Baumann Primordial Cosmology

TASI Lectures on
Primordial Cosmology
Daniel Baumann
Institute of Theoretical Physics, University of Amsterdam,

Science Park, 1098 XH Amsterdam, The Netherlands
Contents
1 Introduction 1
I Relics from the Hot Big Bang 4
2 Big Bang Cosmology 5

2.1 Geometry and Dynamics 5
2.2 Thermal History 8
2.3 Structure Formation 10
2.4 Initial Conditions 10
3 Afterglow of the Big Bang 12

3.1 CMB Anisotropies 12
3.2 CMB Power Spectrum 17
4 Cosmic Sound Waves 20

4.1 Photon-Baryon Fluid 20
4.2 Acoustic Oscillations 21
5 Light Relics 25
5.1 Dark Radiation 25
5.2 Imprints in the CMB 27
5.3 EFT of Light Species 31
II Relics from Inflation 35
6 Inflationary Cosmology 36
6.1 Horizon Problem 36
6.2 Slow-Roll Inflation 38
6.3 Effective Field Theory 40
7 Quantum Initial Conditions 42

7.1 Quantum Fluctuations 42
7.2 Curvature Perturbations 50
7.3 Gravitational Waves 51
8 Primordial Interactions 55
8.1 Non-Gaussianity 55
8.2 In-In Formalism 57
8.3 Gravitational Floor 59
1
9 Heavy Relics 63
9.1 Massive Fields in de Sitter 63
9.2 Local and Non-Local 65
9.3 Cosmological Collider Physics 68
References 71
1 Introduction
There are many reasons to believe that our current understanding of fundamental physics is
incomplete. For example, the nature of dark matter and dark energy are still unknown, the
stability of the Higgs mass remains unsolved, the origin of neutrino masses is unexplained, the
strong CP problem is still there, the physics of inflation remains elusive, and the origin of the
matter-antimatter asymmetry is still to be discovered. Attempts to address these shortcomings
of the Standard Model (hereafter SM) often involve new degrees of freedom. These new particles
can escape detection in terrestrial experiments either because they are too heavy to be produced
at the energies available or because they are too weakly coupled to be created in sufficiently large
numbers. However, in astrophysical systems and in the early universe the small cross sections can
be compensated for by probing environments with large densities and by following the evolution
over long time scales. For example, the high densities at the cores of stars allow a significant
production of new weakly coupled species. Integrated over the long lifetime of stars this can
lead to large deviations from the standard stellar evolution. Observations of the lifecycles of
stars have therefore put interesting constraints on the couplings of new light species to the SM.
Similarly, in the hot environment of the early universe, even extremely weakly coupled particles
can be produced prolifically and their gravitational influence becomes detectable in the cosmic
microwave background (CMB). Moreover, during inflation even very massive particles can be
created by the rapid expansion of the spacetime. When these particles decay they produce
distinctive signatures in cosmological observables. Observing these effects could teach us a great
deal about the physics driving the inflationary expansion.
New physics can affect cosmological observables in two distinct ways: i) it may affect the
initial conditions of the hot big bang, or ii) it may affect the evolution of these initial conditions
through changes to the thermal history of the universe. These opportunities for probing new
physics with cosmological observations are illustrated in the following schematic:
new physics?
↓
evolution
hζk1 ζk2 . . .i −−−−−−−−−−−−−−−−−−−−→ hO1 O2 . . .i
initial conditions cosmological correlators
↑
new physics?
where ζ denotes the primordial curvature perturbations and O stands for late-time observables,
which may be the temperature variations δT of the CMB or the density fluctuations δρg in the
distribution of galaxies. In these lectures, we will give examples from both of these avenues of
tests for physics beyond the Standard Model (BSM).
The initial fluctuations were drawn from a probability distribution P [ζ] and hence are char-
acterized by the following correlation functions
Z
hζk1 ζk2 . . .i = Dζ P [ζ] ζk1 ζk2 · · · .
1
By tracing the observed correlations hO1 O2 . . .i back in time, cosmologists try to measure the
initial correlations hζk1 ζk2 . . .i and use them to extract information about the initial probability
distribution P [ζ] and about the physics that gave rise to it. If the early universe went through
a period of inflation then these initial correlations were produced dynamically before the hot big
bang. New physics during inflation, such as the existence of new degrees of freedom X, can leave
imprints in the spectrum of primordial perturbations, i.e. P [ζ] → P [ζ, X].
New physics may also affect how the initial correlations evolve into the late-time observables.
For example, the evolution equations for fluctuations in the primordial plasma can be affected
by the presence of new degrees of freedom X:
SM X
∇µ Tµν SM X

Gµν = 8πG Tµν + Tµν , + Tµν = 0.
As we will see, much of the evolution in the early universe is very well understood and probed
by very precise observations. The possibilities for new physics are therefore highly constrained.
The most conservative way to parameterize physics beyond the Standard Model is in terms
of an effective field theory (EFT). The basic input of an EFT are the field content and the
symmetries that are relevant at a given energy scale. The effective Lagrangian is then the sum
of all operators consistent with the symmetries,
X
Leff ⊂ g OX OSM ,
where OSM denotes operators made from the SM degrees of freedom and OX stands for opera-
tors constructed from any additional fields. The couplings g parameterize the strengths of the
interactions between the fields X and the SM. Deviations from the SM predictions then scale
with the size of the couplings g. In these lectures, I will show how cosmological observations put
constraints on these couplings.
Outline The goal of these lectures is to show that cosmology is becoming an increasingly
sensitive probe of BSM physics. The presentation is divided into two parts: In Part I, we
study the production of new light particles in the hot big bang and describe their effects on the
anisotropies of the cosmic microwave background. In Part II, we investigate the possibility of
very massive particles being created during inflation and determine their imprints in higher-order
cosmological correlations.
Notation and conventions Throughout these lectures, we will use natural units, c = ~ = 1,
with reduced Planck mass Mpl 2 = 1/8πG. Our metric signature is (− + ++). Greek letters will
stand for spacetime indices, µ, ν, . . . = 0, 1, 2, 3, and Latin letters for spatial indices, i, j, . . . =
1, 2, 3. Three-dimensional vectors will be written in boldface, k, and unit vectors will be hatted, k̂.
Overdots and primes will denote derivatives with respect to conformal time τ and physical time t,
respectively. The dimensionless power spectrum of a Fourier mode ζk will be defined as
k3
Pζ (k) ≡ hζk ζ−k i0 ,
2π 2
where the prime on the expectation value indicates that the overall momentum-conserving delta
function has been dropped.
2
Acknowledgements I am grateful to the organizers Igor Klebanov and Mirjam Cvetic for the
invitation to lecture at TASI 2017. Thanks to my students and collaborators for many helpful
discussions on the material presented in these lectures. I am especially grateful to Daniel Green,
Joel Meyers, Benjamin Wallisch and Matias Zaldarriaga for collaboration on topics covered in
Part I, and to Nima Arkani-Hamed, Valentin Assassi, Garrett Goon, Daniel Green, Hayden Lee,
Liam McAllister and Guilherme Pimentel for collaboration on work described in Part II. Finally,
thanks to Benjamin Wallisch and Swagat Saurav Mishra for comments on a preliminary version
of the manuscript.
3
Part I
Relics from the Hot Big Bang
Consider a 2 → N scattering process producing a new particle species X. Assume that X is very
weakly coupled to the Standard Model degrees of freedom. The production of particles X then
carries energy and momentum away from the interaction region. Such missing energy, of course,
famously led to the discovery of neutrinos. However, particles that are more weakly coupled than
neutrinos are produced very rarely in colliders and their missing energy signatures are hard to
detect. Fortunately, the creation of new species may be enhanced in astrophysical systems and
in the early universe. To see this, consider the change in the number densities of the particles
involved in the above interaction. Schematically, this is given by
∆n
∼ nσ × ∆t ,
n
i.e. the fractional change in the number density is equal to the interaction rate, Γ ∼ nσ, times the
interaction time, ∆t. We see that small cross sections σ can be compensated for by high densities
and long time scales. This explains why observations of the lifecycles of stars have put interesting
constraints on the existence of extra species. Taking the typical number density inside of stars to
be n ∼ (1 keV)3 and integrating over the typical lifetime of a star, ∆t ∼ 1016 sec ≈ 1.5×1031 eV−1 ,
we find significant changes in the stellar evolution if
σ > (n∆t)−1 ∼ (1010 GeV)−2 .
Similar constraints can be derived from the observed energy loss in supernova explosions. In that
case, the relevant time scales are much shorter, ∆t ∼ 10 sec, but the densities are much higher,
n ∼ (30 MeV)3 .
These order of magnitude estimates also give us a sense for the power of cosmological con-
straints. In the early universe, the interaction time scales are short, ∆t < 1 sec, but the densities
can be very high, n ∼ T 3 (1 MeV)3 . For temperatures above 104 GeV, we expect cosmological
constraints to be stronger than those from astrophysics. In this part of the lectures, we will show
that primordial cosmology is indeed a highly sensitive probe of new light particles produced in
the hot Big Bang.
We will begin, in Section 2, with a quick review of FRW cosmology. In Section 3, we will
introduce the cosmic microwave background as a tool for precision cosmology. In Section 4, we
will discuss the physics of the acoustic oscillations observed in the CMB anisotropy spectrum.
Finally, in Section 5, we will show that extra relativistic particles leave a unique signature in the
CMB spectrum.
4
2 Big Bang Cosmology
We will begin with a lightning review of some elementary concepts in cosmology. I will assume
that you have seen most, if not all, of this material before, so I will cite many results without
detailed derivations. Further details can be found in my Cosmology course [1] or in any of the
standard textbooks (e.g. [2, 3]).
2.1 Geometry and Dynamics

The Friedmann-Robertson-Walker (FRW) metric of a homogenous and isotropic spacetime is
ds2 = −dt2 + a2 (t)γij dxi dxj , (2.1)
where γij denotes the metric of a maximally symmetric 3-space and a(t) is the scale factor.
Throughout these lectures, we will restrict to the special case of flat spatial slices, i.e. γij = δij ,
and define dx2 ≡ δij dxi dxj . We will also introduce conformal time, dτ = dt/a(t), so that the
metric becomes
ds2 = a2 (τ ) −dτ 2 + dx2 .

(2.2)
We will first discuss the kinematics of particles in an FRW spacetime for an arbitrary scale
factor a(τ ). After that, we will show how the Einstein equations determine a(τ ) in terms of the
matter content of the universe.
Kinematics
Particles in the FRW spacetime evolve according to the geodesic equation
µ
ν µ ν ∂P µ ρ
P ∇ν P = P + Γνρ P = 0, (2.3)
∂xν
where P µ ≡ dxµ /dλ is the four-momentum of the particle. In an expanding spacetime, it is

convenient to write the components of the four-momentum as
P µ = a−1 [E, p] . (2.4)
For massless particles, such as photons, we have the constraint gµν P µ P ν = −E 2 + |p|2 = 0, so
we can write p = E p̂, where p̂ is a unit vector in the direction of propagation.
Exercise.—Show that the non-zero connection coefficients associated with the metric (2.2) are
Γ000 = H , Γ0ij = Hδij , Γij0 = Hδji , (2.5)
where H ≡ ȧ/a is the conformal Hubble parameter.
The µ = 0 component of the geodesic equation (2.3) becomes
dP 0
P0 = −Γ0αβ P α P α . (2.6)
dτ
5
Using (2.4) and (2.5), we get
d −1
(a−1 E) (a E) = −Ha−2 E 2 − Ha−2 |p|2 , (2.7)
dτ
which simplifies to
1 dE 1 da
=− . (2.8)
E dτ a dτ
This describes the redshifting of the photon energy in an expanding spacetime, E ∝ a−1 .
Dynamics
The evolution of the scale factor is determined by the Friedmann equations
3H2 = 8πGa2 ρ̄ , (2.9)

2Ḣ + H2 = −8πGa2 P̄ , (2.10)
where ρ̄ and P̄ are the background density and pressure, respectively.
Exercise.—By substituting (2.5) into
Rµν ≡ ∂λ Γλµν − ∂ν Γλµλ + Γλλρ Γρµν − Γρµλ Γλνρ , (2.11)
show that
R00 = −3Ḣ , Rij = (Ḣ + 2H2 )δij ⇒ a2 R ≡ a2 g µν Rµν = −6(Ḣ + H2 ) . (2.12)
Hence, show that the non-zero components of the Einstein tensor, Gµν ≡ Rµν − 21 Rgµν , are
G00 = 3H2 , Gij = −(2Ḣ + H2 )δij . (2.13)
Use this to confirm that the 00-Einstein equation, G00 = 8πG T00 , implies (2.9) and the ij-Einstein
equation, Gij = 8πG Tij , leads to (2.10).
Combining (2.9) and (2.10), we can write an evolution equation for the density
ρ̄˙ = −3H(ρ̄ + P̄ ) . (2.14)
For pressureless matter (P̄m ≈ 0) this implies ρ̄m ∝ a−3 , while for radiation (P̄r = 31 ρ̄r ) we have
ρ̄r ∝ a−4 .
Exercise.—Derive the continuity equation (2.14) from the conservation of the stress tensor, ∇µ Tµν = 0.
By integrating the Friedmann equation (2.9) for matter and radiation show that
(
τ 2 matter
a(τ ) = . (2.15)
τ radiation
6
Cosmic inventory
The universe is filled with several different species of particles:
baryons (b)
z }| {
photons (γ) neutrinos (ν) electrons (e) protons (p) cold dark matter (c) .
| {z } | {z }
radiation (r) matter (m)
The number density, energy density and pressure of each species a can be written as
d3 p
Z
na = ga fa (x, p) , (2.16)
(2π)3
d3 p
Z
ρa = ga fa (x, p)E(p) , (2.17)
(2π)3
d3 p p2
Z
Pa = ga 3
fa (x, p) , (2.18)
(2π) 3E(p)
where fa (x, p) is the (phase space) distribution function of the species a, and ga is the number of
internal degrees of freedom. In the unperturbed universe, the distribution functions should not
depend on the position and the direction of the momentum, i.e. fa (x, p) → f¯a (E(p)).
At early times, particle interactions were efficient enough to keep the different species in local
equilibrium. They then shared a common temperature T and the distribution functions take the
following maximum entropy form
1
f¯a (E) = , (2.19)
e(Ea −µa )/T ±1
with + for fermions and − for bosons. The chemical potential µa vanishes for photons and is
(likely) small for all other species. We will henceforth set it to zero. When the temperature
drops below the mass of a particle species, T ma , they become non-relativistic and their
distribution function receives an exponential (Boltzmann) suppression, fa → e−ma /T . This means
that relativistic particles (‘radiation’) dominate the density and pressure of the primordial plasma.
By performing the integrals (2.17) and (2.18) in the limit E → p, one finds
(
π2 1 bosons 1
ρ̄a = ga T 4 7 and P̄a = ρ̄a . (2.20)
30 fermions 3
8
The total radiation density is
π2 X 7 X
ρ̄r = g∗ T 4 , where g∗ ≡ ga + ga . (2.21)
30 8
a=b a=f
If equilibrium had persisted until today, all species with masses greater than 10−3 eV would be
exponentially suppressed. This would not be a very interesting world. Fortunately, many massive
particle species (e.g. dark matter) are weakly interacting and decoupled from the primordial
plasma at early times.
7
2.2 Thermal History
The key to understanding the thermal history of the universe is understanding the competition
between the interaction rate of particles, Γ, and the expansion rate of the universe, H. Particles
maintain equilibrium as long as Γ H and freeze out when Γ . H (see Fig. 1).
relativistic non-relativistic
freeze-out
relic density
equilibrium
1 10 100
Figure 1. A schematic illustration of particle freeze-out. At high temperatures, the particle abundance
tracks its equilibrium value. At low temperatures, the particles freeze out and maintain a relic density
that is much larger than the Boltzmann-suppressed equilibrium abundance.
Neutrinos are the most weakly interacting particles of the Standard Model and therefore
decoupled first (around 0.8 MeV or 1 sec after the Big Bang). Shortly after neutrino decoupling,
electrons and positrons annihilated. The energies of the electrons and positrons got transferred
to the photons, but not the neutrinos. The temperature of the photons today is therefore slightly
larger than that of the neutrinos (see insert below). At around the same time, neutron-proton
interactions became inefficient, leading to a relic abundance of neutrons. These neutrons were
essential for the formation of the light elements during Big Bang nucleosynthesis (BBN), which
occurred around 3 minutes after the Big Bang.
Cosmic neutrino background. After the neutrinos decouple, their momenta redshift, pν ∝ a−1 , and
their distribution functions fν maintain their shape. The combination of these two facts requires that
the neutrino temperature evolves as Tν ∝ a−1 . We would like to compare this to the evolution of the
photon temperature Tγ , since this is what has been measured by observations of the CMB. We will
use the fact that the comoving entropy is conserved in equilibrium [1]:
ρ+P 3 2π 2
sa3 = a = g∗ (aT )3 = const . (2.22)
T 45
Since entropy is separately conserved for the thermal bath and the decoupled species, we only need
to consider the change in the effective number of relativistic degrees of freedom in equilibrium with
8
neutrino decoupling
photon heating
electron-positron
annihilation
Figure 2. Thermal history through electron-positron annihilation. Neutrinos are decoupled and their
temperature redshifts simply as Tν ∝ a−1 . The energy density of the electron-positron pairs is transferred
−1/3 −1
to the photon gas whose temperature therefore redshifts more slowly, Tγ ∝ g∗ a .
the photons. Before e+ e− annihilation, i.e. at T+ > me , we have

7 11
g∗ (T+ ) = 2 + × (2 × 2) = , (2.23)
8 2
where we have counted photons, electrons and positrons. After e+ e− annihilation, i.e. at T− < me ,
only the two polarization degrees of freedom of the photons contribute,
g∗ (T− ) = 2 . (2.24)
Comparing g∗ (aTγ )3 = const. to aTν = const., we get

1/3 1/3
g∗ (T− ) 4
Tν = Tγ = Tγ . (2.25)
g∗ (T+ ) 11
Given the measured temperature of the CMB today, Tγ,0 = 2.7 K, this tells us that the present
temperature of the cosmic neutrino background (CνB) is Tν,0 = 1.9 K.
While the neutrinos are still relativistic (i.e. for most of the history of the universe), each species
carries the following energy density
7 4/3
ρν × 2 × Tν4 7 4
= 8 = ≡ aν ≈ 0.227 . (2.26)
ργ 2 × Tγ4 8 11
The Nν = 3 neutrino species of the Standard Model therefore contribute a significant amount to the
total radiation density in the early universe:
P
ρν Nν ρν Nν aν
= = ≈ 41% . (2.27)
ρr Nν ρ ν + ρ γ Nν aν + 1
Although the neutrinos are decoupled, their gravitational effects are significant and have recently been
observed in the CMB [4, 5].
9
Figure 3. Planck measurement of the temperature variations in the CMB sky.
Below about 1 eV, or 380,000 years after the Big Bang, the temperature became low enough
for neutral hydrogen atoms to form through the reaction e− + p+ → H + γ. This is the moment
of recombination. At this point the density of free electrons dropped dramatically. Before re-
combination the strongest coupling between the photons and the rest of the plasma was through
Thomson scattering, e− + γ → e− + γ. The sharp drop in the free electron density after recom-
bination means that this process became inefficient and the photons decoupled. After decoupling
the photons streamed freely through the universe and are observed today as the CMB.
2.3 Structure Formation

The CMB is an almost perfect blackbody with an average temperature of 2.7 K. Small variations
in the CMB temperature across the sky (see Fig. 3), reflect spatial variations in the density of
the primordial plasma, δρa ≡ ρa − ρ̄a (see Section 3), and related perturbations of the spacetime
geometry, δgµν ≡ gµν − ḡµν . During the radiation-dominated phase of the early universe, the
growth of matter perturbations was inhibited by the large pressure provided by the radiation.
Perturbations in the coupled photon and baryon fluids were oscillating with constant amplitude
(see Section 4). Shortly before recombination, however, the universe became matter dominated
and the radiation pressure disappeared, so that density fluctuations could start to grow under
the influence of gravity. This growth of the matter perturbations eventually led to the large-scale
structure (LSS) of the universe.
2.4 Initial Conditions

At sufficiently early times, all scales of interest to current observations were outside the Hubble
radius, k < H. On super-Hubble scales, the evolution of perturbations becomes very simple,
especially for adiabatic initial conditions.
10
Adiabatic perturbations
Adiabatic perturbations have the property that the local state of matter (determined, for example,
by the energy density ρ and the pressure P ) at some spacetime point (τ, x) of the perturbed
universe is the same as in the background universe at some slightly different time τ + δτ (x).
(Notice that the time shift varies with position x.) If the universe is filled with multiple fluids,
adiabatic perturbations correspond to perturbations induced by a common, local shift in time of
all background quantities; e.g. adiabatic density perturbations are defined as
δρa (τ, x) ≡ ρ̄a (τ + δτ (x)) − ρ̄a (τ ) ≈ ρ̄˙ a δτ (x) , (2.28)
where δτ is the same for all species a. This implies that all matter perturbations can be charac-
terized by a single degree of freedom. It also means that we can perform a local time reparam-
eterization to set all matter perturbations to zero, e.g. δρa ≡ 0. In that gauge, the information
about fluctuations is carried by the following perturbation of the metric
gij (τ, x) = a2 (τ ) e2ζ(τ,x) δij , (2.29)
where ζ is called the curvature perturbation. An attractive property of the curvature perturbation
is that it is conserved on super-Hubble scales.
Statistics
The initial conditions for the hot Big Bang are believed to have been created by quantum fluctu-
ations during a period of inflationary expansion [6]. As we will see in Section 7, this mechanism
predicts the statistics of the initial conditions, i.e. it predicts the correlations between the CMB
fluctuations in different directions in the sky, rather than the specific value of the temperature
fluctuation in a specific direction. For Gaussian initial conditions, these correlations are com-
pletely specified by the two-point correlation function
hζ(x)ζ(x0 )i ≡ ξζ (x, x0 ) = ξζ (|x0 − x|) , (2.30)
where the last equality holds as a consequence of statistical homogeneity and isotropy. The
Fourier transform of ζ then satisfies
2π 2
hζ(k)ζ ∗ (k0 )i = Pζ (k) δD (k − k0 ) , (2.31)
k3
where Pζ (k) is the (dimensionless) power spectrum.
Exercise.—Show that
dk
Z
0
ξζ (x, x ) = Pζ (k) sinc(k|x − x0 |) . (2.32)
k
In Section 7, we will explicitly compute the form of Pζ (k) predicted by inflation. However, before
we do that, we will show, in Sections 3 and 4, how generic scale-invariant initial conditions,
Pζ (k) ≈ const., evolve into the anisotropies of the cosmic microwave background. We will first
do this in the context of the Standard Model of particle physics, before asking, in Section 5, what
kind of deviations can arise in theories beyond the SM.
11
3 Afterglow of the Big Bang
Observations of the temperature fluctuations in the cosmic microwave background have played
a pivotal role in establishing the standard cosmological model. We now have a detailed under-
standing of the geometry and composition of the universe, and there is growing evidence that the
primordial fluctuations originated from quantum fluctuations during a period of inflation. In this
section and the next, we will give a simplified analytical treatment of the physics of the CMB
anisotropies. Our goal is to present just enough details to be able to explain how the CMB can
be used as a probe of BSM physics.
3.1 CMB Anisotropies

The first thing one sees when looking at the microwave scy is the motion of the Solar System
with respect to the rest frame of the CMB (cf. Fig. 4).
HOT
COLD
Figure 4. The motion of the Solar System relative to the CMB rest frame produces a dipolar pattern in
the observed CMB temperature.
Consider a photon entering our detectors from a direction n̂. In the rest frame of the CMB, it
has momentum p = −p n̂. Due to the Doppler effect, the observed momentum is
p
p0 (n̂) = ≈ p (1 + n̂ · v) , (3.1)
γ(1 − n̂ · v)
where v is our velocity relative to the CMB rest frame, p is the momentum of the photon in the
CMB rest frame and γ = (1−v 2 )−1/2 is the Lorentz factor. We have also shown an approximation
at leading order in |v| 1. As expected, the momentum is higher if we move towards the photon
(n̂ · v = v) and smaller if we move away from it (n̂ · v = −v). Since the CMB has a blackbody
spectrum, we can relate the change in the observed momentum of photons to a change in the
observed temperature:
δT (n̂) T0 (n̂) − T p0 (n̂) − p

≡ = = n̂ · v = v cos θ . (3.2)
T T p
12
Fitting this dipolar anisotropy to the data, we find that the speed of the Solar System relative
to the CMB is
v = 368 km/s . (3.3)
After subtracting the dipole, we are left with the primordial anisotropy.
Perturbed photon geodesics

Let us trace the life of a photon after decoupling. Its (physical) momentum will redshift due
to the expansion of the universe. In addition, the momentum will change in response to the
inhomogeneities of the universe. We will study these effects by solving the geodesic equation (2.3)
in the perturbed spacetime.
We will treat perturbations in the metric in (conformal) Newtonian gauge
ds2 = a2 (τ ) − (1 + 2Φ)dτ 2 + (1 − 2Ψ)δij dxi dxj ,

(3.4)
where the perturbation Φ ≈ Ψ plays the role of the gravitational potential.
Exercise.—Show that the connection coefficients associated with the metric (3.4) are
Γ000 = H + Φ̇ ,
Γ0i0 = ∂i Φ ,
Γi00 = δ ij ∂j Φ ,
(3.5)
Γ0ij = Hδij − Ψ̇ + 2H(Φ + Ψ) δij ,

Γij0 = H − Ψ̇ δji ,

Γijk = −2δ(j
i
∂k) Ψ + δjk δ il ∂l Ψ .
As we will show in the following insert, the geodesic equation then leads to the following
evolution equation for the photon momentum:
d dΦ ∂(Φ + Ψ)
ln(ap) = − + . (3.6)
dτ dτ ∂τ
In the absence of the source terms on the right-hand side, this implies p ∝ a−1 , which is the
expected redshifting due to the expansion of the universe, cf. eq. (2.8). The inhomogeneous
source terms describe how photons lose (gain) energy as they move out of (into) potential wells.
Derivation.—We will derive eq. (3.6) from the geodesic equation for photons,
dP 0
= −Γ0αβ P α P β , (3.7)
dλ
where P µ = dxµ /dλ is the four-momentum of the photons. We need expressions for the components
of the four-momentum in the presence of metric perturbations.
13
• We first consider the P 0 component. Since photons are massless we have
P 2 = gµν P µ P ν = 0
= −a2 (1 + 2Φ)(P 0 )2 + p2 , (3.8)
where we have substituted the metric (3.4) and defined p2 ≡ gij P i P j . Solving (3.8) for P 0 , we
find
p
P 0 = (1 − Φ) . (3.9)
a
• We then write the spatial component of the four-momentum as
P i ≡ α p̂i . (3.10)
To determine the constant of proportionality α, we use
p2 = gij P i P j = a2 (1 − 2Ψ)δij p̂i p̂j α2

= a2 (1 − 2Ψ)α2 , (3.11)
where the last equality holds because the direction vector is a unit vector. Solving (3.11) for α,
we get α = p(1 + Ψ)/a, or
pp̂i
Pi = (1 + Ψ) . (3.12)
a
Substituting these results into the geodesic equation (3.7), we get

p d hp i
(1 − Φ) (1 − Φ) = −Γ0αβ P α P β , (3.13)
a dτ a
where we have used the standard trick of rewriting the derivative with respect to λ as a derivative
with respect to time multiplied by dτ /dλ = P 0 . We expand out the time derivative to get
dp dΦ a2
(1 − Φ) = Hp(1 − Φ) + p − Γ0αβ P α P β (1 + Φ) . (3.14)
dτ dτ p
Multiplying both sides by (1 + Φ)/p and dropping all quadratic terms in Φ, we find
1 dp dΦ a2
=H+ − Γ0αβ P α P β 2 (1 + 2Φ) . (3.15)
p dτ dτ p
To evaluate the last term on the right-hand side, we use the perturbed Christoffel symbols (3.1). After
a bit of algebra, we get
P αP β ∂Ψ ∂Φ ∂Φ
− Γ0αβ (1 + 2Φ) = −2H + − − 2 p̂i i . (3.16)
p2 ∂τ ∂τ ∂x
Equation (3.15) then becomes

1 dp dΦ ∂Φ ∂Φ ∂(Φ + Ψ)
= −H + −2 + p̂i i + . (3.17)
p dt dτ ∂τ ∂x ∂τ
At leading order, the term in brackets is equal to the total time derivative of Φ:
dΦ ∂Φ dxi ∂Φ ∂Φ ∂Φ
= + = + p̂i i , (3.18)
dτ ∂τ dτ ∂xi ∂τ ∂x
14
where we used
dxi dxi dλ Pi
= = 0 = p̂i (1 + Ψ + Φ) = p̂i + O(1) . (3.19)
dτ dλ dτ P
Substituting (3.18) into (3.17), we get
1 dp 1 da dΦ ∂(Φ + Ψ)
=− − + , (3.20)
p dτ a dτ dτ ∂τ
which confirms the result (3.6).
Line-of-sight solution
By integrating the geodesic equation (3.6) along a line-of-sight, we can relate the observed CMB
temperature anisotropies to the fluctuations at recombination. To simplify matters, we will
therefore work with the idealised approximation of instantaneous recombination. The CMB
photons were then emitted at a fixed time τ∗ . This moment is often called last scattering.
Integrating (3.6) from the time of emission τ∗ to the time of observation τ0 , we then get
Z τ0
∂
ln(ap)0 = ln(ap)∗ + (Φ∗ − Φ0 ) + dτ (Φ + Ψ) . (3.21)
τ∗ ∂τ
To relate this to the temperature anisotropy, we note that

δT
ap ∝ aT̄ 1 + , (3.22)
T̄
where T̄ (τ ) is the mean temperature. Taylor-expanding the logarithms in (3.21) to first order in
δT /T̄ , and keeping in mind that a0 T̄0 = a∗ T̄∗ , we find
Z τ0
δT δT ∂
= + (Φ ∗ − Φ 0 ) + dτ (Φ + Ψ) . (3.23)
T̄ 0 T̄ ∗ τ∗ ∂τ
The term Φ0 only affects the monopole perturbation, so it is unobservable and therefore usually
dropped from the equation. The fractional temperature perturbation at last-scattering can be
expressed in terms of the density contrast of photons, δγ ≡ δργ /ργ , as

δT 1
= (δγ )∗ , (3.24)
T̄ ∗ 4

1
where the factor of 4 arises because ργ ∝ T 4 . Equation (3.23) then reads
Z τ0
δT 1 ∂
= δγ + Φ + dτ (Φ + Ψ) . (3.25)
T̄ 0 4 ∗ τ∗ ∂τ
Each term on the right-hand side has a simple physical interpretation:
• The term 41 δγ can be thought of as the intrinsic temperature variation over the background
last-scattering surface.
15
Figure 5. The motion of electrons at the surface of last-scattering produces an additional temperature
anisotropy.
• The term Φ arises from the gravitational redshift that the photons experience when climbing
out of a potential well at last-scattering. The combination 14 δγ +Φ is often called the Sachs-
Wolfe (SW) term.
• Finally, the integrated Sachs-Wolfe (ISW) term describes the effect of gravitational red-
shifting from evolution of the potentials along the line-of-sight. During matter domination,
Φ̇ ≈ Ψ̇ = 0 and this term vanishes.
So far, we have ignored the motion of the electrons at the surface of last-scattering. Including
this effect leads to an extra Doppler shift in the received energy of photons when referenced to
an observer comoving with the electrons at last-scattering (see Fig. 5),

δT
⊂ (n̂ · ve )∗ . (3.26)
T̄ 0
Putting everything together, we obtain the following important result
Z τ0
δT 1
(n̂) = δγ + Φ + n̂ · ve + dτ (Φ̇ + Ψ̇) , (3.27)
T̄ 4 ∗ τ∗
where we have dropped the subscript ‘0’ on the observed δT /T̄ to avoid clutter. Figure 6 illustrates
the contributions that each of the terms in (3.27) makes to the power spectrum of the CMB
temperature anisotropies (see §3.2). We see that the ISW contribution is subdominant and
that the shape of the power spectrum is mostly determined by the Sachs-Wolfe and Doppler
contributions.
Large scales.—For adiabatic initial conditions, the superhorizon initial condition is δγ ≈ 43 δm ≈ − 83 Ψ.

The Sachs-Wolfe term then becomes
1 2Ψ 1
δγ + Φ = − + Φ ≈ Φ. (3.28)
4 3 3
16
Total
SW
Doppler
ISW
Figure 6. Contributions of the various terms in (3.27) to the (rescaled) power spectrum of CMB
anisotropies, Cl ≡ l(l + 1)Cl .
This shows that, on large scales, an overdense region (Ψ ≈ Φ < 0) appears as a cold spot in the sky.
While the temperature at the bottom of the potential well is hotter than the average (− 23 Ψ), photons
lose more energy (Φ) as they climb out of the potential well, resulting in a cold spot ( 13 Φ < 0).
3.2 CMB Power Spectrum

A map of the cosmic microwave background radiation describes the variation of the CMB tem-
perature as a function of direction, δT (n̂). We will be interested in the statistical correlations
between temperature fluctuations in two different directions n̂ and n̂0 (see Fig. 7), averaged over
the entire sky.
If the initial conditions are statistically isotropic, then we expect these correlations only to
depend on the relative orientation of n̂ and n̂0 . In that case, we can write the two-point correlation
function as
X 2l + 1
δT (n̂)δT (n̂0 ) =

Cl Pl (cos θ) , (3.29)
4π
l
where n̂ · n̂0 ≡ cos θ and Pl are Legendre polynomials. The expansion coefficients Cl are the
angular power spectrum. If the fluctuations are Gaussian, then the power spectrum contains the
entire information of the CMB map.
The right panel of Fig. 7 illustrates the temperature variations created by a single plane
wave inhomogeneity. The CMB anisotropies observed on the sky are a superposition of many
such plane waves with amplitudes that are weighted by the spectrum of primordial curvature
perturbations Pζ (k). In Section 7, we will show that the initial conditions of the primordial
perturbations are expected to be featureless, Pζ (k) ≈ const. The observed features in the CMB
anisotropy spectrum arise from the subhorizon evolution of perturbations in the photon density
17
Figure 7. Left: Illustration of the two-point correlation function of the temperature anisotropy δT (n̂).
Right: Illustration of the temperature anisotropy created by a single plane wave inhomogeneity are recom-
bination.
and the metric. This evolution takes the form of cosmic sound waves (see Section 4). These waves
are captured at recombination and projected onto the sky. The observed oscillations in the CMB
power spectrum are therefore a snapshot of primordial sound waves caught at different phases
in their evolution at the time when photons last scattered off electrons. The beautiful physics of
the CMB fluctuations is described in detail in my Advanced Cosmology course [7].
Substituting the line-of sight solution (3.27) into the definition of the angular power spectrum
(3.29), we find (see insert below)
4π
Z
Cl = d ln k Tl2 (k) Pζ (k) . (3.30)
(2l + 1)2
The transfer function Tl (k) captures both the evolution of the initial fluctuations until recom-
bination and the projection onto the surface of last-scattering. Ignoring the subdominant ISW
contribution, we can write the transfer function as follows:
( 14 δγ + Φ)∗
TSW (k) ≡ ,
ζ(k)
Tl (k) = TSW (k)jl (kr∗ ) + TD (k)jl0 (kr∗ ) , where (3.31)
(ve )∗
TD (k) ≡ − .
ζ(k)
The subscript ∗ denotes quantities evaluated at recombination, with r∗ being the distance to last-
scattering. The Bessel function jl (kr∗ ) and its derivative jl0 (kr∗ ) act almost like delta functions
and map the Fourier modes k to the harmonic moments l ∼ kr∗ . Given that Pζ (k) is expected to
be nearly constant, the angular power spectrum Cl therefore measures the square of the transfer
18
function Tl (k) evaluated at k = l/r∗ :
4π h 2 i
2
Cl ∼ T (k) + T (k) , (3.32)
(2l + 1)2 SW D

k∼l/r∗
where we have dropped the cross term TSW (k) TD (k) because it is negligible. In the next section,
we will discuss the evolution effects that determine the transfer function and hence the CMB
power spectrum.
Projection.—To understand the origin of the Bessel functions in (3.31), let us consider the projection
of the Sachs-Wolfe term onto the surface of last-scattering. Assuming instantaneous recombination,
we can write
Z
δT (n̂) = dr δT (x, n) δD (r − r∗ ) (3.33)
d3 k i(kr∗ )k̂·n̂
Z
= e δT (k) , (3.34)
(2π)3
where we substituted the Fourier expansion of the temperature field in the second line. The exponential
in (3.34) can be written in a Rayleigh plane wave expansion,
X
ei(kr∗ )k̂·n̂ = (−i)l (2l + 1)jl (kr∗ )Pl (k̂ · n̂) . (3.35)
l
The two-point function of temperature anisotropies then becomes
d3 k d3 k 0 X X
Z Z
0
δT (n̂) δT (n̂0 ) = (−i)l+l (2l + 1)(2l0 + 1) jl (kr∗ )jl0 (k 0 r∗ )

(2π) 3 (2π) 3
0 l l
× Pl (k̂ · n̂)Pl0 (k̂0 · n̂0 ) hδT (k) δT (k0 )i . (3.36)
The power spectrum of the temperature field can be written in terms of the power spectrum of the
primordial curvature perturbations, cf. (2.31),
hδT (k)δT (k0 )i = TSW (k)TSW (k 0 )hζ(k) ζ(k0 )i

2 2π 2
= TSW (k) Pζ (k) δD (k − k0 ) . (3.37)
k3
The delta function allows us to trivially perform one of the momentum integrals in (3.36). To evaluate
the angular part of the second momentum integral, we use the following identity
4π
Z
d2 k̂ Pl (k̂ · n̂)Pl0 (k̂ · n̂0 ) = Pl (n̂ · n̂0 ) δll0 . (3.38)
2l + 1
The two-point function then takes the form (3.29) with the angular power spectrum given by
4π
Z
ClSW = d ln k jl2 (kr∗ ) TSW
2
(k) Pζ (k) . (3.39)
(2l + 1)2
↑ ↑ ↑
projection evolution initial conditions
A similar derivation gives the form of the Doppler contribution.
19
4 Cosmic Sound Waves
In the early universe, photons and electrons were strongly interacting, while the electrons were
strongly coupled to protons. The combined system is often called the photon-baryon fluid. In
this section, we will study the evolution of sound waves in this medium. These waves will evolve
in an inhomogeneous spacetime whose perturbations are sourced by all forms of matter in the
universe (see Fig. 8).
ion
d iat
Ra Neutrions
Dark
Photons
Energy
Thomson
Scattering Metric
Dark
Electrons Matter
Co
Sca ulom
tte b
rin
g Protons
Ba er
ryo tt
ns Ma
Figure 8. Interactions between the different forms of matter in the universe.
Our treatment in this section will be rather telegraphic and is just meant to give a flavor
for the beautiful physics underlying the CMB. Further details can be found in the notes of my
Advanced Cosmology course [7], or in the following textbooks [2, 8] and review articles [9–12].
4.1 Photon-Baryon Fluid

Combining the continuity and Euler equations for the photon-baryon fluid leads to an evolution
equation for the photon density perturbations [7]:
HR 4 4HR
δ̈γ + δ̇γ − c2s ∇2 δγ = ∇2 Φ + 4Ψ̈ + Ψ̇ , (4.1)
1+R 3 1+R
↑ ↑ ↑ ↑
friction pressure gravity time dilation
where R ≡ 34 ρ̄b /ρ̄γ is the ratio of the momentum densities of baryons and photons, and the sound
speed of the photon-baryon fluid is defined as
1
c2s ≡ . (4.2)
3(1 + R)
Equation (4.1) is the master equation describing the entire CMB phenomenology. The most
important terms in the equation are the photon pressure term on the left-hand side and the
20
I II III Planck
ACT
SPT
Figure 9. The angular variations of the CMB power spectrum are consequence of the dynamics of sound
waves in the photon-baryon fluid. On large scales (region I), the fluctuations are frozen and we directly
see the spectrum of the initial conditions. At intermediate scales (region II), we observe the oscillations
of the fluid as captured at the moment of last-scattering. Finally, on small scales (region III), fluctuations
are damped because their wavelengths are smaller than the mean free path of the photons.
gravitational forcing term on the right-hand side. In addition, we have a friction term proportional
to the baryon density R on the left-hand side and two terms related to time dilation effects on
the right-hand side. The metric potentials Φ and Ψ are determined by the Einstein equations
(which include important contributions from dark matter).
In practice, the equations describing the many coupled fluctuations in the primordial plasma
have to be solved numerically. To gain some intuition, however, it is nevertheless useful to obtain
approximate analytic results. In the following, we will solve equation (4.1) by making several
(more or less justified) approximations. Our goal is to understand the main features of the CMB
power spectrum shown in Fig. 9.
4.2 Acoustic Oscillations

During radiation domination, the baryon density is subdominant, R 1, so for the moment we
will set R = 0. For simplicity, we will also ignore the time dilation terms in (4.1); we will include
their effects in §5.2. Equation (4.1) can then be written as
Θ̈ − c2s ∇2 Θ = 0 , (4.3)
where c2s ≈ 13 and Θ ≡ 14 δγ + Φ is precisely the combination of the Sachs-Wolfe term appearing
in (3.27). Solving (4.3) for a single Fourier mode, we get
Θ(k, τ ) = Ak cos(cs kτ ) + Bk sin(cs kτ ) , (4.4)
21
where Ak and Bk are parameters that are fixed by the initial conditions. For adiabatic initial
conditions, all perturbations in the limit τ → 0 are analytic functions of k 2 , which is only the
case for the cosine part of the solution above. We therefore set Bk = 0. Moreover, the matching
to the superhorizon initial conditions implies Ak = 3ζk [7] and we get
Θ(k, τ ) = 3ζk cos(cs kτ ) . (4.5)
Evaluating this solution at recombination, τ = τ∗ , gives the Sachs-Wolfe transfer function,

TSW (k) ≡ Θ(k, τ∗ )/ζk = 3 cos(cs kτ∗ ). As we have seen in §3.2, the CMB power spectrum is
roughly given by the square of the transfer function. The oscillatory k-dependence of the transfer
function then maps to the observed oscillations in the CMB power spectrum in harmonic space,
cf. eq. (3.32).
Sound horizon Modes caught at extrema of their oscillations will have enhanced fluctuations
kn = nπ/s∗ , (4.6)
where s∗ ≡ cs τ∗ ≈ √1 τ∗is the sound horizon at recombination. We see that the peaks occur
3 √
at multiples of the fundamental scale k∗ ≡ π/s∗ ≈ 3π/τ∗ . This scale becomes a characteristic
angular scale by simple projection
λ∗
θ∗ = , (4.7)
DA
τ∗
l∗ = k∗ DA ≈ , (4.8)
τ0
where DA is the angular diameter distance (which in a flat universe is DA = τ0 − τ∗ ≈ τ0 ). As-
suming a purely matter-dominated universe after recombination, we have τ ∝ a1/2 and therefore
find
1/2
1
θ∗ ≈ ≈ 2◦ , (4.9)
1100
l∗ ≈ 200 . (4.10)
The presence of dark energy and spatial curvature would slightly change this result. Measure-
ments on the CMB spectrum have now determined θ∗ to better than 0.05%, which puts strong
constraints on the geometry and composition of the universe. Keeping the physical densities of
dark matter, baryons, photons and neutrinos fixed, i.e. Ωi h2 = const., the scale θ∗ is a measure of
the curvature parameter Ωk through its effect on the angular diameter distance to last-scattering.
The observed value of θ∗ is found to be consistent with a flat universe. Allowing the matter den-
sity and the Hubble parameter to vary, one finds θ∗ ≈ f (Ωm h3 ), i.e. there is a specific degeneracy
between variations in Ωm and h. This degeneracy is broken by measurements of the peak mor-
phology of the CMB spectrum and by external data sets (BAO, supernovae, etc.).
Baryon loading Let me briefly comment on the effects of baryons on the CMB spectrum. The
baryon-to-photon ratio increases with time, R ∝ a, reaching an order one value at recombination.
This decreases the sound speed, cf. (4.2). Instead of (4.3), we now have
d2 1
2
(Θ + RΦ) − ∇2 (Θ + RΦ) = 0 , (4.11)
dτ 3
22
where we have ignored the time variation of R relative to the much faster evolution of the acoustic
oscillations. We see that the finite baryon density, R 6= 0, changes the equilibrium point of the
oscillations from Θ = 0 to Θ = −RΦ. Since the CMB spectrum depends on the square of the
solution, the shift of the equilibrium of the oscillating solution leads to odd and even peaks in
the CMB having unequal heights. The relative heights of the CMB spectrum therefore provide
a measure of the baryon density Ωb .
Radiation driving One important effect is not included in our highly simplified treatment.
During the radiation era the gravitational potential Φ decays inside the horizon. Counterin-
tuitively, the decaying potential actually enhances temperature fluctuations through a subtle
resonance effect (see Fig. 10). Since the potential Φ decays after sound horizon crossing, it drives
the first compression of the photon-baryon fluid without a counterbalancing effect on the subse-
quent rarefaction stage. The higher peaks in the CMB spectrum correspond to fluctuations that
began their oscillations in the radiation-dominated era and therefore have enhanced amplitudes.
This effect is sensitive to the ratio of radiation to matter, Ωr /Ωm . Since the radiations density Ωr
is fixed by the observed CMB temperature, measuring the peak heights relative to the Sachs-Wolf
plateau determines the matter density of the universe Ωm .
damp
ing
dri
vin
g
5 10 15 20
Figure 10. Acoustic oscillations with gravitational forcing and diffusion damping. For a mode that
enters the sound horizon during radiation domination, the gravitational potential decays after horizon
crossing and drives the acoustic amplitude higher. As the photon diffusion length increases and becomes
comparable to the wavelength, viscosity σγ is generated and small-scale fluctuations are washed out.
Diffusion damping So far, we have treated photons and baryons as a single perfect fluid,
i.e. we took the mean free path of the photons to be zero. In reality, the coupling between
electrons and photons is imperfect and the photons have a finite mean free path:
1
λC = , (4.12)
ne σ T a
where ne is the electron density and σT is the Thomson cross section. Accounting for this leads
to the damping of small-scale fluctuations: see Fig. 9. As the photons random walk through the
23
baryons, hot and cold regions are mixed. By this process, fluctuations will be erased below the
diffusion length: √ p p
λD = N λC = τ /λC λC = τ λC , (4.13)
which is the geometric mean between the horizon scale and the mean free path. As we will show
in the following insert, the transfer function for the photon density fluctuations will receive an
exponential suppression for modes with k > kD ≡ 2π/λD . As shown in the insert below, the
photon transfer function receives the following correction:
T (k) → D(k)T (k) , (4.14)
2 /k 2
where D(k) = e−k D .
Imperfect fluid.—Diffusion causes heat conduction and generates viscosity in the fluid. Incorporating
these effects into the dynamics leads to a modified oscillator equation [7],
Θ̈ + µ c2s k 2 Θ̇ + c2s k 2 Θ = 0 , (4.15)
where we have ignored the gravitational source terms and defined
R2

16
µ≡ + λC . (4.16)
15 1 + R
Using the WKB ansatz Z

Θ ∝ exp i ω dτ , (4.17)
we get
− ω 2 + µ c2s k 2 iω + c2s k 2 = 0 , (4.18)
which we can write as
h i1/2 h i i
ω = ±cs k 1 + iωµ ≈ ±cs k 1 ± µ cs k . (4.19)
2
Substituting this back into (4.17), we get

1
Θ ∝= e±ikrs exp − (k/kD )2 , (4.20)
2
where we have defined the diffusion wavenumber as

R2

1 16
Z Z
−2 2
kD = dη µ cs = dτ + λC . (4.21)
3(1 + R) 15 1 + R
In the limit R → 0, this becomes

16
Z
−2
kD ≈ dτ λC ∼ τ λC , (4.22)
45
which agrees with our previous estimate (4.13).
Within the Standard Model, the physics of the CMB anisotropies is understood extremely well.
In the next section, we will explore what can be learned about physics beyond the Standard
Model by looking for subtle deviations in the CMB spectrum.
24
5 Light Relics
Future cosmological observations will measure the radiation density of the early universe at
the percent level. In this section, I will show how these observations will probe the physics of
neutrinos, as well as the possible existence of extra light particles that are more weakly coupled to
the SM than neutrinos. Examples of light relics that can be constrained in this way are the QCD
axion [13–15], axion-like particles (ALPs) [16], dark photons [17] and light sterile neutrinos [18].
These particles may be so weakly coupled that they are hard to detect in terrestrial experiments,
but the large number densities in the early universe make their gravitational imprints observable.
5.1 Dark Radiation

Let us assume that some physics beyond the Standard Model adds an extra radiation density ρX
to the early universe. It is conventional to measure this density relative to the density of a SM
neutrino species:
ρX 1 ρX
∆Neff ≡ = , (5.1)
ρν aν ργ
and define Neff = Nν + ∆Neff as the effective number of neutrinos, although ρX may have
nothing to do with neutrinos. Current measurements of the CMB anisotropies and the light
element abundances find
CMB
Neff = 3.04 ± 0.18 , (5.2)
BBN
Neff = 2.85 ± 0.28 , (5.3)
which is consistent with the SM prediction,1 Neff = 3.046. Future CMB observations have the
potential to improve these constraints by an order of magnitude [19].
A natural source for ∆Neff 6= 0 are extra relativistic particles. Let us therefore consider a light
species X as the only additional particle in some BSM theory. Assuming that this species was in
thermal equilibrium with the SM at some point in the history of the universe, we can compute
its contribution to Neff in the same way as we derived the relic density of neutrinos in §2.2.
For concreteness, let us assume that the particles of the species X decouple before neutrino
decoupling, Tdec,X > 10 MeV. Particle-antiparticle annihilations until neutrino decoupling will
lead to a difference between the temperature associated with the species X and that of neutrinos:
1/3
10.75 1/3
1/3
g∗ (Tdec,ν ) 106.75
TX = Tν = Tν
g∗ (Tdec,X ) 106.75 g∗ (Tdec,X )
1/3
106.75
= 0.465 Tν . (5.4)
g∗ (Tdec,X )
After neutrino decoupling, TX and Tν evolve in the same way, with both receiving the same
suppression relative to Tγ from e+ e− annihilation. As long as both X and ν are relativistic, they
1
The predicted value of Neff = 3.046 accounts for plasma corrections of quantum electrodynamics, flavour
oscillations and, in particular, the fact that neutrinos have not fully decoupled when electrons and positrons
annihilated.
25
5
Goldstone boson
Weyl fermion
2 Gauge boson
1
0.5
∆Neff
Current limit (2σ)
0.2
0.1
0.05 0.054
0.047
Future sensitivity 0.027
0.02
0.01
10−5 10−4 10−3 0.01 0.1 1 10 100 103 104 105 106
Tdec [GeV]
Figure 11. Contributions of a single thermally-decoupled Goldstone boson, Weyl fermion or massless
gauge boson to the effective number of neutrinos, ∆Neff , as a function of its decoupling temperature Tdec .
therefore maintain a constant energy ratio

4 4/3
ρX g∗,X TX 106.75
∆Neff ≡ = = 0.027 g∗,X , (5.5)
ρν g∗,ν Tν g∗ (Tdec,X )
where g∗,ν = 74 and g∗,X = {1, 47 , 2, . . .} are the internal degrees of freedom for spin {0, 21 , 1, . . .}
particles. Figure 11 shows the contribution of a single decoupled species as a function of its
decoupling temperature. We see that the contributions asymptote to fixed values for decou-
pling temperatures above the mass of the top quark (the heaviest particle of the SM). Using
g∗ (Tdec,X ) ≤ 106.75 in expression (5.5), we find that the extra species X contributes the follow-
ing minimal amount2


 0.054 gauge boson

∆Neff > 0.027 g∗,X = 0.047 Weyl fermion (5.6)


0.027 Goldstone boson

As we will see, this is an interesting target for future CMB experiments.
2
In deriving this bound, we assumed an extension of the SM in which there is no significant entropy production
after decoupling and that the species X is the only addition to the SM particle content. Additional particles may
both increase ∆Neff is there are light enough, or decrease it if they are relativistic at the decoupling of X, but
become non-relativistic before neutrino decoupling. While entropy production typically dilutes the effects of extra
relativistic species, it can also lead to additional effects that can be looked for in cosmological observations. For a
more detailed discussion of these issues, see [20].
26
5.2 Imprints in the CMB
Adding extra relativistic species will change the shape of the CMB spectrum. Precise measure-
ments of the spectrum therefore make the CMB an accurate tool for probing this type of BSM
physics.
Diffusion damping The main effect of adding radiation density to the early universe is to
increase the damping of the CMB spectrum [21] (see Fig. 12). Increasing Neff increases H∗ , the
expansion rate at recombination. This would change both the damping scale θD and the peak
location θ∗ . Using the estimates presented in §4.2, the ratio of θD and θ∗ scales as
θD 1 1 1/2
= ∝ = H∗ . (5.7)
θ∗ rs,∗ kD −1 1/2
H∗ H∗
Since θ∗ is measured very accurately, we need to keep it fixed. This can be done, for example, by
simultaneously increasing the Hubble constant H0 . Increasing Neff (and hence H∗ ) at fixed θ∗ then
implies larger θD , i.e. the damping kicks in at larger scales reducing the power in the damping tail
(see Fig. 12). By accurately measuring the small-scale CMB anisotropies, observations therefore
put a constraint on the number of relativistic species at recombination, cf. eq. (5.2).
The main limiting factor in these measurements is a degeneracy with the primordial Helium
fraction YP ≡ nHe /nb . At fixed ωb ≡ Ωb h2 , increasing YP decreases the number density of free
electrons. This increases the diffusion length, cf. eq. (4.12), and hence reduces the power in the
damping tail. The parameters YP and Neff are therefore anti-correlated. As we will discuss next,
this degeneracy is broken by a more subtle effect of free-streaming relativistic species on the CMB
spectrum.
Phase shift Recently, CMB experiments have started to become sensitive to perturbations in
the gas of relativistic particles. As we will see, perturbations in the density of free-streaming
103
Cl [µK2 ]
4
Neff
3
102
1
0
2 10 30 500 1000 1500 2000 2500
l
Figure 12. Variation of the CMB spectrum Cl ≡ l(l + 1)Cl as a function of Neff for fixed θ∗ .
27
9.0 7
6.0
3.0 6
0.0
6.0 5
Kl [103 µK2 ]
3.0
0.0 4
Neff
6.0
3
3.0
0.0
0 500 1000 1500 2000 2500 2
5.5
1
4.5
3.5
0
950 1000 1050 1100 1150 1200 1250
l
Figure 13. Variation of the undamped power spectra, Kl ≡ Dl−1 Cl , as a function of Neff . The physical
baryon density ωb , the matter-to-radiation ratio ρm /ρr and the angular size of the sound horizon θ∗ are
held fixed in all panels. The dominant effect in the first panel is the variation of the damping scale θD .
In the second panel, we fixed θD by adjusting the Helium fraction YP . The dominant variation is now
the amplitude perturbation δA. In the third panel, the spectra are normalized at the fourth peak. The
remaining variation is the phase shift ϕ (see the zoom-in in the fourth panel).
relativistic particles (e.g. neutrinos) leave a small imprint in the temporal phase of the acoustic
oscillations and hence a coherent shift in the peak locations of the CMB spectrum (see Fig. 13).
To get some intuition for the physical origin of this effect, let us return to the master equa-
tion (4.1), but now solve it slightly more accurately. Since the effect occurs during the radiation-
dominated era, we can still assume perfect radiation domination with R = 0. Equation (4.1)
then becomes [22]
d¨γ − c2s ∇2 dγ = ∇2 Φ+ , (5.8)
where we have defined dγ ≡ 34 δγ − 3Φ and Φ± ≡ Φ ± Ψ. The solution for a single Fourier mode
can be written as Z y
dγ (y) = A cos(y) − c−2
s dy 0 Φ+ (y 0 ) sin(y − y 0 ) , (5.9)
0
where y ≡ cs kτ and we have dropped the argument k on dγ , A, and Φ+ to avoid clutter. Using
sin(y − y 0 ) = sin(y) cos(y 0 ) − cos(y) sin(y 0 ), we can write this as
dγ (y) = A + c−2 −2

s α(y) cos(y) − cs β(y) sin(y) , (5.10)
where
Z y
α(y) ≡ dy 0 Φ+ (y 0 ) sin(y 0 ) , (5.11)
0
Z y
β(y) ≡ dy 0 Φ+ (y 0 ) cos(y 0 ) . (5.12)
0
28
To obtain the CMB spectrum, we need to evaluate the solution at recombination. For the high-l
modes of the CMB, it is a good approximation to use y → ∞ for the limit of integration. The
solution (5.10) can then be written as
dγ (y) = Ã cos(y + ϕ) , (5.13)
where Ã ≡ A + c−2
s α and
β
sin ϕ ≡ p . (5.14)
β2 + (α + c2s A)2
We see that β 6= 0 corresponds to a constant phase shift of the acoustic oscillations.
To diagnose when such a phase shift can occur, it is useful to write the parameter β as follows
Z ∞
β = dy 0 Φ+ (y 0 ) cos(y 0 )
0
1 +∞ 0 h (s) 0 i iy0
Z
= dy Φ+ (y ) e , (5.15)
2 −∞
where, in the second line, we have analytically continued the integrand and defined the symmetric
(s) (s)
part of the potential, Φ+ (y) ≡ Φ+ (y) + Φ+ (−y). For adiabatic modes, we expect Φ+ (y) to be
an analytic function. Closing the contour in the upper-half plane, we find β = 0 if the contour at
(s)
infinity vanishes. This is the case if Φ+ is sourced by fluctuations that travel at c < cs . Neutrinos
or other free-streaming particles, on the other hand, travel at the speed of light. This induces a
(s)
mode in Φ+ of the form e−ickη = e−i(c/cs )y , with c = 1 > cs , and therefore leads to a finite phase
shift.
Neutrino free-streaming.—We will briefly sketch how free-streaming relativistic particles, such as neu-
trinos, produce a phase shift in the CMB anisotropy spectrum. Details can be found in [5, 22].
The evolution of Φ+ is related to that of Φ− via the following Einstein equation:
4 2
Φ00+ + Φ0+ + Φ+ = Φ00− + Φ0− + 3Φ− ≡ S[Φ− ] , (5.16)
y y
where Φ− is sourced by the anisotropic stress πν created by the neutrinos:
2k 2
Φ− (y) = − fν πν (y) . (5.17)
y2
P
Here, we have introduced the fractional neutrino density fν ≡ ρν /ρr ≈ 0.41. The evolution of πν
follows from the Boltzmann equation for the neutrino distribution function. The solution can be
written as [22]
Z y
k2

2 3
πν (y) ≈ −ζ j2 (c−1
s y) + c−1
s dy 0
Φ + (y 0
) j (c
1 s
−1
(y − y 0
)) − j (c
3 s
−1
(y − y 0
)) . (5.18)
2 0 5 5
We see that the solution involves an integral over Φ+ . Moreover, the solution depends on c−1 s y = kτ ,
i.e. it contains modes travelling at the speed of light. Following [22], the system of equations can be
solved order by order in the fractional neutrino density:
29
• At zeroth order in fν , the potential Φ− vanishes and the homogeneous solution is a function only
of y, i.e. it doesn’t contain modes travelling faster than the sound speed of the photon-baryon
fluid. No phase shift is generated.
• At first order in fν , the potential Φ− is non-zero. Note that the right-hand side of (5.17) is
proportional to fν , so only the zeroth-order solution for πν (and Φ+ ) is needed to determine the
first-order solution for Φ− . Computing the induced first-order correction to Φ+ , one finds [5, 22]
β ≈ 0.60 ζ fν and ϕ ≈ 0.19 π fν . (5.19)
As expected, a finite phase shift is generated. This phase shift has recently been detected in
the CMB data [4, 5]. It has also been measured in the clustering of galaxies via its imprint in
the spectrum of baryon acoustic oscillations [23].
CMB Stage 4 The sensitivity of ground-based CMB experiments can be characterized by the
number of detectors that are mounted onto the telescope. The current generation of experiments
has about 103 detectors, but there are plans to increase the number of detectors by up to two
orders of magnitude [19]. This would lead to a significant improvement in the sensitivity of CMB
experiments (see Fig. 14). What is particularly exciting about this is that it will lead to an order
of magnitude improvement in constraints on Neff , allowing us to probe particles that decoupled
before the QCD phase transition (cf. Fig. 11). If these so-called CMB Stage 4 experiments can
reach the threshold ∆Neff = 0.027, they would be sensitive to any light relics that have ever been
in thermal equilibrium with the Standard Model (modulo the constraints described in footnote 2).
As we will show in the next section, even the absence of a detection would be informative since
it would put strong constraints on the couplings of extra light species to the SM.
WMAP
CMB-S1
Planck
CMB-S2
CMB-S3
CMB-S4
2000 2005 2010 2015 2020

Year
Figure 14. Evolution of the sensitivity of past and future CMB experiments (figure adapted from [19]).
30
5.3 EFT of Light Species
Extra light species arise in many models of physics beyond the Standard Model. Rather than
studying these models one by one, it is more efficient to pass directly to an effective field theory
for the couplings of the new light fields X to the SM,
X
L ⊂ g OX OSM . (5.20)
The strength of the couplings g determines the decoupling temperature, Tdec (g), and hence the
thermal abundance of the species X, cf. Fig. 11.
Following [24], we consider EFTs that are minimal and technically natural. Minimality here
means that the additional particle content is as small as possible, usually consisting of only one
additional elementary particle. Naturalness requires the small masses of the extra particles to
be protected by symmetries, which also restricts the allowed interactions in (5.20). Since the
available symmetries depend on the spin of the new particles, it is convenient to organize the
EFT according to spin.
Spin-0 A particularly well-motivated example of light particles are Goldstone bosons created
by the spontaneous breaking of additional global symmetries. Goldstone bosons are either mass-
less (if the broken symmetry was exact) or naturally light (if it was approximate). Examples of
light pseudo-Nambu-Goldstone bosons (pNGBs) are axions [13–15], familons [25–27], and ma-
jorons [28, 29], associated with spontaneously broken Peccei-Quinn, family and lepton-number
symmetry, respectively. Axion, familon and majoron models are characterized by different cou-
plings in (5.20).
Axions arise naturally in many areas of high-energy physics, the QCD axion being a famous
example. They are a compelling example of a new particle that is experimentally elusive because
of its weak coupling rather than due to kinematic constraints. What typically distinguishes
axions from other pNGBs are their unique couplings to the SM gauge fields. Below the scale
of electroweak symmetry breaking, the shift-symmetric couplings of the axion to the SM gauge
fields are
1 a µν a a µν,a
L=− Fµν F̃ + G G̃ , (5.21)
4 Λγ Λg µν
where Xµν ≡ {Fµν , Gaµν } are the field strengths of photons and gluons, and X̃ µν ≡ 12 µνρσ Xρσ
are their duals. Axion models will typically include couplings to all SM gauge fields, but only
the coupling to gluons is strictly necessary to solve the strong CP problem.
Spin- 21 Light fermions are a natural possibility because Weyl and Dirac mass terms are pro-
tected by chiral and axial symmetries, respectively. A hidden Weyl fermion χ then couples to the
SM through an anapole moment, χ† σ̄ µ χ∂ ν Bµν , or a four-fermion interaction, χ† σ̄ µ χψγµ ψ, while
a Dirac fermion Ψ can couple through a dipole interaction, Ψ̄σ µν ΨBµν .
Spin-1 Massless spin-1 particles are technically natural because they have fewer degrees of
freedom than their massive counterparts. The dominant coupling of dark photons A0µ to the SM
is through the dipole interaction H ψ̄σ µν ψFµν
0 .
31
Spin- 32 The gravitino is a universal prediction of supergravity. Its mass is set by the SUSY
breaking scale, m3/2 = F/Mpl , and can be very small in low-scale SUSY breaking scenarios. The
coupling of the longitudinal component of the gravitino to the SM is equivalent to the Goldstino
coupling χ† σµ ∂ν χ T µν . The strength of the coupling is set by the SUSY-breaking scale F rather
than Mpl .
Spin-2 The graviton interacts only through Planck-suppressed gravitational interactions and
hence has never been in thermal equilibrium with the SM. Its thermal abundance is therefore
negligible.
Cosmic axions Let us illustrate the power of future CMB observations through the example
of axions [30]. To simplify the narrative, I will assume that a future CMB-S4 mission will be
sensitive enough to exclude the minimal abundance of thermal axions, ∆Neff > 0.027. In practice,
this will probably require additional data from large-scale structure surveys [31].
At high energies, the rate of axion production is through the gauge field interactions (5.21)
and can be expressed as [32]
X T3
Γ(Λn , T ) = γn (T ) 2 . (5.22)
n
Λn
The prefactors γn (T ) have their origin in the running of the couplings and are only weakly
dependent on temperature. We will ignore this temperature dependence in the following. We see
that the production rate, Γ ∝ T 3 , decreases faster than the expansion rate during the radiation
era, H ∝ T 2 . To avoid producing a thermal axion abundance requires that the interaction rate
was never larger than the expansion rate. Denoting the reheating temperature of the universe
by TR , this implies
π √ T2
Γ(Λn , TR ) < H(TR ) = √ g∗,R R , (5.23)
90 Mpl
where g∗,R ≡ g∗ (TR ). For a given reheating temperature, this is a constraint on the couplings Λn
in (5.22). Treating the different axion couplings separately, we can write
−1/4
π2

p
Λn > g∗,R γn,R TR Mpl , (5.24)
90
where γn,R ≡ γn (TR ).

The operator that has been most actively investigated experimentally is the coupling to pho-
tons. Photons are easily produced in large numbers in both the laboratory and in many astro-
physical settings which makes this coupling a particularly fruitful target for axion searches. The
couplings in the high-energy theory prior to electroweak symmetry breaking are related to the
photon coupling Λγ through the Weinberg mixing angle. To be conservative, I will present the
weakest constraint which arises when the axion only couples to the U (1)Y gauge field. A specific
axion model is likely to also couple to the SU (2)L sector and the constraint on Λγ would then
be stronger. Using γγ,R ≈ γγ (1010 GeV) = 0.029 and g∗,R = 106.75 + 1, we find
Λγ > 1.4 × 1013 GeV

p
TR,10 , (5.25)
32
Helioscopes (CAST)
Stellar Cooling
ADMX
CMB-S4
n
io
ax
CD
Q
Figure 15. Comparison between current constraints on the axion-photon coupling and the sensitivity of
a future CMB-S4 mission (figure adapted from [33]). The yellow band indicates a range of representative
models for the QCD axion. The future CMB bound is a function of the reheating temperature TR . We
note that ADMX assumes that the axion is all of the dark matter, while all other constraints do not have
this restriction.
where TR,10 ≡ TR /1010 GeV. For a reheating temperature of about 1010 GeV, the bound in (5.25)
is three orders of magnitude stronger than the best current constraints (cf. Fig. 15). Even for a
reheating temperature as low as 104 GeV the bound from the CMB would still marginally improve
over existing constraints.
The coupling to gluons is especially interesting for the QCD axion since it has to be present
in order to solve the strong CP problem. The axion production rate associated with the gluon
interaction in (5.21) is Γg ' 0.41T 3 /Λ2g [32]. The bound (5.24) then implies
Λg > 5.4 × 1013 GeV

p
TR,10 . (5.26)
Laboratory constraints on the axion-gluon coupling are usually phrased in terms of the induced
electric dipole moment (EDM) of nucleons: dn = gd a0 , where a0 is the value of the local axion
field. For the QCD axion, the coupling gd is given by [34, 37]
2π 3.8 × 10−3 GeV−1

gd ≈ × . (5.27)
αs Λg
Constraints on gd (and hence Λg ) are shown in Fig. 16. We see that future CMB-S4 observations
can improve over existing constraints on Λg by up to six orders of magnitude if TR = O(1010 GeV).
Even if the reheating temperature is as low as 104 GeV, the future CMB constraints will be tighter
by three orders of magnitude.
33
M
ED
ic
at
St
SN 1987A
CMB-S4
on
N axi
D
BB QC
Figure 16. Comparison between current constraints on the axion-gluon coupling and the sensitivity of
a future CMB-S4 mission (figure adapted from [34, 35]). We note that the static EDM [34] and BBN
constraints [35] assume that the axion is all of the dark matter, while SN 1987A [36] and the future CMB
constraint do not have this restriction.
Deriving similar constraints for the other axion couplings and for the couplings of fields with
spin is left as an exercise for the reader.
34
Part II
Relics from Inflation
Inflation predicts that the quantum fluctuations of any massless fields get amplified by the rapid
expansion of the spacetime. Two massless fields that are guaranteed to exist in all inflationary
models are the curvature perturbation ζ and the tensor fluctuations γij . In Part I of these lectures,
we assumed a nearly scale-invariant spectrum of curvature perturbations as a source for the
density fluctuations in the late universe. In the following, I will show that these initial conditions
indeed naturally arise from inflation. I will also demonstrate that the same effect produces tensor
fluctuations. The search for this stochastic background of primordial gravitational waves is a
very active area of observational cosmology.
Inflation also excites massive particles as long as their masses aren’t far above the inflationary
Hubble scale. Since the Hubble scale during inflation may be as high as 1014 GeV, this gives us
the opportunity to probe the particle spectrum at energies far beyond the reach of conventional
particle colliders. Once produced, these massive particles quickly decay into the massless modes
ζ and γij , creating higher-order correlations in the inflatonary fluctuations:
As we will see, this effect leads to a characteristic non-locality in cosmological correlators.
We will start with a brief review of slow-roll inflation. In Section 6, we will discuss the dynamics
of the inflationary background, while in Section 7 we will explicitly compute the spectrum of
quantum fluctuations. In Section 8, we will extend this treatment to include interactions and
show that they lead to a characteristic non-Gaussianity in cosmological correlators. In Section 9,
we will discuss the imprints of extra fields, showing how the masses and spins are encoded in the
momentum dependence of higher-order correlation functions.
35
6 Inflationary Cosmology
A key fact about the universe is that on large scales it is described by the FRW metric (2.2).
But why? A naive extrapolation of the radiation-dominated Big Bang cosmology suggests that
the early universe consisted of many causally disconnected regions of space. The fact that these
apparently disjoint patches of space are observed to have nearly the same densities and tempera-
tures is called the horizon problem. In this section, I will explain how inflation—an early period
of accelerated expansion—drives the primordial universe towards homogeneity and isotropy, even
if it started in a more generic initial state.
recombination
singularity
surface of
last-scattering
Figure 17. Illustration of the horizon problem in the conventional Big Bang model. All events that we
currently observe are on our past light cone. The intersection of our past light cone with the spacelike slice
labelled “recombination” corresponds to the “surface of last-scattering”. Any two points on the surface
of last-scattering that are separated by more than 1 degree, appear never to have been in causal contact.
This means that their past light cones do not overlap before the singularity. This is illustrated for opposite
points on the sky labelled p and q.
6.1 Horizon Problem

The particle horizon is the maximal distance that a signal can travel between the time correspond-
ing to the initial singularity, ti ≡ 0, and a later time t. In physical coordinates, this distance is
given by Z t Z a
dt d ln a
D(t) = a(t) = a(t) . (6.1)
0 a(t) 0 a0
If the early universe was filled by ordinary matter, then a00 < 0. In that case, the integral in (6.1)
is dominated by late times and converges to a finite value:
 
 t2/3  3t matter
e.g. a(t) ∝ → D(t) = (6.2)
 t1/2  2t radiation
This leads to a puzzle: because the age of the universe (t0 ) is much larger than the time of
recombination (t∗ ), the CMB naively consists of many causally disconnected patches (see Fig. 17).
36
The following questions immediately arise: Why is the CMB so homogeneous? And, more
importantly, why are the observed CMB fluctuations correlated on large scales and not just
random noise?
The horizon problem is solved if the early universe experienced a sustained period of accel-
erated expansion (= inflation), a00 > 0. In that case, the integral in (6.1) is dominated by early
times and the particle horizon diverges in the past. Signals were therefore able to travel a much
larger distance than suggested by the naive extrapolation of the standard FRW expansion (see
Fig. 18).
BIG BANG
recombination
end of inflation
INFLATION
causal
contact
singularity
Figure 18. Illustration of the inflationary solution to the horizon problem in comoving coordinates (using
conformal time on the vertical axis). The spacelike singularity of the standard Big Bang is replaced by
the reheating surface, i.e. rather than marking the beginning of time it now simply corresponds to the
transition from the end of inflation to the standard Big Bang evolution. All points in the CMB have
overlapping past light cones and therefore originated from a causally connected region of space.
Exercise.—A special case of accelerated expansion is the quasi-de Sitter limit, which is characterized
by a nearly constant expansion rate, H = a0 /a ≈ const., so that a(t) = eH(t−t0 ) , where t0 is some
fiducial time at which a(t0 ) ≡ 1. Show that
1
a(τ ) = − , (6.3)
Hτ
for τ < 0. Notice that the initial singularity has been pushed to τ = −∞ (cf. Fig. 18).
Exercise.—Show that a00 > 0 is equivalent to a slow variation of the Hubble parameter
H0
ε≡− < 1. (6.4)
H2
37
Notice that ε ≈ 0 corresponds to a quasi-de Sitter spacetime with a nearly constant expansion rate,
H ≈ const. Using the Friedmann equations, show that

3 P P 1
ε= 1+ < 1 ⇔ w≡ <− . (6.5)
2 ρ ρ 3
The last condition corresponds to a violation of the strong energy condition.
6.2 Slow-Roll Inflation

As a simple toy model for inflation, let us consider the dynamics of a scalar field, the inflaton
φ(t, x). As indicated by the notation, the value of the field can depend on time t and the position
in space x. Associated with each field value is a potential energy density V (φ) (see Fig. 19). If
the field is dynamical (i.e. changes with time) then it also carries a kinetic energy density. If the
energy density associated with the scalar field dominates the universe, it sources the evolution of
the FRW background. We want to determine under which conditions this can lead to accelerated
expansion.
Figure 19. Example of a slow-roll potential. Inflation occurs in the shaded parts of the potential.
The stress-energy tensor of the scalar field is

1 αβ
Tµν = ∂µ φ∂ν φ − gµν g ∂α φ∂β φ − V (φ) . (6.6)
2
Consistency with the symmetries of the FRW spacetime requires that the background value of
the inflaton only depends on time, φ = φ(t). From the time-time component T 0 0 = −ρφ , we infer
that the energy density of the field is
1 0 2
ρφ = (φ ) + V (φ) . (6.7)
2
We see that this is simply the sum of the kinetic energy density, 12 (φ0 )2 , and the potential energy
density, V (φ). From the space-space component T i j = Pφ δji , we find that the pressure is the
difference of kinetic and potential energy densities,
1 0 2
Pφ = (φ ) − V (φ) . (6.8)
2
38
A field configuration therefore leads to inflation, Pφ < − 31 ρφ , if the potential energy dominates
over the kinetic energy. i.e. if the field rolls slowly.
Exercise.—Using the Einstein equations (2.9) and (2.10), show that


3Mpl2
H 2 = 12 (φ0 )2 + V  1
(φ0 )2
⇒ ε = 2 2 2 < 1, (6.9)
Mpl2
H 0 = − 21 (φ0 )2
 Mpl H
where Mpl = (8πG)−1/2 is the reduced Planck mass.
Equation (6.9) corresponds to the first slow-roll condition:
(φ0 )2 < V . (6.10)
This condition alone, however, does not guarantee successful inflation. We also need to assure
that inflation does not just occur for an instant, but lasts long enough to solve the horizon
problem.
Combining the two Friedmann equations in (6.9) leads to the Klein-Gordon equation for the
evolution of the scalar field
φ00 + 3Hφ0 = −V,φ , (6.11)
where V,φ denotes the derivative of the potential with respect to the field value. To maintain the
slow-roll condition (6.10) for a sufficient period of time, we require that the acceleration of the
field is small. This is quantified by the second slow-roll condition:
φ00 < 3Hφ0 . (6.12)
The evolution of the scalar field is then friction dominated, with the velocity of the field deter-
mined by the slope of the potential.
The two slow-roll conditions (6.10) and (6.12) can be expressed as conditions on the shape of
the inflaton potential:
2
Mpl
2
V,φ 2 V,φφ

εV ≡ < 1, ηV ≡ Mpl < 1. (6.13)
2 V V
We will refer to the parameters in (6.13) as the (potential) slow-roll parameters.
Exercise.—Applying the slow-roll conditions (6.10) and (6.12) to the Friedmann equation (6.9) and
the Klein-Gordon equation (6.11), we get
2
3Mpl H2 ≈ V , (6.14)
0
3Hφ ≈ −V,φ . (6.15)
Use this to show that ε ≈ εV during slow-roll inflation.
39
Inflation ends when the first slow-roll condition is violated, εV (φe ) ≡ 1. The amount of
inflation is measured in terms of ‘e-folds’ of expansion, dN = d ln a. The total number of e-folds
between a point φ on the potential and the end of inflation at φe is
ae te φe
H
Z Z Z
N (φ) ≡ d ln a = H dt = dφ
a t φ φ0
φe φe
1 |dφ| 1 |dφ|
Z Z
= √ ≈ √ . (6.16)
φ 2ε Mpl φ 2εV Mpl
To solve the horizon problem requires at least between 40 and 60 e-folds (with the precise value
depending on the reheating temperature).
Case study: m2 φ2 inflation.—As an example, let us give the slow-roll analysis of arguably the simplest
model of inflation: single-field inflation driven by a mass term
1 2 2
V (φ) = m φ . (6.17)
2
The slow-roll parameters are
2
Mpl
εV (φ) = ηV (φ) = 2 . (6.18)
φ
To satisfy the slow-roll conditions {εV , |ηV |} < 1, we therefore need to consider super-Planckian values
for the inflaton √
φ > 2Mpl ≡ φe . (6.19)
The relation between the inflaton field value and the number of e-folds before the end of inflation is
φ2 1
N (φ) =
4Mpl2 − 2 . (6.20)
Solving the horizon problem then requires that the initial value of the field, φi , satisfies
√
φi > φ60 ≡ 2 60 Mpl ∼ 15 Mpl . (6.21)
We note that the total field excursion is super-Planckian, ∆φ = φi − φe Mpl .
6.3 Effective Field Theory

In the absence of a complete microscopic theory of inflation, we describe inflation in the context
of an effective field theory. We are then obliged to include in the inflationary action all operators
consistent with the assumed symmetries of the inflaton,
1 X φ2n X (∂φ)2n
Leff (φ) = − (∂µ φ)2 − V (φ) + cn V (φ) 2n + dn 4n
+ ··· . (6.22)
2 n
Λ n
Λ
One of the remarkable features of inflation is that it is sensitive even to Planck-suppressed oper-
ators.
40
Eta problem Successful inflation requires that the inflaton mass mφ is parametrically smaller
than the Hubble scale H:
m2φ
ηV = 1. (6.23)
3H 2
It is difficult to protect this hierarchy against high-energy corrections. We know that some new
degrees of freedom must appear at Λ . Mpl to give a UV-completion of gravity. In string theory,
this scale is often found to be significantly below the Planck scale, Λ . Ms Mpl . If φ has
order-one couplings to any massive fields ψ (with mψ ∼ Λ), then integrating out the fields ψ
yields the effective action (6.22) for φ with order-one couplings cn and dn . The above argument
makes us worry that integrating out the massive fields ψ yields corrections to the potential of the
form
φ2
∆V = c1 V (φ) 2 , (6.24)
Λ
where c1 ∼ O(1). If this term arises, then the eta parameter receives the following correction
2
Mpl
2
00 Mpl
∆ηV = (∆V ) ≈ 2c1 > 1, (6.25)
V Λ
where the final inequality follows from Λ . Mpl . Notice that this problem is independent of the
energy scale of inflation. All inflationary models have to address the eta problem.
Large-field inflation The Planck-scale sensitivity of inflation is dramatically enhanced in mod-

els with observable gravitational waves, r & 0.01. In this case, the inflaton field moves over a
super-Planckian range during the last 60 e-folds of inflation, ∆φ > Mpl (see §7.3), and an infinite
number of operators contribute equally to the effective action (6.22). This observation makes an
effective field theorist nervous and a string theorist curious [38].
41
7 Quantum Initial Conditions
One of the most remarkable features of inflation is that it provides a natural mechanism for
producing the initial conditions for the hot big bang. To see this, recall that the evolution of the
inflaton field φ(t) governs the energy density of the early universe ρ(t) and, hence, controls the end
of inflation (see Fig. 20). Essentially, the field φ plays the role of a “clock” reading off the amount
of inflationary expansion still to occur. By the uncertainty principle, arbitrarily precise timing is
not possible in quantum mechanics. Instead, quantum-mechanical clocks necessarily have some
variance, so the inflaton will have spatially varying fluctuations δφ(t, x). There will therefore be
local differences in the time when inflation ends, δt(x), so that different regions of space inflate
by different amounts. These differences in the local expansion histories lead to differences in the
local densities after inflation, δρ(t, x), and to curvature perturbations in comoving gauge, ζ(x).
It is worth remarking that the theory was not engineered to produce these fluctuations, but that
their origin is instead a natural consequence of treating inflation quantum mechanically.
Figure 20. Quantum fluctuations δφ(t, x) around the classical background evolution φ̄(t). Regions acquir-
ing negative fluctuations δφ remain potential-dominated longer than regions with positive δφ. Different
parts of the universe therefore undergo slightly different evolutions. After inflation, this induces density
fluctuations δρ(t, x).
7.1 Quantum Fluctuations

7.1.1 Free Scalar in de Sitter
Before attacking the real problem of interest, namely the quantization of coupled inflaton-metric
fluctuations during inflation, we will consider the simpler case of a free scalar field in de Sitter
space. We will assume that the scalar field carries an insignificant amount of the total energy
density and, hence, doesn’t backreact on the de Sitter geometry. Such a field is sometimes called
a spectator field.
The action of a massless, free scalar field in de Sitter space is
1 √
Z
S = d4 x −g g µν ∂µ ϕ∂ν ϕ
2
1
Z h i
= dτ d3 x a2 ϕ̇2 − (∂i ϕ)2 , (7.1)
2
42
where a(τ ) = −1/(Hτ ). It is useful to define the canonically normalized field v ≡ aϕ, so that

1 ä
Z
S = dτ d3 x v̇ 2 − (∂i v)2 + v 2 . (7.2)
2 a
This action implies the following equation of motion for the Fourier modes of the field

ä
2
v̈k + k − vk = 0 . (7.3)
a
Sometimes this is called the Mukhanov-Sasaki (MS) equation.

We see that the expansion of the universe is captured by the time dependence of the effective
mass of the canonically-normalized field, m2eff (τ ) ≡ ä/a = 2/τ 2 . At early times, −τ k −1 , this
mass is much smaller than the momentum k of the relevant Fourier modes. The dynamics then
reduces to that of a simple harmonic oscillator in flat space,
v̈k + k 2 vk ≈ 0 . (7.4)
In this limit, the quantum fluctuations of the field v therefore follow from the standard quanti-
zation of a simple harmonic oscillator.
Canonical quantization
From the action (7.2), we read off the the momentum conjugate to v:
∂L
π≡ = v̇ . (7.5)
∂ v̇
We promote the fields v(τ, x) and π(τ, x) to quantum operators v̂(τ, x) and π̂(τ, x). These
operators satisfy the equal time commutation relation
[v̂(τ, x), π̂(τ, x0 )] = iδD (x − x0 ) , (7.6)
in units where ~ ≡ 1. The delta function is required by locality: modes at different points in
space are independent and the corresponding operators therefore commute. In Fourier space, this
becomes
Z Z
0 0
[v̂k (τ ), π̂k0 (τ )] = d3 x d3 x0 [v̂(τ, x), π̂(τ, x0 )] e−ik·x e−ik ·x
| {z }
iδD (x − x0 )
Z
0
= i d3 x e−i(k+k )·x
= i (2π)3 δD (k + k0 ) , (7.7)
where the delta function implies that modes with different wavelengths commute.
Note that we are using the Heisenberg picture where operators vary in time while states are
time independent. The operator solution v̂k (τ ) is determined by two initial conditions v̂k (τi )
and π̂k (τi ) = ∂τ v̂k (τi ). Since the evolution equation is linear, the solution is linear in these
43
operators. It is convenient to trade v̂k (τi ) and π̂k (τi ) for a single time-independent non-Hermitian
operator âk , in terms of which the solution can be written as
v̂k (τ ) = vk (τ ) âk + vk∗ (τ )a†−k , (7.8)
where the (complex) mode function vk (τ ) satisfies the classical equation of motion (7.3). Of
course, vk∗ (τ ) is the complex conjugate of vk (τ ) and â†k is the Hermitian conjugate of âk . As
indicated by dropping the vector notation k on the subscript, the mode functions, vk (τ ) and
vk∗ (τ ), are the same for all Fourier modes with k ≡ |k|.3
We choose the normalization of the mode functions, so that vk v̇k∗ − v̇k vk∗ ≡ i. Substituting (7.8)
into (7.7), we then get
[âk , â†k0 ] = (2π)3 δD (k + k0 ) , (7.9)
which is the standard commutation relation for the raising and lowering operators of a harmonic
oscillator. The quantum states in the Hilbert space are constructed by defining the vacuum
state |0i via
âk |0i = 0 , (7.10)
and by producing excited states through repeated application of the creation operators a†k .
Choice of vacuum
The most general solution of equation (7.3) is

i −ikτ i
vk (τ ) = c1 1 − e + c2 1 + eikτ , (7.11)
kτ kτ
where the constraint on the overall normalisation of the mode functions, vk v̇k∗ − v̇k vk∗ ≡ i, implies
1
|c1 |2 − |c2 |2 = . (7.12)
2k
At this point, we still have a one-parameter family of solutions for the mode function vk (τ ). A
change in vk (τ ) could be accompanied by a change in âk that keeps the solution v̂k (τ ) unchanged.
Each such solution corresponds to a different vacuum state, cf. eq. (7.10). However, a special
choice of vk (τ ) is selected if we require the vacuum state |0i to be the ground state of the
Hamiltonian.
To see this, consider the Hamiltonian operator

1 2 1 1 ä 2
Z
3 2
Ĥ = d x π̂ + (∇v̂) + v̂ . (7.13)
2 2 2a
Substituting the mode expansion (7.8) into this, we find
1
Z h i
Ĥ = d3 k ak a−k Fk∗ + a†k a†−k Fk + 2a†k ak + δD (0) Ek , (7.14)
4
3
Since the frequency ωk (τ ) ≡ k2 − ä/a in (7.3) depends only on k ≡ |k|, the evolution does not depend on
direction. The constant operators âk and â†k , on the other hand, define initial conditions which may depend on
direction.
44
where
Ek (τ ) = |v̇k |2 + ωk2 |vk |2 , (7.15)

Fk (τ ) = v̇k2 + ωk2 vk2 . (7.16)
The vacuum expectation value of the Hamiltonian is

1
Z
h0|Ĥ|0i = δD (0) d3 k Ek (τ ) , (7.17)
4
where the divergent factor δD (0) is an artefact of integrating over an infinite volume. The energy
density of the vacuum is
1
Z
εk ≡ d3 k Ek (τ ) . (7.18)
4
At early times, we have
lim vk (τ ) = c1 e−ikτ + c2 eikτ , (7.19)

τ →−∞
lim Ek (τ ) = 4 (|c1 |2 + |c2 |2 ) k 2 . (7.20)

τ →−∞
Given the constraint (7.12), the function in (7.20) is minimized for

1
|c1 | = √ , c2 = 0 . (7.21)
2k
Up to an irrelevant phase, this uniquely determines the Bunch-Davies mode function

1 i
vk (τ ) = √ 1− e−ikτ . (7.22)
2k kτ
Note that (7.21) implies
lim Ek (τ ) = 2k , (7.23)
τ →−∞
lim Fk (τ ) = 0 , (7.24)
τ →−∞
and hence
1
Z
3 †
lim Ĥ = d k ak ak + δD (0) ~ωk , (7.25)
τ →−∞ 2
where we have reinstated Planck’s constant ~. We see that the vacuum state |0i is the state of
minimum energy 12 ~ωk . If any function other than (7.22) had been chosen as the mode function,
then the state annihilated by âk would not be the ground state of the oscillator.
Zero-point fluctuations
Finally, we can predict the quantum statistics of the operator
d3 k h
Z i
∗ †
v̂(τ, x) = v k (τ ) âk + v k (τ )a−k eik·x . (7.26)
(2π)3
45
The expectation value of v̂ vanishes, i.e. hv̂i ≡ h0|v̂|0i = 0. However, the variance of inflaton
fluctuations receives non-zero quantum fluctuations:
h|v̂|2 i ≡ h0|v̂(τ, 0)v̂(τ, 0)|0i

d3 k d3 k 0
Z Z
h0| vk∗ (τ )â†k + vk (τ )âk vk0 (τ )âk0 + vk∗0 (τ )â†k0 |0i

= 3 3
(2π) (2π)
d3 k d3 k 0
Z Z
= v (τ )vk∗0 (τ ) h0|[âk , â†k0 ]|0i
(2π)3 (2π)3 k
d3 k
Z
= |vk (τ )|2
(2π)3
k3
Z
= d ln k |vk (τ )|2 . (7.27)
2π 2
We define the (dimensionless) power spectrum as
k3
Pv (k, τ ) ≡ |vk (τ )|2 . (7.28)
2π 2
We see that the square of the classical solution determines the variance of quantum fluctuations.
Substituting the Bunch-Davies mode function (7.22) into (7.28), we find
2 h 2
Pv (k, τ )

H i
kτ →0 H
Pϕ (k, τ ) = = 1 + (kτ )2 −−−−−→ . (7.29)
a2 (τ ) 2π 2π
Note that in the superhorizon limit, kτ → 0, the dimensionless power spectrum Pϕ approaches
the same constant for all momenta. This is the characteristic of a scale-invariant spectrum.
Massive fields
The above discussion is easily generalized to massive spectator fields. The modified equation of
motion is
2 2 2 ä
v̈k + k + m a − vk = 0 , (7.30)
a
which, in de Sitter space, becomes
ν 2 − 1/4 9 m2

2
v̈k + k − vk = 0 , where ν2 ≡ − . (7.31)
τ2 4 H2
The most general solution of (7.31) is

√ h i
vk (τ ) = −τ c1 Hν(1) (|kτ |) + c2 Hν(2) (|kτ |) , (7.32)
(1) (2)
where Hν and Hν are Hankel functions of the first and second kind. Imposing the Bunch-
Davies initial condition, we find
√
π i(2ν+1) π
c1 = e 4 , c2 = 0 . (7.33)
2
46
For the moment, let us assume m < 32 H, so that ν is real. We then find
2
Pv (k, τ ) kτ →0

H
Pϕ (k, τ ) = −−−−−→ (kτ )3/2−ν . (7.34)
a2 (τ ) 2π
We observe that the superhorizon limit of the spectrum is scale-dependent and evolves in time.
The scale-dependence takes a power law form with the following spectral index
d ln Pϕ 3 mH 1 m2
nϕ ≡ = − ν −−−−−→ . (7.35)
d ln k 2 3 H2
In the limit m → 0, we recover the scale-invariant spectrum of a massless field in de Sitter.
For m > 32 H, the degree of the Hankel function becomes imaginary, ν ≡ iµ, where
p mH
µ≡ m2 /H 2 − 9/4 −−−−−→ m/H . (7.36)
What used to be an irrelevant phase factor in (7.33), now becomes an exponential suppression of
the amplitude of the mode function
√
π −πµ/2 √ (1)
|vk (τ )| = e −τ |Hiµ (|kτ |)| . (7.37)
2
The power spectrum of very massive fields in de Sitter is therefore highly suppressed, Pϕ ∝
e−πm/H , for m 32 H.
7.1.2 Fluctuations during Inflation

We now move to studying the fluctuations in the inflaton field during inflation. These fluctuations
cannot be treated independently from fluctuations in the metric, since the two are coupled by the
Einstein equations. This leads to some technical complications, but conceptually the quantization
of the coupled inflaton-metric fluctuations is the same as before.
Metric fluctuations
We will treat the fluctuations of the metric in the so-called ADM formalism [39]. We start by
writing the perturbed line element as
ds2 = −N 2 dt2 + hij (N i dt + dxi )(N j dt + dxj ) , (7.38)
where N ≡ N (t, x) is the lapse function, N i ≡ N i (t, x) is the shift vector, and hij ≡ hij (t, x) is
the induced metric on three-dimensional hypersurfaces of constant time t. The geometry of the
(3)
spatial slices is characterized by the intrinsic curvature, Rij , i.e. the Ricci tensor of the induced
metric, and by the extrinsic curvature
1 1
h0ij − ∇i Nj − ∇j Ni ≡

Kij ≡ Eij . (7.39)
2N N
The four-dimensional Ricci scalar, R, can be written in terms of the three-dimensional Ricci
scalar, R(3) , and the extrinsic curvature tensor as
R = R(3) + N −2 E ij Eij − E 2 ,

(7.40)
47
where indices are raised with hij , and E ≡ hij Eij .
The inflaton action can then be written as
1 √ h
Z i
S= d4 x −g R − g µν ∂µ φ∂ν φ − 2V
2
1 √
Z h 2 i
hN R(3) − 2V + N −2 E ij Eij − E 2 + N −2 φ0 − N i ∂i φ − hij ∂i φ∂j φ ,

= (7.41)
2
where for the moment we have set Mpl ≡ 1. Note that N and Ni do not have time derivatives
acting on them and are therefore non-dynamical fields that will be fixed by constraint equations.
Indeed, varying the action with respect to N and N i , we find
R(3) − 2V − hij ∂i φ∂j φ − N 2 [Eij E ij − E 2 − (φ0 − N i ∂i φ)2 ] = 0 , (7.42)

∇i [N −1 (Eji − Eδji )] = 0 . (7.43)
Plugging the solutions for N and N i back into the action leaves φ and hij as the only dynamical
variables. We will perform this procedure in perturbation theory. For the time being, we will
focus on scalar perturbations. To fix time and space reparameterizations, we will have to choose
a gauge for φ and hij . We will present the results for two different gauges: spatially flat gauge
and comoving gauge.
Spatially flat gauge

As the name suggests, in spatially flat gauge the induced metric is taken to be unperturbed:
hij = a2 δij . (7.44)
We then consider the perturbations of the inflaton, the lapse and the shift:
φ ≡ φ̄(t) + ϕ(t, x) , N ≡ 1 + α(t, x) , Ni ≡ ∂i β(t, x) . (7.45)
Substituting (7.44) and (7.45) into the constraint equations (7.42) and (7.43), we obtain [40]4
φ̄ 0 (φ̄0 )2 d

H
α= ϕ , ∂2β = − ϕ . (7.46)
2H 2H 2 dt φ̄0
Plugging this solution into the action (7.41), expanding to second order and performing a few
integrations by parts, we find

1 1
Z
3 3 0 2 2
00 2 2
2
S2 = dt d x a (ϕ ) − 2 (∂ϕ) − V − 2(3ε − ε + εη)H ϕ . (7.47)
2 a
Switching to conformal time, the equation of motion for the Fourier components of the canonically
normalized field v ≡ aϕ is

2 2
00 2
ä
v̈k + k + a V − 6εH − vk = 0 , (7.48)
a
4
To find the quadratic action, we only have to solve the constraints to linear order. The second order terms in
N and Ni will multiply the zeroth order constraints which vanish when the background equations of motion are
imposed.
48
where we have dropped terms in the effective mass that are higher order in slow-roll parameters.
Note that the term a2 V 00 − 6εH 2 ∼ O(ε, η)(aH)2 is always smaller than ä/a = 2(aH)2 . How-

ever, it is the only source of time evolution on superhorizon scales and even a small evolution can
accumulate over time. In order not to have to follow the evolving field on superhorizon scales,
it is useful to evaluate the solution at horizon crossing, k = aH, and then map it to a field
that is known not to evolve outside the horizon. This constant mode is the comoving curvature
perturbation, which in spatially flat gauge is defined as
H
ζ=− ϕ. (7.49)
φ̄ 0
The power spectrum of ζ is
2
H
Pζ (k) = Pϕ (k, τ ) . (7.50)
φ̄ 0 k=aH
Dropping the slow-roll suppressed terms in (7.48), the equation of motion reduces to that of a
massless field in de Sitter space. The power spectrum Pϕ (k, η) is therefore given by (7.29) and
we get
2 2
H H
Pζ (k) = 0
. (7.51)
φ̄ 2π k=aH
Another advantage of using ζ rather than ϕ to characterize the initial conditions is that ζ remains
well-defined after inflation.
Comoving gauge
In comoving gauge the inflaton field is taken to be unperturbed:
φ = φ̄(t) . (7.52)
All perturbations are then carried by the metric:
hij = a2 (1 + 2ζ(t, x))δij , N ≡ 1 + α(t, x) , Ni ≡ ∂i β(t, x) , (7.53)
where ζ is the comoving curvature perturbation. Substituting this ansatz into the constraint
equations (7.42) and (7.43), we obtain [40]
ζ0 ∂2ζ (φ̄0 )2 0
α= , ∂2β = − + a2 ζ . (7.54)
H H 2H 2
Plugging this solution into the action (7.41), expanding to second order, performing integrations
by parts and using the background equations of motion, we get a remarkably simple result

1
Z
3 3 0 2 2
S2 = dt d x a ε (ζ ) − 2 (∂ζ) . (7.55)
a
Note that there is no mass term, so ζ is conserved outside the horizon. Consider the equation of
motion
k2
ζk00 + (3 + η)Hζk0 + 2 ζk = 0 . (7.56)
a
49
On super-Hubble scales, k aH, this becomes ζk00 + (3 + η)Hζk0 ≈ 0, which clearly has a constant
mode as a solution.
√
The equation of motion of the canonically normalized field, v ≡ a 2ε ζ ≡ z ζ, is

2 z̈
v̈k + k − vk = 0 , (7.57)
z
where the effective mass can be written as
z̈ ν 2 − 1/4 3 η
= , with ν≈ +ε+ . (7.58)
z τ2 2 2
Recall that we have seen eq. (7.57) before when we considered a massive scalar field in de Sitter
space, cf. eq. (7.31). The Bunch-Davies mode function therefore is
√
π√ kτ →0 2ν Γ(ν) 1
|vk (τ )| = −τ |Hν(1) (−kτ )| −−−−−→ √ √ (−kτ )−ν+1/2 , (7.59)
2 2 π k
and the superhorizon limit of the power spectrum of ζ is
k3 2 1 k3
Pζ (k) ≡ lim |ζk (τ )| = lim |vk (τ )|2
kτ →0 2π 2 z 2 (τ ) kτ →0 2π 2
1 k2 −2ν+1 1 H 2 (τ )
= (−kτ ) = (−kτ )−2ν+3 . (7.60)
2a2 ε 4π 2 16π 2 ε(τ )
Note that the time dependence of H(τ ) and ε(τ ) precisely cancels the time dependence of the
final factor in (7.60), so that the power spectrum is time independent. Let k∗ be a reference scale
that exits the horizon at time τ∗ = −1/k∗ . Equation (7.60) can then be written as
1 1 H∗2 −2ε∗ −η∗

Pζ (k) = 2 (k/k∗ ) , (7.61)
8π 2 ε∗ Mpl
where H∗ ≡ H(τ∗ ) and ε∗ ≡ ε(τ∗ ).
7.2 Curvature Perturbations

For ease of reference, we briefly summarize the results of the previous section. The power spectrum
of the curvature perturbation ζ, cf. eqs. (7.51) and (7.61), takes a power law form
ns −1
k
Pζ (k) = As , (7.62)
k∗
where the amplitude and the spectral index are
1 1 H∗2
As ≡ 2 , (7.63)
8π 2 ε∗ Mpl
ns ≡ 1 − 2ε∗ − η∗ . (7.64)
The observational constraint on the scalar spectral index is ns = 0.9603 ± 0.0073. The observed
percent-level deviation from the scale-invariant value, ns = 1, are the first direct measurement of
time dependence in the inflationary dynamics.
50
Exercise.—Show that for slow-roll inflation, the results (7.63) and (7.64) can be written as
1 1 V
As = 4 , (7.65)
24π 2 εV Mpl
ns = 1 − 6εV + 2ηV , (7.66)
where εV and ηV are the potential slow-roll parameters defined in (6.13). This expresses the amplitude
of curvature perturbations and the spectral index in terms of the shape of the inflaton potential.
7.3 Gravitational Waves

Arguably the cleanest prediction of inflation is a spectrum of primordial gravitational waves.
These are tensor perturbations to the spatial metric,
ds2 = a2 (τ ) −dτ 2 + (δij + 2γij )dxi dxj ,

(7.67)
where γij is transverse and traceless. Substituting (7.67) into the Einstein-Hilbert action and
expanding to second order gives
2 Z
Mpl Mpl2 Z
4 √
dτ d3 x a2 γ̇ij
2
− (∂γij )2 .

S= d x −g R ⇒ S2 = (7.68)
2 8
It is convenient to use rotational symmetry to align the z-axis of the coordinate system with the
momentum of the mode, i.e. k ≡ (0, 0, k), and write
 
v+ v× 0
Mpl 1 
aγij ≡ √  v× −v+ 0  . (7.69)

2 2
0 0 0
The action (7.68) then becomes

1 X ä 2
Z
3 2 2
S2 = dτ d x v̇λ − (∂vλ ) + vλ , (7.70)
2 a
λ=+,×
which is just two copies of the action of a massless scalar field, one for each polarization mode of
the gravitational wave, v+,× . The equation of motion for each polarization is

2 ä
v̈k + k − vk = 0 , (7.71)
a
where the effective mass can be written as
ä ν 2 − 1/4 3
= , with ν ≈ + ε . (7.72)
a τ2 2
The Bunch-Davies mode function is then given by (7.59). The superhorizon limit of the power
spectrum of the tensor flucutations then is
2
k3 k3

2 2
Pγ (k) = 2 × lim |γ k (τ )| = 2 × lim |vk (τ )|2
kτ →0 2π 2 aMpl kτ →0 2π 2
2 H 2 (τ ) −2ν+3
= 2 (−kτ ) , (7.73)
π 2 Mpl
51
where the factor of 2 accounts for the sum over the two polarization modes. Introducing the
reference scale k∗ , this can be written as
2 H∗2 −2ε∗
Pγ (k) = 2 (k/k∗ ) . (7.74)
π 2 Mpl
This result is arguably the most robust and model-independent prediction of inflation. We see
that the form of the tensor power spectrum is again a power law,
nt
k
Pζ (k) = At , (7.75)
k∗
where the amplitude and the spectral index are
2 H∗2
At ≡ 2 , (7.76)
π 2 Mpl
nt ≡ −2ε∗ . (7.77)
Notice that the tensor amplitude is a direct measure of the expansion rate H during inflation.
This is in contrast to the scalar amplitude which depends on both H and ε. The tensor tilt is
a direct measure of ε, whereas the scalar tilt depends both on ε and η. Observationally, a small
value for nt is hard to distinguish from zero. The tensor amplitude is often normalized with
respect to the measured scalar amplitude, As = (2.196 ± 0.060) × 10−9 (at k∗ = 0.05 Mpc−1 ):
At
r≡ = 16ε∗ , (7.78)
As
where r is the tensor-to-scalar ratio. Inflationary models make predictions for (ns , r). The latest
observational constraints on these parameters are shown in Fig. 21.
0.25
N=
Planck + BK14
N=
60
05
0.20
0.15 Co
nve
C on x
cav
e
0.10
0.05
0.00
0.95 0.96 0.97 0.98 0.99 1.00 0.5 1.0
Figure 21. Current constraints on ns and r from CMB measurements of Planck and BICEP [41].
52
Case study: m2 φ2 inflation.—In Section 6, we showed that the slow-roll parameters of m2 φ2 inflation
are 2
Mpl
εV (φ) = ηV (φ) = 2 , (7.79)
φ
and the number of e-folds before the end of inflation is
φ2 1 φ2
N (φ) = 2 − ≈ 2 . (7.80)
4Mpl 2 4Mpl
At the time when the CMB fluctuations crossed the horizon at φ = φ∗ , we have
1
εV,∗ = ηV,∗ ≈ . (7.81)
2N∗
The spectral tilt and the tensor-to-scalar ratio therefore are
2
ns ≡ 1 − 6εV,∗ + 2 ηV,∗ = 1 − ≈ 0.97 , (7.82)
N∗
8
r ≡ 16 εV,∗ = ≈ 0.13 , (7.83)
N∗
where the final equalities are for N∗ ≈ 60. Comparison with Fig. 21 shows that these predictions are
in conflict with the latest CMB data.
The Lyth bound.—During m2 φ2 inflation, the inflaton field value varies by ∆φ ≈ 15 Mpl from the time
the mode observed in the CMB exited the horizon until the end of inflation. This super-Planckian
field variation is, in fact, a general feature of all inflationary models with observable gravitational
waves.
Combining (6.9) and (7.78), the tensor-to-scalar ratio can be related to the evolution of the inflaton:
2
8 dφ
r= 2 , (7.84)
Mpl dN
where dN = Hdt. The total field excursion between the time when CMB fluctuations exited the
horizon at N∗ and the end of inflation at Ne can therefore be written as the integral
Z N∗ r
∆φ r
= dN . (7.85)
Mpl Ne 8
During slow-roll evolution, r(N ) doesn’t evolve much and one may obtain the following approximate
relation, called the Lyth bound:
∆φ r 1/2
= O(1) × , (7.86)
Mpl 0.01
where r ≡ r(N∗ ) is the tensor-to-scalar ratio on CMB scales. Large values of the tensor-to-scalar
ratio, r > 0.01, therefore correlate with super-Planckian field excursions, ∆φ > Mpl .
A major goal of current efforts in observational cosmology is to detect the tensor component of
the primordial fluctuations. Its amplitude depends on the energy scale of inflation and is therefore
not predicted (i.e. it varies between models). While this makes the search for primordial tensor
53
modes difficult, it is also what makes it exciting. Detecting tensors would reveal the energy scale
at which inflation occurred, providing an important clue about the physics driving the inflationary
expansion.
Most searches for tensors focus on the imprint that tensor modes leave in the polarization of
the CMB. Polarization is generated through the scattering of the anisotropic radiation field off
the free electrons just before decoupling. The presence of a gravitational wave background creates
an anisotropic stretching of the spacetime which induces a special type of polarization pattern,
the so-called B-mode pattern (a pattern whose “curl” doesn’t vanish). Such a pattern cannot be
created by scalar (density) fluctuations and is therefore a unique signature of primordial tensors
(gravitational waves). A large number of ground-based, balloon and satellite experiments are
currently searching for the B-mode signal predicted by inflation. A B-mode detection would be
a milestone towards a complete understanding of the origin of all structure in the universe.
54
8 Primordial Interactions
In the previous section, we have computed the two-point function of primordial curvature per-
turbations, or its Fourier equivalent, the power spectrum. If the initial conditions are drawn
from a Gaussian distribution function, then the power spectrum contains all of the information
about the primordial perturbations. In general, however, higher-order correlations can encode a
significant amount of new information. In particular, these correlations are sensitive to nonlinear
interactions while the power spectrum only probes the free theory.
In this section, I will introduce the basic formalism for computing the non-Gaussianity pro-
duced by inflation. I will apply this to single-field slow-roll inflation and show that gravitational
nonlinearities produce a robust (but small) non-Gaussian signature. In Section 9, I will show
that extra massive fields can get excited during inflation and that their decays lead to distinctive
signatures in cosmological correlation functions.
For further reading I highly recommend the classic papers by Maldacena [40] and Wein-
berg [42], as well as the wonderfully clear reviews by Chen [43], Wang [44], Lim [45] and Ko-
matsu [46].
8.1 Non-Gaussianity
The leading non-Gaussian signature is the three-point correlation function, or its Fourier equiv-
alent, the bispectrum
Bζ (k1 , k2 , k3 ) ≡ hζk1 ζk2 ζk3 i . (8.1)
For perturbations around an FRW background, the momentum dependence of the bispectrum
simplifies considerably: Because of homogeneity, or translation invariance, the bispectrum is
proportional to a delta function of the sum of the momenta, Bζ (k1 , k2 , k3 ) ∝ δD (k1 + k2 + k3 ),
i.e. the sum of the momentum three-vectors must form a closed triangle. Because of isotropy, or
rotational invariance, the bispectrum only depends on the magnitudes of the momentum vectors,
but not on their orientations,
Bζ (k1 , k2 , k3 ) = (2π)3 δD (k1 + k2 + k3 ) Bζ (k1 , k2 , k3 ) . (8.2)
It is convenient to define the dimensionless bispectrum as
(k1 k2 k3 )2
Bζ (k1 , k2 , k3 ) ≡ Bζ (k1 , k2 , k3 ) . (8.3)
(2π 2 )2
The amplitude of the non-Gaussianity is then defined as the size of the bispectrum in the equi-
lateral momentum configuration:
5 Bζ (k, k, k)
fNL (k) ≡ , (8.4)
18 Pζ2 (k)
where we have indicated that fNL can in general depend on the overall momentum. On the other
hand, if the fluctuations are scale-invariant, then fNL is a constant and we can write
18
Bζ (k1 , k2 , k3 ) ≡ fNL × S(x2 , x3 ) × Pζ2 , (8.5)
5
55
where x2 ≡ k2 /k1 and x3 ≡ k3 /k1 . The shape function S(x2 , x3 ) is normalized so that S(1, 1) ≡ 1.
As we will discuss in detail below, the shape of the non-Gaussianity contains a lot of information
about the microphysics of inflation. A divergence of the signal for squeezed triangles is a signature
for extra degrees of freedom during inflation (see Fig. 22 and Sec. 9), while a peak in the signal
for equilateral triangles arises from higher-derivative inflaton self-interactions (see Fig. 23).
squeezed
equilateral
1.0
folded
1.0
Figure 22. Bispectrum of local non-Gaussianity. The signal is peaked for squeezed triangles.
squeezed
equilateral
1.0
1.0
1.0
Figure 23. Bispectrum for the interaction ζ̇ 3 . The signal is peaked for equilateral triangles.
56
8.2 In-In Formalism
The problem of computing correlation functions in cosmology differs in important ways from the
corresponding analysis of quantum field theory applied to particle physics.
In particle physics, the main observable is the S-matrix, i.e. the transition probability for a
state |ini in the far past to become some state |outi in the far future,
hout| S|ini = hout(+∞)|in(−∞)i . (8.6)
The scattering particles are taken to be non-interacting at very early and very late times, when
they are far from the interaction region, and the asymptotic states can be taken to be vacuum
states of the free Hamiltonian H0 .
Figure 24. Particle physicists compute in-out transition amplitudes.
In cosmology, on the other hand, the task is to determine the expectation values of products
of operators at a fixed time. Boundary conditions are only imposed at very early times when
their wavelengths are much smaller than the horizon and the interaction picture fields should
have the same form as in Minkowski space. As we have seen in the previous section, this leads to
the definition of the Bunch–Davies vacuum. In this section, we will describe the in-in formalism5
to compute cosmological correlation functions as expectation values in two |ini states.
Figure 25. Cosmologists compute in-in expectation values.
Our goal is to compute n-point functions of the curvature perturbation ζ or the gravitational
wave polarization modes γ × and γ + . We will collectively denote these fluctuations by the field
ψ = {ζ, γ × , γ + } and consider expectation values of operators such as Q̂ = ψ̂k1 ψ̂k2 · · · ψ̂kn ,
hQ̂(τ )i = hin| Q̂(τ ) |ini , (8.7)
where |ini is the vacuum of the interacting theory at some moment τi in the far past, and τ > τi is
some later time, such as horizon crossing or the end of inflation. To compute the matrix element
5
This is also referred to as the Schwinger–Keldysh formalism [47]. The use of the in-in formalism in cosmology
was pioneered in [40, 42, 48, 49] (see also [50, 51]) and is reviewed in [43, 44].
57
in (8.7) we evolve Q(τ ) back to τi using the perturbed Hamiltonian δH. Computing this time
evolution is complicated by the interactions inside of δH = H0 + Hint , which lead to nonlinear
equations of motion. We therefore introduce the interaction picture in which the leading time
dependence of the fields is determined by the quadratic Hamiltonian H0 (or linear equations of
motion). Corrections arising from the interactions are then treated as a power series in Hint .
This leads to the following important result:
Rτ I (τ 0 ) dτ 0
Rτ I (τ 00 ) dτ 00
i Ĥint −i Ĥint
hQ̂(τ )i = h0| T̄ e −∞(1−i) Q̂I (τ ) T e −∞(1+i) |0i , (8.8)
where T (T̄ ) is the (anti-)time-ordering symbol. Note that both Q̂I and Ĥint I are evaluated using
interaction picture operators. The standard i prescription has been used to effectively turn off
the interaction in the far past and project the interacting |ini state onto the free vacuum |0i. By
expanding the exponentials, we can compute the correlation function perturbatively in Ĥint . For
example, at leading order, we find
Z τ
hQ̂(τ )i = −i dτ 0 h0| [Q̂I (τ ), Ĥint
I
(τ 0 )] |0i . (8.9)
−∞
We can use Feynman diagrams to organize the power series, drawing interaction vertices for every
power of Ĥint .
In the following insert, I will derive the in-in master formula (8.8). Readers who are more
interested in applications of the result, may skip this part.
Derivation.—The time evolution of the operators in the Heisenberg picture is determined by
dψ̂ dp̂ψ
= i[Ĥ, ψ̂] , = i[Ĥ, p̂ψ ] , (8.10)
dτ dτ
where Ĥ = Ĥ0 + Ĥint is the perturbed Hamiltonian. This time evolution is complicated by the inter-
actions inside Ĥ, which lead to nonlinear equations of motion. We therefore introduce the interaction
picture in which the leading time dependence of the fields is determined by the quadratic Hamiltonian
Ĥ0 (or, equivalently, by the linear equations of motion):
dψ̂ I dp̂Iψ
= i[Ĥ0 , ψ̂ I ] , = i[Ĥ0 , p̂Iψ ] . (8.11)
dτ dτ
The solution to these equations can be written as
ψ̂kI (τ ) = ψkI (τ ) âk + h.c. , (8.12)
where ψkI (τ ) is the solution to the free-field equation and the operators âk define the free-field vac-
uum |0i. Corrections to the evolution of the operators can then be treated perturbatively in Ĥint .
Relatively straightforward algebraic manipulations of (8.10) and (8.11) allow us to express an operator
in the Heisenberg picture in terms of operators in the interaction picture [42],
Q̂(τ ) = F̂ −1 (τ, τi ) Q̂I (τ ) F̂ (τ, τi ) , (8.13)
where
(τ 00 ) dτ 00
Rτ I
−i Ĥint
F̂ (τ, τi ) ≡ T e τi
. (8.14)
58
We can think of F̂ (τ, τi ) as an operator evolving quantum states in the interaction picture,
|Ω(τ )i = F̂ (τ, τi )|Ω(τi )i , (8.15)
where |Ω(τi )i ≡ |Ωi. We would like to relate the vacuum of the interacting theory, |Ωi, to the vacuum
of the free theory, |0i. Inserting a complete set of energy eigenstates {|Ωi, |ni} of the full theory, where
|ni are the excited states, we have
X
|0i = |ΩihΩ|0i + |nihn|0i , (8.16)
n
and correspondingly
X
e−iĤ(τ −τi ) |0i = e−iĤ(τ −τi ) |ΩihΩ|0i + e−iEn (τ −τi ) |nihn|0i . (8.17)
n
Adding a small imaginary part to the initial time, τi → −∞(1 − i) ≡ −∞− , will project out the
excited states, e−iEn (τ −τi ) → e−∞×En (· · · ) → 0. We are then left with
F̂ (τ, −∞− )|0i

F̂ (τ, −∞− )|Ωi = . (8.18)
hΩ|0i
The i prescription has effectively turned off the interactions in the far past and projected the in-
teracting vacuum |Ωi onto the free vacuum |0i. Setting |hΩ|0i| → 1, we arrive at the in-in master
formula
0 0 00 00
Rτ I
Rτ I
hQ̂(τ )i = h0| T̄ ei −∞+ Ĥint (τ ) dτ Q̂I (τ ) T e−i −∞− Ĥint (τ ) dτ |0i , (8.19)
where ∞± ≡ ∞(1±i). The integration contour goes from −∞(1 − i) to τ (where the correlation
function is evaluated) and back to −∞(1 + i).
8.3 Gravitational Floor

Let us now apply this formalism to the calculation of the bispectrum of curvature perturbations
in single-field slow-roll inflation [40]. The nonlinearities of the gravitational evolution produce a
minimal amount of non-Gaussianity which we will call the “gravitational floor”. Unfortunately,
the amplitude of the signal is too small to be detectable in the foreseeable future.
To compute the bispectrum we need to expand the inflationary action (7.41) to third order in
perturbations. It is convenient to perform the computation in comoving gauge, cf. eqs. (7.52) and
(7.53). In §7.1.2, we solved the lapse N and the shift Ni to first order in ζ. This is sufficient also
for the cubic action since second-order and third-order perturbations in N and Ni will multiply
the first-order and zeroth-order constraint equation, respectively. Substituting (7.54) into (7.41)
and expanding to third order, we find [40]
1
Z
S3 = dt d3 x a3 2ε2 ζ(ζ 0 )2 − 4a−2 ε2 ζ 0 (∂ζ)(∂ζ) + 2a−2 ε2 ζ(∂ζ)2
2
1 δL2
+ a−2 εη 0 ζ 2 ζ 0 + a−2 ε ∂ζ∂β∂ 2 β + a−2 ε ∂ 2 ζ(∂β)2 + 2f (ζ) , (8.20)
2 δζ
59
where we have defined
δL2
≡ (a3 ε ζ 0 )0 − aε∂ 2 ζ , (8.21)
δζ
η ζζ 0 −(∂ζ)2 + ∂ −2 (∂i ∂j (∂i ζ∂j ζ)) ∂ζ∂β − ∂ −2 (∂i ∂j (∂i ζ∂j β))
f (ζ) ≡ ζ 2 + + + . (8.22)
4 H 4a2 H 2 2a2 H 2
Maldacena showed that the term proportional to f (ζ) in (8.20) can be removed by a field redefi-
nition,
ζ → ζ̃ + f (ζ̃) . (8.23)
This field redefinition has the following effect on the correlation function:
η
hζ(x1 )ζ(x2 )ζ(x3 )i = hζ̃(x1 )ζ̃(x2 )ζ̃(x3 )i + hζ̃(x1 )ζ̃(x2 )ihζ̃(x1 )ζ̃(x3 )i + cyclic + · · · . (8.24)
2
The term proportional to f (ζ) in (8.20) therefore leads to a contribution to fNL of order η 1.
The effect of the remaining interactions is computed by expanding the in-in master formula (8.8)
to first order in Hint = −L3 + O(ζ 4 ),
Z 0
hζk1 ζk2 ζk3 i = −i dτ h0| ζ̂k1 ζ̂k2 ζ̂k3 (0), Ĥint (τ ) |0i , (8.25)
−∞
where we have switched to conformal time and taken the superhorizon limit kτ → 0.
Back-of-the-envelope estimate.—Before embarking on a lengthy calculation of the bispectrum, it is

often advisable to perform an order-of-magnitude estimate of the expected size of the signal, i.e. to
estimate (8.25) without explicitly performing the integral. For example, the first term in (8.20), can
be written as Z Z
dτ Hint (τ ) ⊂ − dτ d3 x a2 ε2 ζ ζ̇ 2 . (8.26)
We only need to keep track of factors of H and ε. Any time- and momentum-dependence will work
1/2 √
itself out and only contributes to the shape function. Using a ∝ H −1 and ζ ∝ ζ̇ ∝ Pζ ∼ H/ ε, we
√
estimate that the contribution from the three-point vertex is ∼ H ε. Combining this with estimates
for the size of the three external legs, ζ 3 ∼ H 3 ε−3/2 , we find
H4
Z
hζ i = −i dτ h[ζ̂ 3 , Ĥint (τ )]i ∝
3
∝ O(ε) Pζ2 ∼ fNL Pζ2 . (8.27)
ε
Similar results are obtained for the other interactions in (8.20), fNL ∼ O(ε). We also include the
contribution from the field redefinition in (8.24), fNL ∼ O(η). We conclude that the non-Gaussianity
in slow-roll inflation is slow-roll suppressed,
fNL ∼ O(ε, η) 1 . (8.28)
This small amount of non-Gaussianity will be unobservable in the CMB.
To get the full momentum-dependence of the bispectrum, we actually have to do some real
work and compute the integral in (8.25) using the free-field mode functions for ζ. The final result
60
for the dimensionless bispectrum is [40]
2
Bζ (k1 , k2 , k3 )

ε k1 k1 8 k1 k2
= − + 2 perms. + + 5 perms. + + 2 perms.
Pζ2 2 k2 k3 k2 kt k3
2
η k1
+ + 2 perms. , (8.29)
2 k2 k3
where kt ≡ k1 + k2 + k3 . It is interesting to take the squeezed limit, k3 k1 ≈ k2 , of the result:
Bζ (k1 , k2 , k3 ) k1
lim = (2ε + η) . (8.30)
k3 →0 Pζ2 k3
Note that the coefficient in the squeezed limit equals the deviation from scale invariance of the
power spectrum, 1 − ns = 2ε + η. In fact, this is a general result that applies to all models of
single-field inflation, not necessarily just to slow-roll models. The single-field consistency relation
states that [40, 52]
lim Bζ (k1 , k2 , k3 ) = (1 − ns ) Pζ (k1 )Pζ (k3 ) , (8.31)
k3 →0
i.e. the squeezed limit of the three-point function is suppressed by (1 − ns ) and vanishes for
perfectly scale-invariant perturbations.
Proof.—The squeezed triangle correlates one long-wavelength mode, kL = k3 , to two short-wavelength

modes, kS = k1 ≈ k2 ,
hζk1 ζk2 ζk3 i → h(ζkS )2 ζkL i . (8.32)
Modes with longer wavelengths freeze earlier. The long mode ζkL will therefore already be frozen and
act as a classical background wave when the two short modes ζkS exit the horizon. This provides an
intuitive way to study the correlations between long and short modes.
Why should (ζkS )2 be correlated with ζkL ? The theorem says that it isn’t correlated if ζk is
precisely scale-invariant, but that the short scale power does get modified by the long-wavelength
mode if ns 6= 1. Let’s see why. We decompose the evaluation of (8.32) into two steps:
i) we calculate the power spectrum of short fluctuations hζS2 iζL in the presence of a long mode ζL ;
ii) we then calculate the correlation h(ζS )2 ζL i, i.e. average the short-scale power spectrum over
realizations of the long modes.
The calculation of hζS2 iζL is simplest in real space: When the background mode is homogeneous,
ζL (x) ≡ ζ̄L , it can be reabsorbed simply by a rescaling of the spatial coordinates, x̃i = eζ̄L xi (recall
that ds2 = −dt2 + a2 (t)e2ζ(x,t) dx2 ). After this rescaling, ζ̄L no longer appears in the action, so that
the two-point function in the new coordinates is the same as in the absence of ζ̄L . In other words, in
the limit of constant ζ̄L , we have
hζS (x1 )ζS (x2 )iζ̄L = hζS (x̃1 )ζS (x̃2 )i . (8.33)
When ζL is slowly varying, we can evaluate it at the middle point x+ ≡ 21 (x1 + x2 ) to get
x̃− ' x− + ζL (x+ ) · x− + · · · , (8.34)
where we defined x− ≡ x1 − x2 . The two-point function at linear order in ζL therefore is
hζS (x2 )ζS (x3 )iζL (x) ' ξS (|x− |) + ζL (x+ )[x− · ∇ξS (|x− |)] , (8.35)
61
where
d3 kS
Z
ξS (|x− |) ≡ Pζ (kS ) eikS ·x− . (8.36)
(2π)3
The three-point function then is
hζS (x1 )ζS (x2 )ζL (x3 )i ' hζL (x3 )ζL (x+ )i[x− · ∇ξS (|x− |)] ,
Z 3 Z 3
d kL d kS ikL ·(x3 −x+ ) ∂
= e Pζ (kL )Pζ (kS ) kS · eikS ·x− . (8.37)
(2π)3 (2π)3 ∂kS
Integrating by parts, inserting 1 = d3 k3 δD (k3 + kL ) and using

R
∂ d ln(kS3 Pζ (kS ))
· [kS Pζ (kS )] = Pζ (kS ) , (8.38)
∂kS d ln kS
we get
d3 k3
Z 3 Z 3
d kL d kS −ik1 ·x1 −ikL ·x+ +ikS ·x−
Z
hζS (x1 )ζS (x2 )ζL (x3 )i = − e
(2π)3 (2π)3 (2π)3
d ln(kS3 Pζ (kS ))

× (2π)3 δD (k1 + kL )Pζ (k1 )Pζ (kS ) . (8.39)
d ln kS
Letting kL = k2 + k3 and kS = 12 (k2 − k3 ), we get −ikL · x+ + ikS · x− = −ik2 · x2 − ik3 · x3 . Changing

variables in the integration and Fourier transforming, we get
d ln k33 Pζ (k3 )
lim hζk1 ζk2 ζk3 i = −(2π)3 δD (k1 + k2 + k3 ) Pζ (k1 )Pζ (k3 ) ,
k3 →0 d ln k3
= (2π)3 δD (k1 + k2 + k3 ) (1 − ns ) Pζ (k1 )Pζ (k3 ) . (8.40)
This completes the proof.
The fact that the squeezed limit of the bispectrum for single-field inflation vanished implies that
it can be used as a clean diagnostic for extra fields during inflation. In the next section, we will
expand on this view of the squeezed limit as a particle detector.
62
9 Heavy Relics
As we have seen in the previous sections, computing the quantum correlations in single-field slow-
roll inflation is a well-defined problem leading to clean predictions for the cosmological correlation
functions. In this final section, we will look for deviations from those predictions associated with
the existence of extra particles during inflation.
We will begin, in §9.1, with a review of the allowed spectrum of particles in de Sitter space,
highlighting the qualitative differences to the corresponding results in flat space. In §9.2 and §9.3,
we derive the basic observational imprints that these massive fields create when they are excited
during inflation. For further details we refer the reader to the vast literature on the subject,
e.g. [38, 53–63].
9.1 Massive Fields in de Sitter

Particles in Minkowski space are classified as unitary irreducible representations of the Poincaré
group [64, 65]. The eigenvalues of the Casimir operators of the Poincaré group are related as
follows to the mass m and the spin s of the particles:
C1 ≡ Pµ P µ = m 2 , (9.1)
µ 2
C2 ≡ Wµ W = −s(s + 1)m , (9.2)
where Pµ is the four-momentum and Wµ is the Pauli-Lubanski pseudovector. We distinguish be-

tween massive and massless particles. Massive particles carry 2s+1 degrees of freedom (transverse
and longitudinal polarizations), while massless particles have only two transverse polarizations.
Similarly, particles in de Sitter space are classified as unitary irreducible representations of
the de Sitter group SO(1, 4). The Casimir operators of the de Sitter group have the following
eigenvalues [66, 67]:
1
C1 ≡ MAB M AB = m2 − 2(s − 1)(s + 1)H 2 , (9.3)
2
A 2 2 1 2
C2 ≡ WA W = −s(s + 1) m − s + s − H , (9.4)
2
where MAB are the generators of SO(1, 4), with A, B = {0, . . . , 4}. Notice that this time the
Casimir (9.3) is not just proportional to the mass alone, but also has a term that depends on
spin. For s ≥ 1, the representations of de Sitter space fall into three distinct categories [68–70]:
principal series complementary series discrete series

m2 1 2 m2 1 2 m2

≥ s− s(s − 1) < 2 < s − = s(s − 1) − t(t + 1) ,
H2 2 H 2 H2
for t = 0, 1, 2, ..., s − 1, which is called the depth of the field. Masses that are not associated
with one of the above categories correspond to non-unitary representations and are therefore not
allowed in the spectrum (see Fig. 26). At the specific mass values corresponding to the discrete
series, the system gains an additional gauge invariance and some of the lowest helicity modes
63
Figure 26. Spectrum of spin-0 (top) and spin-4 fields (bottom) in de Sitter space. The green points
correspond to masses in the discrete series.
become pure gauge modes; this phenomenon is called partial masslessness [71]. We see that
unitarity demands the existence of a lower bound
m2 ≥ s(s − 1)H 2 , (9.5)
on the masses of fields except those that belong to the discrete series. For s = 2, this is known
as the Higuchi bound [72].
The Lagrangian for massive fields with arbitrary spin in flat space was constructed by Singh
and Hagen in [73, 74], and generalized to (A)dS spaces in [75]. For massive fields with spin greater
than 2, the action is rather complex and requires introducing auxiliary fields of lower spins. An
alternative, which we will follow, is to use a group theoretical approach to find the equations of
motion directly [66]. In four spacetime dimensions, a massive bosonic spin-s field is described by
a totally symmetric rank-s tensor, σµ1 ···µs , subject to the constraints
∇µ1 σµ1 ···µs = 0 , σ µ µµ3 ···µs = 0 . (9.6)
The conditions in (9.6) project out the components of the tensor which transform as fields with
lower spins. The Casimir eigenvalue equation of the de Sitter group then gives the on-shell wave
equation satisfied by these fields:
− m2s σµ1 ···µs = 0 ,

(9.7)
where ≡ ∇µ ∇µ is the Laplace-Beltrami operator on dS4 and m2s ≡ m2 − (s2 − 2s − 2)H 2 . The
shift in the mass arises from the mismatch between the Casimir and Laplace-Beltrami operators
in de Sitter space and is necessary to describe the correct representations for massless fields.
Explicit solutions to the equation of motion (9.7) can be found in Appendix A of [59]. The
spatial components contain all the physical degrees of freedom, while the other components are
constrained variables or pure gauge modes.6 The late-time behaviour of the solution is
X
lim σi1 ···is (τ, x) = τ ∆± −s σi±1 ···is (x) , (9.8)
τ →0
±
6
This can be seen by solving the equation of motion (9.7) explicitly, where the normalizations of non-physical
components become singular [59].
64
where ∆± is the conformal dimension of the field
s
m2 1 2

3
∆± = ± iµ , µ ≡ − s− . (9.9)
2 H2 2
Note that, for s = 0, the value m = 0 corresponds to a conformally-coupled scalar field. In the
limit m sH, the parameter µ becomes the mass of the particle in Hubble units, µ → m/H.
Particles belonging to the principal series correspond to µ ≥ 0, which covers the largest mass
range. For real µ, the asymptotic scaling is given by a complex-conjugate pair, resulting in a
wavefunction that oscillates logarithmically in conformal time. The complementary series has
imaginary µ and corresponds to the interval −iµ = (0, 1/2). In that case, only the growing mode
survives in the late-time limit.
m=0
0 < m < 32 H
m > 32 H
⌧ 3/2 cos(µ ln k⌧ )
Figure 27. Superhorizon evolution of a massive scalar field. The decay of the field amplitude is determined
by the mass of the field.
9.2 Local and Non-Local

In the remainder of this section, we will study the imprints of massive particles on inflationary
correlations functions. Massive particles during inflation can have two types of effects:
• Particles with masses M H can be integrated out, leading to an EFT of inflaton inter-
actions:
These local interactions create characteristic imprints in the non-Gaussianity of the inflaton
fluctuations. In particular, the soft limits of the resulting correlation functions must satisfy
the single-field consistency relation discussed in the previous section.
• Particles with masses M . few × H cannot be integrated out completely, but are produced
nonperturbatively by the expanding spacetime. When these particles decay, they produce
65
characteristic non-local signatures in cosmological correlators. The single-field consistency
relation will be violated.
Local effects
To illustrate the local effects arising from massive particles, we consider the following two-field
example:
1 1 1 σ(∂µ φ)2
L = − (∂µ φ)2 − V (φ) − (∂µ σ)2 − M 2 σ 2 + , (9.10)
2 2 2 Λ̃
where M H. Integrating out the massive field σ gives
1 1 1
Leff = − (∂µ φ)2 − V (φ) + (∂µ φ)2 (∂µ φ)2 + · · ·
2 8Λ̃2 + M2

1 2 1 4 2 2
= − (∂µ φ) − V (φ) + (∂µ φ) + (∂µ φ) (∂µ φ) + · · · , (9.11)
2 8Λ̃2 M 2 M2
where the second line is written as an expansion in (H/M )2 , with H being the Hubble scale
during inflation. Truncating the expansion of EFT operators at the lowest order, we obtain the
Lagrangian [76]
(∂µ φ)4
Leff = L0 (φ) + , (9.12)
8Λ4
where L0 is the canonical slow-roll Lagrangian and Λ2 ≡ Λ̃M . Following [76], we will compute
the bispectrum induced by the interaction in (9.12).
This time it is useful to work in spatially flat gauge. At leading order in the slow-roll param-
eters, the action for the inflaton perturbation ϕ ≡ φ − φ̄(t) can then be calculated from (9.12)
√
without including the contributions from −gR. The intuitive reason for this simplification is
that in the limit of a flat potential the perturbation ϕ does not induce a perturbation in the
spacetime curvature. Evaluating one of the legs of the interaction (∂µ φ)4 on the background
solution, we get
φ0 0
⇔ Leff,3 = − ϕ (∂µ ϕ)2 . (9.13)
4Λ4
Using this interaction, we can compute the bispectrum hϕϕϕi. We will relate the result to hζζζi
at horizon crossing to avoid the error that would be induced by the superhorizon evolution of ϕ.
To perform this matching, we, in principle, need to know the relation between ϕ and ζ up to
second order. However, the quadratic term is slow-roll suppressed, so it suffices to use the linear
relationship
H
ζ = − 0 ϕ. (9.14)
φ
Feeding Hint = −Leff,3 into the in-in master formula (8.19), we find
Bζ (k1 , k2 , k3 ) 4i (φ0 )2 0
Z
dτ −k12 k22 k33 τ 2 − (k1 · k2 )k32 (1 − ik1 τ )(1 − ik2 τ ) eikt τ

2 = − 4
Pζ k1 k2 k3 Λ −∞
↑ ↑
+ perms. + c.c. , 0 3 (9.15)
(ϕ ) (ϕ )(∂i ϕ)2
0
66
where kt ≡ k1 + k2 + k3 . Evaluating the integral, we get [76]
!
Bζ (k1 , k2 , k3 ) 8 (φ0 )2 1 X X X
2 = ki5 + (2ki4 kj −3ki3 kj2 )+ (ki3 kj kl −4ki2 kj2 kl ) . (9.16)
Pζ k1 k2 k3 Λ4 kt2
i i6=j i6=j6=l
The following features of this result are worth noting:
• The signal peaks in the equilateral configuration (cf. Fig. 23).
• The squeezed limit, limkL →0 hζkS ζkS ζkL i, is an analytic function of kL /kS . This is a conse-
quence of the interaction being local.
• The amplitude is bounded,

35 (φ0 )2
fNL = < 1, (9.17)
108 Λ4
where the final inequality follows from φ0 < Λ2 , as required by consistency of the truncation
used in (9.12).
Nonlocal effects
For M . few × H, there are effects of massive particles that cannot be captured by local inflaton
interactions, i.e. the field cannot be integrated out completely. The non-locality of the effective
inflaton interactions will be reflected in a characteristic non-analyticity in its correlation functions.
As a concrete example, we consider a massive scalar field σ. Its two-point function in de Sitter
space is
π 2
hσk (τ )σk0 (τ 0 )i0 =
H (τ τ 0 )3/2 e−πµ Hiµ (−kτ )Hiµ
∗
(−kτ 0 ) , (9.18)
4
(1) p
where Hiµ ≡ Hiµ is the Hankel function of the first kind and µ ≡ m2 /H 2 − 9/4. For now,
let us focus on particles belonging to the principal series, so that µ is real. The local part of the
two-point function has support only at coincident points in position space, while the non-local
part describes correlations over long distances. In Fourier space, the local and non-local parts
of the two-point function are analytic and non-analytic in the momentum k, respectively. In the
late-time limit, we can split (9.18) into its local and non-local parts
H 2 (τ τ 0 )3/2
τ −iµ
τ iµ
lim hσk (η)σk0 (τ 0 )i0local = Γ(−iµ)Γ(iµ) eπµ 0 + e−πµ 0 , (9.19)
0
τ,τ →0 4π τ τ
H 2 (τ τ 0 )3/2 k 2 τ τ 0 iµ k 2 τ τ 0 −iµ

lim hσk (τ )σk0 (τ 0 )i0non-local = Γ(−iµ)2 + Γ(iµ)2 . (9.20)
τ,τ 0 →0 4π 4 4
Note that Γ(±iµ) → e−πµ/2 for large µ, resulting in an overall suppression of e−πµ of the non-local
contribution (9.20). In the next section, we will see how the non-analyticity of the non-local part
of the two-point function of σ is reflected in the inflationary correlators.
67
Figure 28. The spontaneous creation and decay of a massive particle σ produces a four-point function
for the inflaton field φ. Evaluating one of the legs on the background solution, φ̄, leads to a three-point
function for the inflaton fluctuations ϕ.
9.3 Cosmological Collider Physics

The correlations in the massive field σ become observable when they get converted to inflaton
fluctuations through interactions such as the mixing term in (9.10):
σ(∂µ φ)2
L ⊃ . (9.21)
Λ
Evaluating one of the legs of the inflaton field on the background solution, φ̄, leads to a bispectrum
for the inflaton fluctuations ϕ (and hence of the curvature perturbations ζ); cf. Fig. 28. The
derivation of the bispectrum of curvature perturbations can be found in [58, 59]. Here, we
summarize the main features of the result:
• The non-local effect associated with the massive particle exchange shows up as a non-
analyticity in the squeezed limit, limkL →0 hζkS ζ−kS ζkL i. Between the horizon crossing times
of the long mode ζL and the short modes ζS the amplitude of the massive field oscillates
with a frequency set by the mass of the field. This leads to distinct oscillations in the
bispectrum of curvature perturbations:
3/2
kL M kL
lim hζkS ζ−kS ζkL i ∝ cos ln +δ , (9.22)
kL →0 kS H kS
where δ is a computable phase [58]. We see that the mass M of the field is encoded in the
frequency of the oscillations (see Fig. 29).
• Particles with spin s lead to a unique angular dependence (see Fig. 30):
lim hζkS ζ−kS ζkL i ∝ Ps (cos θ) , (9.23)

kL →0
where cos θ ≡ kS · kL and Ps is a Legendre polynomial of degree s.
There are several effects that determine the amplitude of the signal:
68
Figure 29. Shape function for the bispectrum arising from the exchange of a spin-2 particle with µ = 5,
evaluated in the isosceles-triangle configuration, x2 ≡ k2 /k1 = 1 (figure adapted from [59]). The dotted
line is the non-local part of the signal.
Figure 30. Shape function for the bispectrum arising from the exchange of a spin-2 particle with µ = 5,
as a function of the base angle θ = cos−1 (k̂1 · k̂3 ) for fixed ratios of x3 ≡ k3 /k1 (figure adapted from [59]).
• If we make the conservative assumption that the interaction in (9.21) is only of gravitational
strength [58], then the scale determining the mixing between σ and φ is Λ = Mpl . In
that case, the amplitude of the bispectrum is at most of order the gravitational floor,
fNL . O(ε). However, in principle, the mixing can be much larger. In particular, the
mixing interaction in (9.21) remains perturbative as long as Λ ≥ (φ0 )1/2 , allowing for non-
Gaussianity of order fNL . O(1) [53, 54, 59, 77]. Finally, in the EFT for the inflationary
fluctuations [78] the cutoff Λ can even be below (φ0 )1/2 and the non-Gaussianity can be
−1/2
much larger, fNL . O(Pζ ).
• In addition, the spontaneous production of massive particles in the principal series is ex-
ponentially suppressed. This Boltzmann suppression is inherited by the amplitude of the
bispectrum, fNL ∝ e−πM/H . The creation of particles in the complementary series, on the
69
other hand, does not receive the same suppression and the signal can be large [53, 54].
Instead of oscillations, the squeezed limit in that case has a monotonic scaling (kL /kS )∆ ,
where the scaling dimension ∆ carries the information about the mass of the particle.
• For particles in the principal series, the squeezed limit is suppressed by a factor of (kL /kS )3/2 .
This is to be compared to the squeezed limit for equilateral non-Gaussianity which scales
as (kL /kS )2 . The signal from particles in the complementary series can be larger in the
squeezed limit, (kL /kS )∆ , where 0 < ∆ < 3/2. Finally, partially massless particles do not
decay on superhorizon scales and may therefore induce signals that are not suppressed in
the squeezed limit [63].
Figure 31 is a schematic illustration of current and future constraints on (scale-invariant) pri-

mordial non-Gaussianities. We see that the perturbatively interesting regime spans about seven
orders of magnitude in fNL . Of this regime, three orders of magnitude have been ruled out by
current CMB observations, leaving a window of opportunity of about four orders of magnitude to
be explored. Accessing these low levels of non-Gaussianity will be challenging. Even optimistic
projections for future CMB observations won’t reduce the constraints by more than an order
of magnitude. Digging deeper will require new cosmological probes, such as observations of the
large-scale structure of the universe [79] and the tomography of the 21cm transition of neutral
hydrogen gas [80].
gravitational floor ruled out by non-perturbative

Planck
21cm? accessible with future CMB

future LSS?
Figure 31. Schematic illustration of current and future constraints on (scale-invariant) primordial non-
Gaussianity. The “gravitational floor” denotes the minimal level of non-Gaussianity created by purely
gravitational interactions during inflation [40] (see §8.3).
70
References
[1] D. Baumann, Cosmology (http://cosmology.amsterdam).

[2] S. Dodelson, Modern Cosmology. Academic Press, 2003.
[3] P. Peter and J.-P. Uzan, Primordial Cosmology. Oxford University Press, 2013.
[4] B. Follin, L. Knox, M. Millea, and Z. Pan, “First Detection of the Acoustic Oscillation Phase Shift
Expected from the Cosmic Neutrino Background,” Phys. Rev. Lett. 115 no. 9, (2015) 091301,
arXiv:1503.07863 [astro-ph.CO].
[5] D. Baumann, D. Green, J. Meyers, and B. Wallisch, “Phases of New Physics in the CMB,” JCAP
1601 (2016) 007, arXiv:1508.06342 [astro-ph.CO].
[6] D. Baumann, “Inflation,” in Physics of the Large and the Small, TASI 09, pp. 523–686. 2011.
arXiv:0907.5424 [hep-th].
[7] D. Baumann, Advanced Cosmology (http://cosmology.amsterdam).
[8] R. Durrer, The Cosmic Microwave Background. Cambridge University Press, Cambridge, 2008.
[9] W. Hu, Wandering in the Background: A CMB Explorer. PhD thesis, UC, Berkeley, 1995.
arXiv:astro-ph/9508126 [astro-ph].
[10] W. Hu and S. Dodelson, “Cosmic Microwave Background Anisotropies,” Ann. Rev. Astron.
Astrophys. 40 (2002) 171–216, arXiv:astro-ph/0110414 [astro-ph].
[11] W. Hu, “Lecture Notes on CMB Theory: From Nucleosynthesis to Recombination,”
arXiv:0802.3688 [astro-ph].
[12] W. Hu, “Concepts in CMB Anisotropy Formation,” in The Universe at High-z, Large-Scale
Structure and the Cosmic Microwave Background, pp. 207–239. Springer, 1996.
[13] R. Peccei and H. Quinn, “CP Conservation in the Presence of Pseudoparticles,” Phys. Rev. Lett. 38
(1977) 1440.
[14] S. Weinberg, “A New Light Boson?,” Phys. Rev. Lett. 40 (1978) 223.
[15] F. Wilczek, “Problem of Strong P and T Invariance in the Presence of Instantons,” Phys. Rev. Lett.
40 (1978) 279.
[16] A. Arvanitaki, S. Dimopoulos, S. Dubovsky, N. Kaloper, and J. March-Russell, “String Axiverse,”
Phys. Rev. D81 (2010) 123530, arXiv:0905.4720 [hep-th].
[17] B. Holdom, “Two U (1)’s and Epsilon Charge Shifts,” Phys. Lett. 166B (1986) 196–198.
[18] K. Abazajian et al., “Light Sterile Neutrinos: A White Paper,” arXiv:1204.5379 [hep-ph].
[19] K. Abazajian et al. (CMB-S4 Collaborations), “CMB-S4 Science Book, First Edition,”
[20] B. Wallisch, Cosmological Probes of Light Relics. PhD thesis, University of Cambridge, 2018.
[21] Z. Hou, R. Keisler, L. Knox, M. Millea, and C. Reichardt, “How Massless Neutrinos Affect the
Cosmic Microwave Background Damping Tail,” Phys. Rev. D87 (2013) 083008, arXiv:1104.2333
[astro-ph.CO].
[22] S. Bashinsky and U. Seljak, “Neutrino Perturbations in CMB Anisotropy and Matter Clustering,”
Phys. Rev. D 69 (2004) 083002, arXiv:astro-ph/0310198 [astro-ph].
[23] D. Baumann, F. Beutler, R. Flauger, D. Green, M. Vargas-Magaña, A. Slosar, B. Wallisch, and
71
C. Yèche, “First Measurement of Neutrinos in the BAO Spectrum,” arXiv:1803.10741
[astro-ph.CO].
[24] C. Brust, D. E. Kaplan, and M. Walters, “New Light Species and the CMB,” JHEP 12 (2013) 058,
arXiv:1303.5379 [hep-ph].
[25] F. Wilczek, “Axions and Family Symmetry Breaking,” Phys. Rev. Lett. 49 (1982) 1549.
[26] D. Reiss, “Can the Family Group be a Global Symmetry?,” Phys. Lett. B 115 (1982) 217.
[27] J. Kim, “Light Pseudoscalars, Particle Physics and Cosmology,” Phys. Rep. 150 (1987) 1.
[28] Y. Chikashige, R. Mohapatra, and R. Peccei, “Are There Real Goldstone Bosons Associated with
Broken Lepton Number?,” Phys. Lett. B 98 (1981) 265.
[29] Y. Chikashige, R. Mohapatra, and R. Peccei, “Spontaneously Broken Lepton Number and
Cosmological Constraints on the Neutrino Mass Spectrum,” Phys. Rev. Lett. 45 (1980) 1926.
[30] D. Baumann, D. Green, and B. Wallisch, “New Target for Cosmic Axion Searches,” Phys. Rev.
Lett. 117 no. 17, (2016) 171301, arXiv:1604.08614 [astro-ph.CO].
[31] D. Baumann, D. Green, and B. Wallisch, “Searching for Light Relics with Large-Scale Structure,”
[32] A. Salvio, A. Strumia, and W. Xue, “Thermal Axion Production,” JCAP 1401 (2014) 011.
[33] G. Carosi, A. Friedland, M. Giannotti, M. Pivovaroff, J. Ruz, and J. Vogel, “Probing the
Axion-Photon Coupling: Phenomenological and Experimental Perspectives,”. arXiv:1309.7035
[hep-ph].
[34] P. Graham and S. Rajendran, “New Observables for Direct Detection of Axion Dark Matter,” Phys.
Rev. D 88 (2013) 035023.
[35] K. Blum, R. D’Agnolo, M. Lisanti, and B. Safdi, “Constraining Axion Dark Matter with Big Bang
Nucleosynthesis,” Phys. Lett. B 737 (2014) 30.
[36] G. Raffelt, Stars as Laboratories for Fundamental Physics. University of Chicago Press, 1996.
[37] M. Pospelov and A. Ritz, “Theta-Induced Electric Dipole Moment of the Neutron via QCD Sum
Rules,” Phys. Rev. Lett. 83 (1999) 2526.
[38] D. Baumann and L. McAllister, Inflation and String Theory. Cambridge University Press, 2015.
[39] R. Arnowitt, S. Deser, and C. Misner, “The Dynamics of General Relativity,” Gen. Rel. Grav. 40
(2008) 1997–2027, arXiv:gr-qc/0405109 [gr-qc].
[40] J. Maldacena, “Non-Gaussian Features of Primordial Fluctuations in Single-Field Inflationary
Models,” JHEP 05 (2003) 013, arXiv:astro-ph/0210603 [astro-ph].
[41] P. A. R. Ade et al. (Planck and BICEP2 Collaborations), “Joint Analysis of BICEP2/KeckArray
and P lanck Data,” Phys. Rev. Lett. 114 (2015) 101301, arXiv:1502.00612 [astro-ph.CO].
[42] S. Weinberg, “Quantum Contributions to Cosmological Correlations,” Phys. Rev. D72 (2005)
043514, arXiv:hep-th/0506236 [hep-th].
[43] X. Chen, “Primordial Non-Gaussianities from Inflation Models,” Adv. Astron. 2010 (2010) 638979,
[44] Y. Wang, “Inflation, Cosmic Perturbations and Non-Gaussianities,” Commun. Theor. Phys. 62
(2014) 109–166, arXiv:1303.1523 [hep-th].
72
[45] E. Lim, Part III Advanced Cosmology: Primordial Non-Gaussianities.
[46] E. Komatsu, “Hunting for Primordial Non-Gaussianity in the Cosmic Microwave Background,”
Class. Quant. Grav. 27 (2010) 124010, arXiv:1003.6097 [astro-ph.CO].
[47] J. Schwinger, “Brownian Motion of a Quantum Oscillator,” J. Math. Phys. 2 (1961) 407–432.
[48] E. Calzetta and B. Hu, “Closed Time Path Functional Formalism in Curved Space-Time:
Application to Cosmological Back Reaction Problems,” Phys. Rev. D35 (1987) 495.
[49] R. Jordan, “Effective Field Equations for Expectation Values,” Phys. Rev. D33 (1986) 444–454.
[50] D. Seery, “One-Loop Corrections to a Scalar Field during Inflation,” JCAP 0711 (2007) 025,
arXiv:0707.3377 [astro-ph].
[51] P. Adshead, R. Easther, and E. Lim, “The ’in-in’ Formalism and Cosmological Perturbations,”
Phys. Rev. D80 (2009) 083521, arXiv:0904.4207 [hep-th].
[52] P. Creminelli and M. Zaldarriaga, “Single-Field Consistency Relation for the Three-Point
Function,” JCAP 0410 (2004) 006, arXiv:astro-ph/0407059 [astro-ph].
[53] X. Chen and Y. Wang, “Quasi-Single-Field Inflation and Non-Gaussianities,” JCAP 1004 (2010)
027, arXiv:0911.3380 [hep-th].
[54] D. Baumann and D. Green, “Signatures of Supersymmetry from the Early Universe,” Phys. Rev.
D85 (2012) 103520, arXiv:1109.0292 [hep-th].
[55] X. Chen and Y. Wang, “Quasi-Single-Field Inflation with Large Mass,” JCAP 1209 (2012) 021,
[56] V. Assassi, D. Baumann, and D. Green, “On Soft Limits of Inflationary Correlation Functions,”
JCAP 1211 (2012) 047, arXiv:1204.4207 [hep-th].
[57] T. Noumi, M. Yamaguchi, and D. Yokoyama, “EFT Approach to Quasi-Single-Field Inflation and
Effects of Heavy Fields,” JHEP 06 (2013) 051, arXiv:1211.1624 [hep-th].
[58] N. Arkani-Hamed and J. Maldacena, “Cosmological Collider Physics,” arXiv:1503.08043
[hep-th].
[59] H. Lee, D. Baumann, and G. Pimentel, “Non-Gaussianity as a Particle Detector,” JHEP 12 (2016)
040, arXiv:1607.03735 [hep-th].
[60] L. Bordin, P. Creminelli, M. Mirbabayi, and J. Noreña, “Tensor Squeezed Limits and the Higuchi
Bound,” arXiv:1605.08424 [astro-ph.CO].
[61] R. Flauger, M. Mirbabayi, L. Senatore, and E. Silverstein, “Productive Interactions: Heavy
Particles and Non-Gaussianity,” arXiv:1606.00513 [hep-th].
[62] J.-O. Gong, S. Pi, and M. Sasaki, “Equilateral Non-Gaussianity from Heavy Fields,” JCAP 1311
(2013) 043, arXiv:1306.3691 [hep-th].
[63] D. Baumann, G. Goon, H. Lee, and G. L. Pimentel, “Partially Massless Fields During Inflation,”
arXiv:1712.06624 [hep-th].
[64] E. Wigner, “On Unitary Representations of the Inhomogeneous Lorentz Group,” Annals Math. 40
(1939) 149–204. [Reprint: Nucl. Phys. Proc. Suppl.6,9(1989)].
[65] V. Bargmann and E. Wigner, “Group Theoretical Discussion of Relativistic Wave Equations,” Proc.
Nat. Acad. Sci. 34 (1948) 211.
[66] S. Deser and A. Waldron, “Arbitrary Spin Representations in de Sitter from dS/CFT with
73
Applications to dS Supergravity,” Nucl. Phys. B662 (2003) 379–392, arXiv:hep-th/0301068
[hep-th].
[67] B. de Wit, “Supergravity,” in Unity from duality: Gravity, gauge theory and strings. Proceedings,
NATO Advanced Study Institute, Euro Summer School, 76th session, Les Houches, France, July
30-August 31, 2001, pp. 1–135. 2002. arXiv:hep-th/0212245 [hep-th].
[68] L. Thomas, “On Unitary Representations of the Group of De Sitter Space,” Annals of Mathematics
42 no. 1, (1941) 113–126. http://www.jstor.org/stable/1968990.
[69] T. Newton, “A Note on the Representations of the De Sitter Group,” Annals of Mathematics 51
no. 3, (1950) 730–733. http://www.jstor.org/stable/1969376.
[70] T. Garidi, “What is Mass in de Sitterian Physics?,” arXiv:hep-th/0309104 [hep-th].
[71] S. Deser and A. Waldron, “Gauge Invariances and Phases of Massive Higher Spins in (A)dS,” Phys.
Rev. Lett. 87 (2001) 031601, arXiv:hep-th/0102166 [hep-th].
[72] A. Higuchi, “Forbidden Mass Range for Spin-2 Field Theory in De Sitter Spacetime,” Nucl. Phys.
B282 (1987) 397.
[73] L. Singh and C. Hagen, “Lagrangian Formulation for Arbitrary Spin. 1. The Boson Case,” Phys.
Rev. D9 (1974) 898–909.
[74] L. Singh and C. Hagen, “Lagrangian Formulation for Arbitrary Spin. 2. The Fermion Case,” Phys.
Rev. D9 (1974) 910–920.
[75] Y. Zinoviev, “On Massive High Spin Particles in AdS,” arXiv:hep-th/0108192 [hep-th].
[76] P. Creminelli, “On Non-Gaussianities in Single-Field Inflation,” JCAP 0310 (2003) 003,
arXiv:astro-ph/0306122 [astro-ph].
[77] V. Assassi, D. Baumann, D. Green, and L. McAllister, “Planck-Suppressed Operators,” JCAP
1401 (2014) 033, arXiv:1304.5226 [hep-th].
[78] C. Cheung, P. Creminelli, A. L. Fitzpatrick, J. Kaplan, and L. Senatore, “The Effective Field
Theory of Inflation,” JHEP 03 (2008) 014, arXiv:0709.0293 [hep-th].
[79] M. Alvarez et al., “Testing Inflation with Large Scale Structure: Connecting Hopes with Reality,”
[80] A. Loeb and M. Zaldarriaga, “Measuring the Small-Scale Power Spectrum of Cosmic Density
Fluctuations through 21 cm Tomography Prior to the Epoch of Structure Formation,” Phys. Rev.
Lett. 92 (2004) 211301, arXiv:astro-ph/0312134 [astro-ph].
74

Baumann Primordial Cosmology

Uploaded by

Copyright:

Available Formats

Baumann Primordial Cosmology

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Baumann Primordial Cosmology

Uploaded by

Copyright:

Available Formats

TASI Lectures on

Institute of Theoretical Physics, University of Amsterdam,

I Relics from the Hot Big Bang 4

2 Big Bang Cosmology 5

3 Afterglow of the Big Bang 12

4 Cosmic Sound Waves 20

II Relics from Inflation 35

7 Quantum Initial Conditions 42

σ > (n∆t)−1 ∼ (1010 GeV)−2 .

2.1 Geometry and Dynamics

ds2 = −dt2 + a2 (t)γij dxi dxj , (2.1)

where P µ ≡ dxµ /dλ is the four-momentum of the particle. In an expanding spacetime, it is

P µ = a−1 [E, p] . (2.4)

Γ000 = H , Γ0ij = Hδij , Γij0 = Hδji , (2.5)

where H ≡ ȧ/a is the conformal Hubble parameter.

The µ = 0 component of the geodesic equation (2.3) becomes

3H2 = 8πGa2 ρ̄ , (2.9)

where ρ̄ and P̄ are the background density and pressure, respectively.

Exercise.—By substituting (2.5) into

Rµν ≡ ∂λ Γλµν − ∂ν Γλµλ + Γλλρ Γρµν − Γρµλ Γλνρ , (2.11)

R00 = −3Ḣ , Rij = (Ḣ + 2H2 )δij ⇒ a2 R ≡ a2 g µν Rµν = −6(Ḣ + H2 ) . (2.12)

G00 = 3H2 , Gij = −(2Ḣ + H2 )δij . (2.13)

ρ̄˙ = −3H(ρ̄ + P̄ ) . (2.14)

The total radiation density is

the photons. Before e+ e− annihilation, i.e. at T+ > me , we have

Comparing g∗ (aTγ )3 = const. to aTν = const., we get

2.3 Structure Formation

2.4 Initial Conditions

δρa (τ, x) ≡ ρ̄a (τ + δτ (x)) − ρ̄a (τ ) ≈ ρ̄˙ a δτ (x) , (2.28)

gij (τ, x) = a2 (τ ) e2ζ(τ,x) δij , (2.29)

hζ(x)ζ(x0 )i ≡ ξζ (x, x0 ) = ξζ (|x0 − x|) , (2.30)

3.1 CMB Anisotropies

δT (n̂) T0 (n̂) − T p0 (n̂) − p

Perturbed photon geodesics

ds2 = a2 (τ ) − (1 + 2Φ)dτ 2 + (1 − 2Ψ)δij dxi dxj ,

where the perturbation Φ ≈ Ψ plays the role of the gravitational potential.

To determine the constant of proportionality α, we use

p2 = gij P i P j = a2 (1 − 2Ψ)δij p̂i p̂j α2

Substituting these results into the geodesic equation (3.7), we get

which confirms the result (3.6).

To relate this to the temperature anisotropy, we note that

Each term on the right-hand side has a simple physical interpretation:

Large scales.—For adiabatic initial conditions, the superhorizon initial condition is δγ ≈ 43 δm ≈ − 83 Ψ.

3.2 CMB Power Spectrum

The two-point function of temperature anisotropies then becomes

hδT (k)δT (k0 )i = TSW (k)TSW (k 0 )hζ(k) ζ(k0 )i

A similar derivation gives the form of the Doppler contribution.

Figure 8. Interactions between the different forms of matter in the universe.

4.1 Photon-Baryon Fluid

4.2 Acoustic Oscillations

Θ(k, τ ) = Ak cos(cs kτ ) + Bk sin(cs kτ ) , (4.4)

Θ(k, τ ) = 3ζk cos(cs kτ ) . (4.5)

Evaluating this solution at recombination, τ = τ∗ , gives the Sachs-Wolfe transfer function,

Θ̈ + µ c2s k 2 Θ̇ + c2s k 2 Θ = 0 , (4.15)

where we have ignored the gravitational source terms and defined

Using the WKB ansatz  Z 

where we have defined the diffusion wavenumber as

In the limit R → 0, this becomes

which agrees with our previous estimate (4.13).

Using the WKB ansatz Z

We note that the total field excursion is super-Planckian, ∆φ = φi − φe Mpl .

fNL ∼ O(ε, η) 1 . (8.28)