Está en la página 1de 20

Notes on Inflation

Ben Craps

Theoretische Natuurkunde, Vrije Universiteit Brussel, and


International Solvay Institutes, Pleinlaan 2, B-1050 Brussels, Belgium

Ben.Craps@vub.be

ABSTRACT

These notes complement lectures on the theory of inflation and its implications for obser-
vations of the cosmic microwave background, taught as part of the course “Early Universe
Cosmology”, Spring 2018.
1 Introduction
The purpose of these lectures is to give a coherent discussion of how inflation can be
realized using a slowly rolling scalar field, how quantum fluctuations give rise to cosmological
perturbations, and how these perturbations are visible in the cosmic microwave background.
I will introduce many of the central concepts and the key equations governing them. Time
constraints will not allow me to derive all key equations from first principles, but I will
very often point the reader to the relevant sections of references where more details can be
found. While I do not expect you to study all missing derivations in detail, I do hope that
the references will help satisfy your curiosity if you want to have a better idea of where the
equations come from.
The main reference for this course on “Early Universe Cosmology” as a whole is Dodelson
[1], and I will mostly use his notation and conventions; in this reference, interested readers
will find many of the computations that are behind the equations we will discuss. Part of
the lecture on inflation is inspired by the brief, informal discussion in Senatore [2], which
nicely emphasizes some of the key conceptual points. For the link with observations, a
similar role is played by Lesgourgues [3]. Some background material on Gaussian random
fields is based on Mukhanov [4]. Both Baumann [5] and Mukhanov offer much more detailed
discussions of inflationary perturbations than we can cover in this course.
These notes are aimed to complement the lectures, by providing an overview of the main
topics that are covered, specific references to the literature, additional background material
and sometimes more details. (They are not intended to be exhaustive or to replace the
lectures.) An example of a point I treat more thoroughly than Dodelson is the role of
the action in the computation of inflationary quantum fluctuations, the reason being that
this allows us to determine canonical commutation relations and thereby the normalization
of the quantum fluctuations from first principles. (The classical equations of motion are
insufficient for this.)
Following Dodelson, in these notes an overdot will denote a derivative with respect to
conformal time η. Derivatives with respect to the time t are written explicitly. In our units,
c = ~ = 1.

2 Inflation
2.1 Reminder: the horizon and flatness problems of standard big bang cos-
mology
[Dodelson, sections 6.2 and 6.3.1; Senatore, section 1.2]
See the previous lectures.

2.2 Slow-roll inflation and slow-roll parameters


[Dodelson, sections 6.3.2 and 6.3.3]
Inflation corresponds to accelerated expansion of the universe. The second Friedmann

1
equation
d2 a/dt2 4πG
=− (ρ + 3P ) (1)
a 3
implies that the expansion of a universe dominated by matter (P = 0) or radiation (P =
ρ/3) always decelarates. During inflation, the universe must have been dominated by a
type of energy density satisfying P < −ρ/3. Vacuum energy has the equation of state
P = −ρ and leads to accelerated expansion (for ρ > 0). While the present-day stage
of accelerated expansion observed in our universe may be explained by vacuum energy
(or equivalently a cosmological constant), the energy density driving inflation in the early
universe was not vacuum energy – otherwise, inflation would not have ended. Therefore, we
need a dynamical form of energy that leads to accelerated expansion for a limited amount
of time and then decays away. The simplest inflationary models use a scalar field, called
the “inflaton”. Fundamental scalar fields play an important role in particle physics models,
most notably in models of spontaneous symmetry breaking. Readers unfamiliar with scalar
fields are referred to the appendix to this section for an introduction.
The starting point for discussing a scalar field in a curved spacetime is the action

Z  
4 1 µν
S = d x −g − g ∂µ φ∂ν φ − V (φ) , (2)
2

where g is the determinant of the metric, φ(xµ ) is the scalar field and V (φ) is its potential.
The energy-momentum tensor is obtained by varying the action with respect to the inverse
metric:
2 δS 1
Tµν ≡ − √ µν
= ∂µ φ∂ν φ − gµν g ρσ ∂ρ φ∂σ φ − gµν V (φ). (3)
−g δg 2
For a homogeneous scalar field (∂i φ = 0, i = 1, 2, 3) in a Friedmann-Lemaı̂tre-Robertson-
Walker (FLRW) spacetime, which we take to be spatially flat for simplicity (the curvature
quickly becomes unimportant after inflation starts),

ds2 = −dt2 + a2 (t)dx2 , (4)

the energy-momentum tensor (3) is that of a perfect fluid with


 2
1 dφ
ρ = + V (φ),
2 dt
 2
1 dφ
P = − V (φ). (5)
2 dt

The first Friedmann equation reads


"  2 #
8πG 1 dφ
H2 = V (φ) + , (6)
3 2 dt

2
with H = (da/dt)/a. The equation of motion for the scalar field is obtained by extremizing
the action (2). For a homogeneous scalar field in the spacetime (4), we find

d2 φ dφ
2
+ 3H = −V 0 (φ). (7)
dt dt
From (5), we see that the condition for accelarated expansion, P < −ρ/3, is equivalent with
(dφ/dt)2 < V (φ). So a universe dominated by a scalar field undergoes inflation whenever
the potential energy dominates. In particular, for a “slowly rolling” scalar field satisfying
 2

 V (φ), (8)
dt

H is approximately constant, so that

a(t) ≈ eHt (9)

as long as the scalar field is slowly rolling. The condition that the field be slowly rolling
requires the potential to be sufficiently flat. Eventually, the slow-roll phase will end and
the potential energy of the inflaton φ will be converted into particles making up the matter
and radiation of the subsequent standard cosmological evolution.
To quantify slow roll, one introduces the slow roll parameters
 
d 1
≡ (10)
dt H

and
1 d2 φ/dt2
δ≡ , (11)
H dφ/dt
which are both small in most models of inflation. The parameter δ is often denoted by
η, but in Dodelson’s (and our) conventions the latter symbol is used for the conformal
time. The slow roll parameters are very important because crucial features of inflation
(including deviations from a scale-invariant spectrum of perturbations, and the production
of gravitational waves) are proportional to them. To lowest order, they can be expressed
in terms of the potential and its derivatives:
 0 2
1 V 1 V 00
= , δ =− . (12)
16πG V 8πG V

2.2.1 Appendix: action principle for classical scalar fields


First consider a system with N degrees of freedom, for which we introduce generalized coor-
dinates q(t) ≡ {q1 (t), q2 (t), . . . , qN (t)}. Classical trajectories extremize an action functional
Z t2
S[q(t)] = dt L(t, q(t), q̇(t)). (13)
t1

3
In this appendix, in which no confusion with conformal time is possible (it coincides with
t in Minkowski space) an overdot simply denotes a derivative with respect to t. According
to the Euler-Lagrange equation, the extremality condition δS = 0 is equivalent to

∂L(t, q, q̇) d ∂L(t, q, q̇)


− = 0. (14)
∂q dt ∂ q̇
Note that in general the first order variation of a functional can be written as
Z
δS
δS = δq(t) dt, (15)
δq(t)

where δS/δq(t) is called a functional derivative. For the action functional S[q], we can
therefore write
δS ∂L(t, q, q̇) d ∂L(t, q, q̇)
= − . (16)
δq(t) ∂q dt ∂ q̇
A scalar field φ(x, t) describes an infinite number of degrees of freedom: q(t) ≡ {φx (t)}.
An example is a free scalar field with mass m in Minkowski space-time, described by the
action
Z Z
1 4 µν 2 2 1
S[φ] = d x [−η ∂µ φ∂ν φ − m φ ] = dt d3 x [φ̇2 − (∇φ)2 − m2 φ2 ]. (17)
2 2
The corresponding equation of motion is
δS
0= = φ̈(x, t) − ∆φ(x, t) + m2 φ(x, t). (18)
δφ(x, t)

2.3 Cosmological perturbations


[This section is partly based on Mukhanov’s book [4].]
Historically, the motivation for inflation was its ability to solve certain problems of
the standard big bang cosmology, as we have discussed before. Today, however, the most
important feature of inflation is that it produces interesting spectra of density perturbations
and gravitational waves. Even though inflation tends to “wash away” inhomogeneities,
quantum fluctuations ensure that the resulting spacetime is not perfectly homogeneous.
In particular, a nearly scale-invariant spectrum of density perturbations is predicted. Not
only do these density perturbations have the right properties to grow into the observed
large scale structure of our present universe (as will be discussed later in the course), they
also predict small anisotropies in the cosmic microwave background that are in spectacular
agreement with observations. Similarly, inflation predicts a spectrum of gravitational waves
that might leave imprints on the polarization spectra of the CMB – whether these effects
will be observable depends on the specific models of inflation, though. Before turning to
the mechanism that produces nearly scale invariant density perturbations in an inflationary
universe, we first discuss how to characterize perturbations.

4
One way to describe inhomogeneities is by specifying energy density perturbations δρ/ρ
(or related quantities such as the gravitational potential) as a function of space x at a given
time t (or, equivalently, conformal time η). It turns out to be convenient to treat these
quantities as random fields, which we generically denote as f (x). Observationally, we deter-
mine f (x) in our universe (for instance density perturbations at the time of recombination
can be inferred from measurements of temperature anisotropies in the CMB). Any given
theoretical model (for instance, slow-roll inflation with a certain potential V (φ)) predicts a
probability distribution function for f (x) (for instance by giving a probability distribution
for each Fourier component or spherical harmonic). Our observed universe is then claimed
to be a particular realization of this theoretical probability distribution; statistical analy-
sis then determines whether this claim is consistent with the data, i.e. whether the model
works.
Note that this is very reminiscent of quantum mechanics, where the theory predicts a
probability distribution on the possible outcomes of measurements. While it is impossible
to make a sharp prediction for the outcome of any single experiment, it is of course possible
to test quantum mechanics very accurately by doing many experiments. Similarly, we can
test cosmological models quite accurately by observing CMB photons from many distinct
patches in the sky (the accuracy being limited by “cosmic variance”, the fact that there
is only one sky to take patches from). In fact, this similarity is not accidental: we shall
see that the cosmological perturbations predicted by inflation have a quantum mechanical
origin.
Define the Fourier components of f (x) by

d3 k
Z
f (x) = fk eik·x . (19)
(2π)3

They satisfy f−k = fk∗ . The simplest inflationary models predict Gaussian perturbations,
with probability density formally given by

|fk |2
 
1
p(fk ) = exp − (20)
πV σk2 V σk2

for every Fourier coefficient, where V is the volume of space (which we take to be infinite)
and where σk only depends on k = |k| (not to be confused with the parameter k in the
spatial curvature!). For the expectation value of the product of two Fourier coefficients, we
find
hfk fk0 i = V σk2 δk,−k0 = (2π)3 σk2 δ(k + k0 ) ≡ (2π)3 Pf (k) δ(k + k0 ), (21)
where Pf (k) is the power spectrum. (Cf. (6.53) of Dodelson.) Readers unfamiliar with
Gaussian probability distributions are referred to the appendix to this section for more
details.
Another way to characterize a Gaussian random field is by its two-point correlation
function
ξf (x, y) ≡ hf (x)f (y)i. (22)

5
For the homogeneous, isotropic Gaussian distribution (20), we compute, using (21) and the
fact that σk only depends on |k|,
Z 3 0
d3 k
Z
dk 0
ξf (x, y) = 3 3
hfk fk0 ieik·x eik ·y
(2π) (2π)
Z 3
d k ik·(x−y) 2
= e σk
(2π)3
dk sin(kr) k 3 σk2
Z
= , (23)
k kr 2π 2

where r ≡ |x − y|. In the last step, we have performed the integral over the angles.1

2.3.1 Appendix: Gaussian perturbations


The simplest inflationary models predict nearly Gaussian density perturbations. Gaus-
sian density perturbations are consistent with observations of the Cosmic Microwave Back-
ground, for instance by the WMAP and Planck satellites. Let us therefore develop the
formalism to describe Gaussian perturbations.
Consider a real field f (x) (which could for instance be a density perturbation at the
time of recombination). In a region with finite volume V , this field can be Fourier expanded
as
1 X
f (x) = √ fk eik·x . (25)
V k
Reality of f implies that f−k = fk∗ . Writing fk in terms of its real and imaginary parts,
fk = ak + ibk , we find a−k = ak and b−k = −bk .
One way to think about probability distributions is to imagine having a very large
number N of such spatial regions, and that the coefficients ak and bk can take different
values in different regions. In general, a probability distribution function P ({ak }, {bk }) can
be defined by Y
dN = N P ({a0k }, {b0k }) (dak dbk ), (26)
k

where dN denotes the number of regions with a0k


< ak < a0k +dak and b0k < bk < b0k +dbk for
all k. We are interested in situations where the various modes are uncorrelated, meaning
that Y
P ({ak }, {bk }) = p(ak , bk ). (27)
k

1
A popular alternative definition of the power spectrum is

k 3 σk2
δf2 (k) ≡ , (24)
2π 2
which is a dimensionless quantity (if f (x) itself is dimensionless); cf. Mukhanov.

6
The simplest inflationary models predict homogeneous and isotropic Gaussian distribu-
tions,  2  2
1 ak b
p(ak , bk ) = 2
exp − 2 exp − k2 , (28)
πσk σk σk
where σk only depends on k ≡ |k|. Now we can compute correlation functions, such as

hfk fk0 i = hak ak0 i − hbk bk0 i + ihak bk0 i + ihbk ak0 i = σk2 δk,−k0 , (29)

where the Kronecker δk,−k0 equals 1 if k = −k0 and 0 otherwise. While we leave this
computation as an exercise for the reader, let us mention one ingredient as an example:
Z ∞  2  2
1 ak b σ2
2
hak i = dak dbk 2 exp − 2 exp − k2 a2k = k . (30)
−∞ πσk σk σk 2

To obtain formulae valid in the infinite volume limit, V → ∞, one performs the substi-
tutions
Z
X V
→ 3
d3 k,
k
(2π)
fkold → V −1/2 fknew ,
pold (ak , bk ) → V pnew (ak , bk ), (31)

so that
d3 k
Z
f (x) = fk eik·x ,
(2π)3 
|fk |2

1
p(fk ) = exp − ,
πV σk2 V σk2
hfk fk0 i = V σk2 δk,−k0 = (2π)3 σk2 δ(k + k0 ), (32)

where δ(k + k0 ) is the Dirac delta function. These are the expressions used in the main
text. Note that in the V → ∞ limit the expression for p(fk ) is only formally defined, but
the correlator hfk fk0 i remains well-defined.

2.4 Quantum fluctuations of a free scalar field in de Sitter space


[See also Dodelson, section 6.5.1; see also Senatore, sections 2.1 and 2.2 for heuristic deriva-
tions]
In this section we compute the spectrum of scalar field perturbations in an exponentially
expanding universe. In section 2.5, these scalar field perturbations will be related to metric
perturbations after inflation, which, in turn, will be related to observations in section 3.
First we introduce the “conformal time” coordinate
1 −Ht
η=− e , (33)
H

7
defined such that the metric takes the form
1
ds2 = a2 (η)(−dη 2 + dx2 ), a(η) = − . (34)

The action (2) now reads
a2  2
Z   
3 2 4
S= dηd x φ̇ − (∇φ) − a V (φ) , (35)
2
where dot denotes differentiation with respect to the conformal time coordinate η.
We split the scalar field φ in a homogeneous classical background φ(0) (η) and small
fluctuations δφ(η, x):
φ = φ(0) (η) + δφ(η, x). (36)
The classical background satisfies the equation of motion

φ̈(0) + 2 φ̇(0) = −a2 V 0 (φ(0) ), (37)
a
while the fluctuations satisfy to linear order

δ φ̈ + 2 δ φ̇ − ∇2 δφ = −a2 V 00 (φ(0) )δφ ≈ 0, (38)
a
where in the last step we have approximated the slow-roll potential V by a constant.
To eliminate the damping term in the equation of motion (38), we introduce a new
fluctuation field
χ = a δφ, (39)
in terms of which the action for the fluctuation has a canonical kinetic term:
Z  
(0) 1 4 2 2 ä 2
S = S[φ ] + d x χ̇ − (∇χ) + χ . (40)
2 a
The equation of motion for χ is

χ̈ = ∇2 χ + χ. (41)
a
The field χ can be Fourier expanded as
d3 k 
Z 
ik·x † ∗ −ik·x
χ= ak χk (η)e + ak χk (η)e . (42)
(2π)3/2
The “mode function” χk satisfies the equation of motion
 
2 2
χ̈k = − k χk (43)
η2
and behaves as a positive frequency mode at early times η → −∞:
 
1 i
χk = √ 1− e−ikη . (44)
2k kη

8
Classically, ak and a†k are complex conjugate c-numbers.
Quantum mechanically, the field χ as well as its conjugate momentum π ≡ δS/δ χ̇ = χ̇
are promoted into operators satisfying the canonical commutation relations
[χ(η, x), χ(η, y)] = [π(η, x), π(η, y)] = 0,
[χ(η, x), π(η, y)] = iδ(x − y). (45)
In (42), ak and a†k are now hermitean conjugate operators satisfying the commutation
relations
[ak , al ] = 0, [ak , a†l ] = δ(k − l). (46)
The incoming vacuum state |0i is defined to be annihilated by the “annihilation operators”
ak :
ak |0i = 0. (47)
The “creation operator” a†k creates a particle with comoving momentum k (the physical
momentum is k/a). Repeated action of the creation operators on the vacuum state builds up
a “Fock space” of states with arbitrary numbers of particles. Readers unfamiliar with (free)
quantum field theory may want to refer to the appendix to this section for an introduction.
Having defined the space of states of the quantum field theory, let us comment on the
form of the mode functions (44). For early times (kη  −1), χk oscillates like a positive
frequency mode in Minkowski space. Therefore, the state |0i is what an early time observer
would call the vacuum. When kη ≈ −1, the second term in parenthesis in (44) starts taking
over: the mode “freezes” (stops oscillating) and starts growing.
To determine the power spectrum of the scalar field fluctuations, we compute the spatial
two-point function
d3 k d3 k 0
Z Z
1 0
h0|δφ(η, x)δφ(η, y)|0i = 2 3/2 3/2
χk (η)χ∗k0 (η)eik·x e−ik ·y δ(k − k0 )
a (η) (2π) (2π)
d k ik·(x−y) |χk (η)|2
3
Z
= e . (48)
(2π)3 a2 (η)
In the first equality, we used the commutation relations (46) as well as the definition (47)
of the vacuum state. Comparing with (23) and using (21), (44) and (34), we read off
H2 H2
 
2 1 2 2
Pδφ (k) = σk = H η + 2 ≈ 3, (49)
2k k 2k
where in the last step we ignored the first term. To justify this, note that given a distance
r ≡ |x − y|, the integral (48) will mainly get contributions from k ≤ 1/r (the high k
contributions tend to cancel each other because of the oscillatory integrand). For the
distance scales of observational interest today (e.g., x and y two points on the surface of
last scattering that can be resolved by CMB observations), the corresponding momenta k
were well inside the “frozen” regime at the end of inflation, so that the approximation in
(49) is excellent. A power spectrum that scales like 1/k 3 is called scale invariant.2
2
The alternatively defined power spectrum (see footnote 1) corresponding to (49) is k-independent:

2 H2
δδφ (k) = . (50)
4π 2

9
R
Incidentally, one might worry about the fact that (48) exhibits an integral dk/k over
low momenta, which is logarithmically divergent. This is indeed the case for a spacetime
that has undergone exponential expansion forever. In inflationary models, though, one
assumes that inflation only started in some finite region of space, which provides an infrared
cutoff (corresponding to the inverse size of the region), eliminating the divergence.

2.4.1 Gaussianity
[Senatore, section 2.6]
The simplest inflationary models predict nearly Gaussian perturbations. The basic
intuition is simple: a free quantum field is equivalent to a collection of independent harmonic
oscillators (one for each wave number k), and the ground state of a harmonic oscillator is
a Gaussian wave function. Senatore computes the Gaussian wave function for a given
wave number, starting at early times for which the mode is deep inside the horizon and
the expansion of the universe can be ignored (i.e., the evolution is adiabatic), and taking
into account the “freezing” that happens when the mode leaves the horizon. The resulting
two-point function (Senatore’s (108)) agrees with that computed in section 2.4. (Note that
Senatore does not keep track of numerical factors.)

2.4.2 Appendix: quantum fields


As discussed before, the action for a free scalar field φ(x, t) in Minkowski space-time is
given by Z
1
S[φ] = dt d3 x [φ̇2 − (∇φ)2 − m2 φ2 ] (51)
2
and the equation of motion is
φ̈(x, t) − ∆φ(x, t) + m2 φ(x, t) = 0. (52)
If we assume the three space directions to be periodic with length L, so that the volume of
space is V = L3 , the field can be Fourier decomposed as follows:
1 X
φ(x, t) = √ φk (t) eik·x , (53)
V k

with kx , ky , kz ∈ (2π/L)Z. In terms of the Fourier modes, the equation of motion becomes
φ̈k + (k 2 + m2 )φk = 0. (54)
These are the equations of motion of a collection of harmonic oscillators (one for every value
of k) with frequencies ωk = (k 2 + m2 )1/2 .
Let us remind ourselves how to quantize a harmonic oscillator. Starting from the La-
grangian
1 1
L = q̇ 2 − ω 2 q 2 , (55)
2 2

10
we construct the canonical momentum p ≡ ∂L/∂ q̇ = q̇ and the Hamiltonian

p2 ω 2 q 2
H= + . (56)
2 2
To quantize the system, we promote impose q and p to operators q̂ and p̂ and impose the
canonical commutation relation [q̂, p̂] = i, where we work in units with ~ = 1. In terms of
annihilation and creation operators
r r
ω i ω i
a≡ (q̂ + p̂), a† ≡ (q̂ − p̂), (57)
2 ω 2 ω

which satisfy the commutation relation [a, a† ] = 1, the Hamiltonian can be usefully rewritten
as  
† 1
H =ω a a+ . (58)
2
The ground state |0i is defined by a|0i = 0 and has energy E0 = ω/2. The n’th excited
state is proportional to (a† )n |0i and has energy En = ω(n + 1/2).
Now we extend this quantization prescription to free fields. Start from the Lagrangian
Z
1
L[φ] = d3 x [φ̇2 − (∇φ)2 − m2 φ2 ], (59)
2
define the canonical momentum
δL[φ]
π(x, t) ≡ = φ̇(x, t) (60)
δ φ̇(x, t)
and the Hamiltonian
Z Z
3 1
H = d x π(x, t) φ̇(x, t) − L = d3 x [π 2 + (∇φ)2 + m2 φ2 ]. (61)
2
To quantize the system, promote π and φ to operators and impose the canonical commu-
tation relations

[φ̂(x, t), π̂(y, t)] = iδ(x − y),


[φ̂(x, t), φ̂(y, t)] = 0,
[π̂(x, t), π̂(y, t)] = 0. (62)

Expanding φ̂(x, t) in a basis of solutions to the classical equations of motion,

d3 k
Z
1 h −iωk t ik·x iωk t −ik·x †
i
φ̂(x, t) = √ e e a k + e e a k , (63)
(2π)3/2 2ωk
the reader is invited to show as an exercise that the canonical commutation relations are
equivalent to
[ak , a†k0 ] = δ(k − k0 ), [ak , ak0 ] = 0 = [a†k , a†k0 ]. (64)

11
The quantum field theory vacuum |0i is defined to be annihilated by all annihilation oper-
ators ak :
ak |0i = 0 for all k. (65)
The state a†k |0i is a one-particle state with momentum k. The state a†k a†l |0i is a two-
particle state with particle momenta k and l, etc. The reader is invited to check that the
Hamiltonian can be written as
Z Z  
3 ωk † † 3 † 1 (3)
H= dk (ak ak + ak ak ) = d k ωk ak ak + δ (0) . (66)
2 2
It follows that the ground state energy is given by
Z
(3) 1
E0 = δ (0) d3 k ω k . (67)
2
The factor δ (3) (0) can be interpretedR as a volume factor V /(2π)3 , so that we find a vac-
uum energy density proportional to d3 k ωk , which would diverge if we integrated up to
arbitrarily high momenta. In a quantum field theory not coupled to gravity (such as the
free theory discussed in this appendix), the zero of the energy can be chosen arbitrarily,
so the infinite energy density can simple be discarded. In a theory coupled to gravity, the
vacuum energy is important (since it induces curvature of space-time), but in that case one
does not expect quantum field theory to be valid up to arbitrarily high energies (certainly
not beyond the Planck scale, where quantum gravity effects are important). However, with
a cutoff at the Planck scale or even at the TeV scale, this is still a large contribution to
the vacuum energy, much larger than the value suggested by observations. This is the
cosmological constant problem.

2.5 From scalar field perturbations to density perturbations


[Dodelson, section 6.5.2; Senatore, section 2.3]
When connecting the theory of inflation to observations, we will be interested in metric
perturbations long after inflation has ended, for instance at the time of recombination,
when the CMB photons were emitted. At the end of inflation, the energy stored in the
potential energy of the inflaton is converted into radiation and matter, a process known
as reheating. The details of this process are poorly understood, which raises the question
whether the “initial conditions” produced by inflation can be reliably propagated to later
times. As we will see, this is indeed possible by identifying a quantity that is conserved for
superhorizon modes, independently of the details of reheating. By computing this quantity
during inflation (at the time the mode leaves the horizon), we learn what the quantity
will be equal to at later times, until it enters the horizon during the standard big bang
cosmology (which for modes of observational interest happens safely within the radiation
or matter dominated eras).
There is an obvious way in which fluctuations in the inflaton gives rise to metric fluctu-
ations already during inflation; it will turn out, however, that the resulting metric fluctu-
ations are negligibly small, making this mechanism irrelevant. Consider first a mode deep

12
inside the horizon during slow roll inflation. In this regime, the gravitational potential Ψ
is determined by the Poisson equation

k2
− Ψ = 4πG δρ, (68)
a2
and to estimate orders of magnitude we can continue to use this equation up to the time
of horizon crossing, k 2 ∼ a2 H 2 . For a mode crossing the horizon, we thus estimate
δρ δρ
Ψ∼G 2
∼ , (69)
H ρ
and before horizon crossing Ψ will be even smaller. For a slowly rolling scalar field, the
energy density is dominated by the potential, ρ = V (φ). The scalar field fluctuations give
rise to density perturbations

δρ V 0 (φ(0) ) √
δρ = V 0 (φ(0) ) δφ, = δφ ∼ G δφ, (70)
ρ V (φ(0) )

which thus also have a scale invariant power spectrum. It can be checked that the resulting
metric perturbation (69) can be ignored in the equation of motion of the scalar field, which
justifies a posteriori our computation in the previous section, where we considered a scalar
field in an exponentially expanding homogeneous spacetime, ignoring the backreaction of
the scalar on the metric.
Intuitively, the dominant mechanism by which fluctuations of the inflaton give rise to
metric perturbations after inflation is as follows. (It will be made more precise in sec-
tion 2.5.2.) Let us assume that inflation ends when the inflaton φ reaches a certain value
φend , where the potential ceases to be flat enough. Space-dependent fluctuations δφ of the
inflaton cause different spatial regions to reach this critical value of the inflaton at different
times, the time delay being
δφ(x)
δη(x) ∼ . (71)
φ̇(0)
The fact that some regions of space inflate longer than others causes space to curve, and,
as we will see in section 2.5.2, this curvature will lead to perturbations in the gravitational
potential Ψ after inflation has ended.

2.5.1 Adiabatic density perturbations


[Dodelson, section 6.1; Lesgourgues, section 1.4]
Modes of observational interest were superhorizon when inflation ended, and inflation
provides initial conditions for their subsequent evolution. Superhorizon modes undergo very
simple evolution, unaffected by local interactions. In single-field inflation, the time-shifting
(71) is the only degree of freedom that determines superhorizon modes. Initial conditions
of this type are called adiabatic.

13
Assuming for simplicity that the universe contains a number of uncoupled perfect fluids
(labeled by an index x) with sound speeds c2x = δPx /δρx , we find for superhorizon modes

ρx (η, x) = ρ(0) (0) (0)


x (η + δη(x)) = ρx (η) + ρ̇x (η) δη(x), (72)
Px (η, x) = Px(0) (η + δη(x)) = Px(0) (η) + Ṗx(0) (η) δη(x). (73)

Combining this with the background energy conservation equation


ȧ (0)
ρ̇(0) (0)
x = −3 (ρx + Px ), (74)
a
we find
δρx ȧ δρx ȧ
(0) (0)
= −3 (0)
= −3 δη(x), (75)
ρx + Px a ρ̇x a
which is independent of the index x that labels species. Using the equations of state for
matter and radiation, this implies (still for superhorizon modes)
3 3
δ = δb = δν = δγ . (76)
4 4
Einstein’s equations can be used to show that, for superhorizon modes during radiation
domination,
1
Ψ = − δγ = constant. (77)
2
In general, superhorizon metric and density perturbations are constant except when the
equation of state changes.

2.5.2 ζ conservation for superhorizon modes


[See also Senatore, section 2.3.1 for a heuristic argument]
In order to relate fluctuations generated during inflation to the metric perturbations
they generate at later times, it is very useful to consider a quantity that remains constant
for superhorizon modes, even when the equation of state changes. It can be shown that
this is the case for the quantity ζ defined3 by (6.86) of Dodelson:

1 δρ
ζ ≡ −Ψ + . (78)
3ρ+P
During slow-roll inflation, for modes inside the horizon and those which have just left
the horizon, Ψ can be neglected. Using δρ = V 0 (φ(0) )δφ and dρ/dt = −3H(ρ + P ), we then
find
δφ
ζ = −aH (during inflation). (79)
φ̇(0)
3
This does not quite agree with Dodelson’s definition (6.77), but it does agree with Baumann’s definition
(136). And it does the job.

14
After the end of inflation, when the universe is dominated by radiation, it follows from (78)
and (77) that
3
ζ = − Ψ (after inflation). (80)
2
The conservation of ζ for superhorizon modes then leads to the following relation between
the post-inflation power spectrum of Ψ and the power spectrum of δφ at horizon crossing
during inflation:  2
4 aH
PΨ (k) = Pδφ aH=k . (81)
9 φ̇(0)
Using
4πG(φ̇(0) )2 = a2 H 2 , (82)
which you are asked to derive in the homework, and (49), this becomes

16πG 8πGH 2
PΨ (k) = Pδφ aH=k = . (83)
9 9k 3 aH=k
Comparing with (69) and (70), we see that  is now in the denominator rather than the
numerator, making the gravitational potential induced by the time delay much larger than
the metric perturbations during inflation. In the intuitive language used around (71), we
can say that the flatter the potential, the longer it takes to cover an additional distance in
φ space and the more curvature is generated by additional inflation.

2.6 Tensor perturbations


[Dodelson, section 6.4; Senatore, section 2.8]
Dodelson’s (5.63) gives the equation of motion of tensor perturbations in an FLRW
spacetime. This evolution equation is essentially identical to (38), which governs the fluctu-
ations of a free scalar field. The same mechanism that gave rise to quantum fluctuations of
the inflaton gives rise to quantum fluctuations of tensor modes. This leads to a stochastic
background of gravitational waves with a nearly scale-invariant spectrum:

8πGH 2
Ph (k) = aH=k
. (84)
k3
Comparing with (83), we see that the slow-roll parameter  is now absent. This implies
that tensor modes have much smaller amplitude than scalar metric perturbations, making
them much harder to detect. But if and when tensor modes are detected, they will directly
determine H 2 and thereby, via the first Friedmann equation, the energy scale of inflation.

2.7 Scale invariance and spectral indices


[Dodelson, section 6.6; Senatore, section 2.4]
A scale-invariant spectrum is one for which k 3 P (k) is k-independent. The spectra (83)
and (84) are not quite scale-invariant because the quantities H 2 and  are to be evaluated

15
at the time of horizon crossing, aH = k, which depends on k. We will now show that the
deviations from scale invariance can be expressed in terms of the slow-roll parameters.
The spectral index n of the scalar perturbations, often denoted as ns and often called
the tilt, is defined by
k 3 PΨ (k) ∼ k n−1 . (85)
The tensor spectral index nT is defined by

k 3 Ph (k) ∼ k nT . (86)

With these standard conventions, scale-invariant spectra of scalar and tensor perturbations
would correspond to n = 1 and nT = 0, respectively.
Let us first compute the tensor spectral index, which according to (84) equals

d ln H 2 2k dH dη
nT = = . (87)
d ln k H dη dk aH=k

Taking into account that  = −Ḣ/(aH 2 ) (see (10)) and aH ≈ −1/η (see (34)), this becomes

2k
nT = (−aH 2 )η 2 aH=k = −2. (88)
H
An important observation is that the same parameter  controls both the tensor spectral
index and the tensor-to-scalar ratio Ph (k)/PΨ (k); this is a quite robust prediction of infla-
tion.
The scalar spectral index is given by
d 
n−1= ln H 2 − ln  aH=k , (89)
d ln k
where the first term in parentheses will again give −2. In the homework, you are asked to
prove
d
= 2aH( + δ), (90)

which implies that the second term in parentheses will give −2( + δ). The scalar spectral
index thus becomes
n = 1 − 4 − 2δ. (91)

2.8 The energy scale of inflation


[Senatore, section 2.5; Dodelson, section 6.6]
Measuring the amplitude of tensor perturbations would directly determine H and, via
the first Friedmann equation, the energy scale of inflation. Currently, however, we only
know the amplitude and spectral index of the scalar perturbations. From (83), we see that
then  is required to determine H (and therefore the energy scale of inflation). But (91)
tells us that the spectral index n by itself does not determine . If we assume, however,

16
that δ ∼ , then  is of the same order of magnitude as 1 − n, which has been measured to
be a few percent. Taking into account the measured PΨ ∼ 10−10 , this leads to
H
∼ 10−6 , H ∼ 1013 GeV, V ∼ (1015 GeV)4 , (92)
MPl
which are obviously extremely high energy scales. As is clear from (12), measuring the
slow-roll parameters amounts to probing the structure of this potential.

2.9 Homework 4
The homework is due on Thursday April 19, 2018 at 10:30 am; you can either hand it in at
the beginning of class, or email it to your assistant with me in cc. You are welcome (and
even encouraged) to discuss the problems with each other, but please make sure to write
up your own version of the solutions. (Copying solutions from each other or from another
source is not allowed.)
1. Derive (3) by varying (2) with respect to g µν . (You can use the formula δg =
−g gµν δg µν for varying the determinant of the metric.) Then show that for a ho-
mogeneous scalar field in an FLRW spacetime (4) it reduces to that of a perfect fluid
with energy density and pressure given by (5).
2. Derive (7) by varying the action (2) with respect to the scalar field.
3. Show that (12) holds. (This is Exercise 14 of Dodelson’s section 6.)
4. Derive (40). (When using integration by parts, feel free to assume that boundary
terms can be ignored.)
5. Show that (45) and (46) are equivalent. (First verify the useful identity χk χ̇∗k −
χ∗−k χ̇−k = i. To show that (46) implies (45), compute (45) using (42) and (46). To
show the reverse, apply the inverse Fourier transformation to (42) and the analogous
expression for π (obtained by taking a time derivative) and take appropriate linear
combinations to find expressions for ak and a†k in terms of χ and π.)
6. Verify (79) en (80).
7. Verify (82). (This is Exercise 12(b) of Dodelson’s section 6.)
8. Verify (90). (This is Exercise 12(c) of Dodelson’s section 6, with a minus sign cor-
rected.)
9. Please go through Wayne Hu’s excellent online tutorial [6], where the material of the
next lecture is covered with less equations but more pictures and animations. (No
need to hand in anything.)
10. For a quick introduction to recent CMB observations at the level of the general public,
please read the press releases [7] and [8] by the Planck collaboration. (No need to
hand in anything.)

17
3 Contact with CMB observations
3.1 Photon scattering
[Lesgourgues, section 2.1]

3.2 The Boltzmann equation for photons


[Lesgourgues, section 2.2; see also Dodelson, sections 4.2–4.4 and 8.3.1]

3.3 Temperature anisotropy in a given direction


[Lesgourgues, section 2.3]

3.3.1 Large scale anisotropy


3.4 Spectrum of temperature anisotropies
[Lesgourgues, section 2.4; see also Dodelson, section 8.5]

3.5 Acoustic oscillations


[Lesgourgues, section 2.5; Dodelson section 8.3]

3.6 Polarization
[Lesgourgues, sections 2.7–2.8]

3.7 Cosmological parameters


[Dodelson, section 8.7; see also Lesgourgues, section 2.6]

3.8 Homework 5
The homework is due on Thursday May 3, 2018 at 10:30 am; you can either hand it in at
the beginning of class, or email it to your assistant with me in cc. You are welcome (and
even encouraged) to discuss the homework with each other, but please make sure to write
up your own version.

1. Explain clearly how CMB observations can be used to test specific inflationary models.
Given a slow-roll potential V (φ), how does one relate it to cosmological parameters,
and how can these parameters be measured? (Please focus on explaining the main
ideas clearly, no need to write a very long or detailed text.) What is the current
observational status of the simple inflationary model with potential V (φ) ∼ φ2 ? (For
a technical paper which contains relevant information but which I certainly do not
expect you to read completely, see [9].)

18
References
[1] S. Dodelson, “Modern Cosmology,” Amsterdam, Netherlands: Academic Pr. (2003)
440 p

[2] L. Senatore, “Lectures on Recent Developments in Cos-


mology”, Asian String School, Puri, India (2014),
http://icts.res.in/media/uploads/Talk/Slides/Lectures on inflation India.pdf

[3] J. Lesgourgues, “Cosmological Perturbations,” arXiv:1302.4640 [astro-ph.CO].

[4] V. Mukhanov, “Physical Foundations of Cosmology,” Cambridge University Press


(2005).

[5] D. Baumann, “Inflation,” arXiv:0907.5424 [hep-th].

[6] W. Hu, “Ringing in the New Cosmology. Intermedi-


ate Guide to the Acoustics Peaks and Polarization”,
http://background.uchicago.edu/∼whu/intermediate/intermediate.html.

[7] ESA/Planck press release, “Simple but challenging: the universe according to Planck”,
21 March 2013, http://sci.esa.int/planck/51551-simple-but-challenging-the-universe-
according-to-planck/

[8] ESA/Planck press release, “Planck: gravitational waves remain elusive”, 30 January
2015, http://sci.esa.int/planck/55362-planck-gravitational-waves-remain-elusive/

[9] P. A. R. Ade et al. [Planck Collaboration], “Planck 2015 results. XX. Constraints on
inflation,” arXiv:1502.02114 [astro-ph.CO].

19

También podría gustarte