Cosmic

Download as pdf or txt
Download as pdf or txt
You are on page 1of 59

version 3

An overview of cosmology
Julien Lesgourgues
Theoretical Physics Division, CERN, CH-1211 Genève 23, Switzerland
LAPTH, Chemin de Bellevue, B.P. 110, F-74941 Annecy-Le-Vieux Cedex, France
Preprint number CERN-TH/2002-249

These notes follow from a series of lectures given at the


CERN Student Summer School
(August 2003)

Please don’t hesitate to report any mistake or misprint to


Julien.Lesgourgues@cern.ch

What is the difference between astrophysics and cosmology?

While astrophysicists study the surrounding celestial bodies, like planets,


stars, galaxies, clusters of galaxies, gas clouds, etc., cosmologists try to describe
the evolution of the Universe as a whole, on the largest possible distances and
time scales. While purely philosophical in the early times, and still very specu-
lative at the beginning of the XX-th century, cosmology has gradually entered
into the realm of experimental science over the past 80 years. Today, as we will
see in chapter 2, astronomers are even able to obtain very precise maps of the
surrounding Universe a few billion years ago, i.e. over distances ranging up to
billions of light-years.
Cosmology has raised some fascinating questions like: is the Universe static
or expanding ? How old is it and what will be its future evolution ? Is it flat,
open or closed ? Of what type of matter is it composed ? How did structures
like galaxies form ? In this course, we will try to give an overview of these
questions, and of the partial answers that can be given today.
In the first chapter, we will introduce some fundamental concepts, in partic-
ular from General Relativity. Along this chapter, we will remain in the domain
of abstraction and geometry. In the second chapter, we will apply these concepts
to the real Universe and deal with concrete results, observations, and testable
predictions.
Contents

1 The Expanding Universe 4


1.1 The Hubble Law . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.1 The Doppler effect . . . . . . . . . . . . . . . . . . . . . . 4
1.1.2 The discovery of the galactic structure . . . . . . . . . . . 5
1.1.3 The Cosmological Principle . . . . . . . . . . . . . . . . . 5
1.1.4 Hubble’s discovery . . . . . . . . . . . . . . . . . . . . . . 6
1.1.5 Homogeneity and inhomogeneities . . . . . . . . . . . . . 7
1.2 The Universe Expansion from Newtonian Gravity . . . . . . . . . 9
1.2.1 Newtonian Gravity versus General Relativity . . . . . . . 9
1.2.2 The rate of expansion from Gauss theorem . . . . . . . . 9
1.2.3 The limitations of Newtonian predictions . . . . . . . . . 11
1.3 General relativity and the Friemann-Lemaı̂tre model . . . . . . . 11
1.3.1 The curvature of space-time . . . . . . . . . . . . . . . . . 12
1.3.2 Building the first cosmological models . . . . . . . . . . . 14
1.3.3 Our Universe is curved . . . . . . . . . . . . . . . . . . . . 14
1.3.4 Comoving coordinates . . . . . . . . . . . . . . . . . . . . 15
1.3.5 Bending of light in the expanding Universe . . . . . . . . 17
1.3.6 The Friedmann law . . . . . . . . . . . . . . . . . . . . . . 22
1.3.7 Relativistic matter and Cosmological constant . . . . . . 22

2 The Standard Cosmological Model 24


2.1 The Hot Big Bang scenario . . . . . . . . . . . . . . . . . . . . . 24
2.1.1 Various possible scenarios for the history of the Universe . 24
2.1.2 The matter budget today . . . . . . . . . . . . . . . . . . 27
2.1.3 The Cold and Hot Big Bang alternatives . . . . . . . . . . 28
2.1.4 The discovery of the Cosmic Microwave Background . . . 30
2.1.5 The Thermal history of the Universe . . . . . . . . . . . . 30
2.1.6 A recent stage of curvature or cosmological constant dom-
ination? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.1.7 Dark Matter . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Cosmological perturbations . . . . . . . . . . . . . . . . . . . . . 33
2.2.1 Linear perturbation theory . . . . . . . . . . . . . . . . . 33
2.2.2 The horizon . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.3 Photon perturbations . . . . . . . . . . . . . . . . . . . . 37
2.2.4 Observing the CMB anisotropies . . . . . . . . . . . . . . 40
2.2.5 Matter perturbations . . . . . . . . . . . . . . . . . . . . . 42
2.2.6 Hierarchical structure formation . . . . . . . . . . . . . . 44
2.2.7 Observing the matter spectrum . . . . . . . . . . . . . . . 45
2.3 Measuring the cosmological parameters . . . . . . . . . . . . . . 47
2.3.1 Abundance of primordial elements . . . . . . . . . . . . . 47
2.3.2 CMB anisotropies . . . . . . . . . . . . . . . . . . . . . . 48
2.3.3 Age of the Universe . . . . . . . . . . . . . . . . . . . . . 50
2.3.4 Luminosity of Supernovae . . . . . . . . . . . . . . . . . . 50

2
2.3.5 Large Scale Structure . . . . . . . . . . . . . . . . . . . . 52
2.4 The Inflationary Universe . . . . . . . . . . . . . . . . . . . . . . 53
2.4.1 Problems with the Standard Cosmological Model . . . . . 53
2.4.2 An initial stage of inflation . . . . . . . . . . . . . . . . . 54
2.4.3 Scalar field inflation . . . . . . . . . . . . . . . . . . . . . 56
2.4.4 Quintessence ? . . . . . . . . . . . . . . . . . . . . . . . . 58

3
Chapter 1

The Expanding Universe

1.1 The Hubble Law


1.1.1 The Doppler effect
At the beginning of the XX-th century, the understanding of the global structure
of the Universe beyond the scale of the solar system was still relying on pure
speculation. In 1750, with a remarkable intuition, Thomas Wright noticed that
the luminous stripe observed in the night sky and called the Milky Way could
be a consequence of the spatial distribution of stars: they could form a thin
plate, what we call now a galaxy. At that time, with the help of telescopes,
many faint and diffuse objects had been already observed and listed, under the
generic name of nebulae - in addition to the Andromeda nebula which is visible
by eye, and has been known many centuries before the invention of telescope.
Soon after the proposal of Wright, the philosopher Emmanuel Kant suggested
that some of these nebulae could be some other clusters of stars, far outside
the Milky Way. So, the idea of a galactic structure appeared in the mind of
astronomers during the XVIII-th century, but even in the following century
there was no way to check it on an experimental basis.
At the beginning of the nineteenth century, some physicists observed the
first spectral lines. In 1842, Johann Christian Doppler argued that if an ob-
server receives a wave emitted by a body in motion, the wavelength that he will
measure will be shifted proportionally to the speed of the emitting body with
respect to the observer (projected along the line of sight):

∆λ/λ = ~v .~n/c (1.1)

where c is the celerity of the wave (See figure 1.1). He suggested that this
effect could be observable for sound waves, and maybe also for light. The later
assumption was checked experimentally in 1868 by Sir William Huggins, who
found that the spectral lines of some neighboring stars were slightly shifted
toward the red or blue ends of the spectrum. So, it was possible to know the
projection along the line of sight of star velocities, vr , using

z ≡ ∆λ/λ = vr /c (1.2)

where z is called the redshift (it is negative in case of blue-shift) and c is the
speed of light. Note that the redshift gives no indication concerning the distance
of the star. At the beginning of the XX-th century, with increasingly good
instruments, people could also measure the redshift of some nebulae. The first
measurements, performed on the brightest objects, indicated some arbitrary
distribution of red and blue-shifts, like for stars. Then, with more observations,

4
n
v
Figure 1.1: The Doppler effect

it appeared that the statistics was biased in favor of red-shifts, suggesting that
a majority of nebulae were going away from us, unlike stars. This was raising
new questions concerning the distance and the nature of nebulae.

1.1.2 The discovery of the galactic structure


In the 1920’s, Leavitt and Shapley studied some particular stars, called the
cepheids, known to have a periodic time-varying luminosity. They could show
that the period of cepheids is proportional to their absolute luminosity L (the
absolute luminosity is the total amount of light emitted by unit of time, i.e., the
flux integrated on a closed surface around the star). They were also able to give
the coefficient of proportionality. So, by measuring the apparent luminosity, i.e.
the flux l per unit of surface through an instrument pointing to the star, it was
easy to get the distance of the star r from

L = l × 4πr2 .

(1.3)

Using this technique, it became possible to measure the distance of various


cepheids inside our galaxies, and to obtain the first estimate of the characteristic
size of the stellar disk of the Milky Way (known today to be around 80.000 light-
years).
But what about nebulae? In 1923, the 2.50m telescope of Mount Wilson (Los
Angeles) allowed Edwin Hubble to make the first observation of individual stars
inside the brightest nebula, Andromeda. Some of these were found to behave like
cepheids, leading Hubble to give an estimate of the distance of Andromeda. He
found approximately 900.000 light-years (but later, when cepheids were known
better, this distance was established to be around 2 million light-years). That
was the first confirmation of the galactic structure of the Universe: some nebulae
were likely to be some distant replicas of the Milky Way, and the galaxies were
separated by large voids.

1.1.3 The Cosmological Principle


This observation, together with the fact that most nebulae are redshifted (ex-
cepted for some of the nearest ones like Andromeda), was an indication that on
the largest observable scales, the Universe was expanding. At the beginning,
this idea was not widely accepted. Indeed, in the most general case, a given
dynamics of expansion takes place around a center. Seeing the Universe in ex-
pansion around us seemed to be an evidence for the existence of a center in the
Universe, very close to our own galaxy.
Until the middle age, the Cosmos was thought to be organized around
mankind, but the common wisdom of modern science suggests that there should

5
A B C A B C

Figure 1.2: Homogeneous expansion on a two-dimensional grid. Some equally-


spaced observers are located at each intersection. The grid is plotted twice. On
the left, the arrays show the expansion flow measured by A; on the right, the
expansion flow measured by B. If we assume that the expansion is homogeneous,
we get that A sees B going away at the same velocity as B sees C going away.
So, using the additivity of speeds, the velocity of C with respect to A must
be twice the velocity of B with respect to A. This shows that there is a linear
relation between speed and distance, valid for any observer.

be nothing special about the region or the galaxy in which we leave. This in-
tuitive idea was formulated by the astrophysicist Edward Arthur Milne as the
“Cosmological Principle”: the Universe as a whole should be homogeneous, with
no privileged point playing a particular role.
Was the apparently observed expansion of the Universe a proof against the
Cosmological Principle? Not necessarily. The homogeneity of the Universe is
compatible either with a static distribution of galaxies, or with a very special
velocity field, obeying to a linear distribution:
~v = H ~r (1.4)
where ~v denotes the velocity of an arbitrary body with position ~r, and H is a
constant of proportionality. An expansion described by this law is still homo-
geneous because it is left unchanged by a change of origin. To see this, one can
make an analogy with an infinitely large rubber grid, that would be stretched
equally in all direction: it would expand, but with no center (see figure 1.2).
This result is not true for any other velocity field. For instance, the expansion
law
~v = H |~r| ~r (1.5)
is not invariant under a change of origin: so, it has a center.

1.1.4 Hubble’s discovery


So, a condition for the Universe to respect the Cosmological Principle is that the
speed of galaxies along the line of sight, or equivalently, their redshift, should
be proportional to their distance. Hubble tried to check this idea, still using the
cepheid technique. He published in 1929 a study based on 18 galaxies, for which
he had measured both the redshift and the distance. His results were showing
roughly a linear relation between redshift and distance (see figure 1.3). He
concluded that the Universe was in homogeneous expansion, and gave the first
estimate of the coefficient of proportionality H, called the Hubble parameter.
This conclusion has been checked several time with increasing precision and
is widely accepted today. It can be considered as the starting point of exper-
imental cosmology. It is amazing to note that the data used by Hubble was

6
Figure 1.3: The diagram published by Hubble in 1929. The labels of the hori-
zontal (resp. vertical) axis are 0, 1, 2 Mpc (resp. 0, 500, 1000 km.s−1 ). Hubble
estimated the expansion rate to be 500 km.s−1 Mpc−1 . Today, it is known to be
around 70 km.s−1 Mpc−1 .

so imprecise that Hubble’s conclusion was probably a bit biaised... Anyway,


current data leaves no doubt about the proportionality, even if there is still an
uncertainty concerning the exact value of H. The Hubble constant is generally
parametrized as
H = 100 h km s−1 Mpc−1 (1.6)
where h is the dimensionless “reduced Hubble parameter”, currently known to
be in the range h = 0.71 ± 0.04, and Mpc denotes a Mega-parsec, the unity
of distance usually employed for cosmology (1 Mpc ' 3 × 1022 m ' 3 × 106
light-years; the proper definition of a parsec is “the distance to an object with a
parallax of one arcsecond”; the parallax being half the angle under which a star
appears to move when the earth makes one rotation around the sun). So, for
instance, a galaxy located at 10 Mpc goes away at a speed close to 700 km s −1 .

1.1.5 Homogeneity and inhomogeneities


Before leaving this section, we should clarify one point about the “Cosmological
Principle”, i.e., the assumption that the Universe is homogeneous. Of course,
nobody has ever claimed that the Universe was homogeneous on small scales,
since compact objects like planets or stars, or clusters of stars like galaxies
are inhomogeneities in themselves. The Cosmological Principle only assumes
homogeneity after smoothing over some characteristic scale. By analogy, take
a grid of step l (see figure 1.4), and put one object in each intersection, with a

7
smoothing
radius

Figure 1.4: We build an inhomogeneous distribution of objects in the following


way: starting from each intersection of the grid, we draw a random vector and
put an object of random mass at the extremity of the vector. Provided that
all random vectors and masses obey to the same distributions of probability,
the mass density is still homogeneous when it is smoothed over a large enough
smoothing radius (in our example, the typical length of the vectors is smaller
than the step of the grid; but our conclusion would still apply if the vectors were
larger than the grid step, provided that the smoothing radius is even larger).
This illustrates the concept of homogeneity on a given scale, like in the Universe.

randomly distributed mass (with all masses obeying to the same distribution of
probability). Then, make a random displacement of each object (again with all
displacements obeying to the same distribution of probability). At small scales,
the mass density is obviously inhomogeneous for three reasons: the objects
are compact, they have different masses, and they are separated by different
distances. However, since the distribution has been obtained by performing a
random shift in mass and position, starting from an homogeneous structure, it
is clear even intuitively that the mass density smoothed over some large scale
will remain homogeneous again.
The Cosmological Principle should be understood in this sense. Let us sup-
pose that the Universe is almost homogeneous at a scale corresponding, say, to
the typical intergalactic distance, multiplied by thirty or so. Then, the Hub-
ble law doesn’t have to be verified exactly for an individual galaxy, because
of peculiar motions resulting from the fact that galaxies have slightly different
masses, and are not in a perfectly ordered phase like a grid. But the Hubble
law should be verified in average, provided that the maximum scale of the data
is not smaller than the scale of homogeneity. The scattering of the data at a
given scale reflects the level of inhomogeneity, and when using data on larger
and larger scales, the scattering must be less and less significant. This is exactly
what is observed in practice. An even better proof of the homogeneity of the
Universe on large scales comes from the Cosmic Microwave Background, as we
shall see in section 2.2.
We will come back to these issues in section 2.2, and show how the forma-
tion of inhomogeneities on small scales are currently understood and quantified
within some precise physical models.

8
1.2 The Universe Expansion from Newtonian Grav-
ity
It is not enough to observe the galactic motions, one should also try to explain
it with the laws of physics.

1.2.1 Newtonian Gravity versus General Relativity


On cosmic scales, the only force expected to be relevant is gravity. The first
theory of gravitation, derived by Newton, was embedded later by Einstein into
a more general theory: General Relativity (thereafter denoted GR). However,
in simple words, GR is relevant only for describing gravitational forces between
bodies which have relative motions comparable to the speed of light1 . In most
other cases, Newton’s gravity gives a sufficiently accurate description.
Precisely, the speed of neighboring galaxies is always much smaller than
the speed of light. So, a priori, Newtonian gravity should be able to explain
the Hubble flow. One could even think that historically, Newton’s law led to
the prediction of the Universe expansion, or at least, to its first interpretation.
Amazingly, and for reasons which are more mathematical than physical, it hap-
pened not to be the case: the first attempts to describe the global dynamics
of the Universe came with GR, in the 1910’s. In this course, for pedagogical
purposes, we will not follow the historical order, and start with the Newtonian
approach.
Newton himself did the first step in the argumentation. He noticed that
if the Universe was of finite size, and governed by the law of gravity, then all
massive bodies would unavoidably concentrate into a single point, just because
of gravitational attraction. If instead it was infinite, and with an approximately
homogeneous distribution at initial time, it could concentrate into several points,
like planets and stars, because there would be no center to fall in. In that case,
the motion of each massive body would be driven by the sum of an infinite
number of gravitational forces. Since the mathematics of that time didn’t allow
to deal with this situation, Newton didn’t proceed with his argument.

1.2.2 The rate of expansion from Gauss theorem


In fact, using Gauss theorem, this problem turns out to be quite simple. Suppose
that the Universe consists in many massive bodies distributed in an isotropic and
homogeneous way (i.e., for any observer, the distribution looks the same in all
directions). This should be a good modelization of the Universe on sufficiently
large scales. We wish to compute the motion of a particle located at a distance
r away from us. Because the Universe is assumed to be isotropic, the problem is
spherically symmetric, and we can employ Gauss theorem on a sphere centered
on us, and of radius r (see figure 1.5). So, the acceleration of any particle on
the surface of the sphere reads

GM (r)
r̈ = − (1.7)
r2
where G is Newton’s constant and M (r) is the mass contained inside the sphere
of radius r. In other words, the particle feels the same force as if it had a
two-body interaction with the mass of the sphere concentrated at the center.
Note that r varies with time, but M (r) remains constant: because of spherical
1 Going a little bit more into details, it is also relevant when an object is so heavy and so

close that the speed of liberation from this object - defined like the speed of liberation from
the Earth - is comparable to the speed of light.

9
r

Figure 1.5: Gauss theorem applied to the local Universe.

symmetry, no particles can enter or leave the sphere, which contains always the
same mass.
Since Gauss theorem allows us to make completely abstraction of the mass
outside the sphere, we can make an analogy with the motion e.g. of a satellite
ejected vertically from the Earth. We know that this motion depends on the
initial velocity, compared with the speed of liberation from the Earth: if the
initial speed is large enough, the satellites goes away indefinitely, otherwise it
stops and falls down. We can see this mathematically by multiplying equation
(1.7) by ṙ, and integrating it over time:
ṙ2 GM (r) k
= − (1.8)
2 r 2
where k is a constant of integration. We can replace the mass M (r) by the
volume of the sphere multiplied by the homogeneous mass density ρmass , and
rearrange the equation as
 2
ṙ 8πG k
= ρmass − 2 . (1.9)
r 3 r
The quantity ṙ/r is called the rate of expansion. Since M (r) is time-independent,
the mass density evolves as ρmass ∝ r−3 (i.e., matter is simply diluted when the
Universe expands). The behavior of r(t) depends on the sign of k. If k is posi-
tive, r(t) can grow at early times but it always decreases at late times, like the
altitude of the satellite falling back on Earth: this would correspond to a Uni-
verse expanding first, and then collapsing. If k is zero or negative, the expansion
lasts forever.
In the case of the satellite, the critical value, which is the speed of liberation
(at a given altitude), depends on the mass of the Earth. By analogy, in the case
of the Universe, the important quantity that should be compared with some
critical value is the homogeneous mass density. If at any time, ρmass it is bigger
than the critical value
3(ṙ/r)2
ρmass = (1.10)
8πG
then k is positive and the Universe will re-collapse. Physically, what it means
is that the gravitational force wins against inertial effects. In the other case,
the Universe expands forever, because the density is too small with respect
to the expansion velocity, and gravitation never takes over inertia. The case
k = 0 corresponds to a kind of equilibrium between gravitation and inertia.
The Universe expands forever, following a power–law: r(t) ∝ t2/3 .

10
k<0

k=0

r
k>0

Figure 1.6: The motion of expansion in a Newtonian Universe is equivalent to


that of a body ejected from Earth. It depends on the initial rate of expansion,
compared with the equivalent of the speed of liberation. When the parameter
k is zero or negative, the expansion lasts forever, otherwise the Universe re-
collapses (r → 0).

1.2.3 The limitations of Newtonian predictions


In the previous calculation, we cheated a little bit: we assumed that the Universe
was isotropic around us, but we didn’t check that it was isotropic everywhere
(and therefore homogeneous). Following what we said before, homogeneous
expansion requires proportionality between speed and distance. Looking at
equation (1.9), we see immediately that this is true at a given time only when k =
0. So, it seems that the other solutions are not compatible with the Cosmological
Principle.
This argument shouldn’t be taken seriously, because the link that we made
between homogeneity and linear expansion was based on the additivity of speed
(look for instance at the caption of figure 1.2), and therefore, on Newtonian
mechanics. But Newtonian mechanics cannot be applied at large distances,
where v becomes large and comparable to the speed of light. This would occur
for a characteristic scale called the Hubble radius RH :

RH = cH −1 , (1.11)

at which the Newtonian expansion law gives v = HRH = c.


So, the full problem has to be formulated in relativistic terms. In the GR
results, we will see again some solutions with k 6= 0, but they will remain
compatible with the homogeneity of the Universe.

1.3 General relativity and the Friemann-Lemaı̂tre


model
It is far beyond the scope of this course to introduce General Relativity, and to
derive step by step the relativistic laws governing the evolution Universe. We
will simply write these laws, asking the reader to admit them - and in order
to give a perfume of the underlying physical concepts, we will comment on the
differences with their Newtonian counterparts.

11
Figure 1.7: Measuring the curvature of some two-dimensional spaces. By walk-
ing four times along a distance d, turning 90 degrees left between each walk, a
small man on the plane would find that he is back at his initial point. Doing
the same thing, a man on the sphere would cross his trajectory, and would stop
way from his departure point. Instead, a man on the hyperboloid would not
close his trajectory.

1.3.1 The curvature of space-time


When Einstein tried to build a theory of gravitation compatible with the invari-
ance of the speed of light, he found that the minimal price to pay was :
• to abandon the idea of a gravitational potential, related to the distribution
of matter, and whose gradient gives the gravitational field in any point.
• to assume that our four-dimensional space-time is curved by the presence
of matter.
• to impose that free-falling objects describe geodesics in this space-time.
What does that mean in simple words?
First, let’s recall briefly what a curved space is, first with only two-dimensional
surfaces. Consider a plane, a sphere and an hyperboloid. For us, it’s obvious
that the sphere and the hyperboloid are curved, because we can visualize them
in our three-dimensional space: so, we have an intuitive notion of what is flat
and what is curved. But if there were some two-dimensional people living on
these surfaces, not being aware of the existence of a third dimension, how could
they know whether they leave in a flat or a in curved space-time?
There are several ways in which they could measure it. One would be to
obey the following prescription: walk in straight line on a distance d; turn 90
degrees left; repeat this sequence three times again; see whether you are back at
your initial position. The people on the three surfaces would find that they are,
as long as they walk along a small square, smaller than the radius of curvature.
But a good test is to repeat the operation on larger and larger distances. When
the size of the square will be of the same order of magnitude as the radius of
curvature, the habitant of the sphere will notice that before stopping, he crosses
the first branch of his trajectory (see figure 1.7). The one on the hyperboloid
will stop without closing his trajectory.
It is easy to think of the curvature of a two-dimensional surface because we
can visualize it embedded into three-dimensional space. Getting an intuitive
representation of a three-dimensional curved space is much more difficult. A 3-
sphere and a 3-hyperboloid could be defined analytically as some 3-dimensional
spaces obeying to the equation a2 + b2 + c2 ± d2 = R2 inside a 4-dimensional
Euclidian space with coordinates (a, b, c, d). If we wanted to define them by
making use of only three dimensions, the problem would be exactly like for
drawing a planisphere of the Earth. We would need to give a map of the spaces,

12
C

B
y

x
A

Figure 1.8: Gravitational lensing. Somewhere between an object C and a ter-


restrial observer A, a massive object B - for instance, a galaxy - curves its sur-
rounding space-time. Here, for simplicity, we only draw two spatial dimensions.
In absence of gravity and curvature, the only possible trajectory of light between
C and A would be a straight line. But because of curvature, the straight line
is not anymore the shortest trajectory. Photons prefer to follow two geodesics,
symmetrical around B. So, the observer will not see one image of C, but two
distinct images. In fact, if we restore the third spatial dimension, and if the
three points are perfectly aligned, the image of C will appear as a ring around
B. This phenomenon is observed in practice.

together with a crucial information: the scale of the map as a function of the
location on the map - the scale on a planisphere is not uniform! This would
bring us to a mathematical formalism, called Riemann geometry, that we don’t
have time to introduce here.
That was still four three dimensions. Finally, the curvature of a four-
dimensional space-time is impossible to visualize intuitively, first because it has
even more dimensions, and second because even in special/general relativity,
there is a difference between time and space (for the readers who are familiar
with special relativity, what is referred here is the negative signature of the
metric).
The Einstein theory of gravitation says that four-dimensional space-time is
curved, and that the curvature in each point is given entirely in terms of the
matter content in this point. In simple words, this means that the curvature
plays more or less the same role as the potential in Newtonian gravity. But the
potential was simply a function of space and time coordinates. In GR, the full
curvature is described not by a function, but by something more complicated -
like a matrix of functions obeying to particular laws - called a tensor.
Finally, the definition of geodesics (the trajectories of free-falling bodies)
is the following. Take an initial point and an initial direction. They define a
unique line, called a geodesic, such that any segment of the line gives the shortest
trajectory between the two points (so, for instance, on a sphere of radius R,
the geodesics are all the great circles of radius R, and nothing else). Of course,
geodesics depend on curvature. All free-falling bodies follow geodesics, including
light rays. This leads for instance to the phenomenon of gravitational lensing
(see figure 1.8).
So, in General Relativity, gravitation is not formulated as a force or a field,
but as a curvature of space-time, sourced by matter. All isolated systems follow
geodesics which are bent by the curvature. In this way, their trajectories are

13
affected by the distribution of matter around them: this is precisely what gravity
means.

1.3.2 Building the first cosmological models


After obtaining the mathematical formulation of General Relativity, around
1916, Einstein studied various testable consequences of his theory in the solar
system (e.g., corrections to the trajectory of Mercury, or to the apparent diam-
eter of the sun during an eclipse). But remarkably, he immediately understood
that GR could also be applied to the Universe as a whole, and published some
first attempts in 1917. However, Hubble’s results concerning the expansion
were not known at that time, and most physicists had the prejudice that the
Universe should be not only isotropic and homogeneous, but also static – or
stationary. As a consequence, Einstein (and other people like De Sitter) found
some interesting cosmological solutions, but not the ones that really describe
our Universe.
A few years later, some other physicists tried to relax the assumption of sta-
tionarity. The first was the russo-american Friedmann (in 1922), followed closely
by the Belgian physicist and priest Lemaı̂tre (in 1927), and then by some amer-
icans, Robertson and Walker. When the Hubble flow was discovered in 1929, it
became clear for a fraction of the scientific community that the Universe could
be described by the equations of Friedmann, Lemaı̂tre, Roberston and Walker.
However, many people – including Hubble and Einstein himself – remained re-
luctant to this idea for many years. Today, the Friedmann – Lemaı̂tre model is
considered as one of the major achievements of the XXth century.
Before giving these equations, we should stress that they look pretty much
the same as the Newtonian results given above - although some terms seem
to be identical, but have a different physical interpretation. This similarity
was noticed only much later. Physically, it has to do with the fact that there
exists a generalization of Gauss theorem to GR, known as Birkhoff theorem.
So, as in Newtonian gravity, one can study the expansion of the homogeneous
Universe by considering only matter inside an sphere. But in small regions,
General Relativity admits Newtonian gravity as an asymptotic limit, and so the
equations have many similarities.

1.3.3 Our Universe is curved


The Friemann-Lemaı̂tre model is defined as the most general solution of the laws
of General Relativity, assuming that the Universe is isotropic and homogeneous.
We have seen that in GR, matter curves space-time. So, the Universe is curved
by its own matter content, along its four space and time dimensions. However,
because we assume homogeneity, we can decompose the total curvature into two
parts:
• the spatial curvature, i.e., the curvature of the usual 3-dimensional space
(x, y, z) at fixed time. This curvature can be different at different times.
It is maximally symmetric, i.e., it makes no difference between the three
directions (x, y, z). There are only three maximally symmetric solutions:
space can be Euclidean, a 3-sphere with finite volume, or a 3-hyperboloid.
These three possibilities are referred to as a flat, a closed or an open
Universe. A priori, nothing forbids that we leave in a closed or in an open
Universe: if the radius of curvature was big enough, say, bigger than the
size of our local galaxy cluster, then the curvature would show up only
in long range astronomical observations. In the next chapter, we will see
how recent observations are able to give a precise answer to this question.

14
• the two-dimensional space-time curvature, i.e., for instance, the curvature
of the (t, x) space-time. Because of isotropy, the curvature of the (t, y)
and the (t, z) space-time have to be the same. This curvature is the one
responsible for the expansion. Together with the spatial curvature, it fully
describes gravity in the homogeneous Universe.
The second part is a little bit more difficult to understand for the reader who
is not familiar with GR, and causes a big cultural change in one’s intuitive
understanding of space and time.

1.3.4 Comoving coordinates


In the spirit of General Relativity, we will consider a set of three variables that
will represent the spatial coordinates (i.e., a way of mapping space, and of giving
a label to each point), but NOT directly a measure of distance!
The laws of GR allow us to work with any system of coordinates that we
prefer. For simplicity, in the Friedmann model, people generally employ a par-
ticular system of spherical coordinates (r, θ, φ) called “comoving coordinates”,
with the following striking property: the physical distance dl between two in-
finitesimally close objects with coordinates (r, θ, φ) and (r + dr, θ + dθ, z + dz)
is not given by
dl2 = dr2 + r2 (dθ2 + sin2 θ dφ2 ) (1.12)
as in usual Euclidean space, but

dr2
 
dl2 = a2(t) + r 2
(dθ 2
+ sin 2
θ dφ 2
) (1.13)
1 − kr2

where a(t) is a function of time, called the scale factor, whose time-variations
account for the curvature of two-dimensional space-time; and k is a constant
number, related to the spatial curvature: if k = 0, the Universe is flat, if k > 0,
it is closed, and if k < 0, it is open. In the last two cases, the radius of curvature
Rc is given by
a(t)
Rc (t) = p . (1.14)
|k|
When the Universe is closed, it has a √ finite volume, so the coordinate r is defined
only up to a finite value: 0 ≤ r < 1/ k.
If k was equal to zero and a was constant in time, we could redefine the
coordinate system with (r 0 , θ0 , φ0 ) = (ar, θ, φ), find the usual expression (1.12)
for distances, and go back to Newtonian mechanics. So, we stress again that
the curvature really manifests itself as k 6= 0 (for spatial curvature) and ȧ 6= 0
(for the remaining space-time curvature).
Note that when k = 0, we can rewrite dl 2 in Cartesian coordinates:

dl2 = a2(t) dx2 + dy 2 + dz 2 .



(1.15)

In that case, the expressions for dl contains only the differentials dx, dy, dz,
and is obviously left unchanged by a change of origin - as it should be, because
we assume a homogeneous Universe. But when k 6= 0, there is the additional
factor 1/(1 − kr 2 ), where r is the distance to the origin. So, naively, one could
think that we are breaking the homogeneity of the Universe, and privileging
a particular point. This would be true in Euclidean geometry, but in curved
geometry, it is only an artifact. Indeed, choosing a system of coordinates is
equivalent to mapping a curved surface on a plane. By analogy, if one draws a
planisphere of half-of-the-Earth by projecting onto a plane, one has to chose a
particular point defining the axis of the projection. Then, the scale of the map

15
Figure 1.9: Analogy between the Friedman coordinates for an open/closed Uni-
verse, and the mapping of a sphere by parallel projection onto a plane. Two
maps are shown, with different axes of projection. On each map, the scale de-
pends on the distance to the center of the map: the scale is smaller at the center
than on the borders. However, the physical distance between too point on the
sphere can be computed equally well using the two maps and their respective
scaling laws.

is not the same all over the map. However, if one reads too maps with different
projection axes (see figure 1.9), and computes the distance between Geneva and
Rome from the two maps, using the appropriate scaling laws, he will of course
find the same physical distance. This is exactly what happens in our case: in
a particular coordinate system, the expression for physical lengths depends on
the coordinate with respect to the origin, but physical lengths are left unchanged
by a change of coordinates. In figure 1.10 we push the analogy and show how
the factor 1/(1 − kr 2 ) appears explicitly for the parallel projection of a 2-sphere
onto a plane. The rigorous proof that equation (1.13) is invariant by a change
of origin would be more complicated, but essentially similar.
Free-falling bodies follow geodesics in this curved space-time. Among bodies,
we can distinguish between photons, which are relativistic because they travel at
the speed of light, and ordinary matter like galaxies, which are non-relativistic
(i.e., an observer sitting on them can ignore relativity at short distance, as we
do on Earth).
The geodesics followed by non-relativistic bodies in space-time are given by
dl = 0, i.e., fixed spatial coordinates. So, galaxies are at rest in comoving coor-
dinates. But it doesn’t mean that the Universe is static, because all distances
grow proportionally to a(t): so, the scale factor accounts for the homogeneous

16
r
dl dr
θ
0
R

Figure 1.10: We push the analogy between the Friedmann coordinates in a closed
Universe, and the parallel projection of a half–sphere onto a plane. The physical
distance dl on the surface of the sphere
p corresponds to the coordinate interval
dr, such that dl = dr/cos θ = dr/ 1 − sin2 θ = dr/ 1 − (r/R)2 , where R is
p

the radius of the sphere. This is exactly the same law as in a closed Friedman
Universe.

expansion. A simple analogy helps in understanding this subtle concept. Let


us take a rubber balloon and draw some points on the surface. Then, we inflate
the balloon. The distances between all the points grow proportionally to the
radius of the balloon. This is not because the points have a proper motion on
the surface, but because all the lengths on the surface of the balloon increase
with time.
The geodesics followed by photons are straight lines in 3-space, but not in
space-time. Locally, they obey to the same relation as in Newtonian mechanics:
c dt = dl, i.e.,
dr2
 
2
c2 dt2 = a2(t) + r 2
(dθ 2
+ sin θ dφ 2
) . (1.16)
1 − kr2
So, on large scales, the distance traveled by light in a given time interval is
not ∆l = c∆t like in Newtonian mechanics, but is given by integrating the in-
finitesimal equation of motion. We can write this integral, taking for simplicity
a photon with constant (θ, φ) coordinates (i.e., the origin is chosen on the tra-
jectory of the photon). Between t1 and t2 , the change in comoving coordinates
for such a photon is given by
Z r2 Z t2
dr c
√ = dt. (1.17)
t1 a(t)
1 − kr 2
r1

This propagation equation of light is extremely important - probably, one of


the two most important of cosmology, together with the Friedmann equation,
that we will give soon. It is with this equation that we are able today to measure
the curvature of the Universe, its age, its acceleration, and other fundamental
quantities.

1.3.5 Bending of light in the expanding Universe


Lets give a few examples of the implications of equation (1.16), which give the
bending of the trajectories followed by photons in our curved space-time, as
illustrated in figure 1.11.

17
to

te θe r
re
Figure 1.11: An illustration of the propagation of photons in our Universe. The
dimensions shown here are (t, r, θ): we skip φ for the purpose of representa-
tion. We are siting at the origin, and at a time t0 , we can see a the light of
a galaxy emitted at (te , re , θe ). Before reaching us, the light from this galaxy
has traveled over a curved trajectory. In any point, the slope dr/dt is given
by equation (1.16). So, the relation between re and (t0 − te ) depends on the
spatial curvature and on the scale factor evolution. The trajectory would be a
straight line in space-time only if k = 0 and a = constant, i.e., in the limit of
Newtonian mechanics in Euclidean space. The ensemble of all possible photon
trajectories crossing r = 0 at t = t0 is called our “past light cone”, visible here
in orange. Asymptotically, near the origin, it can be approximated by a linear
cone with dl = cdt, showing that at small distance, the physics is approximately
Newtonian.

The redshift.
First, a simple calculation based on equation (1.16) - we don’t include it here
- gives the redshift associated with a given source of light. Take two observers
sitting on two galaxies (with fixed comoving coordinates). A light signal is sent
from one observer to the other. At the time of emission t1 , the first observer
measures the wavelength λ1 . At the time of reception t2 , the second observer
will measure the wavelength λ2 such that

δλ λ2 − λ 1 a(t2 )
z= = = −1 . (1.18)
λ λ1 a(t1 )

So, the redshift depends on the variation of the scale-factor between the time
of emission and reception, but not on the curvature parameter k. This can
be understood as if the scale-factor represented the “stretching” of light in out

18
curved space-time. We can check that when the Universe expands (a(t2 ) >
a(t1 )), the wavelengths are enhanced (λ2 > λ1 ), and the spectral lines are
red-shifted (contraction would lead to blue-shift, i.e., to a negative redshift
parameter z < 0).
In real life, what we see when we observe the sky is never directly the dis-
tance, size, or velocity of an object, but the light traveling from it. So, by study-
ing spectral lines, we can easily measure the redshifts. This is why astronomers,
when they refer to an object or to an event in the Universe, generally mention
the redshift rather than the distance or the time. It’s only when the function
a(t) is known - and as we will see later, we almost know it in our Universe -
that a given redshift can be related to a given time and distance.
The importance of the redshift as a measure of time and distance comes
from the fact that we don’t see our full space-time, but only our past line-
cone, i.e., a three-dimensional subspace of the full four-dimensional space-time,
corresponding to all points which could emit a light signal that we would receive
today on Earth. If we remove the coordinate φ and draw the past line-cone
of a given observer, it does look like a cone, of course curved like all photon
trajectories (see figure 1.11).
Also note that in Newtonian mechanics, the redshift was defined as z = v/c,
and seemed to be limited to |z| < 1. The true GR expression doesn’t have such
limitations, since the ratio of the scale factors can be arbitrarily large without
violating any fundamental principle. And indeed, observations do show many
objects - like quasars - at redshifts of 2 or even bigger. We’ll see later that we
also observe the Cosmic Microwave Background at a redshift of approximately
z = 1000!

The Hubble parameter in General Relativity.


In the limit of small redshift, we expect to recover the Newtonian results,
and to find a relation similar to z = v/c = Hr/c. To show this, let’s assume
that t0 is the present time, and that our galaxy is at r = 0. We want to compute
the redshift of a nearby galaxy, which emitted the light that we receive today
at a time t0 − dt. In the limit of small dt, the equation of propagation of light
shows the physical distance between the galaxy and us is simply

dl = c dt (1.19)

while the redshift of the galaxy is

a(t0 ) 1 ȧ(t0 )
z= −1= ȧ(t0 )
−1= dt . (1.20)
a(t0 − dt) 1 − a(t0 ) dt a(t0 )

By combining these two relations we obtain

ȧ(t0 ) dl
z= . (1.21)
a(t0 ) c

So, at small redshift, we recover the Hubble law, and the role of the Hubble
parameter is played by ȧ(t0 )/a(t0 ). In the Friedmann Universe, we will directly
define the Hubble parameter as the expansion rate of the scale factor:

ȧ(t)
H(t) = . (1.22)
a(t)

The present value of H is generally noted H0 :

H0 = 100h km s−1 Mpc−1 , 0.5 < h < 0.8. (1.23)

19
Angular diameter – redshift relation.
When looking at the sky, we don’t see directly the size of the objects, but
only their angular diameter. In Euclidean space, i.e. in absence of gravity, the
angular diameter dθ of an object would be related to its size dl and distance r
through
dl
dθ = . (1.24)
r
Recalling that z = v/c and v = Hr, we easily find an angular diameter – redshift
relation valid in Euclidean space:
H dl
dθ = . (1.25)
cz
In General Relativity, because of the bending of light by gravity, the steps of the
calculation are different. Using the definition of infinitesimal distances (1.13),
we see that the physical size dl of an object is related to its angular diameter
dθ through
dl = a(te ) re dθ (1.26)
where te is the time at which the galaxy emitted the light ray that we observe
today on Earth, and re is the comoving coordinate of the object. The equation
of motion of light gives a relation between re and te :
Z 0 Z t0
−dr c
√ = dt (1.27)
re 1 − kr 2
te a(t)
where t0 is the time today. So, the relation between re and te depends on a(t)
and k.
If we knew the function a(t) and the value of k, we could calculate the
integral (1.27) and re-express equation (1.26) in the form
H dl
dθ = f (z) (1.28)
cz
where f (z) would be a function of redshift, accounting for general relativistic
corrections, and depending on the curvature. We conclude that in the Fried-
mann Universe, for an object of fixed size and redshift, the angular diameter
depends on the curvature - as illustrated graphically in figure 1.12. Therefore,
if we know in advance the physical size of an object, we can simply measure
its redshift, its angular diameter, and immediately obtain some informations on
the geometry of the Universe.
It seems very difficult to know in advance the physical size of a remote object.
However, we will see in the next chapter that some beautiful developments of
modern cosmology enabled physicists to know the physical size of some objects
visible at a redshift of z ' 1000 (namely, the anisotropies of the Cosmological
Microwave Background). So, the angular diameter – redshift relation has been
used in the past decade in order to measure the spatial curvature of the Universe.
We will show the results in the last section of chapter two.

Apparent Luminosity – redshift relation.


We have seen that in Newtonian mechanics, the absolute luminosity of an
object and the apparent luminosity l that we measure on Earth per unit of
surface are related by
L LH 2
l= 2
= . (1.29)
4πr 4πc2 z 2
So, if for some reason we know independently the absolute luminosity of a
celestial body (like for cepheids), and we measure its redshift, we can obtain the
value of H, as Hubble did in 1929.

20
to

dl

te r
re’
re θe
θ’e

Figure 1.12: Angular diameter – redshift relation. We consider an object of fixed


size dl and fixed redshift, sending a light signal at time te that we receive at
present time t0 . All photons travel by definition with θ =constant. However, the
bending of their trajectories in the (t, r) plane depends on the spatial curvature
and on the scale factor evolution. So, for fixed te , the comoving coordinate of
the object, re , depends on curvature. The red lines are supposed to illustrate
the trajectory of light in a flat Universe with k = 0. If we keep dl, a(t) and
te fixed, but choose a positive value k > 0, we know from equation (1.27) that
the new coordinate re 0 has to be smaller. But dl is fixed, so the new angle
dθ0 has to be bigger, as easily seen on the figure for the purple lines. So, in a
closed Universe, objects are seen under a larger angle. Conversely, in an open
Universe, they are seen under a smaller angle.

But we would like to extend this technique to very distant objects (in par-
ticular, supernovae, that are observed up to a redshift of two, and are thought
to have a calibrated absolute luminosity, like cepheids). So, one needs to recom-
pute this relation in the framework of general relativity. Because of the bending
of light rays, the trajectory of the photons traveling from an object located at
a given redshift depends both on the curvature and the scale factor. So, today,
the light will be dispersed on a sphere of variable size.
If we know several objects with the same absolute magnitude, and we plot
them on an apparent magnitude - redshift diagram, we will get a curve whose
shape at small redshift is given by (1.29), and at large redshift by General
Relativity corrections (depending on the spatial curvature and the scale factor.)
We will see in the next chapter that such an experiment has been performed for
supernovae, leading to one of the most intriguing discovery of the past years.

21
In summary of this section, according to General Relativity, the homoge-
neous Universe is curved by its own matter content, and the curvature can be
specified with just one number and one function: the curvature k, and the scale
factor a(t). We should be able to relate these two quantities with what causes
the curvature: the matter density.

1.3.6 The Friedmann law


The Friedmann law relates the scale factor a(t), the spatial curvature parameter
k and the homogeneous energy density of the Universe ρ(t):
 2
ȧ 8πG ρ kc2
= − . (1.30)
a 3 c2 a2

Together with the propagation of light equation, this law is the key ingredient
of the Friedmann-Lemaı̂tre model.
In special/general relativity, the total energy of a particle is the sum of its
rest energy E0 = mc2 plus its momentum energy. So, if we consider only non-
relativistic particles like those forming galaxies, we get ρ = ρmass c2 . Then,
the Friedmann equation looks exactly like the Newtonian expansion law (1.9),
excepted that the function r(t) (representing previously the position of material
points) is replaced by the scale factor a(t). Of course, the equations look the
same, but are not equivalent . First, we have already seen that only at very
small distance, the distinction between the scale factor a(t) and the classical
position r(t) is irrelevant; on large distances – like the Hubble radius – the
difference of interpretation between the two is crucial. Second, we have seen
that the term proportional to k seems to break the homogeneity of the Universe
in the Newtonian formalism, while in General Relativity, when it is correctly
interpreted as the spatial curvature term, it is perfectly consistent with the
Cosmological Principle.
Actually, there is a third crucial difference between the Friedmann law and
the Newtonian expansion law. In the previous paragraph, we only considered
non-relativistic matter like galaxies. But the Universe also contains relativistic
particles traveling at the speed of light, like photons (and also neutrinos if their
mass is very small). A priori, their gravitational effect on the Universe expansion
could be important. How can we include them?

1.3.7 Relativistic matter and Cosmological constant


The Friedmann equation is true for any types of matter, relativistic or non-
relativistic; if there are different species, the total energy density ρ is the sum
over the density of all species.
What depends on the type of matter is the relation between ρ and a, i.e.,
the way in which energy gets diluted by the expansion.
For non-relativistic matter, the answer is obvious. Take a distribution of
particles with fixed comoving coordinates. Their energy density equals their
mass density times c2 . Look only at the matter contained into a comoving
sphere centered around the origin, of comoving radius r. If the sphere is small
with respect to the curvature, the physical volume inside the sphere is just V =
4π 3
3 (a(t)r) . Since both the sphere and the matter particles have fixed comoving
coordinates, no matter can enter or leave from inside the sphere during the
expansion. Therefore, the mass (or the energy) inside the sphere is conserved.
We conclude that ρV is constant and that ρ ∝ a−3 .

22
For ultra–relativistic matter like photons, the energy is not given by the rest
mass but by the frequency ν or the wavelength λ:

E = hν = hc/λ. (1.31)

But we knows that physical wavelengths are stretched proportionally to the


scale factor a(t). So, if we repeat the argument of the sphere, assuming now
that it contains a homogeneous bath of photons with equal wavelength, we
see that the total energy inside the sphere evolves proportionally to a−1 . So,
the energy density of relativistic matter is proportional to ρ ∝ a−4 . If the
scale factor increases, the photon energy density decreases faster than that of
ordinary matter.
This result could be obtained differently. For any gas of particles with a
distribution of velocities, one can define a pressure (corresponding physically to
the shock of particles against the borders if the gas was enclosed into a box).
This can be extended to a gas of relativistic particles, for which the speed equals
the speed of light. A calculation based on statistical mechanics gives the famous
result that the pressure of a relativistic gas is related to its density by p = ρ/3.
In the case of the Friedmann universe, General Relativity provides several
equations: the Friedmann law, and also, for each specie, the equation of conser-
vation:

ρ̇ = −3 (ρ + p) . (1.32)
a
This is consistent with what we already said. For non-relativistic matter, like
galaxies, the pressure is negligible (like in a gas of still particles), and we get


ρ̇ = −3 ρ ⇒ ρ ∝ a−3 . (1.33)
a
For relativistic matter like photons, we get
ȧ 1 ȧ
ρ̇ = −3 (1 + )ρ = −4 ρ ⇒ ρ ∝ a−4 . (1.34)
a 3 a
Finally, in quantum field theory, it is well–known that the vacuum can have an
energy density different from zero. The Universe could also contain this type
of energy (which can be related to particle physics, phase transitions and spon-
taneous symmetry breaking). For vacuum, one can show that p = −ρ. This
means that the vacuum energy is never diluted and remains constant. This con-
stant energy density was called by Einstein – who introduced it in a completely
different way and with other motivations – the Cosmological Constant. We will
see that this term is probably important in our Universe.

23
Chapter 2

The Standard Cosmological


Model

The real Universe is not homogeneous: it contains stars, galaxies, clusters of


galaxies...
In cosmology, all quantities – like the density and pressure of each specie –
are decomposed into a spatial average, called the background, plus some inho-
mogeneities. The later are assumed to be small with respect to the background
in the early Universe: so, they can be treated like linear perturbations. As a
consequence, the evolution of each Fourier mode is independent of the others.
During the evolution, if the perturbations of a given quantity become large, the
linear approximation breaks down, and one has to employ a full non–linear de-
scription, which is very complicated in practice. However, for many purposes in
cosmology, the linear theory is sufficient in order to make testable predictions.
In section 2.1, we will describe the evolution of the homogeneous background.
Section 2.2 will give some hints about the evolution of linear perturbations –
and also, very briefly, about the final non-linear evolution of matter perturba-
tions. Altogether, these two sections provide a brief summary of what is called
the standard cosmological model, which depends on a few free parameters. In
section 2.3, we will show that the main cosmological parameters have already
been measured with quite good precision. Finally, in section 2.4, we will in-
troduce the theory of inflation, which provides some initial conditions both for
the background and for the perturbations in the very early Universe. We will
conclude with a few words on the so-called quintessence models.

2.1 The Hot Big Bang scenario


A priori, we don’t know what type of fluid or particles gives the dominant
contributions to the energy density of the Universe. According to the Friedmann
equation, this question is related to many fundamental issues, like the behavior
of the scale factor, the spatial curvature, or the past and future evolution of the
Universe...

2.1.1 Various possible scenarios for the history of the Uni-


verse
We will classify the various types of matter that could fill the Universe according
to their pressure-to-density ratio. The three most likely possibilities are:

24
1. ultra–relativistic particles, with v = c, p = ρ/3, ρ ∝ a−4 . This
includes photons, massless neutrinos, and eventually other particles that
would have a very small mass and would be traveling at the speed of
light. The generic name for this kind of matter, which propagates like
electromagnetic radiation, is precisely “radiation”.

2. non-relativistic pressureless matter – in general, simply called “matter”


by opposition to radiation – with v  c, p ' 0, ρ ∝ a−3 . This applies
essentially to all structures in the Universe: planets, stars, clouds of gas,
or galaxies seen as a whole.
3. a possible cosmological constant, with time–invariant energy density and
p = −ρ, that might be related to the vacuum of the theory describing
elementary particles, or to something more mysterious. Whatever it is, we
leave such a constant term as an open possibility. Following the definition
given by Einstein, what is actually called the “cosmological constant” Λ
is not the energy density ρΛ , but the quantity

Λ = 8πGρΛ /c2 (2.1)

which has the dimension of the inverse square of a time.

We write the Friedmann equation including these three terms:


 2
ȧ 8πG 8πG kc2 Λ
H2 = = 2
ρR + 2
ρM − 2 + (2.2)
a 3c 3c a 3

where ρR is the radiation density and ρM the matter density. The order in
which we wrote the four terms on the right–hand side – radiation, matter,
spatial curvature, cosmological constant – is not arbitrary. Indeed, they evolve
with respect to the scale factor as a−4 , a−3 , a−2 and a0 . So, if the scale factors
keeps growing, and if these four terms are present in the Universe, there is a
chance that they all dominate the expansion of the Universe one after each other
(see figure 2.1). Of course, it is also possible that some of these terms do not
exist at all, or are simply negligible. For instance, some possible scenarios would
be:
• only matter domination, from the initial singularity until today (we’ll come
back to the notion of Big Bang later).
• radiation domination → matter domination today.
• radiation dom. → matter dom. → curvature dom. today
• radiation dom. → matter dom. → cosmological constant dom. today
But all the cases that do not respect the order (like for instance: curvature
domination → matter domination) are impossible.
During each stage, one component strongly dominates the others, and the
behavior of the scale factor, of the Hubble parameter and of the Hubble radius
are given by:
1. Radiation domination:
ȧ2 1
∝ a−4 , a(t) ∝ t1/2 , H(t) = , RH (t) = 2ct. (2.3)
a2 2t
So, the Universe is in decelerated power–law expansion.

25
2
H

a−4

a−3
−2
a
cst

Figure 2.1: Evolution of the square of the Hubble parameter, in a scenario in


which all typical contributions to the Universe expansion (radiation, matter,
curvature, cosmological constant) dominate one after each other.

2. Matter domination:
ȧ2 2 3
∝ a−3 , a(t) ∝ t2/3 , H(t) = , RH (t) = ct. (2.4)
a2 3t 2
Again, the Universe is in power–law expansion, but it decelerates more
slowly than during radiation domination.
3. Negative curvature domination (k < 0):

ȧ2 1
∝ a−2 , a(t) ∝ t, H(t) = , RH (t) = ct. (2.5)
a2 t
An open Universe dominated by its curvature is in linear expansion.
4. Positive curvature domination: if k > 0, and if there is no cosmological
constant, the right–hand side finally goes to zero: expansion stops. After,
the scale factor starts to decrease. H is negative, but the right–hand side
of the Friedmann equation remains positive. The Universe recollapses. We
know that we are not in such a phase, because we observe the Universe
expansion. But a priori, we might be living in a closed Universe, slightly
before the expansion stops.
5. Cosmological constant domination:

ȧ2 p
→ constant, a(t) ∝ exp(Λt/3), H = c/RH = Λ/3. (2.6)
a2
The Universe ends up in exponentially accelerated expansion.
So, in all cases, there seems to be a time in the past at which the scale factor
goes to zero, called the initial singularity or the “Big Bang”. The Friedmann
description of the Universe is not supposed to hold until a(t) = 0. At some time,
when the density reaches a critical value called the Planck density, we believe

26
that gravity has to be described by a quantum theory, where the classical notion
of time and space disappears. Proposals for such theories exist, called “string
theories”. Sometimes, they try to address the initial singularity problem, and to
build various scenarios for the origin of the Universe. Anyway, this field is still
very speculative, and of course, our understanding of the origin of the Universe
will always break down at some point. A reasonable goal is just to go back as
far as possible, on the basis of testable theories.
The future evolution of the Universe heavily depends on the existence of a
cosmological constant. If the later is exactly zero, then the future evolution
is dictated by the curvature (if k > 0, the Universe will end up with a “Big
Crunch”, where quantum gravity will show up again, and if k ≤ 0 there will be
eternal decelerated expansion). If instead there is a positive cosmological term,
then the Universe necessarily ends up in eternal accelerated expansion.

2.1.2 The matter budget today


In order to know the past and future evolution of the Universe, it would be
enough to measure the present density of radiation, matter and Λ, and also to
measure H0 . Then, thanks to the Friedmann equation, it would be possible to
extrapolate a(t) at any time. Let us express this idea mathematically. We take
the Friedmann equation, evaluated today, and divide it by H02 :
8πG kc2 Λ
1= 2 (ρ R0 + ρ M0 ) − + . (2.7)
3H0 c2 a20 H02 3H02
where the subscript 0 means “evaluated today”. Since by construction, the
sum of these four terms is one, they represent the relative contributions to the
present Universe expansion. These terms are usually noted
8πG
ΩR = ρR0 , (2.8)
3H02 c2
8πG
ΩM = ρM0 , (2.9)
3H02 c2
kc2
Ωk = , (2.10)
a20 H02
Λ
ΩΛ = , (2.11)
3H02
(2.12)

and the “matter budget” equation is

ΩR + ΩM − Ωk + ΩΛ = 1. (2.13)

The Universe is flat provided that

Ω0 ≡ Ω R + Ω M + Ω Λ (2.14)

is equal to one. In that case, as we already know, the total density of matter,
radiation and Λ is equal at any time to the critical density
3c2 H 2 (t)
ρc (t) = . (2.15)
8πG
Note that the parameters Ωx , where x ∈ {R, M, Λ}, could have been defined as
the present density of each specie divided by the present critical density:
ρx0
Ωx = . (2.16)
ρc0

27
So far, we conclude that the evolution of the Friedmann Universe can be de-
scribed entirely in terms of four parameters, called the “cosmological parame-
ters”:
Ω R , Ω M , Ω Λ , H0 . (2.17)
One of the main purposes of cosmology is to measure the value of the cosmo-
logical parameters.

2.1.3 The Cold and Hot Big Bang alternatives


Curiously, after the discovery of the Hubble expansion and of the Friedmann
law, there were no rapid progresses in cosmology for a few decades. The most
likely explanation is that most physicists were not considering seriously the
possibility of studying the Universe in the far past, near the initial singularity,
because they thought that it would always be impossible to test experimentally
any model of this type.
Nevertheless, a few pioneers tried to think about the origin of the Universe.
At the beginning, for simplicity, they assumed that the expansion of the Uni-
verse is always dominated by a single component, the one forming galaxies, i.e.,
pressureless matter. Since going back in time, the density of matter increases
as a−3 , matter had to be very dense at early times. This was formulated as the
“Cold Big Bang” scenario.
According to Cold Big Bang, in the early Universe, the density is so high
that matter has to consist in a gas of nucleons and electrons. Then, when the
density falls below a critical value, some nuclear reactions form the first nuclei -
this era is called nucleosynthesis. But later, due to the expansion, the dilution of
matter is such that nuclear reactions are suppressed (in general, the expansion
freezes out all processes when their characteristic time–scale becomes smaller
than the so–called Hubble time–scale H −1 ). So, only the lightest nuclei have
time to form in a significant amount. After nucleosynthesis, matter consists in a
gas of nuclei and electrons, with electromagnetic interactions. When the density
becomes even smaller, they finally combine into atoms – this second transition
is called the recombination. At late time, any small density inhomogeneity
in the gas of atoms is enhanced by gravitational interactions. The atoms get
accumulated into clumps like stars and planets - but this is a different story.
In the middle of the XX-th century, a few particle physicists tried to build
the first models of nucleosynthesis – the era of nuclei formation. In particular,
four groups – each of them not being aware of the work of the others – reached
approximately the same negative conclusion: in the Cold Big Bang scenario,
nucleosynthesis does not work properly, because the formation of hydrogen is
strongly suppressed with respect to that of heavier elements. But this conclusion
is at odds with observations: using spectrometry, astronomers know that there
is a lot of hydrogen in stars and clouds of gas. The groups of the Russo-
American Gamow in the 1940’s, of the Russian Zel’dovitch (1964), of the British
Hoyle and Tayler (1964), and of Peebles in Princeton (1965) all reached this
conclusion. They also proposed a possible way to reconcile nucleosynthesis with
observations. If one assumes that during nucleosynthesis, the dominant energy
density is that of photons, the expansion is driven by ρR ∝ a−4 , and the rate
of expansion is different. This affects the kinematics of the nuclear reactions is
such way that enough hydrogen can be created.
In that case, the Universe would be described by a Hot Big Bang scenario,
in which the radiation density dominates at early time. Before nucleosynthesis
and recombination, the mean free path of the photons is very small, because
they are continuously interacting – first, with electrons and nucleons, and then,
with electrons and nuclei. So, their motion can be compared with the Brownian

28
2
H
a−4

a−3

equality a now

γ O
pling
decou
esis
synth
nucleo

Figure 2.2: On the top, evolution of the square of the Hubble parameter as a
function of the scale factor in the Hot Big Bang scenario. We see the two stages
of radiation and matter domination. On the bottom, an idealization of a typical
photon trajectory. Before decoupling, the mean free path is very small due to the
many interactions with baryons and electrons. After decoupling, the Universe
becomes transparent, and the photon travels in straight line, indifferent to the
surrounding distribution of electrically neutral matter.

motion in a gas of particles: they form what is called a “black–body”. In any


black–body, the many interactions maintain the photons in thermal equilibrium,
and their spectrum (i.e., the number density of photons as a function of wave-
length) obeys to a law found by Planck in the 1890’s. Any “Planck spectrum”
is associated with a given temperature.
Following the Hot Big Bang scenario, after recombination, the photons do
not see any more charged electrons and nuclei, but only neutral atoms. So,
they stop interacting significantly with matter. Their mean free path becomes
infinite, and they simply travel along geodesics – excepted a very small fraction
of them which interacts accidentally with atoms, but since matter gets diluted,
this phenomenon rapidly becomes negligible. So, essentially, the photons travel
freely from recombination until now, keeping the same energy spectrum as they
had before, i.e., a Planck spectrum, but with a temperature that decreases with
the expansion. This is an effect of General Relativity: the wavelength of an
individual photon is proportional to the scale factor; so the shape of the Planck
spectrum is conserved, but the whole spectrum is shifted in wavelength. The
temperature of a black–body is related to the energy of an average photon with
average wavelength: T ∼< E >= hc/ < λ >. So, the temperature decreases like
1/ < λ >, i.e., like a−1 (t).
The physicists that we mentioned above noticed that these photons could
still be observable today, in the form of a homogeneous background radiation
with a Planck spectrum. Following their calculations – based on nucleosynthesis

29
– the present temperature of this cosmological black–body had to be around a
few Kelvin degrees. This would correspond to typical wavelengths of the order
of one millimeter, like microwaves.

2.1.4 The discovery of the Cosmic Microwave Background


These ideas concerning the Hot Big Bang scenario remained completely un-
known, excepted from a small number of theorists.
In 1964, two American radio–astronomers, A. Penzias and R. Wilson, de-
cided to use a radio antenna of unprecedented sensitivity – built initially for
telecommunications – in order to make some radio observations of the Milky
Way. They discovered a background signal, of equal intensity in all directions,
that they attributed to instrumental noise. However, all their attempts to elim-
inate this noise failed.
By chance, it happened that Penzias phoned to a friend at MIT, Bernard
Burke, for some unrelated reason. Luckily, Burke asked about the progresses
of the experiment. But Burke had recently spoken with one of his colleagues,
Ken Turner, who was just back from a visit Princeton, during which he had
followed a seminar by Peebles about nucleosynthesis and possible relic radiation.
Through this series of coincidences, Burke could put Penzias in contact with
the Princeton group. After various checks, it became clear that Penzias and
Wilson had made the first measurement of a homogeneous radiation with a
Planck spectrum and a temperature close to 3 Kelvins: the Cosmic Microwave
Background (CMB). Today, the CMB temperature has been measured with
great precision: T0 = 2.728 K.
This fantastic observation was a very strong evidence in favor of Hot Big
Bang scenario. It was also the first time that a cosmological model was checked
experimentally. So, after this discovery, more and more physicists realized that
reconstructing the detailed history of the Universe was not purely science fiction,
and started to work in the field.
The CMB is present in our everyday life: fortunately, it is not as powerful
as a microwave oven, but when we look at the background noise on the screen
of a TV set, one fourth of the power comes from the CMB!

2.1.5 The Thermal history of the Universe


Between the 1960’s and today, a lot of efforts have been made for studying with
increasing precision the various stages of the Hot Big Bang scenario. Today,
some epochs in the history of Universe are believed to be well understood, and
are confirmed by observations; some other remain very speculative.
For the earliest stages, there are still many competing scenarios – depending,
for instance, on what string theory looks like. Following the most conventional
picture, gravity becomes a classical theory (with well–defined time and space
dimensions) at a time called the Planck time 1 : t ∼ 10−36 s, ρ ∼ (1018 GeV)4 .
Then, there would be a stage of inflation (see section 2.4), possibly related
with GUT (Grand Unified Theory) symmetry breaking at t ∼ 10−32 s, ρ ∼
(1016 GeV)4 . The EW (electroweak) symmetry breaking would occur at t ∼
10−6 s, ρ ∼ (100 GeV)4 . The generation of the baryons out of some more
fundamental particles would be related to one of these two symmetry breakings
(there are various possible scenarios). Then, at t ∼ 10−4 s, ρ ∼ (100 MeV)4 , the
quarks would combine into hadrons.
1 By convention, the origin of time is chosen by extrapolating the scale-factor to a(0) = 0.

Of course, this is only a convention, it has no physical meaning.

30
After these stages, we enter into a series of phase transitions that are much
better understood, and well constrained by the observations. These are:
1. at t = 1 − 100 s, T = 109 − 1010 K, ρ ∼ (0.1 − 1 MeV)4 , nucleosynthesis:
formation of light nuclei, in particular hydrogen, helium and lithium. By
comparing the theoretical predictions with the observed abundance of light
elements in the present Universe, it is possible to give a very precise esti-
mate of the total density of baryons in the Universe: ΩB h2 = 0.020±0.002.
2. at t ∼ 104 yr, T ∼ 104 K, ρ ∼ (1 eV)4 , the radiation density equals the
matter density: the Universe goes from radiation domination to matter
domination.
3. at t ∼ 105 yr, T ∼ 2500 K, ρ ∼ (0.1 eV)4 , recombination of atoms and
decoupling of photons. After that time, the Universe is transparent: the
photons free–stream along geodesics. So, by looking at the CMB, we
obtain a picture of the Universe at decoupling. Nothing has changed in
the distribution of photons between 105 yr and today, excepted for an
overall redshift of all wavelengths (implying ρ ∝ a−4 , and T ∝ a−1 ).
4. after recombination, the small inhomogeneities of the smooth matter dis-
tribution are amplified. This leads to the formation of stars and galaxies,
as we shall see later.
The success of the Hot big bang Scenario relies on the existence of a radiation–
dominated stage followed by a matter–dominated stage. However, an additional
stage of curvature or cosmological constant domination is not excluded.

2.1.6 A recent stage of curvature or cosmological constant


domination?
If today, there was a very large contribution of the curvature and/or cosmological
constant to the expansion (ΩM  |Ωk | or ΩM  |ΩΛ |), the deviations from
the Hot Big Bang scenario would be very strong and incompatible with many
observations. However, nothing forbids that Ωk and/or ΩΛ are of order one. In
that case, they would play a significant part in the Universe expansion only in
a recent epoch (typically, starting at a redshift of one or two). Then, the main
predictions of the conventional Hot Big Bang scenario would be only slightly
modified. For instance, we could leave in a close or open Universe with |Ωk | ∼
0.5: in that case, there would be a huge radius of curvature, with observable
consequences only at very large distances.
In both cases Ωk = O(1) and ΩΛ = O(1), the recent evolution of a(t) would
be affected (as clearly seen from the Friedmann equation). So, there would be
a modification of the apparent magnitude–redshift relation, and of the angular
diameter–redshift relation. We will see in section 2.3 how this has been used in
recent experiments.
Another consequence of a recent k or Λ domination would be to change the
age of the Universe. Our knowledge of the age of the Universe comes from the
measurement of the Hubble parameter today. Within a given cosmological sce-
nario, it is possible to calculate the function H(t) (we recall that by convention,
the origin of time is such that a → 0 when t → 0). Then, by measuring the
present value H0 , we obtain immediately the age of the Universe t0 . The result
does not depend very much on what happens in the early Universe. Indeed,
equality takes place very early, at 104 yr. So, when we try to evaluate roughly
the age of the Universe in billions of years, we are essentially sensitive to what
happens after equality.

31
In a matter–dominated Universe, H = 2/(3t). Then, using the current
estimate of H0 , we get t0 = 2/(3H0 ) ∼ 9 Gyr. If we introduce a negative
curvature or a positive cosmological constant, it is easy to show (using the
Friedmann equation) that H decreases more slowly in the recent epoch. In that
case, for the same value of H0 , the age of the Universe is obviously bigger. For
instance, with ΩΛ ' 0.9, we would get t0 ∼ 18 Gyr.
We leave the answer to these intriguing issues for section 2.3.

2.1.7 Dark Matter


We have many reasons to believe that the non-relativistic matter is of two kinds:
ordinary matter, and dark matter. One of the well-known evidences for dark
matter arises from galaxy rotation curves.
Inside galaxies, the stars orbit around the center. If we can measure the
redshift in different points inside a given galaxy, we can reconstruct the dis-
tribution of velocity v(r) as a function of the distance r to the center. It is
also possible to measure the distribution of luminosity I(r) in the same galaxy.
What is not directly observable is the mass distribution ρ(r). However, it is
reasonable to assume that the mass distribution of the observed luminous mat-
ter is proportional to the luminosity distribution: ρlum (r) = b I(r), where b is
an unknown coefficient of proportionality called the bias. From this, we can
compute the gravitational potential Φlum generated by the luminous matter,
and the corresponding orbital velocity, given by ordinary Newtonian mechanics:

ρlum (r) = b I(r), (2.18)


∆Φlum (r) = 4πG ρlum (r), (2.19)
2 ∂
vlum (r) = r Φlum (r). (2.20)
∂r

So, vlum (r) is known up to an arbitrary normalization factor b. However, for
many galaxies, even by varying b, it is impossible to obtain a rough agreement
between v(r) and vlum (r) (see figure 2.3). The stars rotate faster than expected
at large radius. We conclude that there is some non–luminous matter, which
deepens the potential well of the galaxy.
Apart from galactic rotation curves, there are many arguments – of more cos-
mological nature – which imply the presence of a large amount of non–luminous
matter in the Universe, called dark matter. For various reasons, it cannot con-
sist in ordinary matter that would remain invisible just because it is not lighten
up. Dark matter has to be composed of particle that are intrinsically uncoupled
with photons – unlike ordinary matter, made up of baryons. Within the stan-
dard model of particle physics, a good candidate for non-baryonic dark matter
would be a neutrino with a small mass. Then, dark matter would be relativistic
(this hypothesis is called Hot Dark Matter or HDM). However, HDM is excluded
by some types of observations: dark matter particles have to be non-relativistic,
otherwise galaxy cannot form during matter domination. Non–relativistic dark
matter is generally called Cold Dark Matter (CDM).
There are a few candidates for CDM in various extensions of the standard
model of particle physics: for instance, some supersymmetric partners of gauge
bosons (like the neutralino or the gravitino), or the axion of the Peccei-Quinn
symmetry. Despite many efforts, these particles have never been observed di-
rectly in the laboratory. This is not completely surprising, given that they are
– by definition – very weakly coupled to ordinary particles.
In the following, we will decompose ΩM in ΩB + ΩCDM . This introduces one
more cosmological parameter. With this last ingredient, we have described the

32
r
o

redshift luminosity

v(r) v (r)
lum
b

r r
Figure 2.3: A sketchy view of the galaxy rotation curve issue. The genuine
orbital velocity of the stars is measured directly from the redshift. From the
luminosity distribution, we can reconstruct the orbital velocity under the as-
sumption that all the mass in the galaxy arises form of the observed luminous
matter. Even by varying the unknown normalization parameter b, it is impos-
sible to obtain an agreement between the two curves: their shapes are different,
with the reconstructed velocity decreasing faster with r than the genuine ve-
locity. So, there has to be some non–luminous matter around, deepening the
potential well of the galaxy.

main features of the Standard Cosmological Model, at the level of homogeneous


quantities. We will now focus on the perturbations of this background.

2.2 Cosmological perturbations


2.2.1 Linear perturbation theory
All quantities, such as densities, pressures, velocities, curvature, etc., can be
decomposed into a spatial average plus some inhomogeneities:

ρx (t, ~r) = ρ̄x (t) + δρx (t, ~r). (2.21)

We know that the CMB temperature is approximately the same in all directions
in the sky. This proves that in the early Universe, at least until a redshift
z ∼ 1000, the distribution of matter was very homogeneous. So, we can treat the
inhomogeneities as small linear perturbations with respect to the homogeneous

33
background. For a given quantity, the evolution becomes eventually non–linear
when the relative density inhomogeneity

δρx (t, ~r)


δx (t, ~r) = (2.22)
ρ̄x (t)

becomes of order one.


In linear perturbation theory, it is very useful to make a Fourier transforma-
tion, because the evolution of each Fourier mode is independent of the others:
Z
~ ~
δx (t, k) = d3~r e−ik.~r δx (t, ~r). (2.23)

Note that the Fourier transformation is defined with respect to the comoving
coordinates ~r. So, k is the comoving wavenumber, and 2π/k the comoving
wavelength. The physical wavelength is

2πa(t)
λ(t) = . (2.24)
k
This way of performing the Fourier transformation is the most convenient one,
leading to independent evolution for each mode. It accounts automatically for
the expansion of the Universe. If no specific process occurs, δx (t, ~k) will remain
constant, but each wavelength λ(t) is nevertheless stretched proportionally to
the scale factor.
Of course, to a certain extent, the perturbations in the Universe are ran-
domly distributed. The purpose of the cosmological perturbation theory is not
to predict the individual value of each δx (t, ~k) in each direction, but just the
spectrum of the perturbations at a given time, i.e., the root–mean–square of all
δx (t, ~k) for a given time and wavenumber, averaged over all directions. This
spectrum will be noted simply as δx (t, k).
The full linear theory of cosmological perturbation is quite sophisticated.
Here, we will simplify considerably. The most important density perturbations
to follow during the evolution of the Universe are those of photons, baryons,
CDM, and finally, the curvature perturbations. As far as curvature is con-
cerned, the homogeneous background is fully described by the Friedmann ex-
pression for infinitesimal distances, i.e., by the scale factor a(t) and the spatial
curvature parameter k. We need to define some perturbations around this aver-
age homogeneous curvature. The rigorous way to do it is rather technical, but
at the level of this course, the reader can admit that the most relevant curvature
perturbation is a simple function of time and space, which is almost equivalent
to the usual Newtonian gravitational potential Φ(t, ~r) (in particular, at small
distance, the two are completely equivalent).

2.2.2 The horizon


Now that we have seen which types of perturbations need to be studied, let us
try to understand their evolution. In practice, this amounts in integrating a
complicated system of coupled differential equations, for each comoving Fourier
mode k. But if we just want to understand qualitatively the evolution of a given
mode, the most important is to compare its wavelength λ(t) with a characteristic
scale called the horizon. Let us define this new quantity.
Suppose that at a time t1 , two photons are emitted in opposite directions.
The horizon dH at time t2 relative to time t1 is defined as the physical distance
between the two photons at time t2 (see figure 2.4). In order to compute it, we

34
t dH

t2 r2 r

γ γ

t1
Figure 2.4: The horizon dH (t1 , t2 ) is the physical distance at time t2 between
two photons emitted in opposite directions at time t1 .

note that if the origin r = 0 coincides with the point of emission, the horizon
can be expressed as
Z r2 Z r2
dr
dH (t1 , t2 ) = 2 dl = 2 a(t2 ) √ . (2.25)
0 0 1 − kr2
But the equation of propagation of light, applied to one of the photons, also
gives Z t2 Z r2
c dt dr
= √ . (2.26)
t1 a(t) 0 1 − kr2
Combining the two equations, we get
t2
c dt
Z
dH (t1 , t2 ) = 2a(t2 ) . (2.27)
t1 a(t)

Physically, since no information can travel faster than light, the horizon rep-
resents the maximal scale on which a process starting at time t1 can have an
impact at time t2 . In the limit in which t1 is very close to the initial singularity,
dH represents the causal horizon, i.e. the maximal distance at which too points
in the Universe can be in causal contact. In particular, perturbations with wave-
lengths λ(t2 ) > dH (t1 , t2 ) are acausal, and cannot be affected by any physical
process – if they were, it would mean that some information has traveled faster
than light.

During radiation domination, when a(t) ∝ t1/2 , it is straightforward to

35
compute
1/2 1/2 1/2
dH (t1 , t2 ) = 4 c t2 [t2 − t1 ]. (2.28)
In the limit in which t1  t2 , we find

dH (t1 , t2 ) → 4 c t2 = 2 RH (t2 ). (2.29)

So, during radiation domination, the causal horizon coincides with the Hubble
radius.

During matter domination, when a(t) ∝ t2/3 , we get


2/3 1/3 1/3
dH (t1 , t2 ) = 6 c t2 [t2 − t1 ]. (2.30)

When t1  t2 this becomes

dH (t1 , t2 ) → 6 c t2 = 4 RH (t2 ). (2.31)

So, even during a matter–dominated stage following a radiation–dominated


stage, the causal horizon remains of the same order of magnitude as the Hubble
radius.

Since each wavelength λ(t) = 2πa(t)/k evolves with time at a different rate
than the Hubble radius RH (t), a perturbation of fixed wavenumber k can be
super–horizon (λ  RH ) at some initial time, and sub–horizon (λ  RH ) at
a later time. This is illustrated on figure 2.5. While the scale factor is always
decelerating, the Hubble radius grows linearly: so, the Fourier modes enter one
after each other inside the causal horizon. The modes which enter earlier are
those with the smallest wavelength.
What about the perturbations that we are able to measure today? Can they
be super–horizon? In order to answer this question, we first note that the largest
distance that we can see today in our past–light–cone is the one between two
photons emitted at decoupling, and reaching us today from opposite direction
in the sky. Because the Universe was not transparent to light before decoupling,
with current technology, we cannot think of any observation probing a larger
distance.
So, in order to compute the maximal wavelength accessible to observation,
we can make a short calculation, quite similar to that of the horizon, but with
reversed photon trajectories: the two photons are emitted at (t1 , r1 ), and they
reach us today at (t0 , 0) from opposite directions. The propagation–of–light
equation gives
Z t0 Z 0
c dt −dr
= √ . (2.32)
t1 a(t) r1 1 − kr2
But today, two points located at coordinate r1 in opposite directions are sepa-
rated by the physical distance
Z r1 Z r1 Z t0
dr c dt
2 dl = 2 a(t0 ) √ = 2a(t0 ) = dH (t1 , t2 ). (2.33)
0 0 1 − kr 2
t1 a(t)

We conclude that the causal horizon evaluated today gives an upper bound on
the distances accessible to any observation (whatever the time of decoupling
is). Our knowledge of the Universe is limited to a finite sphere of physical
radius RH (t0 ). So, all observable cosmological perturbations are sub–horizon
today, while in the past, they were super–horizon (see figure 2.5). The largest

36
d RH

λα a

t eq t

Figure 2.5: The wavelengths of some cosmological perturbations compared with


the Hubble radius, during radiation domination and matter domination. Be-
cause all wavelengths λ(t) = 2πa(t)/k grow with negative acceleration, and
RH (t) grows linearly, the modes enter one after each other inside the horizon.
The smallest perturbations enter during radiation domination, the others dur-
ing matter domination. The largest cosmological perturbation observable today
(in CMB observations) has λ(t0 ) = RH (t0 ).

observable scale has entered inside the horizon very recently, with a wavelength
approximately equal to

c 3 × 108 m s−1
RH (t0 ) = ' ' 5 × 103 Mpc. (2.34)
H0 70 km s−1 Mpc−1

All these general considerations will allow us to understand qualitatively the


evolution of the different perturbations, and especially those of photons and
CDM, which are the most interesting. We will try to visualize the evolution
on a three-dimensional diagram, like the one of figure 1.6. This representation
gives the perturbation amplitude δx (t, k) for a given specie x. In other words, it
gives the evolution of each Fourier mode with respect to time. The purple line
shows the limit between causal (sub–horizon) perturbations on the right, and
acausal (super–horizon) perturbations on the left.

2.2.3 Photon perturbations


In good approximation, the photons keep a Planck spectrum during all the
relevant evolution. As a consequence, the photon density perturbations δγ (t, k)
can be reformulated as some temperature fluctuations around the mean black–
body temperature: δT /T̄ (t, k) (where T̄ is the spatial average of the temperature
at time t). The evolution is very different before and after the time of photon
decoupling.
Indeed, before decoupling, photons are strongly coupled to baryons through
electromagnetic interactions, and have a vanishing mean–free–path. As a con-
sequence, the baryons and the photons are almost indistinguishable: they share

37
perturbation

lar k
ge radiation domination

sm
wa

all
v

wa
ele

AC

v
ng

AU

ele
th

ng
AL

CA

th
t eq

US
AL
matter domination

t 2πa
RH

Figure 2.6: For any specie, this type of diagram can be used for understanding
qualitatively the evolution of the Fourier spectrum as a function of time. Each
Fourier mode of fixed comoving wavenumber k enters into the causal horizon
when it crosses the purple line (which corresponds to λ(t) = RH (t), i.e., to
k = 2πa(t)/RH (t)).

the same density perturbations, and can be viewed as single fluid. In addi-
tion, since baryons have a mass, they feel gravitational forces: so, the entire
baryon–photon fluid is coupled with the gravitational potential.
So, in each well of gravitational potential, the baryon–photon fluid tends to
be compressed by gravity, but the relativistic photon pressure p = 31 ρ tends to
resist to this compression. The dynamics of the fluid is governed by two antago-
nist forces, as for any oscillator. So, for each Fourier mode, any departure from
equilibrium – such as a non–zero initial perturbation – will lead to oscillations.
These oscillations propagate through the fluid exactly like sound waves in the
air: due to this analogy, the perturbations in the baryon–photon fluid before
decoupling are called acoustic oscillations.
Of course, the oscillations can occur only for causal wavelengths: because of
causality, wavelengths bigger than the Hubble radius do not feel any force. So, a
given mode with a given non–zero initial amplitude starts oscillating only when
it enters into the horizon (see figure 2.7). The initial amplitude for each mode
at initial time is given by the primordial spectrum of perturbations. Of course, if
the Universe was perfectly homogeneous from the very beginning (with no initial
perturbations), there would be no acoustic oscillations – and no formation of
structure either, so we must assume some primordial perturbations. At this

38
γ perturbation ( δ T/T )

decoupling

t 2πa
RH

Figure 2.7: Evolution of the Fourier spectrum of photon temperature fluctu-


ations, as a function of time. We start from a flat primordial spectrum, as
predicted by inflation and as confirmed by observations. When a given Fourier
mode enters inside the horizon (i.e., crosses the dashed purple line), it starts
to oscillate. At decoupling, the temperature distribution is frozen (later, all
temperatures are redshifted by a factor 1/a(t), so the relative fluctuation δT / T̄
remains constant). So, the Fourier spectrum measured today is the same as at
the time of decoupling. It exhibits a flat plateau for wavelengths bigger than the
Hubble radius at decoupling, and then a series of peaks called acoustic peaks.

stage, the primordial spectrum has to be assumed arbitrarily. Only in the last
section, we will see a precise mechanism called “inflation” leading to a given
primordial spectrum. On figure 2.7, we start from a flat spectrum, independent
of k – in good approximation, this corresponds to what is observed today, and
also to what is predicted by the theory of inflation.
When the photons decouple from the matter, their mean–free–path grows
rapidly from zero ton infinity, and they travel in straight line with essentially no
interaction. During this free-streaming, all their characteristics – momentum,
energy, wavelength, spectrum, temperature – do not evolve, excepted under the
effect of homogeneous expansion. So, δT and T̄ are both shifted proportionally
to 1/a(t), but the distribution of δT /T̄ (t, k) observed today is exactly the same
as at the time of decoupling. Observing the temperature anisotropies in the sky
amounts exactly in taking a picture of the Universe at the time of decoupling.
So, in order to know the present Fourier spectrum of temperature fluctua-

39
tions, it is enough to follow the acoustic oscillations until decoupling. As eas-
ily understood from figure 2.7, the Fourier modes which are still outside the
Hubble radius at decoupling have no time to oscillate. Smaller wavelengths
have enough time for one oscillation, even smaller ones for two oscillations, etc.
The Fourier spectrum of temperature fluctuations consists in a flat plateau for
k < 2πa(tdec)/RH (tdec ), and then a first peak, a second peak, a third peak, etc.
Note that according to this model, the perturbations oscillate, but they
keep the same order of magnitude (there is no large amplification). Therefore,
if we start from small primordial perturbations, we are able to describe all the
evolution of temperature fluctuations with the linear theory of perturbations.

2.2.4 Observing the CMB anisotropies


The measurement of the CMB by Penzias and Wilson showed no evidence for
perturbations – in fact, we should say anisotropies, because we can only map
the temperature along different directions. However, some CMB anisotropies
must exist, because our Universe is not homogeneous today. The current in-
homogeneities have to grow from fluctuations in the past, and at the time of
decoupling there must be already at least some small fluctuations in the den-
sities, and therefore, in the temperature. The order of magnitude of CMB
anisotropies was predicted many years before being measured. By extrapolat-
ing from the present inhomogeneous structure back to the time of decoupling,
many cosmologists in the 80’s expected δT /T̄ to be at least of order 10−6 –
otherwise, clusters of galaxies could not have formed today.
Many experiments were devoted to the detection of these anisotropies. The
first successful one was COBE-DMR, an American satellite carrying an inter-
ferometer of exquisite sensitivity. In 1992, COBE mapped the anisotropies all
over the sky, and found an average amplitude δT /T̄ ∼ 10−5 (see figure 2.8).
This was in perfect agreement with the theoretical predictions – another big
success for cosmology. The COBE experiment had an angular resolution of a
few degrees: so, anisotropies seen under one degree or less were smoothed by the
detector. In a Fourier decomposition, it means that COBE could only measure
the spectrum of wavelengths larger than the Hubble radius at decoupling. So,
it was not probing the acoustic oscillations, but only the flat plateau.
COBE was the first experimental evidence in favor of the linear theory of cos-
mological perturbations. The next step was to increase the angular resolution,
in order to probe smaller wavelengths, check the existence of the acoustic peaks,
and compare them with the theory. In the 90’s, there were some considerable
efforts in this direction.
First, on the theoretical side, the spectrum of CMB anisotropies was inves-
tigated with greater and greater precision. Some people developed numerical
codes computing the exact spectrum for any choice of cosmological parameters.
Indeed, the details of the peaks depend on many quantities. For instance, the
amplitude of the acoustic oscillations is governed by the density of baryons and
photons. The whole final shape of the spectrum depends on the primordial spec-
trum of perturbations. Last but not least, the position of the peaks depends
on the Hubble radius at decoupling. Fortunately, the theory is able to predict
the physical size of RH (tdec ), and the redshift of decoupling – so, by measuring
the angular size associated to the wavelength of the acoustic peaks, it should be
possible to use the angular diameter–redshift relation (see section 1.3.5), and to
measure the curvature of the Universe! We conclude that if the acoustic peaks
could be measured, the comparison with the theoretical spectrum would reveal
the values of the most crucial cosmological parameters.
On the experimental side, a decisive progress was made with Boomerang,

40
Figure 2.8: The first genuine “picture of the Universe” at the time of decou-
pling, 100 000 years after the initial singularity, and 15 billion years before
the present epoch. Each blue (resp. red) spot corresponds to a slightly colder
(resp. warmer) region of the Universe at that time. This map, obtained by the
American satellite COBE in 1992, covers the entire sky: so, it pictures a huge
sphere centered on us. The fluctuations are only of order 10−5 with respect to
the average value T0 = 2.728 K. They are the “seeds” for the present structure
of the Universe: each red spot corresponds to a small overdensity of photons
and baryons at the time of decoupling, that has been enhanced later, leading to
galaxies and clusters of galaxies today.

a US–Italian–Canadian balloon, carrying some detectors called bolometers. In


2001, Boomerang published the map of figure 2.9. It focuses on a small patch
of the sky, but with much better resolution than COBE (a few arc–minutes).
The Fourier decomposition of the Boomerang map clearly shows the first three
acoustic peaks (see figure 2.10). The data points are perfectly fitted with a
theoretical curve obtained numerically, for a simple choice of cosmological pa-
rameters (figure 2.11).
At the beginning of this year, the NASA satellite WMAP published a full-
sky CMB map (see figure 2.12); it is the second one after that of COBE, but
with a resolution increased by a factor 30 (see figure 2.13). The corresponding
Fourier spectrum is extremely precise on the scale of the first and second acoustic
peaks, and perfectly compatible with the theory (see figure 2.14). This beautiful
agreement with some predictions made several decades before the experiment is
one of the major successes of cosmology. By fitting theoretical curves to this
data, it is possible to measure various cosmological parameters with already a
good precision. For instance, the position of the first peak is strongly in favor of
k = 0, i.e., a spatially flat Universe. We will come back to these measurements
in section 2.3.

41
Figure 2.9: The map of CMB anisotropies obtained by the balloon experiment
Boomerang in 2001. Unlike COBE, Boomerang only analyzed a small patch of
the sky, but with a much better angular resolution of a few arc–minutes. The
black (resp. yellow) spots correspond to colder (resp. warmer) regions.

There are still many CMB experiments going on. The satellite WMAP
goes on acquiring data in order to reach even better resolution. In 2007, the
European satellite “Planck” will be launched. Planck is expected to perform
the best possible measurement of the CMB temperature anisotropies, with such
great precision that most cosmological parameters should be measured with
only one per cent uncertainty.

2.2.5 Matter perturbations


We now focus on the evolution of the matter perturbations. Because we believe
that the Universe contains much more CDM than baryons, in first approxima-
tion, we can forget the baryons, and do as if all the matter was in the form of
CDM, with absolutely no electromagnetic interactions. This is crucial, because
in absence of any coupling with the photons, the CDM fluid has a vanishing
pressure: so, it tends to concentrate in each well of gravitational potential –
instead of oscillating like the baryons before decoupling. This is called gravita-
tional collapse.
Of course, as all causal mechanisms, gravitational collapse can only affect
modes inside the Hubble radius. However, there is one more restriction. During
matter domination, when a potential well appears, CDM concentrates in it, and
deepens the gravitational potential even more. So, both CDM and gravitational
potential perturbations are quickly amplified. On the other hand, during radi-
ation domination, the gravitational potential follows the perturbations of the
dominant energy density: that of radiation. So, there is no significant coupling

42
Figure 2.10: The Fourier decomposition of the Boomerang map reveals the
structure of the first three acoustic oscillations. These data points account for
the spectrum of (δT )2 in (µK)2 – we know from COBE that δT is roughly of
order 10−5 T0 ' 30 µK. The spectrum is not shown as a function of wavenumber
k, but of multipole number l. This is because the anisotropies are not observed
in 3-dimensional space, but on a two-dimensional sphere: the sphere of redshift
z ∼ 1000 centered on us. For purely geometric reasons, a spherical map is
not expandable in Fourier modes, but in Legendre multipoles. However, the
interpretation is the same. The first, second and third peak correspond to the
modes that could oscillate one, two or three times between their entry into the
Hubble radius and the time of decoupling.

between the CDM inhomogeneities and the gravitational potential. As a conse-


quence, the amplification of CDM perturbations is efficient only after equality
and inside RH . If we look at the diagram of figure 2.6, this corresponds to
the region on the right of the purple curve, and below the yellow line. So,
even today, the Fourier spectrum of CDM perturbations has a characteristic
scale corresponding to the Hubble radius at equality. The same type of numer-
ical code which compute the spectrum of CMB anisotropies can also compute
the linear matter spectrum with great precision, for any choice of cosmological
parameters.
But linear perturbation theory is valid only as long as the perturbations
are small, δCDM (t, k)  1. After, it is no longer interesting to use a Fourier
decomposition, because the evolution of each mode is not independent of the
others: so, everything becomes much more complicated, and we can make some
predictions only from numerical simulations of the gravitational collapse in real
space. These are called an N-body simulations.
It is crucial to know at what time a given mode enters into the non-linear
regime, δCDM (t, k) ∼ 1. If there was absolutely no growth of CDM perturba-

43
Figure 2.11: The spectrum of the Boomerang map is perfectly fitted by the
theoretical predictions. For a given choice of cosmological parameters – such as
ΩB , ΩCDM , ΩΛ , ΩR , H0 , the parameters describing the primordial spectrum,
etc.– the theoretical calculation is performed by a numerical code of typically
2000 lines. By comparing with the experimental data, one can extract the
parameter values, as we will see in section 2.3.

tions inside the Hubble radius during radiation domination, this time would be
the same for all small wavelengths, because they would grow at the same rate,
starting from the same value. But because there is a very slow amplification of
perturbations during this epoch, the most amplified wavelength are the smallest
ones – since they entered earlier into the Hubble radius. So, the smallest wave-
lengths are the first ones to become non-linear. As a consequence, the smallest
structures in the Universe are the older ones: this is called hierarchical structure
formation.

2.2.6 Hierarchical structure formation


Let’s go back to equality, when no efficient amplification of the matter pertur-
bations has started: the matter density in the Universe is still a continuous
function, with small inhomogeneities of order 10−5 . Then, when the smallest
wavelengths become non-linear (δCDM (t, k) ∼ 1), some small compact clumps
start to form. But at that time, if we smooth the density distribution with a
smoothing radius corresponding to the largest non-linear scale, we find again
a continuous distribution (remember the analogy in figure 1.4), with a Fourier
spectrum obeying to the predictions of the linear theory of perturbations.
As time goes on, some larger and larger scales become non-linear and start to
collapse. So, some small overdense regions (protogalaxies) are expected to form
first; then, they merge into bigger and bigger galaxies, and later clusters of galax-

44
Figure 2.12: The full-sky map of CMB anisotropies obtained by the satellite
WMAP in 2003. The blue (resp. red) spots correspond to colder regions. This
map has been cleaned from any foreground contamination, including that of the
Milky Way: it only shows the CMB contribution.

ies will also appear. Finally, in the present Universe, the scale of non-linearity
is expected to be around 20 or 40 Mpc. This is one order of magnitude bigger
than clusters of galaxies: so, even the clusters are not smoothly distributed. In
fact, they tend to organize themselves along huge filaments, separated by voids
of size 20 to 40 Mpc. The structure of the Universe on these huge scales looks
very much like a spounge... But by smoothing the present distribution of matter
over scales larger than 40 Mpc, one finds again a continuous distribution, with
small inhomogeneities described by the linear Fourier spectrum of cosmological
perturbation theory.

2.2.7 Observing the matter spectrum


So, if we have a large three–dimensional map of the galaxy distribution in the
Universe, we can smooth it on very large scales, compare with the linear spec-
trum predicted by the theory, and measure the cosmological parameters on
which this spectrum depends. Of course, what we observe in the sky is only lu-
minous baryonic matter, not CDM. But astrophysicists have very good reasons
to believe that on large scales, the distribution of luminous matter “traces” the
distribution of mass. This means that the inhomogeneities in luminous mat-
ter would be proportional to the total density inhomogeneities – i.e., in good
approximation, to the CDM inhomogeneities. The unknown coefficient of pro-
portionality is called the bias b:

δluminous (t, k) ' b δCDM (t, k). (2.35)

Over the past decades, astronomers could build some very large three-
dimensional maps of the galaxy distribution – even larger than the scale of
non-linearity. On figure 2.15, we can see two thin “slices” of the surrounding
Universe. Inside these slices, each point corresponds to a galaxy seen by the
“2dF Galaxy Redshift Survey”. The radial coordinate is the redshift, or the
distance between the object and us. Even at redshift z < 0.1, we can see some

45
Figure 2.13: Comparison between the COBE and WMAP anisotropy maps. In
both figure, the red stripe in the middle is the foreground contamination from
the Milky Way. The orientation of the two maps is the same in order to allow
for direct visual comparison. The resolution of WMAP is 30 times better, but
it is clear that on large angular scale, the two maps reveal the same structure.

darker regions, corresponding to the voids - the largest non-linear structures in


the Universe. At z > 0.1, the average density seems to decrease smoothly, but
this is simply because the survey is limited in sensitivity – at small distance
it sees all the galaxies, while at large distance it sees only the brightest ones.
By taking this selection effect into account, and by smoothing over distances of
order of 50 Mpc, one recovers a very homogeneous distribution of matter.
On figure 2.16, we see the Fourier spectrum reconstructed from such a red-
shift survey. Each experimental point is obtained by smoothing over smaller
wavelengths. The data points with k < 0.2h Mpc−1 can be directly compared
with the linear spectrum predicted by the linear theory of cosmological pertur-
bations. The red line shows one of the best-fit theoretical curves. From such a
comparison, one can extract some information on the value of the cosmological
parameters. This is similar to the parameter extraction from CMB anisotropies,

46
Figure 2.14: The spectrum of WMAP map is perfectly fitted by the theoretical
predictions, for a given choice of cosmological parameters – namely, for the
model showed here, ΩB , ΩCDM , ΩΛ , the so-called “optical depth to reionization”
(a quantity not defined in this course), and the two parameters describing a
power-law primordial spectrum.

but it provides some independent constraints.


On the extreme left of figure 2.16, we see that the theoretical curve reaches
a maximum (around k ∼ 0.01h Mpc−1 , i.e., λ ∼ 400 Mpc). This corresponds to
the scale which entered into the Hubble radius at the time of equality between
radiation and matter. Future redshift surveys are expected to give better data
around this scale. Observing the turnover of the matter spectrum with good
precision will improve the measurement of the cosmological parameters.

2.3 Measuring the cosmological parameters


There has been some continuous progress in measuring the cosmological param-
eters over the past three decades; however, the most spectacular improvements
have been made in the last five years. Many techniques can be employed for
constraining the cosmological scenarios – probably around twenty. Instead of
giving an exhaustive review, we will focus only on five important methods.

2.3.1 Abundance of primordial elements


In sections 2.1.3 and 2.1.5, we have seen that the theory of Nucleosynthesis can
predict the abundance of light elements formed in the early Universe, when the
energy density was of order ρ ∼ (1MeV)4 . This theory has reached today a
great level of precision: it is believed that each relevant nuclear reaction are
taken carefully into account. In the standard model of Nucleosynthesis, there
are only two free parameters: the baryon density ρB , or equivalently, ΩB h2 ;

47
Figure 2.15: The distribution of galaxies in two thin slices of the neighboring
Universe, obtained by the 2dF Galaxy Redshift Survey. The radial coordinate
is the redshift, or the distance between the object and us.

and the number of neutrino families Nν . For each value of the baryon density
and of Nν , the theory predicts a given abundance of hydrogen, helium, lithium,
etc. So, by observing the abundance of these elements in the Universe – for
instance, using spectroscopy – it is possible to constrain the baryon density.
Current observations lead to the prediction

ΩB h2 = 0.020 ± 0.002. (2.36)

Moreover, they are perfectly compatible with Nν = 3, which is known from


accelerator physics to be the number of neutrino families.

2.3.2 CMB anisotropies


The precise dynamics of the acoustic oscillations before decoupling depends on
almost all the cosmological parameters: so, the spectrum of CMB anisotropies
can give a wide variety of constraints. This type of experiment is considered as

48
Figure 2.16: The Fourier spectrum of matter perturbations, reconstructed from
a large galaxy redshift survey (here, the Sloan Digital Sky Survey). The data
are compared with a theoretical spectrum, obtained by integrating numerically
over a system of differential equations, accounting for the evolution of linear
perturbations. In the calculation, the cosmological parameters were chosen in
order to obtain a maximum likelihood fit to the data. Comparing the data
with the linear spectrum is meaningful only for the largest wavelengths, k <
0.2h Mpc−1 , which have not been affected by the non–linear evolution. For
smaller scales, the agreement between the data and the theoretical prediction is
only a coincidence.

the most complete and precise cosmological test. Future CMB experiments like
MAP and Planck should be able to measure each parameter with great precision.
At the moment, some particular parameters or combinations of parameters are
already measured unambiguously, using the CMB anisotropy map of WMAP
(and also of other experiments like ACBAR and CBI). One example is the
baryon density, which is related in an interesting way (that we will not detail
here) to the relative amplitude of the acoustic peaks. The WMAP data gives
ΩB h2 = 0.022 ± 0.001. (2.37)
This constraint is as precise as that from Nucleosynthesis, and they are in perfect
agreement with each other – which is really striking, since these two measure-
ments are done with completely different techniques: one uses nuclear physics,

49
applied to the Universe when it had a density of order (1 Mev)4 , while the other
uses fluid mechanics, applied to a much more recent era with density (0.1 eV)4 .
The perfect overlap between the two constraints indicates that our knowledge
of the Universe is impressively good, at least during the epoch between t ∼ 1 s
and t ∼ 100 000 yr.
We have seen that the position of the acoustic peak is given by the Hubble
radius at decoupling, which value is predicted by the theory. So, using the
angular diameter–redshift relation (see section 1.3.5), it is possible to extract
some information both on the spatial curvature and on the evolution of the
scale factor a(t). A detailed study shows that the leading effect is that of
spatial curvature. So, the position of the acoustic peak gives a measurement of
Ωk = 1 − Ω0 . The result is

Ω0 = 1.03 ± 0.05. (2.38)

So, according to current CMB experiments, the Universe seems to be spatially


flat. This result is extremely useful, but it doesn’t say how the present energy
density is distributed between matter and a possible cosmological constant2 .
Current CMB experiments also give other interesting constraints (for in-
stance, on the primordial spectrum of perturbations) that we will not explain
here.

2.3.3 Age of the Universe


Observations can set a reliable lower bound on the age of the Universe: if an
object is seen at a given redshift, corresponding to a given age, the Universe has
to be older than this object. During the 1990’s, it has been established that the
most distant observable quasars are at least 11 Gyr old. On the other hand, a
spatially flat matter–dominated Universe would be only 9 Gyr old (see section
2.1.6). So, distant quasars seem to indicate that the Universe is either open or
Λ–dominated today.

2.3.4 Luminosity of Supernovae


The evidence for a non–flat Universe and/or a non–zero cosmological constant
has increased considerably in 1998, when two independent groups studied the
apparent luminosity of distant type Ia supernovae (SNIa). For this type of
supernovae, astronomers believe that there is a simple relation between the
absolute magnitude and the luminosity decay rate. In other words, by studying
the rise and fall of the luminosity curve during a few weeks, one can deduce the
absolute magnitude of a given SNIa. Therefore, it can be used in the same way
as cepheid, as a probe of the apparent luminosity–redshift relation. In addition,
Supernovae are much brighter that cepheids, and can be observed at much
larger distances (until redshifts of order one). While observable cepheids only
probe short distances, where the apparent luminosity–redshift relation only gives
the Hubble law (the proportionality between distance and redshift), the most
distant observable SNIa’s are in the region where general relativity corrections
are important: so, they can provide a measurement of the spatial curvature and
of the scale factor evolution.
On figure 2.17, the various curves represent the apparent magnitude–redshift
relation, computed for various choices of ΩM and ΩΛ . For a given value of H0 ,
all the curves are asymptotically equal at short distance. Significant differences
show up only at redshifts z > 0.2. Each red data point corresponds to a single
2 As far as radiation is concerned, we know that today, it is negligible with respect to
matter, so that Ω0 ' ΩM + ΩΛ = ΩB + ΩCDM + ΩΛ .

50
(Further back in time)
FAINTER
(Farther)
Perlmutter, et al. (1998) (ΩΜ,ΩΛ) =
26 ( 0, 1 )
(0.5,0.5) (0, 0)
( 1, 0 ) (1, 0)
24 (1.5,–0.5) (2, 0)
Supernova

Λ=0
Flat
Cosmology
Project
22
effective mB

20
Calan/Tololo
(Hamuy et al,
A.J. 1996)
18

16

14
0.02 0.05 0.1 0.2 0.5 1.0
redshift z MORE REDSHIFT
(More total expansion of universe
since the supernova explosion)

In flat universe: ΩM = 0.28 [± 0.085 statistical] [± 0.05 systematic]


Prob. of fit to Λ = 0 universe: 1%

Figure 2.17: The results published by the “Supernovae Cosmology Project” in


1998. The various curves represent the apparent magnitude–redshift relation,
computed for various choices of ΩM and ΩΛ . The solid black curves account for
three examples of a closed/flat/open Universe with no cosmological constant.
The dashed blue curves correspond to three spatially flat universes with different
values of ΩΛ . For a given value of H0 , all the curves are asymptotically equal
at short distance, probing only the Hubble law. The yellow points are short–
distance SNIa’s: we can check that they are approximately aligned. The red
points, at redshifts between 0.2 and 0.9, show that distant supernovae are too
luminous to be compatible with a flat matter-dominated Universe (ΩM , ΩΛ )
=(1,0).

supernovae, from the 1998 results of the “Supernovae Cosmology Project”. Even
if it is not very clear visually from the figure, a detailed statistical analysis reveals
that a flat matter–dominated Universe (with ΩM = 1, ΩΛ = 0) is excluded. This
result has been confirmed by more recent data at larger redshift. According to

51
Figure 2.18: Combined constraints on (ΩM , ΩΛ ) from CMB and Supernovae
experiments. The simple case of a spatially flat Universe with no cosmological
constant corresponds to the lower right corner of the plot. In blue, the CMB
allowed region favors a spatially flat Universe with ΩM + ΩΛ ' 1. In orange, the
Supernovae constraint is orthogonal to the CMB one. By combining the two
results, one obtains the favored region limited by the black ellipses, centered
near (ΩM , ΩΛ ) = (0.3, 0.7).

statistical analysis, the allowed region in the (ΩM , ΩΛ ) parameter space is


−1 < ΩM − ΩΛ < 0. (2.39)
When this observation is combined with the CMB results (figure 2.18), it
appears that our Universe is spatially flat and has a non–zero cosmological
constant – a case corresponding to accelerated expansion today! The favored
values are (ΩM , ΩΛ ) = (0.3,0.7).

2.3.5 Large Scale Structure


We have seen that the Fourier spectrum of matter perturbations on very large
scales (figure 2.16) depends on the cosmological parameters. Current galaxy
redshift surveys provide some constraints which are essentially independent from
those of the previous sections. It is remarkable that they converge toward the
same values as the previous techniques.

52
In conclusion, current experiments agree on the following matter budget in
the present Universe:
Ω0 = ΩB + ΩCDM + ΩΛ ' 1
ΩΛ ' 0.73
ΩB ' 0.04
So, the Universe seems to be accelerating today, and 96% of its energy density
is non–baryonic and of essentially unknown nature! The measured value of
the cosmological constant shows that the Universe started to be Λ–dominated
only recently: at redshifts z ≥ 2, the Universe was still matter–dominated. We
seem to be leaving precisely during the period of transition between matter–
domination and Λ–domination!

2.4 The Inflationary Universe


2.4.1 Problems with the Standard Cosmological Model
According to the modern version of the standard cosmological model, called
the ΛCDM model, the Universe is spatially flat, with three stages: radiation–
domination before z ∼ 104 , matter–domination, and now, transition to Λ–
domination. This scenario is impressively successful in explaining a wide va-
riety of observations, as we could see in section 2.3. However, it also raises a
few problems.

The flatness problem.


The present contribution of the spatial curvature to the expansion is given by
Ωk = ρ0 (t0 )/ρc (t0 ) − 1. But this contribution has evolved with time. From the
Friedmann law, it is straightforward to see that
c2 |k| c2 |k|
|Ωk (t)| = |ρ0 (t)/ρc (t) − 1| = = . (2.40)
a2 H 2 ȧ2
During the domination of radiation (resp. matter), the factor ȧ−2 grows like
t (resp. t2/3 ). Therefore, a spatially flat Universe is unstable: if initially, k
is exactly zero, Ωk will remain zero; but if it is only close to zero, Ωk will
increase with time! In order to have no significant curvature–domination today
(|Ωk (t0 )| ≤ 0.1), we find that near the Planck time, the curvature was incredibly
small, roughly |Ωk (t)| ≤ 10−60 ... The standard cosmological model provides no
explanation for obtaining such a small number in the early Universe: the fact
that the Universe is spatially flat – even approximately – appears like the most
unnatural possibility! This is called the “flatness problem”.

The horizon problem.


We have seen that during radiation and matter domination, the causal horizon is
equal to the Hubble radius (times a factor 2 or 4). From the observation of CMB
anisotropies, we know that the Hubble radius at the time of decoupling is seen
today under an angle of one degree in the sky. So, on CMB maps, two points
seen under an angle bigger than one degree seem to be be causally disconnected.
In other words, the CMB maps seem to be composed of approximately 103
independent regions, which could not exchange any information in the past.
But then, how can they be at the same temperature – up to 10−5 ? Given these
theoretical arguments, we would rather expect each of these regions to have its
own temperature. Then, today we would observe inhomogeneities of order one.
Measuring approximately the same temperature T0 = 2.73 K all over the sky
appears like a big mystery called the “horizon problem”.

53
The origin of fluctuations.
The horizon problem can be reformulated in a slightly different way. We have
seen on figure 2.5 that during radiation and matter domination, the Hubble
radius (and the causal horizon) grow faster than the physical wavelength of
each perturbation. So, all the wavelengths observable today on cosmological
scales were larger than the horizon in the early Universe. So far, we did not
propose any mechanism for the generation of primordial perturbations. But we
have the following problem: if primordial perturbations are acausal, how can
they be generated without violating causality (i.e., the fact that no information
can travel faster than the speed of light)?

The Cosmic Coincidence problem.


We have seen that the cosmological constant starts to dominate the energy
density of the Universe only in the very near past. Why are we leaving just at
1/4
the time of the transition? If ΩΛ is of order one, then ρΛ is of order 10−3 eV.
Any vacuum energy density resulting from a spontaneous symmetry breaking
in the standard model of particle physics should be at least of the order of the
electroweak scale, near 1 TeV. Eventually, some unknown symmetry could force
the vacuum energy to vanish exactly. But why should it be as small as 10−3 eV?
So, particle physics seems to be easily compatible with a Universe that would
be either completely matter-dominated or completely Λ–dominated today. But
the subtle balance observed between ordinary matter and the cosmological con-
stant appears as extremely unnatural and mysterious.

2.4.2 An initial stage of inflation


So, the standard cosmological model raises a series of problem – the previous
list is not exhaustive. We will see that many of them can be solved in an elegant
way.
Let us assume that in the early Universe, before radiation domination, there
has been a stage of accelerating expansion: ä > 0. So far, we don’t say which
mechanism could be responsible for this period, called inflation. If ä is positive,
ȧ increases, and according to equation (2.40), |Ωk | decreases. The contribution
of the spatial curvature to the expansion is driven to zero. This provides a
possible solution to the flatness problem: during the initial inflationary stage,
|Ωk | becomes extremely small. Then, it grows again, but today it might still be
very far from order one.
The simple relation that we found between the causal horizon and the Hub-
ble radius is not true anymore during inflation. As a particular example of
accelerated expansion, let us assume that a(t) grows exponentially:

a(t) = eHi t . (2.41)

In that case, the Hubble radius is constant:


 −1
ȧ c
RH = c = , (2.42)
a Hi
while the causal horizon grows exponentially:
Z t2
c dt c H(t2 −t1 )
dH (t1 , t2 ) = 2a(t2 ) −→ e when t 2  t1 . (2.43)
t1 a(t) H i

So, at the end of an inflation, the causal horizon is much bigger than the Hubble
radius. After, during radiation and matter domination, they grow at the same
rate, and the causal horizon remains larger. So, an initial stage of inflation can

54
d
RH

λα a

inflation radiation matter t

Figure 2.19: Comparison of the Hubble radius with the physical wavelength
of a few cosmological perturbations. During the initial stage of accelerated
expansion (called inflation), the Hubble radius grows more slowly than each
wavelength. So, cosmological perturbations find their origin inside RH . Then,
each mode exits from the Hubble radius during inflation and re-enters during
radiation or matter domination.

solve the horizon problem: at the time of decoupling, if the causal horizon is
at least on thousand times larger than the Hubble radius, it encompasses the
whole observable Universe. Then, all the points in the CMB map are in causal
contact, and may have acquired the same temperature in the early Universe.
In the inflationary scenario, it is important not to make confusion between
the different causal horizons. An horizon is always defined with respect to an
initial time. We have seen that if the horizon is defined with respect to the
initial singularity, or to the beginning of inflation, then it can be very large.
But the horizon defined with respect to any time after the end of inflation
is still close to RH . So, the causal distance for a physical process starting
after inflation is unchanged. In particular, the characteristic scale that appears
in the evolution of cosmological perturbations is still the Hubble radius. In
inflationary cosmology, one must be very careful because there are two different
causal horizons, which are relevant for different problems.
Let’s come back finally to the origin of fluctuations. We can update
figure 2.5 by adding a preliminary stage of inflation. In general, the ratio of an
arbitrary physical wavelength over the Hubble radius is given by

λ(t) 2πa(t) ȧ(t) 2π ȧ(t)


= = (2.44)
RH (t) k c a(t) ck

So, under the simple assumption that ä > 0, we see that this ratio increases
during inflation: each wavelength grows faster than the Hubble radius. We
conclude that each Fourier mode observable today on cosmological scales started
inside the Hubble radius, crossed it during inflation, and re-entered later (see
figure 2.19). In addition, all these wavelengths could have been smaller than
the causal horizon at any time. This helps a lot in finding a realistic mechanism
for the generation of fluctuations, as we will see in the next section.

All these arguments – and also some other ones – led a few cosmologists like
Guth, Starobinsky and Hawking to introduce the theory of inflation in 1979.
Today, it is one of the corner stones of cosmology.

55
2.4.3 Scalar field inflation
So, the first three problems of section 2.4.1 can be solved under the assumption
of a long enough stage of accelerated expansion in the early Universe. How can
this be implemented in practice?
First, by combining the Friedman equation (1.30) in a flat Universe with the
conservation equation (1.32), it is easy to find that

ä > 0 ⇒ ρ + 3p < 0. (2.45)

What type of matter corresponds to such an unusual relation between density


and pressure? A positive cosmological constant can do the job:

pΛ = −ρΛ ⇒ ρΛ + 3pΛ = −2ρΛ < 0. (2.46)

But since a cosmological constant is... constant, it cannot be responsible for


an initial stage of inflation: otherwise this stage would go on forever, and there
would be no transition to radiation domination.
Let us consider instead the case of a scalar field (i.e., a field of spin zero,
represented by a simple function of time and space, and invariant under Lorentz
transformations). In general, for a homogeneous scalar field ϕ(t), the density
and pressure read
1 2
ρ = ϕ̇ + V (ϕ), (2.47)
2
1 2
p = ϕ̇ − V (ϕ), (2.48)
2
where V is the scalar potential. So, the pressure is equal to the kinetic energy
minus the potential energy. We see that in the limit – called slow–roll – where
the kinetic energy is subdominant, the pressure is close to −ρ, and the field
behaves almost like a cosmological constant. However, the field is still slowly
rolling toward the minimum of its potential – which should have a small slope,
otherwise the kinetic energy would be leading.
So, a particular way to obtain a stage of accelerated expansion in the early
Universe is to introduce a scalar field, with a flat enough potential. Scalar field
inflation has been proposed in 1979 by Guth. During the 80’s, most important
aspects of inflation were studied in details, by Guth, Starobinsky, Hawking,
Linde and other people. Finally, during the 90’s, many ideas and models were
proposed in order to make contact between inflation and particle physics. The
purpose of scalar field inflation is not only to provide a stage of accelerated
expansion in the early Universe, but also, a mechanism for the generation of
matter and radiation particles, and another mechanism for the generation of
primordial cosmological perturbations. Let us summarize how it works in a
very sketchy way.

Slow-roll.
First, let us assume that just after the initial singularity, the energy density
is dominated by a scalar field, with a potential flat enough for slow–roll. In any
small region where the field is approximately homogeneous and slowly–rolling,
accelerated expansion takes place: this small region becomes exponentially large,
encompassing the totality of the present observable Universe. Inside this region,
the causal horizon becomes much larger than the Hubble radius, and any initial
spatial curvature is driven almost to zero – so, some of the main problems
of the standard cosmological model are solved. After some time, when the
field approaches the minimum its potential, the slow-roll condition breaks, and
inflation ends: the expansion becomes decelerated again (this is not true for all

56
potentials, it depends on additional assumptions; for instance, with any simple
potential of the form V (ϕ) = ϕ2n , slow–roll has to break in a vicinity of the
minimum).

Reheating.
At the end of inflation, the kinetic energy of the field is bigger than the
potential energy; in general, the field is quickly oscillating around the minimum
of the potential. According to the laws of quantum field theory, the oscillating
scalar field will decay into fermions and bosons. This could explain the origin
of all the particles of matter and radiation. Because the particles generally end
up with a thermal distribution, this stage is called “reheating”.

Generation of primordial perturbations.


Finally, the theory of scalar field inflation also explains the origin of cosmo-
logical perturbations – the ones leading to CMB anisotropies and large scale
structure formation. Using again quantum field theory in curved space–time, it
is possible to compute the amplitude of the small quantum fluctuations of the
scalar field. The physical wavelength of these fluctuations grows quickly, like in
figure 2.19. So, they are initially inside the Hubble radius, where quantum me-
chanics is important, and imposes the initial normalization of each mode; then,
they grow larger than the Hubble radius. The perturbations of the scalar field
also generate some perturbations of the space–time curvature – in simple words,
of the gravitational potential. When the wavelength is bigger than the Hubble
radius, the perturbations of the gravitational potential remain frozen: they are
not affected by the decay of the scalar field during reheating. But when the
radiation and matter particles are formed during reheating, they are sensitive
to the gravitational potential, and more particles accumulate in the potential
wells. So, the gravitational potential behaves like a mediator between the scalar
field perturbations during inflation and the radiation/matter perturbations in
the radiation/matter–dominated Universe. Trough this mechanism, the primor-
dial perturbations of the photons, the baryons and the CDM originate from the
quantum fluctuations of the scalar field at the beginning of inflation.
Of course, the perturbations that we observe today are not of quantum
nature, and we don’t need to employ wave functions and operators when we
compute the evolution CMB anisotropies. This is consistent with another nice
property of inflation, which can be deduced rigorously from the laws of quan-
tum field theory, applied to the inflationary space-time in accelerated expansion.
During inflation, and after exiting the Hubble radius, each perturbation under-
goes a quantum–to–classical transition. In other words, quantum mechanics is
unavoidable in order to describe them initially, but later, they become gradually
equivalent to random classical fluctuations.
It is far beyond the level of these notes to compute the evolution of primor-
dial perturbations during inflation. However, we should stress that it can be
studied mathematically, in a very precise way. This leads to generic predictions
for the primordial perturbations, and in particular, for their Fourier spectrum,
which is related to a few free parameters – like the derivatives of the scalar po-
tential. The biggest success of inflation is that the CMB anisotropies observed
for instance by Boomerang are in perfect agreement with the predictions of
inflation concerning the primordial perturbations (in technical words, they are
coherent, Gaussian, adiabatic, and approximately scale–invariant: we will not
explain these concepts here). A priori, such a nice agreement between theory
and observations was far from obvious. It is possible to build some mechanisms
with an initial stage of accelerated expansion, which solve the flatness and hori-
zon problem, but fail in explaining the spectrum of primordial perturbations.

57
Current observations of CMB anisotropies are the best evidence in favor of in-
flation, and with future CMB experiments, it should be possible to reconstruct
the shape of the inflationary potential to some extent; that would be really fan-
tastic: some observations made today of the Universe at decoupling would give
us details about what happened just after the initial singularity – at energies of
order ρ1/4 ∼ 1016 GeV !

Inflation is still a very active field of research, especially as far as the con-
nection with particle physics is concerned. First, the scalar field invoked for
inflation is ad hoc, it is introduced just for cosmological purposes. There are
several proposals concerning the role of this field in particle physics model. For
instance, it could be a Higgs boson associated with the GUT symmetry, or it
could have to do with extra dimensions in string theory. Some people try to
think how these assumptions could be tested, at least in some indirect way
(e.g., through the shape of the primordial spectrum of perturbations). Second,
the theory of reheating is still being investigated in details, in a more an more
realistic way from the point of view of particle physics.

2.4.4 Quintessence ?
It is a pity that the theory of inflation, which is able to solve so many crucial is-
sues in cosmology, doesn’t say anything about the Cosmic Coincidence problem.
Inflation predicts that ΩM + ΩΛ = 1, in agreement with what is observed today,
but it doesn’t give any hint on the splitting between matter and cosmological
constant – while the hint from particle physics is either ΩΛ = 0 or ΩΛ = 1, at
odds with observations.
As long as we explain the supernovae data with a cosmological constant, it
appears as an incredible coincidence that ρΛ is of the same order of magnitude
as ρM just in the present Universe, while in the early Universe it was incredibly
smaller...
We have seen that during slow-roll inflation, the pressure of the scalar field
is negative, almost like for a cosmological content. In the past five years, many
people have suggested to explain the supernovae data with a slow-rolling scalar
field. So, there would be two stages of scalar field inflation, one in the early
Universe, and one today. A priori, they would be caused by two different scalar
fields, called respectively the inflaton and the quintessence field (although some
model try to unify the two). The big advantage with these models is that the
energy density responsible for the cosmic acceleration today doesn’t have to
be constant in the past. In the early Universe, it could very well be of the
same order of magnitude as that of radiation, and only slightly smaller. During
radiation domination, the density of quintessence and radiation would decay
at the same rate, and after radiation–matter equality, quintessence would enter
into a stage of slow–roll. Then, its energy density would remain almost constant,
and it would start to dominate today, leading to the acceleration of the Universe.
Such classes of models are under investigation. However, they have their
own problems, and they don’t solve the Cosmic Coincidence problem in a fully
convincing way. In fact, a wide variety of models have been proposed in order
to explain the supernovae data, but none of them succeeds in explaining while
the Universe starts to accelerate only in the very near past. Together with the
nature of dark matter, the nature of the cosmological term is the biggest enigma
of modern cosmology – and these two questions are far from being small details,
since altogether, Λ and CDM account for 96% of the present energy density in
the Universe! So, cosmology has done some very impressive progresses in the
past decades, but it seems that many other exciting discoveries could be waiting

58
for us in the near future!

Acknowledgments
I would like to thank all the participants of the school for their stimulating
interest, disturbing questions, and interesting comments!

59

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy