is the peculiar acceleration field given by Eq. (2.45) and ρb , ρ̄b , vb and p
are, respectively, the baryonic mass density, mean mass density, peculiar velocity and pressure.
These equations must be used in conjunction with a third continuity equation for the energy
or for the entropy. Outside shocks these can be written as:
∂u 1 p 1
+ vb · ∇u = − ∇ · vb + (Q − Λ) (2.50)
∂t a aρb ρb
∂S 1 1
+ vb · ∇S = (Q − Λ) , (2.51)
∂t a p
where u and S are the gas thermal energy and entropy per unit of mass. For a perfect gas
u = p/[(γ − 1)ρb ] and S = (γ − 1)−1 ln(pρ−γ
b ) where γ is the ratio of the specific heats (γ = 5/3
for a monotonic perfect gas). The quantities Q and Λ are the heating and cooling rates of
the gas. They are important in non-adiabatic calculations and depend on several parameters
such as ionization, gas chemistry and radiative energy transfer (see Section 2.5.4).
The strong nonlinearity of these equations makes the test of numerical algorithms of gas
dynamics very hard. The accuracy and convergence of these methods are mainly tested by
comparing the results from different codes or testing their predictions against simple one–
dimensional problems.
Smoothed Particle Hydrodynamics (SPH) was invented by Gingold & Monaghan, Ref. [38], and
Lucy, Ref. [68], in 1977 (more recent reviews can be found in Refs. [71, 102]). It is a ‘particle-
tracking’ Lagrangian method for integrating the fluid equations (2.48)–(2.50) or (2.51), which
involves the estimation of various fields (such as baryon density, velocity, temperature, pressure
gradient, etc) from a discrete set of particles with a fixed mass The way SPH tackles this
problem is by smoothing the particles over a finite volume, which is set by the profile of a
spherically–symmetric smoothing kernel, W . For example, according to this method, the value
of a scalar quantity A(x) at the position x, is estimated by the formula
8 Ai
⟨A(r)⟩ = mi W (|r − ri | , hi ) , (2.52)
i
ρi

where mi , ρi , hi and Ai are, respectively, the mass, mass density, smoothing length and the
scalar quantity A of the particle labelled with the index i. The shape of the smoothing kernel,
W (|r − ri | , hi ), is generally taken to be a spline, such as the normalised spherically–symmetric
B2 –spline kernel (see Ref. [72]) below,
⎧
4 − 6x2 + 3x3 , 0 ≤ x ≤ 1
1 ⎨
W (x, hi ) = (2 − x)3 , 1<x≤2 (2.53)
4πh3i ⎩
0, x>2
−1/3 −1/3
where x = |r − ri | /hi . The smoothing length, hi , is usually taken to vary with ρi ∝ ρb , in
order to have a fixed amount of particles (typically 30–100) included in the sum of Eq. (2.52).
Particles in low density regions have therefore larger smoothing lengths then particles in high
density regions.
Other methods exist to solve the gas dynamical equations. Methods like the Eulerian Grid and
the adaptive Grid algorithms are among the ‘competitors’. These are grid-based hydrodynamic
methods which provide numerical solutions of the Eulerian field equations for fluid dynamics.
A discussion of these and some other gas dynamics algorithms is given in Refs. [8, 117].

Additional physics: Cooling and non-gravitational heating

Besides gravity and pressure forces there are several other physical processes which can have
considerable impact on the hydrodynamics of the gas. Processes such as radiative cooling,
non-gravitational heating and energy feedback from star formation, are among the possible
effects that can be included in the simulations to achieve a better description of the structure
formation phenomenon. These effects can be modelled by specifying the appropriate heating
and cooling rates in equations (2.50), and (2.51).
The lack of the Gunn–Peterson trough (see Ref. [39]) in the spectra of high-redshift quasars
indicates that the intergalactic medium (IGM) is already in a very high state of ionization by
redshifts as high as z ≃ 5 (see also Ref. [42, 67]). This means that a considerable amount of
energy is required to reionize the universe from the neutral state it acquires after recombina-
tion.
Possible IGM heating mechanisms include sources of astrophysical nature, such as energy
injection from stars, winds from supernovae explosions and energy release due to gravitational
accretion into Black Holes (e.g. quasars and active galactic nuclei (AGN)), or other more
exotic mechanisms such as energy release by late-decaying particles (e.g. decaying neutrinos).
All these processes can induce changes in the thermal state of the gas and therefore influence
its dynamics. For example, as shown in Ref. [26], photoionization heating can strongly prevent
gas from cooling into the shallow potential wells at high redshifts.
Non-gravitational heating effects are generally difficult to model due to their localized nature
and lack of information on the relevant energy release mechanisms. The physical scales on
which they occur are also usually below the resolution limit of large-scale structure simulations.
In such cases, heating of the IGM can only be implemented at a phenomenological level.
Radiative cooling is believed to play a central role for the condensation and survival of galax-
ies within larger virialized structures. Unlike heating mechanisms, cooling effects are well
understood at least in the case of optically-thin media with primordial gas composition, see
Refs. [10, 78, 93]. There are several ways in which gas can dissipate energy. These include
processes such as Bremsstrahlung emission from hot electrons, recombinative cooling due to
collisional ionization followed by electron recombination, and collisional excitation cooling re-
sulting from the emission of radiation by neutral atoms after excitation due to collisions with
free electrons. In all cases the emitted radiation escapes from the gas if the medium is optically
thin.
Since all the above process are collisional, i.e. result from two-body encounters, the net
cooling rate of the gas scales as Λ = ne np Λc , where ne and np are the number densities of
the electrons and atoms/ions. The computation of the net cooling function, Λc , requires a
detailed study of all possible chemical reactions among the different particle species present
in the gas. The results are usually tabulated as a function of the temperature and metallicity
of the gas. Fig. 2.1 shows the net cooling function obtained from the tables presented in
Ref. [105]. The calculation assumes collisional ionization equilibrium and an optically-thin
gas with primordial chemical composition. At low temperatures the curves are dominated
by collisional excitation cooling from neutral hydrogen and helium. As metallicity builds up
collisional excitation from heavier elements starts to dominate. All curves converge at high
temperature as the gas becomes fully ionised and Bremsstrahlung emission from hot electrons
is the dominant cooling process.
An effect which strongly affects the cooling properties of the gas is photoionization heating
due to ultraviolet (UV) radiation, such as that produced by quasars and early generations
of massive stars, see Ref. [26]. This can ionize low-density gas at temperatures as low as
104 Kelvin and therefore totally suppress collisional excitation cooling, which is the dominant
cooling mechanism at these temperatures. As a result, the peaks observed at low temperatures
in Fig. 2.1 can be strongly suppressed if a significant UV background is present, see Ref. [102].
An extra factor proportional to ne (T −TCMB ) can be added to Λ to take into account Compton
cooling of the ionised gas by cosmic microwave background photons, known as the Sunyaev–
Zel’dovich (SZ) effect. The SZ effect acts as a cooling mechanism in which thermal energy
is depleted from the gas and transfered into the CMB. In practice this process is inefficient
at low redshifts (z " 8). This is because below this redshift the characteristic time scale for
Compton cooling, tcc , is larger then the Hubble time, tH (tcc /tH ≃ 180/(1 + z)5/2 ). Above
z ≃ 8 the efficiency of the effect is also small because the SZ distortions become very weak at
high redshifts, see e.g. Ref. [10].
Figure 2.1: Dependence on metallicity of the gas cooling function for an optically thin-medium
assuming collisional ionization equilibrium. Lines are obtained from interpolation of the cool-
ing rates tabulated in Ref. [105].

In the absence of significant photoionization heating, the function in Fig. 2.1 provides a simple
and accurate description of the cooling mechanisms of the baryonic gas.

Numerical integration and timestepping

The integration of the equations of motion for gas and dark matter is done numerically,
assuming a given timestepping scheme. A popular timestepping scheme is the PEC (Predict,
Evaluate, Correct) algorithm described in Ref. [20]. This uses only positions and velocities
(plus densities and thermal energies for the gas) to evaluate the forces on the particles and
allows arbitrary changes in the integration time step to account for rapid variations of the forces
with time. Though variable during the time integration process, the timestep is spatially
constant throughout the simulation box. The timestep is variable (i.e. changes from one
integration step to another), but spatially constant throughout the simulation box. It is
determined by evaluating different timescales, such as the largest particle acceleration (∆ta ),
the largest particle speed (∆tv ), and the Hubble time (∆tH = 1/H(t)), at various stages of the
force calculation. The ∆tH = 1/H(t) scale ensures that adiabatic cooling due to the Hubble
expansion is followed accurately. The timescales, ∆ta and ∆tv , depend on the value of the
assumed force softening. At early times the effect of ∆tH dominates (see e.g. Refs. [20, 59]).
Simulation output

Simulation outputs (also known as simulation snapshots) are files containing a set of particles
with masses, positions, velocities, densities, temperatures and smoothing lengths. Other phys-
ical properties, such as gas metallicities and gravitational potential at the particles positions,
are also possible outputs in public simulation codes. The spatial distribution of the particles
gives a representation of the gas and dark matter density fields inside the simulation volume.
Figure 2.2 shows the zero redshift output of a ΛCDM simulation made with the public Hydra
code, Refs. [19, 20]. The green and red points represent the gas and dark matter particles,
respectively. The physical size of the box at z = 0 is 200 Mpc. Equal number of gas and dark
matter particles are used in these plots. The predominance of green points is because gas
particles were plotted after the dark matter particles. The lack of visible red points indicates
that gas follows the dark matter distribution well. The bottom panels show the aspect of the
box when viewed face on from its X-Y and Y-Z planes.
Cosmological simulations usually assume periodical boundary conditions. This implies that
structures near one side of the box are in fact close to particles on the opposite side of the
box. The boundary periodicity allows us to randomly shift the positions of all particles
(simultaneously) in a given direction, provided that particles leaving the simulation volume
are forced to re-enter the box from the opposite direction.
Figure 2.2: The top panel shows the gas (green) and dark matter (red) particles in a simulation
box, of volume 2003 Mpc3 , at redshift z = 0. The bottom panels result from viewing the box
face on from the X-Y and Y-Z planes, respectively. The gas particles were plotted after the
dark matter particles, which explains the predominance of the green points.
Chapter 3

Cosmic Microwave Background

Radiation

3.1 The intensity spectrum

The shape of the cosmic microwave background spectrum depends on the physical processes
which mediate the interaction between matter and radiation in the early universe. After the
epoch of particle–antiparticle annihilation (z ≃ 109 ) these processes are essentially Compton
scattering, double Compton scattering and free-free (or Bremsstrahlung) emission. The so
called double Compton scattering differs from the usual Compton effect because the photon–
electron interaction gives rise to an extra photon. The Bremsstrahlung interaction consists of
the emission of photons by accelerated electrons moving in the electrical potential created by
positively charged ions. Contrary to what happens with the Compton scattering, neither of
these processes preserves the total number of photons. In all cases the energy of the photons
is changed. The effectiveness of these processes in shaping the CMB spectrum depends on the
characteristic time-scale on which they occur compared to the Hubble time at a given epoch.
It’s possible to show that kinetic equilibrium between matter and radiation can be achieved
via Compton scattering alone, if the radiation spectrum is not too different from a blackbody
spectrum after the particle–antiparticle annihilation epoch (see e.g. Refs. [15,78]). This is true
at least until a critical redshift zc = 2.2 × 104(ΩB h2 )−1/2 , below which the electron and photon
densities become too low. During this period the kinetic equilibrium ensures a Bose–Einstein
spectrum for radiation field,
$−1
2hν 3
! ' (
hν
I(ν) = 2 exp + µ(ν) − 1 , (3.1)
c kB T
where µ(ν) is the chemical potential, T the temperature and h and kB are the Planck and
Boltzmann constants, respectively. The dependence of the chemical potential on frequency
can be modelled as µ = µ0 exp(−2x0 /x) where x = hν/kB T , x0 = 5 × 10−2 (Ω0 h2 )7/8 and
µ0 is the chemical potential at x ≫ x0 [118]. Its effect is stronger on the Rayleigh–Jeans
spectral region, which is where (3.1) shows larger deviations from a blackbody distribution –

47
Figure 3.1: CMBR spectral distortion caused by energy release in the early universe. The
observed CMBR will deviate from a blackbody spectrum (solid line) if energy is released in
the redshift range zth ≃ 2 × 106 ! z ! zc ≃ 3 × 105 . The radiation field will then follow a
Bose-Einstein spectrum (dashed line). In both cases the temperature of the radiation field
was taken to be T = 2.725 K. The chemical potential for the Bose–Einstein spectrum was
µ0 = 0.1 and Ω0 h2 = 0.17 (see text).

see Figure 3.1.

For thermal equilibrium to be achieved at some epoch, interaction processes that can “create”
and “destroy” photons must operate so that a blackbody spectrum can be produced. We know
for instance that Bremsstrahlung emission is an efficient process to generate photons at low
frequencies (note that the Bremsstrahlung cross section scales with the square of the photons’
wavelength). This helps the radiation spectrum to approach the blackbody form. Detailed
studies on how radiation can acquire a Planckian shape indicate that double Compton scatter-
ing is enough to generate a blackbody spectrum for redshifts zth ≥ 1.5 − 2.5 × 106 (ΩB h2 )−α ≃
2 × 106 , where α = 0.34 − 0.4, see Ref. [15]. This is also possible with Bremsstrahlung alone
if ΩB h2 > 0.1.
Therefore, the CMBR we observe today will have a thermal spectrum, unless energy is
added to the radiation field after the redshift zth (energy released before zth would simply
be rethermalized into a higher temperature blackbody spectrum). Energy added in the in-
terval, zth ≃ 2 × 106 ! z ! zc ≃ 3 × 105 , would result in a Bose–Einstein CMB spectrum,
as kinetic but not thermal equilibrium can occur during this period. At epochs later then zc
it’s also impossible to establish kinetic equilibrium if more energy is added to the radiation
field. Possible mechanisms for this energy release include: particle decay into a much lighter
particle plus a photon; line emission at recombination; Far infrared emission from dust; and
inverse Compton scattering by hot plasma, known as the SZ effect. The spectral distortion
produced by the later mechanism is the subject of this thesis.
The intensity spectrum of the CMB radiation is now known with high accuracy over a wide
range of frequencies. It presents no evident departure from a blackbody spectrum, although
somewhat larger deviations are consistent with the data at wavelengths larger then ∼ 1 cen-
timetre. The experiment FIRAS (Far Infrared Absolute Spectrophotometer) on board of the
COBE satellite measured the difference between the CMB spectral intensity and a precise
blackbody spectrum. Spectral deviations from a Planck law were found to be very small and
only consistent with an upper limit of |µ| < 9 × 10−5 (95 % confidence level [CL]) for the
chemical potential of a Bose-Einstein distribution and an upper limit of |y| < 1.5 × 10−5 (95
% CL) for the Kompaneets y–parameter describing Comptonized spectra via the SZ effect,
see Ref. [33]. The CMB blackbody temperature from FIRAS was computed in Ref. [69] to
be T = 2.725 ± 0.002 K. This implies a CMB thermal spectrum characterised by a specific
intensity maximum of Imax ∼ 3.7 × 10−18 J m−2 s−1 Hz−1 rad−2 at the frequency νmax ∼ 160
GHz and a photon number and energy densities given by nγ ∼ 4 × 108 photons m−3 and
ργ ∼ 5 × 10−31 kg m−3 , respectively.

3.2 The angular power spectrum

The natural way of expanding the CMBR temperature fluctuation field, Θ(n̂) = ∆T /T (n̂) =
∆T /T (θ, φ), on the celestial sphere is in terms of spherical harmonics:
∞ 8
8 ℓ
Θ(n̂) = ∆T /T (θ, φ) = aℓm Yℓm (θ, φ). (3.2)
ℓ=0 m=−ℓ

where n̂ denotes a direction in the sky specified by the spherical angles θ, φ and aℓm are the
multi-polar moments, given by

∆T ′ ′
.
∗
aℓm = Yℓm (θ′ , φ′) (θ , φ )dΩ′ , (3.3)
T

where dΩ′ is the solid angle element. In general, the aℓm are complex quantities. Assuming
Θ(n̂) as a random Gaussian field (as predicted by inflation) the temperature fluctuations are
fully specified by their angular correlation function:

∆T ∗ ∆T ′
; < ; <
′ ∆T ∆T ′
C(n̂, n̂ ) ≡ (n̂) (n̂ ) = (n̂) (n̂ ) =
T T T T
=> ?> ?@
88 88
= a∗lm Ylm∗
(n̂) al′ m′ Yl′m′ (n̂′ ) =
l m l′ m′
88
= ⟨a∗lm al′ m′ ⟩ Ylm
∗
(n̂)Yl′ m′ (n̂′ ) (3.4)
l l′ m m′
where the angle brackets denote averages over an ensemble of universes with statistically
equivalent perturbations and n̂ and n̂′ are two unit vectors pointing to arbitrary directions in
the sky. If we assume rotational symmetry (isotropy) it is conventional to write

⟨a∗ℓm aℓ′ m′ ⟩ = Cℓ δℓ ℓ′ δm m′ , Cl ≡ |aℓm |2 ,

6 7
(3.5)

(i.e. the aℓm are not correlated) where Cℓ is the angular power spectrum of the temperature
fluctuation field. Inserting Eq. (3.5) into Eq. (3.4) we find
88
C(n̂, n̂′ ) = ∗
Cl δl l′ δm m′ Ylm (n̂)Yl′ m′ (n̂′ ) =
l l′ m m′
8 8
∗
= Cl Ylm (n̂)Ylm (n̂′ ) =
l m
8 (2l + 1)
= Cl Pl (cos ϑ) = C(cos ϑ) (3.6)
l
4π

with cos ϑ = n̂.n̂′ . Using (3.3) and (3.5) one can express the angular power spectrum as,
∆T ∗
; <
∆T ′ ′
.
27 ′ ∗
(θ′ , φ′ ).
6
Cl = |alm | = dΩdΩ Ylm (θ, φ) (θ, φ) (θ , φ ) Ylm (3.7)
T T
It is through Cℓ that theoretical models and the observations can be confronted.
Random fields defined on compact manifolds, such as the sphere S 2 , are not ergodic. This is
because compact manifolds have finite volumes and the ergodic hypothesis only holds accu-
rately when the spatial averages are done over an infinite volume of space. Therefore, even
if we could reduce experimental errors to zero, there’s always a limit to the precision we can
achieve by converting all the above ensemble averages into spatial averages. This effect is
known as cosmic variance and it can severely reduce the accuracy with which the angular
power spectrum can be measured on large scales. The uncertainty on each multipole is given
by ∆Cℓ = Cℓ /(ℓ + 1/2)1/2 , [14]. For each ℓ there are 2ℓ + 1 independent aℓm , to be drawn from
a distribution of variance Cℓ . For larger ℓ’s this corresponds to a larger number of available
samples of the distribution, which implies a lower cosmic variance and better knowledge of
the angular power spectrum Cℓ .
Assuming a primordial power spectrum of the form of Eq. (2.31) and accounting only for scalar
(density) perturbations, it is possible to show (see Ref. [11]) that the angular power spectrum
is only dependent on the quadrupole normalization amplitude, ⟨Q⟩, and the spectral index, n,
for small values of ℓ (ℓ ≤ 20)
n−1
" # " 9−n #
2 4π Γ ℓ + 2 Γ 2
Cℓ = ⟨Q⟩ " 5−n
# " 3+n #. (3.8)
5 Γ ℓ+ 2 Γ 2

Equating n = 1 (Harrison–Zel’dovich power spectrum) and using the fact that Γ(x+1) = xΓ(x)
one easily obtains
! $1/2 2
ℓ(ℓ + 1)Cℓ 12
= ⟨Q⟩ = const. (3.9)
2π 5
Figure 3.2: CMBR angular power spectra for three CDM models investigated in this thesis:
a ΛCDM model with Ω0 = 0.35, ΩΛ = 0.65 (blue solid line); a critical density model τ CDM
with Ω0 = 1 (red dashed line); and an open model OCDM with Ω0 = 0.35, ΩΛ = 0 (green
dotted line). In each case the shape parameter of the matter power spectrum was chosen to be
Γ = 0.21, which provides a good fit to the large scale structure observations. For the critical
density model, this implies a first Doppler peak located at much higher ℓ then that found in
the (flat) ΛCDM case (see e.g. Ref. [113]), being the later in much better agreement with
present CMB observations. All curves were generated with the cmbfast code [1, 98]).

This result explains the Sachs–Wolfe [95] plateau in Figure 3.2. The quantity ⟨Q⟩ is the
best fit to the angular power spectrum normalization, given the observed data, and should
not be confused with the measured value of the quadrupole amplitude, Qrms . Assuming a
scale-invariant primordial spectrum and using the COBE data set, the best fit of the power
spectrum normalization is ⟨Q⟩ = 1.8 ± 1.6µK, [7].
General features of the CMBR angular power spectra in Figure 3.2 are: (1) a Sachs-Wolfe
plateau, for ℓ " 30, resulting from perturbations larger then the size of the Hubble horizon at
the present; (2) a series of peaks, named acoustic, with a dominant peak at ℓp ≈ 220 followed
by secondary peaks and valleys; (3) and “cut off” from ℓ ! 1000 due to the damping of
perturbations on small scales (Silk damping and finite LSS thickness effects, which cause a
mixture of photons coming from different points inside the LSS). The peaks observed in the
intermediate multipole region, (angular scales inferior to 1o ) reflect the acoustic oscillations of
the baryon-photon fluid at the recombination epoch. (see e.g. Ref. [51]).
 Primárias Gravidade
Doppler
Flutuações de densidade
Diffusion Damping
Defeitos Topológicos
Secundárias Gravidade Early ISW
Late ISW
Rees-Sciama
Lensing
Reionização Local SZ cinético
SZ térmico
Reionização Global Supressão
Novo doppler
Vishniac
“Terciárias” Extragaláctios Fontes Rádio pontuais
(foregrounds) Fontes IR pontuais
Galácticos Poeira
Free-free
Synchroton
Locais Sistema Solar
Atmosfera
Ruı́do, etc.

Table 3.1: Processos que geram anisotropias na CMBR. Adaptado de [107].

3.3 Fontes geradores de anisotropias CMB: “A primer”

Nesta secção faz-se um resumo dos processos que podem gerar anisotropias na radiação cósmica
de fundo (para reviews ver Tegmark 1996 [107], Hu et al. 1997 [53], Barreiro 1999 [5]).
Estes processos são normalmente catalogados de acordo com a sua origem. Designam-se por
anisotropias primárias as flutuações de temperatura porvocadas pelos processos que ocorreram
até ou durante o perı́odo de desacoplamento matéria-radiação. Anisotropias secundárias são
as anisotropias geradas posteriormente, por fenómenos que alteram o espectro da radiação de
fundo à medida que esta se propaga desde a última superfı́cie de scattering (LSS) até nós. Para
além destes efeitos existe todo um conjunto de outros efeitos, tais como emisso galatica e fontes
pontuais extragalacticas que se adicionam e contaminam o espectro de anisotropia do CMB.
Estas fontes de contaminação so normalmente denominadas de anisotropias “terciárias” do
CMB. Na tabela 3.1 apresenta-se um resumo dos processos que podem originar ou contaminar
o espectro de anisotropias da radiação de fundo.
3.3.1 Anisotropias Primárias
Gravidade, Doppler e flutuações de densidade

Os fotões CMB que hoje observamos interagiram pela ultima vez com a matéria para um red-
shift z ≃ 103 , quando o universo tinha cerca de 300 mil anos de idade e quando se encontravam
a uma distância de ∼ 6000h−1 Mpc. Consigo transportam diferentes “impressões digitais” da
região aonde sofreram o último scattering. Essas impressões podem ser modeladas pelo po-
tencial local (φ); a velocidade radial peculiar da matéria (vr ); e a flutuação de densidade do
fluı́do (δ). Estes três campos influênciam o comportamento da radiação da seguinte forma:

1. Fotões que sofreram o último scattering num poço de potencial (φ < 0) experimentam
um redshift gravitacional à medida que sobem esse poço.

2. Fotões dispersos pela matéria cuja velocidade peculiar está a afastar-se de nós (vr > 0),
sofrem redshift.

3. Fotões que saem duma região de sobre-densidade (δ > 0) possuem temperaturas mais
elevadas, porque regiões mais densas são intrinsecamente mais quentes.

Estes três efeitos correspondem às três primeiras entradas da tabela 3.1 e podem ser equa-
cionados do seguinte modo:
∆T 1
(r̂) = φ(⃗r ) − r̂.⃗v (⃗r) + δ(⃗r ) (3.10)
T 3
onde ⃗r é a distância própria à superfı́cie de último scattering, r̂ é o versor na direccção ⃗r e os
campos φ, ⃗v e δ devem ser calculados no instante de recombinação (z ∼ 103 ; r ∼ 6000h−1 Mpc).
A forma do espectro de potência angular, Fig. 3.2, pode ser intrepretado à luz desta equação:

• Efeito Sachs-Wolfe: Para escalas angulares superiores à escala que subentende o hor-
1/2
izonte na altura da recombinação, ϑ ≫ 1.7o Ω0 , as flutuações dos campos φ, ⃗v e δ
estão “congeladas”. Como, nos molelos adiabáticos, as condições iniciais fazem coincidir
os locais de sobre-densidade com os poços de potencial, o terceiro termo em (3.10) é
aproximadamente dado por δ ≈ −2φ e cancela parcialmente com o primeiro termo do
segundo membro, resultando:
∆T 1
(r̂) = φ(⃗r ) − r̂.⃗v (⃗r). (3.11)
T 3
Este efeito é denominado efeito Sachs-Wolfe e é responsável pela parte plana do espectro
de potência para os baixos valores de l (l ≪ 100. Ver gráfico 3.2).

• Oscilações acústicas: Para escalas angulares inferiores à escala que do horizonte do

1/2
som na última superfı́cie de scattering, ϑ ≪ 1.7o Ω0 (ver Eq. (1.37)), as flutuações em
φ, ⃗v e δ tiveram tempo para entrar em oscilações “acústicas” antes de se iniciar a re-
combinação. Isto dá origem aos picos “doppler” que se observam na Fig. 3.2. A posição
 destes picos é essencialmente determinada pela geometria do universo (Kamionkowski et
al. 1994 [58], Frampton et al. 1998 [34]), uma vez que uma mesma escala fı́sica suben-
tende diferentes escalas angulares para universos com diferentes curvaturas. De acordo
com (1.37) o ângulo subentendido por uma dada escala diminui com Ω0 . Deste modo os
picos doppler do gráfico 3.2 deslocam-se no sentido das pequenas escalas angulares (l’s
elevados) à medida que Ω0 diminui.

Diffusion Damping
1/2
Para pequenas escalas, ϑ < 0.1o Ω0 (l > 103 ), as flutuações de temperatura começam a
esbater-se. Isto acontece porque o processo de desacoplamento da radiação com a matéria
não é instantâneo e a última superfı́cie de scattering não é infinitamente fina. Com efeito,
como o perı́odo de decoupling decorre durante um intervalo de redshifts ∆z ≃ 100 ( [57, 62]),
a temperatura CMB que se observa numa dada direcção do céu é na realidade uma “média
ponderada”, correspondendo a uma mistura de fotões vindos das partes mais próximas e mais
afastadas da LSS. Este efeito é designado de efeito de “damping” e apaga/destroi quaisquer
flutuações em escalas inferiores à espessura da LSS.
Para além do ‘diffusion damping’ existem outros mecanismos de difusão que destroiem e
apagam a assinatura das anisotropias na região das pequenas escalas. Um destes mecanismos
é o chamado Silk damping (Silk 1968 [100]) e consiste numa diminuição das flutuações da
radiação e da matéria bariónica na LSS, por via da difusão de fotões que arrastam consigo
bariões das regiões de sobredensidade. Como a radiação não interage com matéria escura
não bariónica, este efeito não afecta as flutuações de densidade dos constituintes deste tipo
de matéria. No entanto existe um outro mecanismo de difusão (free–streeming) das regiões
de sobredensidade para as regiões menos densas, que pode diminuir as flutuações de densi-
dade de partı́culas não interactivas (como é o caso dos constituintes da matéria escura). De
acordo com Bond & Szalay 1983 [?], este mecanismo depende essencialmente da velocidade
e da massa das partı́culas em causa. Partı́culas “frias”, cuja velocidade é baixa (por exem-
plo matéria escura fria - CDM), praticamente não são afectadas. Partı́culas susceptı́veis de
sofrerem free–streeming (como por exemplo matéria escura quente - HDM) podem apresentar
fortes diminuições nas flutuações de densidade a partir de uma certa escala fı́sica. A escala
fı́sica a partir da qual este efeito se torna mais significativo, depende da massa das partı́culas.
Por exemplo para neutrinos com massa mν ≃ 10 eV, as flutuações de densidade correspon-
dentes podem ser fortemente suprmidas para escalas inferiores ou da ordem do tamanho actual
dos super enxames de galáxias [?] (alguns Mpc h−1 ).

Defeitos Topológicos

Apesar de actualmente existirem modelos hı́bridos de defeitos topológicos e inflação (Jean-

nerot 1996 [?], Linde et al 1997 [?], Avelino et al 1998 [?]), tradicionalmente os modelos
topológicos são considerados como uma teoria alternativa ao paradigma inflacionário para
gerar perturbações de densidade no universo primordial. Cordas cósmicas (strings), texturas
e monopólos constam entre os exemplos de defeitos topológicos que, de acordo com a teoria,
 Figure 3.3: Efeito da variação de φ na energia dos fotões CMBR

podem ter sido formados por mecanismos de transição de fase associados a quebras de simetria
do universo primordial. Contrariamente ao que acontece nos modelos inflacionários standard,
as anisotropias CMB geradas por defeitos topológicos apresentam um caracter não gausseano1 .
Esta importante caracterı́stica pode ser utilizada como meio de descriminar entre estes dois
tipos de cenário. Até ao momento as observações da radiação de fundo não dão indicações de
fugas significativas à gausseanidade.

3.3.2 Anisotropias Secundárias

Classificam-se de anisotropias secundárias todas as flutuações de temperatura originadas por
processos que actuam sobre os fotões CMB após o último scattering. Enquanto a radiação se
propaga livre pelo espaço apenas efeitos gravı́ticos e de rescattering (dispersão) podem gerar
novas anisotropias na radiação de fundo. O estudo destes efeitos é normalmente utilizado
como uma fonte de informação preciosa sobre a evolução do universo após o decoupling do
CMB.

Efeitos Gravı́ticos

Em teoria, quaisquer perturbações na métrica após o último scattering induzem anisotropias

secundárias na radiação de fundo. A acção da gravidade sobre os fotões CMB pode ocorrer de
várias formas, sendo algumas delas manifestações de um mesmo efeito: O efeito Sachs Wolfe
Integrado (ISW). Este efeito dá conta das flutuações de temperatura originadas pela variação
do potencial gravı́tico ao longo das geodésicas dos fotões. Admitindo apenas perturbações
1
No seio dos modelos inflacionários é igualmente possı́vel originar anisotropias de natureza não gausseana
(Salopek 1992 [?], Peebles 1999 [?, ?])
escalares na métrica, a sua magnitude é dada por
∆T
.
= φ̇ [(r̂(η), η] dη (3.12)
T
onde η é o tempo conforme, φ̇ = ∂φ/∂η e o integral é feito ao longo da linha de voo do fotões.
Se um fotão atravessa um poço de potencial sempre com φ̇ = 0, o blueshift que adquire na
queda é compensado, com o redshift que sofre quando “sobe” o poço de potencial (situação
(a) da figura 3.3). Se o potencial ficar mais ou menos profundo enquanto o fotão se encontra
no poço, o blueshift da descida e o redshift da subida já não se cancelam totalmente. Se
φ̇ > 0 tem-se um blueshift residual (situação (b) da figura 3.3). Se φ̇ < 0 tem-se um redshift
residual (situação (c) da figura 3.3). Em regime linear, a partir da altura em que o universo
fica dominado pela matéria, o potencial gravı́tico peculiar φ permanece constante e portanto
não existe qualquer efeito Sachs-Wolfe integrado.
Para além do efeito ISW, a gravidade pode actuar sobre a radiação de fundo como uma lente
gravitacional. Neste caso a energia dos fotões permanece constante (em primeira ordem de
aproximação), mas as suas trajectorias variam. Em resumo, os principais efeitos gravı́ticos
geradores de anisotropias secundárias são (os três primeiros items da lista tratam-se diferentes
manifestações do efeito ISW):

• Efeito Sachs-Wolfe integrado inicial (Early ISW)

Para modelos tı́picos, a época de igualdade matéria radiação ocorre um pouco antes
da recombinação. Sendo assim, logo após o último scattering, a densidade de radiação
contribui ainda de forma não negligı́vel para a densidade total do universo. À medida
que o universo se expande e se torna dominado pela matéria isto provoca um decaimento
no potencial φ, logo após o último scattering, a que se dá o nome de efeito ISW inicial
(early). Tendo em conta que a densidade dos fotões está fixa pela temperatura actual
do CMB, este efeito torna-se tanto mais importante quanto menor for a densidade total
do universo, Ω.
• Efeito Sachs-Wolfe integrado tardio (Late ISW)
Se Λ > 0 o universo eventualmente acabará por ser dominado pelo vácuo. Se k ̸ = 0
(Ω + ΩΛ ̸ = 1) o universo poderá passar por uma fase em que é dominado pela curvatura
(ver secção 1.2.1). Em ambas as situações o universo experimenta mudanças bruscas na
taxa de expansão o que acarreta variações em φ para baixos valores de z (z < Ω−1 ).
• Efeito Rees-Sciama
Em regime linear de perturbações, após o domı́nio da matéria, φ̇ = 0. No entanto
à medida que se formam estruturas não lineares (enxames e galáxias) o regime linear
deixa de ser válido e φ̇ deixa de ser nulo. Este é um efeito igualmente “tardio” uma vez
que a formação de estruturas não lineares, por colapso gravitacional, inicia-se somente
para baixos redshifts. Apesar de ser ainda um assunto em aberto, tudo indica que a
intensidade relativa do efeito Rees-Sciama (Rees & Sciama 1968 [?], Sanz et al 1996, [?])
é pequena face à amplitude das anisotropias primárias. Nos modelos CDM standart, por
exemplo, o efeito é em geral negligı́vel, à excepção de escalas angulares muito pequenas.
 • Lensing Gravitacional
É um efeito complementar ao efeito ISW, também relacionado com a variação de φ ao
longo das geodésicas dos fotões. No efeito ISW os fotões perdem ou ganham energia por
troca de pequenos impulsos de momento com o gradiente do campo gravı́tico paralelo à
linha de voo, ∇φ⊥ . No entanto a componente perpendicular a essa mesma linha, ∇φ|| ,
provoca igualmente alterações no momento dos fotões, que em primeira ordem não altera
a energia da radiação mas provoca deflecções na sua trajectória. Se θ for a distância
angular entre dois fotões sem deflecções e θ + ∆θ o ângulo real com deflecções, é possı́vel
mostrar que ∆θ/θ ≪ 1 (Seljak 1996 [?]). Para modelos tı́picos ∆θ/θ ≃ 0.1 − 0.2.
Isto significa que este é essencialmente um efeito de lensing fraco. Se imaginarmos as
flutuações CMB (na ausência deste efeito) pintadas sobre a superfı́cie de uma esfera, a
acção do lensing constituiria um esticar e deformar da esfera, de forma aleatória, tal
como acontece com a imagem distorcida dos espelhos dos parques de diversão. O efeito
resultante traduz-se por um deformar da imagem da LSS. No entanto como ∆θ/θ << 1,
estas distorções constituem sempre mapas injectivos (one-to-one) e por isso em nenhuma
região a imagem se sobrepõe sobre si própria. Na figura 3.4 (gráfico da esquerda) pode
observar-se uma estimativa da acção do lensing gravı́tico sobre o espectro de potência
angular. As linhas a tracejado e a cheio representam curvas com e sem lensing do CMB.O
resultado final deste efeito é pequeno (tipicamente introduz variações inferiores a poucos
pontos percentuais nas regiões das pequenas escalas angulares) e consiste essencialmente
em redistribuir potência dos picos para os vales. Algumas das futuras experiências CMB
poderão vir a detectar este efeito.

• Ondas gravitacionais
Ondas gravitacionais podem gerar anisotropias secundárias através do efeito ISW. A sua
acção sobre o espectro da radiação, a observar-se, afecta as escalas angulares superiores
à escala do horizonte à época da recombinação (Crittenden et al. 1994, [?]).

Reonização e rescattering

Se o universo se reoniza após o último scattering, electrões livres podem voltar a dispersar os
fotões da radiação cósmica de fundo (resacttering). A existência de rescattering faz apagar as
anisotropias primárias e cria novas anisotropias. Isto pode acontecer localmente, por exemplo
no interior de enxames de galáxias, ou de uma forma global por todo o universo.

• Reionização Global
As anisotropias primárias podem ser fortemente suprimidas se o universo se reonizou de
uma forma global para um redshift suficientemente elevado (zr ∼ 100). Na região das
pequenas escalas angulares essa supressão pode ser total. Existindo reionização global,
os fotões provenientes de uma dada região do céu não tem necessariamente que provir
dessa mesma direcção, desde a última superficie de scattering até nós. Na figura 3.5
vemos que um fotão que sofra rescattering, por exemplo, para zr = 10 pode ter tido
Figure 3.4: Efeitos do lensing gravitacional (à esquerda), reionização global e efeito Vishniac
(à direita) sobre o espectro CMBR. Imagens reimpressas de [?] e [52], respectivamente.

origem em qualquer ponto da região a escuro na LSS. Sendo assim num universo aonde
ocorra reonização global, a temperatura medida numa dada direcção do céu corresponde
a uma média ponderada de temperaturas de uma fração da LSS. Na figura 3.4 (gráfico
da direita) pode observar-se os efeitos da reonização global no espectro de potência
angular.As diferentes curvas correpondem a diferentes redshifts zr a partir do quais o
universo subitamente se reionizou e assim permaneceu até à actualidade. As principais
conclusões a retirar são: (i) Na região das grandes escalas angulares (pequenos l) o
espectro praticamente não se altera. (ii) Na região das9pequenas escalas (grandes l) o
espectro é suprimido por um factor e−2τ , onde τ = σT ne dη é a profundidade óptica
do universo (Tegmark & Silk 1995 [?]):
' (%
−1/2 hΩB zr &3/2
τ≈ Ω0 (3.13)
0.06 92

(ne é a densidade de electrões e σT é a secção eficaz de Thomson).

• Efeito Vishniac
É um efeito associado a uma reonização global do universo, que cria novas flutuações
de temperatura na região das pequenas escalas angulares (Ostriker & Vishniac 1986 [?],
Vishniac 1987 [?]). A sua origem está relacionada com o acoplamento das flutuações
de velocidade e de densidade dos electrões livres. É na realidade um efeito cinemático
da mesma natureza que o efeito SZ cinético (ver abaixo) relacionado com a evolução
linear das perturbações de densidade. Conforme se pode observar no gráfico da direita
da figura 3.4 a intensidade deste efeito é pequena.

• Reionização Local. Efeito Sunyaev Zel’dovich (SZ).

Figure 3.5: Cone de luz do passado num universo plano com Ω = 1. Um fotão que sofra
rescattering para z = 10 pode ter tido origem em qualquer ponto dentro da região a escuro na
LSS. Neste diagrama o eixo vertical é medido em tempo conforme η e as dimensões espaciais
em coordenadas comoveis. Imagem reimpressa de [107].

Figure 3.6: À esquerda: Distorção espectral provocada pelo efeito SZ térmico, com y = 0.1
(valor muito exagerado). O campo de radiação ganha energia ao ser disperso, através do efeito
compton inverso, por gas a elevada temperatura no interior de enxames de galáxias. Como o
número total fotões permanece constante, fotões na região de Rayleigh-Jeans ganham energia
e deslocam-se para a região espectral de Wien. À direita: Distorção espectral provocada pelo
efeito SZ cinético, com τ = 0.9 e vr = ∓ 0.1c. Estas quantidades estão igualmente exageradas
para se poder observar melhor a acção deste efeito sobre o espectro térmico CMBR.Em ambos
os gráficos as linhas a tracejado e a cheio representam, respectivamente, o espectro CMB com
e sem distorções SZ, [?].
Figure 3.7: Funções de distorção espectral g(x) (linha a cheio) e h(x) (linha a tracejado) dos
efeitos SZ térmico e cinético. A função de distorção do efeito térmico anula-se para x = 3.83,
o que, para Tcmb = 2.728, corresponde à frequência ν = 217 GHz. Deste modo a distorção SZ
total no canal 217 GHz é apenas devida ao efeito cinético. Esta é uma propriedade importante
do efeito SZ, que poderá permitir separar as anisotropias geradas pelo efeito cinético do efeito
térmico, [?].

Ao contrário do que se passa com a reionização global, é um dado adquirido que existem
fenómenos de reionização local no universo. O colapso gravitacional de gas para os poços
de potêncial faz elevar a temperatura das partı́culas ao ponto de estas se reionizarem.
Nos enxames de galáxias, por exemplo, a temperatura do gas pode atingir as várias
dezenas de milhões de graus Kelvin, temperatura que é manifestamente suficiente para
manter o hidrogénio e o hélio ionozados. A acção da reonização local sobre o espectro da
radiação de fundo manifesta-se de uma forma caracterı́stica, através do efeito Sunyaev
Zel’dovich (Sunyaev & Zel’dovich 1970,1972 [?, ?]). Este efeito, que consite na variação
de energia da radiação de fundo quando esta sofre dispersão de compton pelos electrões
livres existentes no interior das massas de gas, actua localmente ao nı́vel das pequenas
escalas angulares e pode ser separado em duas componentes:
Efeito SZ - Térmico: É o efeito SZ dominante. É devido ao movimento térmico dos
electrões, cuja destribuição de velocidades é aproximadamente isotrópica em torno do
centro de massa. Como os electrões estão a uma temperatura mais elevada do que os
fotões CMB, a radiação ganha energia (por efeito de Compton inverso), sem que haja
variação do número de fotões. Conforme se pode observar no gráfico da esquerda da
figura 3.6 isto dá origem a uma sobre-população do espectro térmico da radiação na
região de Wien (altas frequências) e um decréscimo na região espectral de Rayleigh-
Jeans. A distorção espectral do efeito SZ térmico é dada por:

kB σT
.
∆Ith = I0 g(x) Te ne dl (3.14)
me c2
 onde I0 = 2(kB Tcmb )3 /(hc)2 é uma constante, g(x) é uma função da frequência adimen-
sional x = hν/kB Tcmb (ver figura 3.7) e o integral (feito ao longo da linha de visão) é
o chamado parametro de comptonização y. Conforme se pode ver a partir de (3.14), o
efeito SZ térmico é tanto mais intenso quanto maior for a densidade e a temperatura dos
electrões na direcção da linha de observação. A sua dependência espectral caracterı́stica
(ver figura 3.7) permite-nos separa-lo das restantes anisotropias CMB. Experiências, que
observam em múltiplos canais de frequência, como é o caso do satélite Planck, permitem
já observar um grande número (mais de um milhar) de fontes SZ térmico (essencialmente
enxames de galáxias). Estas observações são de grande importância, porque permitem,
em conjunção com observações em raios X e no óptico, impor constrangimentos adicionais
em parametros cosmológicos importantes como a normalização do espectro de potência
da matéria e, h e Ω0 . Por outro lado permite igualmente estudar as caracterı́sticas da
distribuição de matéria no interior dos enxames de galáxias.
Efeito SZ - Cinético: É devido à velocidade do centro de massa do gas relativamente
ao referêncial do observador. Se a massa de gas se está a aproximar/afastar do obser-
vador, o scattering dos fotões CMB provoca um efeito de doppler, que se traduz num
blueshift/redshift dos fotões observados na direcção da massa de gas. A variação de
intensidade espectral correspondente é dada por:
vr
∆Ik = −I0 h(x) τ (3.15)
c
e pode ser observada no diagrama da direita da figura 3.6. Nesta expressão h(x) é a
função de distorção espectral do efeito cinético (ver figura 3.7), τ é a profundidade óptica
e vr é a velocidade do centro de massa do gas (positiva em caso de afastamento, negativa
em caso de aproximação) na diracção de observação. O efeito SZ cinético é tipicamente
mais fraco que o efeito térmico (para valores tı́picos de y, vr e τ a razão de intensidade
entre os efeitos é geralmente inferior a 0.1). Este facto, associado a uma dependência
espectral idêntica à dependência das anisotropias primárias, faz com que o efeito SZ
cinético seja de difı́cil observação.

• Reionização não Homogénia

Fenómenos de reionização não homogénia, provocados por estrelas e quasars formados a
elevados redshifts, podem gerar (via efeito doppler) anisótropias secundárias na radiação
de fundo.

3.3.3 Anisotropias Terciárias

Não se tratam de anisotropias CMBR no sentido convencional. São antes foregrounds e ruı́dos,
que se sobrepoem, na mesma região espectral, às anisotropias da radiação cósmica de fundo
- ver figura 3.9. Para se poder ter acesso à informação contida nas anisotropias primárias e
secundárias do CMB, torna-se imprescindı́vel estimar e remover, da melhor forma possı́vel,
todos os tipos de contaminação espectral, que possam estar presentes no sinal CMB que
se observa. Qualquer tentativa credı́vel para determinação de parâmetros cosmológicos ou
informação acerca do universo permordial, que use a radiação de fundo, exige trabalho aturado
e um estudo profundo de todas estas contaminações. É convencional catalogar os vários
foregrounds quanto à proximidade da sua origem. Estes podem ter origem,

• Extra-galáctica: Essencialmente podem-se destinguir dois tipos de contaminações com

origem extra-galática. A baixas frequências, ν < 300 GHz, as contaminações são domi-
nadas por populações de fontes radio emissoras (como Nucleos Galácticos Activos com-
pactos, blazers, radio loud QSOs). A frequências elevadas, ν > 300 GHz, dominam fontes
emissoras no infra-vermelho longı́nquo (exemplo, galáxias espirais inactivas). Pelo facto
de se encontrarem a grandes distancias, estas fontes surgem nos mapas como fontes pon-
tuais (point sources). O seu impacto sobre o fundo de mico-ondas tem maior expressão
nas experiências de elevada resolução anglar, como é o caso da missão Planck.

• Galáctica: As princı́pais fontes de contaminação do CMB têm origem na nossa galáxia.

São elas poeira, bremsstrahlung (free-free) e emissão de sincrotrão. No primeiro caso,
os grãos de poeira interstelar são aquecidos por radiação visı́vel e re-emitem energia no
infra-vermelho longı́nquo. Isto afecta essencialmente frequências acima de ν > 90 GHz
(grãos de poeira com rotação podem igualmente contaminar frequências entre os ∼ 10
GHz e os ∼ 100 GHz, ver [5] e referências aı́ incluidas). A emissão bremsstrahlung
tem origem quando electrões livres, a elevada temperatura (T > 104 ), interagem com
o potencial dos iões existentes nas massas de gas interestelar, emitindo radiação. A
faixa de frequências afectada é ν ∼ 25 − 75 GHz. Finalmente, a emissão de sincrotrão
ocorre sempre que electrões relativistas são acelerados em campos magnéticos e domina
as restantes contaminações galácticas na região das baixas frequências, ν < 20 GHZ.

• Contaminações locais: O Sol, a Lua a Terra e outros corpos planetários, que emi-
tam ou reflectem na região das micro-ondas, devem de ser tidos em conta por qualquer
dispositivo experimental que pretenda medir flutuações de temperatura da ordem dos
10−5 K. A esta escala de temperaturas a própria emisão térmica do aparato experi-
mental introduz ruı́do “electrónico” nas medições. Experiências baseadas em terra ou
em balões, sofrem igualmente das consequências da emissão atmosférica. Estes são al-
guns dos muitos problemas que podem afectar as observações CMB e são naturalmente
tratados caso-a-caso, uma vez que estão directamente dependentes das caracterı́sticas
do aparato experimental.

Na figura 3.8 faz-se uma sı́ntese da sensibilidade angular e de frequência de algumas ex-
periências CMB. As zonas a tracejado representam as regiões do plano ν−l onde os foregrounds
dominam o o fundo de radiação de micro-ondas.
As anisotropias secundárias devidas a fenómenos de reionização são normalmente incluidas
nesta lista de foregrounds. O facto de terem origem extra–galáctica faz com que elas próprias
contenham geralmente informação de caracter cosmológico e por isso mesmo sejam tratadas em
separado. O efeito Sunyaev Zel’dovich, por exemplo, (responsável pelas últimas duas camadas
de foregrounds da figura 3.9) apresenta um espectro de potência angular muito semelhante
a ruı́do gausseano (i.e. Cl constante com l), que importa “isolar” para se poder estudar
Figure 3.8: As formas poligonais representam as regiões de frequências e multipolos analizadas
por diferentes experiências CMB. As zonas a tracejado representam as regiões do plano ν − l
onde se espera que os foregrounds excedam as anisotropias CMBR nas 20% regiões mais limpas
do céu. O esquema de cores adoptado para os foregrounds é o seguinte: Poeira (vermelho),
point-sources (verde), sincrotrão (magenta) e emissão free-free (azul claro). A linha a tracejado
azul clara indica o lugar do plano onde a contribuição total dos foregrounds é minima para
cada multipolo (Tegmark & Efstathiou 1996, [?])

separadamente a informação contida nas anisotropias primárias do CMB e nas anisotropias

introduzidas pelo próprio efeito SZ.
Figure 3.9: Visão esquemática dos vários componentes do fundo de micro-ondas (“sandwich
cósmica”). O sinal CMB que se mede no céu contém, para além das anisótropias CMB
primárias, todo um conjunto de foregrounds e ruı́dos que importa remover. Cada uma das
camadas corresponde a um foreground que se adiciona ao sinal cosmológico CMB (na base).
Os mapas têm uma extensão angular de 100 e estão representados na vertical consoante a prox-
imidade da sua origem. No topo estão representados os efeitos de ruido (striping) intrı́nsecos à
experiência. À medida que “descemos na sandwich” encontramos os foregrounds com origem
na nossa galáxia, poeira (DUST), emissão de sincrotrão (SYNCHROTRON), bremsstrahlung
(FREE-FREE) e com origem extra–galáctica, emissão em infravermelho e rádio de galáxias
(GALAXIES) e contaminação devida a enxames (efeitos Sunyaev Zeldovich cinético, ∆T /T ,
e térmico, Y-SX).
Bibliography

[1] CMBFAST web site: physics.nyu.edu/matiasz/CMBFAST/cmbfast.html.

[2] J. M. Bardeen. Phys. Rev., D22:1882, 1980.

[3] J. M. Bardeen, J. R. Bond, N. Kaiser, and A. S. Szalay. 304:15, 1986.

[4] J. M. Bardeen, P. J. Steinhardt, and M. S. Turner. Phys. Rev., D28:679, 1983.

[5] R. Belén Barreiro. The cosmic microwave background: State of the art. astro-
ph/9907094, 1999.

[6] F. Benardeau. 427, 51 1994.

[7] C. L. Bennett, A. J. Banday, K. M. Gorski, G. Hinshaw, P. Jackson, P. Keegstra,

A. Kogut, G. F. Smoot, D. T. Wilkinson, and E. L. Wright. Four-year cobe dmr cosmic
microwave background observations: Maps and basic results. 464, L1 1996.

[8] E. Bertschinger. Simultations of structure formation in the universe. Ann. Rev. Astron.
Astrophys., 36:599–654, 1998.

[9] G. D. Birkhoff. Relativity and Modern Physics. Harvard University Press, Cambridge,
Mass., 1923.

[10] A. Blanchard. The physics of baryons in cosmology. In David Valls-Gabaud et al.,

editors, ASP Conf. Ser. 126: From Quantum Fluctuations to Cosmological Structures,
volume 1 of 329-347. Astronomical Society of the Pacific, 1997.

[11] J. R. Bond and G. Efstathiou. The statistics of cosmic background radiation fluctuations.
MNRAS, 226:655, 1987.

[12] J. R. Bondi and A. S. Szalay. 276:443, 1983.

[13] G. L. Bryan and M. L. Norman. Ap. J, 495:80, 1998.

[14] E. F. Bunn. Calculation of cosmic background radiation anisotropies and implications.

In NATO ASIC Proc. 502: The Cosmic Microwave Background, pages 135+, 1997.
astro-ph/960788.

65
[15] C. Burigana, L. Danese, and G. De Zotti. Formation and evolution of early distortions
of the microwave background spectrum: a numerical study. Astron. Astrophys., 246:49,
1991.

[16] S. M. Carroll, W. H. Press, and E. L. Turner. The cosmological constant. Annu. Rev.
Astron. Astrophys., 30:499–542, 1992.

[17] S. Cole and C. Lacey. MNRAS, 281:716, 1996.

[18] P. Coles and F. Lucchin. Cosmology – The Origin and Evolution of Cosmic Structure.
John Wiley & Sons, 1995.

[19] H. M. P. Couchman. Mesh-refined p3m - a fast adaptive n-body algorithm. 368:L23–L26,

1991.

[20] H. M. P. Couchman, P. A. Thomas, and F. R Pearce. Hydra: an adaptive-mesh imple-

mentation of p 3m-sph. 452:797, 1995.

[21] P. de Bernardis, P. A. R. Ade, J. J. Bock, J. R. Bond, J. Borrill, A. Boscaleri, K. Coble,

C. R. Contaldi, B. P. Crill, G. De Troia, P. Farese, K. Ganga, M. Giacometti, E. Hivon,
V. V. Hristov, A. Iacoangeli, A. H. Jaffe, W. C. Jones, A. E. Lange, L. Martinis, S. Masi,
P. Mason, P. D. Mauskopf, A. Melchiorri, T. Montroy, C. B. Netterfield, E. Pascale,
F. Piacentini, D. Pogosyan, G. Polenta, F. Pongetti, S. Prunet, G. Romeo, J. E. Ruhl,
and F. Scaramuzzi. Multiple Peaks in the Angular Power Spectrum of the Cosmic
Microwave Background: Significance and Consequences for Cosmology. 564:559–566,
January 2002.

[22] R. D’Inverno. Introducing Einstein Relativity. Clarendon Press, Oxford, 1992.

[23] Ray D’Inverno. Introducing Einstein Relativity. Clarendon Press, Oxford, 1992.

[24] S. Dodelson, G. Gyuk, and M. S. Turner. Is a massive tau neutrino just what cold dark
matter needs? Phys. Rev. Lett., 72:3754–3757, 1994. astro-ph/9402028.

[25] R. Durrer, M. Kunz, and A. Melchiorri. Cosmic structure formation with topological
defects. astro-ph/0110348, 2001.

[26] G. Efstathiou. Suppressing the formation of dwarf galaxies via photoionization. MNRAS,
256:43–47, 1992.

[27] G. Efstathiou, M. Davis, C. S. Frenk, and S. D. M. White. Numerical techniques for

large cosmological n-body simulations. Ap. J. Suppl., 57:241–260, 1985.

[28] G. Efstathiou and J. W. Eastwood. On the clustering of particles in an expanding

universe. MNRAS, 194:503–525, 1981.

[29] D. G. Eisenstein and W. Hu. 511:5, 1999.

[30] V. R. Eke, S. Cole, and C. S. Frenk. Cluster evolution as a diagnostic for omega.
MNRAS, 282:263–280, 1996.

[31] V. R. Eke, J. F. Navarro, and C. S. Frenk. The evolution of x-ray clusters in a low-density
universe. 503:569, 1998.

[32] A. E. Evrard. Biased cold dark matter theory - trouble from rich clusters? 341:L71–L74,
1989.

[33] D. J. Fixsen, E. S. Cheng, J. M. Gales, Mather J. C., R. A. Shafer, and E. L. Wright.

The cosmic microwave background spectrum from the full cobe firas data set. 473:576,
1996.

[34] P. Frampton, Y Jack Ng, and Ryan Rohm. Cosmic background radiation temperature
anisotropy: Position of the first doppler peak. astro-ph/9806118, 1998.

[35] W. L. Freedman. The measure of cosmological parameters. astro-ph/0202006.

[36] W. L. Freedman. Determination of the hubble constant. In N. Turok, editor, Critical

Dialogs in Cosmology, Singapore, 1997. World Scientific. astro-ph/9612024.

[37] W. L. Freedman, B. F. Madore, B. K. Gibson, L. Ferrarese, D. D. Kelson, S. Sakai, J. R.

Mould, R. C. Kennicutt, H. C. Ford, J. A. Graham, J. P. Huchra, S. M. G. Hughes,
G. D. Illingworth, L. M. Macri, and P. B. Stetson. Final Results from the Hubble Space
Telescope Key Project to Measure the Hubble Constant. 553:47–72, May 2001.

[38] R. A. Gingold and J. J. Monaghan. Smoothed particle hydrodynamics - theory and

application to non-spherical stars. MNRAS, 181:375–389, 1977.

[39] J. E. Gunn and B. A. Peterson. On the density of neutral hydrogen in intergalactic

space. 142:1633–1636, 1965.

[40] A. H. Guth. Phys. Rev., D23:347, 1981.

[41] A. H. Guth and S.-Y. Pi. Phys. Rev. Lett., 49:1110, 1982.

[42] Z. Haiman and L. Knox. Reionization of the intergalactic medium and its effect on the
cmb. In A. de Oliveira-Costa & M. Tegmark, editor, Microwave Foregrounds, volume
181 of ASP Conference Series. ASP Conference Series, 1999.

[43] R. Harrison. Phys. Rev., D1:2726, 1970.

[44] S. M. Harun-or-Rashid and M. Roos. Statistical evaluation of the observational infor-

mation on ωm and ωΛ . Astron. and Astrophys., 373:369–376, July 2001.

[45] S. W. Hawking. Phys. Lett., B115:295, 1982.

[46] P. Helbig. Constraints in the λ0 and ω0 plane from gravitational lensing. In IAU
Symposium, page 295, 2000. astro-ph/0011031.
[47] M. A. Hendry and R. J. Tayler. Contemp. Phys., 37:263, 1996.

[48] J. P. Henry and K. A. Arnaud. A measurement of the mass fluctuation spectrum from
the cluster x-ray temperature function. 372:410–418, 1991.

[49] M. B. Hindmarsh and T. W. B. Kibble. Rep. Prog. Phys., 58:477, 1995.

[50] R. W. Hockney and J. W. Eastwood. Computer Simulation Using Particles. Adam

Hilger, Bristol, 1988.

[51] W. Hu. Wandering in the Background: A Cosmic Microwave Background Explorer. PhD
thesis, University of California at Berkeley, 1995. astro-ph/9508126.

[52] Wayne Hu. Wandering in the Background: A Cosmic Microwave Background Explorer.
PhD thesis, University of California at Berkeley, 1995. astro-ph/9508126.

[53] Wayne Hu, Naoshi Sugiyama, and Joseph Silk. The physics of microwave background
anisotropies. Nature, 386:37–43, 1997. astro-ph/9604166.

[54] G. H. Jacoby, D. Branch, R. Clardullo, R. L. Davies, W. E. Harris, M. J. Pierce, C. J.

Pritchet, J. L. Tonry, and D. L. Welch. A critical review of selected techniques for
measuring extragalactic distances. Pub. Astron. Soc. Pac., 104:599–662, August 1992.

[55] J. H. Jeans. The stability of spiral nebula. Phill. Trans. R. Soc., 199A:1, 1902.

[56] S. Jha, P. M. Garnavich, R. P. Kirshner, P. Challis, A. M. Soderberg, L. M. Macri, J. P.

Huchra, P. Barmby, E. J. Barton, P. Berlind, W. R. Brown, N. Caldwell, M. L. Calkins,
S. J. Kannappan, D. M. Koranyi, M. A. Pahre, K. J. Rines, K. Z. Stanek, R. P. Stefanik,
A. H. Szentgyorgyi, P. Väisänen, Z. Wang, J. M. Zajac, A. G. Riess, A. V. Filippenko,
W. Li, M. Modjaz, R. R. Treffers, C. W. Hergenrother, E. K. Grebel, P. Seitzer, G. H.
Jacoby, P. J. Benson, A. Rizvi, L. A. Marschall, J. D. Goldader, M. Beasley, W. D.
Vacca, B. Leibundgut, J. Spyromilio, B. P. Schmidt, and P. R. Wood. The Type IA
Supernova 1998BU in M96 and the Hubble Constant. Ap. J. Supp., 125:73–97, November
1999.

[57] B. Jones and R. Wyse. A&A, 149(14), 1985.

[58] M. Kamionkowski, D. Spergel, and N. Sugiyama. ApJ, 426(57), 1994.

[59] S. Kay. Modelling Properties of Galaxies and Clusters. PhD thesis, University of
Durham, 2000.

[60] T. W. B. Kibble. J. Phys., A9:1387, 1976.

[61] E. Kolb and M. Turner. The Early Universe, volume 69 of Frontiers in Physics. Addison-
Wesley Publishing Company, 1990.

[62] Edward Kolb and Michael Turner. The Early Universe, volume 69 of Frontiers in
Physics. Addison-Wesley Publishing Company, 1990.
[63] A. R. Liddle. The early universe. In David Valls-Gabaud et al., editors, ASP Conf.
Ser. 126: From Quantum Fluctuations to Cosmological Structures, volume 1 of 31-62.
Astronomical Society of the Pacific, 1997.

[64] A. R Liddle and D. H. Lyth. The cold dark matter density perturbation. Phys Reports,
231:1, 1993.

[65] A. R. Liddle and D. H. Lyth. Cosmological Inflation and Large Scale Structure. Cam-
bridge University Press, 2000.

[66] E. M. Lifshitz. On the gravitational instability of the expanding universe. Sov. Phys.
JETP, 10:116, 1946.

[67] A. Loeb and R. Barkana. The reionization of the universe by the first stars and quasars.
Ann. Rev. Astron. Astrophys., 39:19–66, 2001.

[68] L. C. Lucy. A numerical approach to the testing of the fission hypothesis. Astron. J.,
82:1013–1024, 1977.

[69] J. C. Mather, D. J. Fixsen, R. A. Shafer, C. Mosier, and D. T. Wilkinson. Calibrator

Design for the COBE Far-Infrared Absolute Spectrophotometer (FIRAS). 512:511–520,
February 1999.

[70] W. Mattig. Astron. Nach., 284:109, 1958.

[71] J. J. Monaghan. Smoothed particle hydrodynamics. Ann. Rev. Astron. Astrophys.,

30:543–574, 1992.

[72] J. J. Monaghan and J. C. Lattanzio. Further studies of a fragmentation problem. Astron.

and Astrophys., 158:207–211, 1986.

[73] V. F. Mukhanov, H. A. Feldman, and R. H. Brandenberger. The theory of cosmological

perturbations. Phys. Rep., 215:203–333, 1992.

[74] C. B. Netterfield, P .A. R. Ade, J. J. J.J. Bock, J. R. Bond, J. Borrill, A. Boscaleri,

K. Coble, C. R. Contaldi, B. P. Crill, P. de Bernardis, P. Farese, K. Ganga, M. Gia-
cometti, E. Hivon, V. V. Hristov, A. Iacoangeli, A.H. Jaffe, W.C. Jones, A. E. Lange,
L. Martinis, S. Masi, P. Mason, P.D. Mauskopf, A. Melchiorri, T. Montroy, E. Pas-
cale, F. Piacentini, D. Pogosyan, F. Pongetti, S. Prunet, G. Romeo, J. E. Ruhl, and
F. Scaramuzzi. A measurement by BOOMERANG of multiple peaks in the angular
power spectrum of the cosmic microwave background. astro-ph/0104460.

[75] T. Padmanabham. Structure formation in the Universe. Cambridge University Press,

Cambridge, UK, 1993.

[76] J. A. Peacock. Cosmological Physics. Cambridge University Press, Cambridge, UK,

1999.
[77] J. A. Peacock and S. J. Dodds. Reconstructing the linear power spectrum of cosmological
fluctuations. MNARS, 267:1020, 1994.

[78] P. J. E. Peebles. Physical Cosmology. Princeton University Press, Princeton, 1971.

[79] P. J. E. Peebles. The Large-Scale Structure of the Universe. Princeton University Press,
Princeton, 1980.

[80] P. J. E. Peebles. Principles of Physical Cosmology. Princeton University Press, Prince-

ton, 1993.

[81] Ue-Li Pen. Normalizing the temperature function of clusters of galaxies. Ap. J. S.,
498:60, 1998.

[82] Ue-Li Pen. Brief note: Analytical fit to the luminosity distance for flat cosmologies with
a cosmological constant. Ap. J. S., 120:49, 1999.

[83] S. Perlmutter, G. Aldering, M. della Valle, S. Deustua, R. S. Ellis, S. Fabbro, A. Fruchter,

G. Goldhaber, D. E. Groom, I. M. Hook, A. G. Kim, M. Y. Kim, R. A. Knop, C. Lidman,
R. G. McMahon, P. Nugent, R. Pain, N. Panagia, C. R. Pennypacker, P. Ruiz-Lapuente,
B. Schaefer, and N. Walton. Discovery of a supernova explosion at half the age of the
universe. Nature, 391:51, January 1998.

[84] S. Perlmutter, G. Aldering, G. Goldhaber, R. A. Knop, P. Nugent, P. G. Castro,

S. Deustua, S. Fabbro, A. Goobar, D. E. Groom, I. M. Hook, A. G. Kim, M. Y. Kim,
J. C. Lee, N. J. Nunes, R. Pain, C. R. Pennypacker, R. Quimby, C. Lidman, R. S. Ellis,
M. Irwin, R. G. McMahon, P. Ruiz-Lapuente, N. Walton, B. Schaefer, B. J. Boyle, A. V.
Filippenko, T. Matheson, A. S. Fruchter, N. Panagia, H. J. M. Newberg, W. J. Couch,
and The Supernova Cosmology Project. Measurements of Omega and Lambda from 42
High-Redshift Supernovae. 517:565–586, June 1999.

[85] E. Pierpaoli, D. Scott, and M. White. Power-spectrum normalization from the local
abundance of rich clusters of galaxies. MNRAS, 325:77–88, 1999.

[86] W. H. Press and P. Schechter. Formation of galaxies and clusters of galaxies by selfsimilar
gravitational condensation. 187:425–438, 1974.

[87] J. R. Primack. Cosmological Parameters. In Sources and Detection of Dark Matter and
Dark Energy in the Universe, pages 3–25, 2001. astro-ph/0007187.

[88] C. Pryke, N. W. Halverson, E. M. Leitch, J. Kovac, J. E. Carlstrom, W. L. Holzapfel, and

M. Dragovan. Cosmological Parameter Extraction from the First Season of Observations
with the Degree Angular Scale Interferometer. Ap J., 568:46–51, March 2002.

[89] A. G. Riess, A. V. Filippenko, P. Challis, A. Clocchiatti, A. Diercks, P. M. Garnavich,

R. L. Gilliland, C. J. Hogan, S. Jha, R. P. Kirshner, B. Leibundgut, M. M. Phillips,
D. Reiss, B. P. Schmidt, R. A. Schommer, R. C. Smith, J. Spyromilio, C. Stubbs, N. B.
Suntzeff, and J. Tonry. Observational Evidence from Supernovae for an Accelerating
 Universe and a Cosmological Constant. Astronomical Journal, 116:1009–1038, Septem-
ber 1998.

[90] A. G. Riess, W. H. Press, and R. P. Kirshner. A Precise Distance Indicator: Type IA

Supernova Multicolor Light-Curve Shapes. 473:88, December 1996.

[91] M. Roos and S. M. Harun-or Rashid. How flat is the universe? astro-ph/0003040.

[92] M. Rowan-Robinson. The Cosmological Distance Ladder. Freeman, New York, 1985.

[93] G. B. Rybicki and A. P Lightman. Radiative Processes in Astrophysics. John Wiley and
Sons, New York, 1979.

[94] R. Sachs and A. Wolfe. ApJ, 147(1):73–90, 1967.

[95] R. K. Sachs and A. M. Wolfe. Perturbations of cosmological model and angular variations
of the microwave background. 147:73–90, 1967.

[96] A. R. Sandage, R. G. Korn, and M. S. Longair. The Deep Universe. Saas-Fee Advanced
Course 23. Springer-Verlag, 1993.

[97] B. Schutz. A first course in General Relativity. Cambridge University Press, Cambridge,
1990.

[98] U. Seljak and M. Zaldarriaga. A line-of-sight integration approach to cosmic microwave

background anisotropies. 469:437–444, 1996.

[99] S. F. Shandarin and Ya. B. Zel’dovich. Rev. Mod. Phys., 61:185, 1989.

[100] J. Silk. ApJ, 151:459, 1968.

[101] A. A. Starobinsky. Phys. Lett. B, 117:175, 1982.

[102] M. Steinmetz. Simulating galaxy formation. In A. Provenzale S. Bonometto, J. Primack,

editor, International School of Physics Enrico Fermi - Course CXXXII: Dark Matter in
the Universe. IOP, 1995. Vardena.

[103] R. Stompor, M. Abroe, P. Ade, A. Balbi, D. Barbosa, J. Bock, J. Borrill, A. Bosca-

leri, P. de Bernardis, P. G. Ferreira, S. Hanany, V. Hristov, A. H. Jaffe, A. T. Lee,
E. Pascale, B. Rabii, P. L. Richards, G. F. Smoot, C. D. Winant, and J. H. P. Wu.
Cosmological Implications of the MAXIMA-1 High-Resolution Cosmic Microwave Back-
ground Anisotropy Measurement. 561:L7–L10, November 2001.

[104] N. Sugiyama. Ap. J. Supp., 100:281, 1995.

[105] R. S. Sutherland and M. A. Dopita. Cooling functions for low-density astrophysical

plasmas. Ap. J. S., 88:253–327, 1993.

[106] M. Tegmark. Probes of the Early Universe. PhD thesis, University of California at
Berkeley, April 1994.
[107] Max Tegmark. Doppler peaks and all that: Cmb anisotropies and what can they tell
us. astro-ph/9511148, 1996.

[108] P. T. P. Viana and A. R. Liddle. The cluster abundance in flat and open cosmologies.
MNRAS, 281:323, 1996.

[109] P. T. P. Viana and A. R. Liddle. Galaxy clusters at 0.3 < z < 0.4 and the value of ω0 .
MNRAS, 303:535, 1999.

[110] A. Vilenkin and E. P. S. Shellard. Cosmic Strings and Other Topological Defects. Cam-
bridge University Press, Cambridge, 1994.

[111] S. Weinberg. Gravitation and Cosmology. John Wiley & Sons, 1972.

[112] M. White, J. E. Carlstrom, M. Dragovan, and W. L. Holzapfel. Interferometric obser-

vation of cosmic microwave background anisotropies. 514:12–24, 1999.

[113] M. White, G. Gelmini, and J. Silk. Structure formation with decaying neutrinos. Phys.
Rev. D, 51:2669–2676, March 1995.

[114] Martin White and Wayne Hu. The sachs-wolfe effect. astro-ph/9609105, 1996.

[115] Martin White, Douglas Scott, and Joseph Silk. Anisotropies in the cosmic microwave
background. Ann. Rev. Astron. & Astrophys., 32:319–370, 1994.

[116] S. D. M. White, G. Efstathiou, and C. S. Frenk. The amplitude of mass fluctuations in

the universe. MNRAS, 262:1023–1028, 1993.

[117] G. Yepes. Cosmological numerical simulations. In David Valls-Gabaud et al., editors,

ASP Conf. Ser. 126: From Quantum Fluctuations to Cosmological Structures, pages
279–311. Astronomical Society of the Pacific, 1997.

[118] Y. B. Zeldovich. My universe. Selected reviews. Chur: Harwood Academic Publishers,

1992, edited by Zeldovich, Boris Ya.; Sazhin, M.V., 1992.

[119] Ya. B. Zel’dovich. Gravitational instability: an approximate theory for large density
perturbations. Astron. Astrophys., 5:84, 1970.