Generating Curvature Perturbations
Generating Curvature Perturbations
Generating Curvature Perturbations
a contracting universe
Aaron M. Levy
A Dissertation
Presented to the Faculty
of Princeton University
in Candidacy for the Degree
of Doctor of Philosophy
April 2017
c Copyright by Aaron M. Levy, 2017.
This thesis studies bouncing cosmologies in which the present-day expansion of the
universe was preceded not by a “big bang”– before which time and space ceased to
have meaning– but by a contracting phase that then bounced. We discuss two com-
peting paradigms for generating the observed, scale-invariant spectrum of primordial
density perturbations during the contracting phase: “the matter bounce scenario”
and “ekpyrosis.” First, we discuss the matter bounce scenario, and in particular, its
fine-tuning instability to the growth of anisotropic stress. Then, we examine ekpyro-
sis. In the best-understood ekpyrotic models, one scalar field drives the background
evolution of the universe while another (entropic) scalar field generates the density
perturbations. We study the stability of these models, showing that in contrast to
previous theorems, the simplest (as measured by parameters and degrees of freedom),
observationally viable realizations are dynamical attractors. Finally, we present a new
mechanism called “warm ekpyrosis,” which eliminates altogether the need for the sec-
ond (entropic) scalar field. Rather, a single field falls down its ekpyrotic potential,
smoothing and flattening the universe, while simultaneously, through couplings to
lighter degrees of freedom, decaying into hot, ultrarelativistic matter. This decay
allows both for the production of a scale-invariant density perturbation and for a
possible mechanism of reheating.
iii
Acknowledgements
I would like to thank my advisor, Paul Steinhardt. I would also like to express my
sincerest gratitude to my senior colleague, Lasha Berezhiani, for helpful discussions
and insights; David Spergel for his consistently sound advice; Frans Pretorius and
Chris Tully for serving on my prethesis committee; Herman Verlinde for his valuable
counsel as Director of Graduate Studies; Lyman Page for his indispensable role as
Department Chair; Peter Meyers for being my experimental advisor; Joaquin Turiaci
for our fun collaboration; and my friends and family for their unwavering love and
encouragement.
iv
Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
1 Introduction 1
1.1 Olbers’ paradox and the beginning of the Universe . . . . . . . . . . . 1
1.2 The cosmic microwave background . . . . . . . . . . . . . . . . . . . 2
1.3 Big bang problems and the inflationary proposal . . . . . . . . . . . . 3
1.4 Bouncing cosmologies . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
v
3.2 Scaling solutions, scale-invariance and instability . . . . . . . . . . . . 26
3.3 Ekpyrosis and scale-invariance without fine tuning . . . . . . . . . . . 32
3.3.1 V (ψ) = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.2 V (ψ) 6= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.3 Further generalizations . . . . . . . . . . . . . . . . . . . . . . 40
3.4 Constructing new models . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4.1 Fast growth: |f,σ /f | & 1 . . . . . . . . . . . . . . . . . . . . . 44
3.4.2 Slow growth: |f,σ /f | . 1 . . . . . . . . . . . . . . . . . . . . . 46
3.4.3 “Just-so” growth: |f,σ /f | ∼ 1 . . . . . . . . . . . . . . . . . . 46
3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.6 Appendix A: Deriving the coordinate transformation . . . . . . . . . 49
3.7 Appendix B: Deriving the properties . . . . . . . . . . . . . . . . . . 56
4 Warm ekpyrosis 60
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.3 Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3.1 Cold ekpyrosis . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.2 Warm ekpyrosis . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.5 Appendix A: Perturbation equations . . . . . . . . . . . . . . . . . . 78
4.6 Appendix B: Locality of fluid response . . . . . . . . . . . . . . . . . 83
4.7 Appendix C: Thermal contribution to the scalar spectrum . . . . . . 85
Bibliography 94
vi
Chapter 1
Introduction
verse
The darkness of the night sky immediately presents a paradox [1], named after the
German astronomer Heinrich Wilhelm Olbers, that sheds. . . well. . . light on the na-
ture of the cosmos. For if the Universe were eternal and static and filled more or less
uniformly with stars like the sun, then the intensity of light observed at the Earth
would be infinite; that is, the night sky would be bright, not dark. To see this, note
that in a spherical shell around the Earth of width dr at radius r containing a number
density n of stars, there would exist 4πr2 ndr stars. If each one emitted at luminosity
L, then an Earth-bound observer would measure an intensity, dI, from this shell that
is independent of radius, i.e.,
L
dI = (4πr2 n) dr = nLdr. (1.1)
4πr2
R R∞
Therefore, the total intensity would diverge, i.e., I = dI = nL 0
dr = ∞, indicat-
ing an infinitely bright sky.
1
The resolution, encoded in Einstein’s equations of general relativity [2], is that
the Universe is not static. Space is not merely a fixed background through which
particles move, but rather stretches and bends in response to them. In our Universe,
as a result of this stretching, distant regions grow ever more distant over time. As a
corollary, in the past, the Universe was smaller and denser. Extrapolating Einstein’s
equations back through this high density regime reveals a singularity approximately
13.8 billion years ago, an event Fred Hoyle famously derided in 1949 as a “big bang.”
Ever since then, light, which travels at finite speed, can have traversed only a finite
distance. In turn, this means that Earth-bound observers today can see only a finite
distance away and therefore finitely many stars [3]. Therefore, the total intensity is
finite; the night sky is dark.
When we observe an object that is one light year away, it appears to us as it was
one year ago, since this is when the light was emitted, only having reached us after
a year of transit. By extension, the farther out we look in distance, the further back
we are looking in time. By observing very distant light sources, we are in effect
studying the Universe in its nascency. Unfortunately, there is a limit to how far we
can see. The farthest light we can detect originates from the time when the Universe,
having burst into existence hot and dense after the big bang, had stretched and cooled
enough that free electrons, from which photons had previously Thomson scattered,
bound to protons to form neutral hydrogen. Over the last 13.8 billion years, these old
photons have been free streaming through space, carrying a snapshot of the Universe
as it was 380,000 years after the big bang. First detected in 1964 by Penzias and
Wilson [4] and measured today by precision ground-based [5, 6, 7, 8], balloon-borne
[9, 10] and satellite telescopes [11, 12, 13], this “cosmic microwave background” (CMB)
2
radiation has a nearly ideal blackbody spectrum at a temperature of 2.73 K and is
isotropic except for minuscule anisotropic fluctuations of order but 10−5 (ignoring the
dipole anisotropy induced by the peculiar motion of the Earth with respect to the
cosmological rest frame of the CMB). According to the theory of general relativity,
these fluctuations grew via gravitational instability to form the foam-like cosmic web
of large-scale structure we observe today.
Given the great success of explaining the growth of large-scale structure from these
initial fluctuations, it is natural to ask what caused them. In searching for an ex-
planation, in 1969, Charles Misner realized something troubling, a conundrum he
referred to as the “horizon paradox” [14]. Due to the finite time since the big bang
and the finite speed of light, there is a greatest possible distance– a particle horizon–
over which physical information can travel. Cosmic microwave background photons
that subtend an angle of more than ≈ 2◦ on the sky were beyond each other’s particle
horizon, meaning they were out of causal contact, at the time they decoupled from
matter. Nevertheless, the CMB radiates as a nearly perfect isotropic blackbody. It is
as though distant regions achieved thermal equilibrium, despite insufficient time for
any physical mechanism to induce it. Therefore, the Universe began with a density
profile that is inexplicably smooth. In that same year, Robert Dicke noticed a similar
problem: the Universe began with a spatial geometry that is inexplicably flat [15].
Reflecting on these problems, Misner remarked, “Things that you don’t under-
stand can be a constant to ten or twenty percent, but to one percent, it requires
an explanation.” Since 1981, the conventional explanation for the initial smooth-
ness and flatness of the Universe has been the inflation scenario [16, 17, 18]. In its
ideal formulation, typical conditions arising from the big bang allow small, causally
3
connected regions of the Universe to undergo a period of rapid, accelerated expan-
sion, i.e., inflation, before nucleosynthesis. The accelerated expansion smoothes and
flattens the local geometry, while simultaneously stretching quantum fluctuations to
cosmological scales. These quantum fluctuations [19, 20, 21, 22, 23] imprint the ob-
served anisotropies onto the CMB and form the density perturbations that seed the
formation of large-scale structure.
However, this idealization has been criticized for several weaknesses. First, al-
though it was introduced to explain the required smoothness and flatness, inflation
requires special initial conditions of its own, conditions certain entropic arguments
suggest are extremely, exponentially unlikely [24]. For example, applying the canon-
ical Liouville measure to homogeneous, isotropic universe, the authors of Ref. [25]
demonstrate that such universes are exponentially less likely to result from long pe-
riods of inflation than from no inflation at all. Concordantly, the quantum argument
given in Ref. [26] concludes that constituent quanta at sub-Planckian energy densities
are exponentially unlikely to initiate inflation, even when weighted by volume.
Second, the inflationary paradigm is versatile to the point of being unfalsifiable.
By adding parameters and adjusting their values, inflationary models can accommo-
date nearly any cosmological observables [27]. Unfortunately, for the models most fa-
vored by observation– those involving a single field with a flat, plateau-like potential–
initiating inflation requires smoothness over scales much greater than the Hubble ra-
dius [28], the very condition inflation was supposed to provide. Without it, gradients
and inhomogeneities rapidly dominate the evolution of the Universe, preventing in-
flation from ever starting. Thus, “inflation does not predict smoothness and flatness–
it assumes it” [29].
Third, given a particular model, inflation leads almost unavoidably to the multi-
verse, according to which any outcome is possible [30, 31, 32, 33, 34, 35]. In most
regions, quantum fluctuations of the inflaton field exert small effects– the ones respon-
4
sible for the temperature anisotropy observed in the CMB. In these places, inflation
terminates when the classically evolving inflaton field falls far enough down its poten-
tial function. But in rare regions, the quantum fluctuations deliver large kicks back
up the potential, impeding the classical descent of the inflaton, thereby prolonging
the exponential expansion. In short order, these rare eternally inflating regions dom-
inate the volume of space, and the process repeats anew. Far from the uniformity
inflation was intended to provide, the universe obtains a fractal-like structure as small
pocket universes terminate inflation at different times and with different cosmologi-
cal properties. Some are flat and smooth with a scale-invariant spectrum of density
fluctuations like ours, but an infinite number are not. According to Guth, “Anything
that can happen will happen; in fact, it will happen infinitely many times” [36]. With
no selection rule– no probability measure– to favor pocket universes like our own, the
multiverse renders inflation unpredictive.
Finally, inflating universes have been shown to be geodesically past-incomplete
[37]. This implies that inflation is fundamentally unable to explain the nature of the
big bang singularity. Rather, some unknown physics is required. These issues have
motivated the search for better answers.
One approach is to replace the initial big bang singularity with a big bounce. In
contrast to inflation, one-time bouncing cosmologies (and their cyclic cousins) can
be past-eternal [38], meaning the universe has no creation event because it always
exists. As a happy consequence, the pre-bounce contracting phase offers plenty of time
for resolving Misner’s horizon paradox: in a past-eternal universe, particle horizons
are infinite, so distant regions enjoy perennial causal contact. Moreover, bouncing
5
cosmologies resolve the flatness problem since contraction, like inflation, is a natural
flattener.
The earliest cyclic solution to Einstein’s equations predates inflation by nearly
60 years, way back to Friedmann’s “periodic world” of 1922 [39]. In it, Friedmann
argued that a homogeneous universe with a cosmological constant and a supercritical
matter density executes eternal, regular periods of expansion and contraction with
bouncing transitions in between. These ideas generated significant interest amongst
the founders of the big bang theory (e.g., Einstein [40, 41] and Lemaître [42]) until
Tolman emphasized a serious thermodynamic flaw [43, 44]. He argued that as a direct
consequence of the second law of thermodynamics, the regular periods of expansion
and contraction should increase in duration over time. Backtracking, then, the periods
universe achieves zero size a finite time in the past, thereby reintroducing the creation
event the model was designed to avoid.
Another problem for contracting universes, first shown by Belinksii, Khalatnikov,
and Lifshitz (BKL) in the 1960s, is their instability to the growth of anisotropic
stresses [45]. In expanding universes, anisotropies are usually ignored, since they di-
lute much faster than other stress-energy components and therefore quickly become
negligible. In contracting universes, however, BKL showed that the opposite is true:
as space contracts, anisotropies grow more important. In short order, they drive the
universe towards a Kasner-like evolution in which one spatial direction expands while
the other two contract. If spatial curvature is present, then every so often, sudden
jumps between Kasner-like solutions switch the directions and rates of contraction in
a manner reminiscent of a kitchen mixer [14]. Over time, inhomogeneities in the cur-
vature induce asynchronies between jumps at different locations, leading ultimately to
a highly inhomogeneous crunch that is incompatible with the large-scale homogeneity
observed today.
6
As a result of Tolman’s entropy problem and the BKL instability, cyclic cosmolo-
gies fell into disfavor until Steinhardt and Turok proposed an innovation that resolves
both. Initially formulated in the language of M-theory, their new cyclic model [46]
posited that the observable Universe occupies one of two parallel 3-branes whose reg-
ular collisions punctuate long periods of intra-brane stretching and manifest as big
crunch/big bang transitions. The stretching of the branes dilutes the entropy density,
thereby circumventing Tolman’s problem. Meanwhile, the inter-brane potential re-
sponsible for the periodic collisions generates an ultra-slow contracting phase, called
ekpyrosis [47], that evades the BKL instability before the bounce [48]. Over time, this
new cyclic model was understood to admit an effective 3+1-dimensional description
using the same ingredients, i.e., scalar fields, as inflation. In this framework, the ekpy-
rotic phase derives from an effective scalar whose pressure exceeds its energy density,
so that the field dominates all other stress-energy components, even anisotropies, and
the Universe contracts homogeneously.
In light of these successes, bouncing cosmologies have resurged in recent years.
Many mechanisms have been proposed for achieving a bounce, both singular (see,
e.g., [49, 50, 51, 52, 53, 54])– in which quantum gravity effects introduce bounce-
inducing deviations from conventional general relativity– and nonsingular (e.g., [55,
56, 57, 58, 59, 60, 61, 62])– in which the density remains sub-Planckian and the
transition from contraction to expansion evolves classically via a violation of the null
energy condition. Both approaches face difficult challenges, although recent progress
on cubic Galileon theories [63, 64] has provided classically stable nonsingular bounces
that avoid many pathologies plaguing other models, e.g., ghosts, gradient instabilities,
super-luminal sound speeds, etc.
Whether singular or nonsingular, any bouncing model must ultimately explain
the nearly scale-invariant spectrum of primordial fluctuations observed in the CMB.
Since fluctuations on length scales of cosmological interest originate in the contracting
7
phase preceding the bounce, predictions for primordial spectra depend critically on the
gravitational dynamics, i.e., the equation of state of the dominant energy component,
during the contracting phase. In this thesis, we consider two popular contracting
scenarios, examining the stability and spectra of each: the matter bounce scenario
(chapter 2) and ekpyrosis (chapters 3 and 4). We will assess to what extent these
scenarios address the standard fine-tuning problems without introducing additional
tunings of their own.
This thesis is organized as follows: In chapter 2, we show that the BKL insta-
bility poses a serious fine-tuning challenge for any bouncing cosmology involving a
matter-like phase of contraction to generate the observed spectrum of perturbations.
In chapter 3, we argue that, in contrast to previous theorems, precisely the opposite
is true of the simplest, observationally viable, two-field models of ekpyrosis. In par-
ticular, we prove that these models are strong attractors for a wide range of initial
conditions. Finally, in chapter 4, we propose an ekpyrotic mechanism for generat-
ing the observed spectrum that is also an attractor. Termed “warm ekpyrosis,” the
model involves a single ekpyrotic field coupled to a thermal bath of radiation into
which it continuously decays. The decay allows a scale-invariant spectrum to be gen-
erated without the need for a second field. It also provides a possible mechanism for
reheating the post-bounce universe without recourse to stringy, high energy physics.
Chapters 2, 3, and 4 are based on work published in Refs. [65, 66, 67], respectively.
8
Chapter 2
2.1 Introduction
Matter bounce models were introduced to provide a simple mechanism for generating
a scale-invariant spectrum of adiabatic perturbations in accord with observations
[68, 69, 70, 71]. The basic idea is that quantum fluctuations naturally generate a scale-
invariant spectrum of adiabatic curvature perturbations during a contracting phase
if the dominant density is a pressureless (i.e., matter-like) fluid, such as a scalar field
rolling along an exponential potential [72, 73, 74]. If the matter-like phase is followed
by a nonsingular bounce, say, the scale-invariant spectrum can be preserved after
the bounce and provide an explanation of the observed fluctuations in the microwave
background and of the large-scale structure of the universe. This scenario is referred
to as “matter bounce” [75]. It resolves the horizon problem (and in some incarnations,
the flatness problem) of standard Big Bang cosmology [76], generates the observed
9
perturbations, and avoids the multiverse problem [77, 31, 32, 33, 34, 35] of inflation
[78, 17, 18].
One well-known problem with the matter bounce scenario is its overproduction
of tensor fluctuations during the matter-like contracting phase. This issue has been
studied extensively [62, 79, 80] with various proposed resolutions [81, 82, 83, 84, 85].
But perhaps the biggest problem for the matter bounce scenario is the instability
of the contracting matter-like phase to anisotropy. If unchecked, anisotropy rapidly
dominates the energy budget of the universe– spoiling the equation of state respon-
sible for the scale-invariant spectrum– and ultimately leading to chaotic Belinskii-
Khalatnikov-Lifshitz (BKL) behavior [45].
Most works attempting to resolve the anisotropy problem focus on avoiding BKL
instability after the matter-like contracting phase and thus after the generation of
scale-invariant superhorizon perturbations [62, 86, 79]. They argue that if anisotropy
is subdominant by the end of the matter-like contraction, then an ensuing phase of
ekpyrosis will render it negligible thereafter. Suppressing the anisotropy after the
matter-like phase is far too late, though, as we emphasize in this chapter. Unless the
anisotropy is exponentially suppressed before the matter-like phase begins, it rapidly
overtakes the energy density of matter before a sufficient number of scale-invariant
modes have been generated.
In this chapter, we examine the anisotropy problem and demonstrate that it re-
quires extreme, exponential tuning of the initial conditions– in some cases, expo-
nentially more tuning than required to resolve the flatness problem, for example. In
Sec. 2.2, we summarize the anisotropy problem along lines similar to those in Ref. [87].
In Sec. 2.3, we argue that suppressing the anisotropy requires a protracted isotropiz-
ing phase prior to the matter-like phase, and moreover that the degrees of freedom
responsible for the matter-like phase must be coupled to those driving the isotropiz-
ing phase. In Sec. 4.4, we construct an example of this sort involving a canonical
10
scalar field with a specially constructed potential. In Sec. 2.5, we show how resolving
the anisotropy problem requires extreme fine-tuning of this potential. In Sec. 2.6, we
conclude, arguing that this extreme fine-tuning is a generic property of the matter
bounce scenario.
In what follows, we employ reduced Planck units (8πG = 1) and metric signature
(− + ++).
1/
af Hi
= , (2.1)
ai Hf
where subscripts i and f denote initial and final values. In the above, overdots
denote derivatives with respect to coordinate time, t. The ratio of the energy density
in anisotropy (∝ a−6 ) to that in X (∝ a−2 ), f ≡ ρσ /ρX , scales as
2(−3) 2(1− 3 )
ff af Hi
= = , (2.2)
fi ai Hf
where the second equality follows from Eq. (2.1). A perturbation with comoving
wavenumber k will exit the horizon when, k = a|H|. As the universe evolves, a|H|
grows (for > 1), taking shorter and shorter wavelengths outside the horizon. That
11
is, between times tf and ti
af Hf 1 Hf
N ≡ ln = 1− ln (2.3)
ai Hi Hi
e-foldings of scales will have exited the horizon. Combining Eqs. (2.3) and (2.2) yields
ff −3
= exp −2N . (2.4)
fi −1
During ekpyrosis, > 3, so the right side of Eq. (2.4) decreases exponentially with N ,
reflecting the isotropizing power of ekpyrosis. By contrast, during matter-dominated
contraction, = 3/2, so the right side of Eq. (2.4) grows exponentially with N ,
12
matter bounce scenario is many, many orders of magnitude worse than the flatness
problem of standard Big Bang cosmology: it involves a factor of e−214 more tuning.
Without a powerful isotropizing phase before matter domination, it is clear from the
last section that anisotropy quickly spoils the generation of scale-invariant modes. But
suppressing anisotropy before matter domination is impossible unless the degrees of
freedom responsible for the isotropizing phase are coupled to those responsible for the
matter-like phase, as we now show.
Consider a universe with three components: anisotropy, pressureless matter, and
a third stress-energy component, X, which will be used to suppress anisotropy. Since
anisotropy grows faster than matter, suppressing anisotropy with X requires that X
grows faster than both anisotropy and matter. For example, X might be an ekpyrotic
field. Thereafter, this component, which begins greater than matter and grows faster
than matter must somehow give way to matter. If X is decoupled from the matter,
such a transition is impossible. Either X must decay directly into matter, or else
it must drive the matter-like phase itself. In either case, suppressing the anisotropy
imposes extreme fine-tuning requirements on the Hubble parameter, as we will show
below.
The decay scenario suffers additionally from the tight constraint that the decay
products must gravitate like nonrelativistic matter and nothing stiffer that might
spoil a matter-like background. For example, from the reasoning of the previous
section, any relativistic species, produced even in modest amounts, will grow faster
than matter by a factor of exp(2N ), quickly spoiling the matter-like phase. Therefore,
namely in ensuring these modes are today cosmological. That is, it is not enough to generate 7.6
e-foldings of scale-invariant modes; these modes must also be pushed far outside the horizon before
the bounce– but after the matter-like phase– by some mechanism that does not amplify anisotropy,
i.e., ekpyrosis, but that pushes them to just the right distance so that they re-enter the horizon
today. This tuning can exceed that captured by Eq. (2.5).
13
we will focus on the scenario without decay, presenting one realization in which X
is a scalar field whose potential is specially constructed to produce both phases, first
(stiff) ekpyrotic- and then (soft) matter-like contraction. We show that suppressing
the anisotropy is possible only if the potential is extremely fine-tuned. Thus, the
tuning of the anisotropy is traded for a tuned potential, and hence a tuned Hubble
parameter.
In this section, we present a toy model in which the universe undergoes a phase of
ekpyrotic contraction before transitioning into matter-like contraction. The stiff-to-
soft transition is possible because both phases are driven by the same scalar field, φ,
whose potential energy density, V (φ), pictured schematically in Fig. 2.1, is specially
constructed to obtain this behavior. As φ moves from the far right of Fig. 2.1 to the
left, the universe contracts with an ekpyrotic equation of state ek > 3 until it crosses
the kink in the middle of the figure. Thereafter, the field runs up the potential, driven
by Hubble anti friction, and the universe contracts with the same equation of state
as pressureless matter, md ≡ 3/2. As discussed, the purpose of the ekpyrotic phase
is to suppress the anisotropy so that the succeeding phase remains matter-like and
thereby generates a scale-invariant spectrum of adiabatic perturbations.
1 1
L = R − (∂φ)2 − V (φ), (2.6)
2 2
14
where R is the Ricci scalar,
φ − φe
V (φ) = Vmd (φ) + Θ × (Vek (φ) − Vmd (φ)) , (2.7)
∆φ
√
Vek (φ) ≡ −Vekf exp − 2ek (φ − φe ) ,
(2.8)
i
√
Vmd (φ) ≡ Vmd exp − 2md (φ − φe ) , (2.9)
Vekf > 0 is the magnitude of the potential energy density at the end of ekpyrosis, Vmd
i
>
0 is the magnitude of the potential energy density at the onset of matter domination,
φe is the field value at which ekpyrosis transitions into matter domination, and ∆φ
sets the width of the transition. We will consider the limit of a rapid transition,
namely ∆φ → 0, so that the changeover from ekpyrosis to matter domination can be
approximated by a Heaviside θ function, i.e., Θ( φ−φ
∆φ
e
) → θ(φ − φe ). In this limit,
Vek (φ)
for φ > φe
V (φ) ≈ (2.11)
Vmd (φ)
for φ < φe ,
so the scalar field generates ekpyrotic contraction to the right of the kink and matter-
like contraction to the left of the kink. The solution to the equations of motion,
2 1 1 2
H = φ̇ + V (φ) , (2.12)
2 2
0 = φ̈ + 3H φ̇ + V,φ , (2.13)
is given by
s 1/ek
Vekf
aek (t) = ae 1 + (te − t) ek (2.14)
ek − 3
√
aek (t)
φek (t) = φe + 2ek ln (2.15)
ae
15
V(ϕ)
ϕ
Figure 2.1: This shows the potential in Eq. (2.7). The field moves from right to
left across the figure, transitioning from ekpyrotic to matter-like contraction after the
kink.
s 1/md
i
Vmd
amd (t) = ae 1 + (te − t) md (2.16)
3 − md
√
amd (t)
φmd (t) = φe + 2md ln , (2.17)
ae
i
for t > te , where te is the time at which φ = φe , ae ≡ a(te ), and Vmd = 3Vekf /(2(ek −3)).
Figure 2.2 shows excellent agreement between this analytic solution and a numerical
solution to the equations of motion. Since the rest of this section is devoted to a
derivation of this solution, the casual reader may skip to Sec. 2.5 with no loss of
continuity.
We can gain insight into these dynamics by analyzing the equations of mo-
tion in the dimensionless “Ω-variables” (or more properly their square roots)
16
ϕ(t)
t
-2.5 -2. -1.5 -1.
-0.5
-1.0
-1.5
Figure 2.2: This shows a numerical solution to Eq. (2.13) superimposed over the
i
analytic result in Eqs. (2.15) and (2.17) (where φe = te = −Vmd = −1, ek = 6,
and ∆φ = 10−4 ). The two curves are indistinguishable. For t < te , the solution is
ekpyrotic, and for t > te it is matter-like.
√
|V |
(x, y) ≡ ( √φ̇6H , − √3H ), characterizing respectively the fractional kinetic and po-
tential energy density in the φ field. In these variables, the Friedmann equation,
p
Eq. (4.10), takes the simple form y = ±(x2 − 1), where the upper sign corresponds
to the case V < 0 as in the ekpyrotic phase and the lower sign corresponds to
V > 0 as in the matter-like phase. Thus, during ekpyrosis, x > 1, and during matter
domination, x < 1. In either case, the scalar field equation, Eq. (2.13), can be
rewritten as
r
dx 2
= 3(x − 1) x − . (2.18)
d ln a 3
The function on the right side of Eq. (2.18) is plotted in Fig. 2.3. There is a fixed-
p
point, scaling solution at x = /3. At first, when = ek > 3, this solution
corresponds to the red dot in Fig. 2.3(a). If the transition is rapid (which we can
ensure by taking ∆φ small), then the function plotted in Fig. Fig. 2.3(a), which
applies during the ekpyrotic phase, changes rapidly into that shown in Fig. 2.3(b),
which applies during the matter-dominated phase. Since in the matter-dominated
17
dx
d ln a
5
4
3
2
1
x
-1 1 2 3
(a)
dx
d ln a
5
4
3
2
1
x
-1 1 2 3
(b)
Figure 2.3: This shows the right side of Eq. (2.18) during (a) ekpyrosis (for the choice
= 6)
pand (b) matter domination ( = 3/2). In both cases, the fixed-point solution
x = /3 is indicated by a red dot. In (a), the blue curve slopes upward at the fixed
point, indicating stability. In (b), the blue curve slopes downward at the fixed point,
indicating instability. 18
phase, = md = 3/2, the fixed-point solution corresponding to the red dot is now at
√
x = 1/ 2. As explained in Fig. 2.3, the fixed-point solution is an attractor during
ekpyrosis and a repeller during matter domination. Thus, to ensure that matter
domination lasts long enough to generate 60 e-foldings of scale-invariant modes, the
√
transition must leave the system very close to x = 1/ 2. We now show how to
achieve this.
Recall that te is the time at which φ crosses the kink in the potential separating
ekpyrosis from matter domination. To find the matching conditions at te for the
solutions in the two regimes, we first multiply Eq. (2.13) by φ̇ to obtain
d 1 2
φ̇ + V = −3H φ̇2 , (2.19)
dt 2
where the last term on the left side follows from the identity V̇ = V,φ φ̇. Now we
integrate over time from te − δ to te + δ and take the limit δ → 0. Assuming the right
side is finite (though possibly discontinuous) in some neighborhood of the kink, this
yields “conservation of energy” across the kink, i.e.,
−
where ∆(F ) ≡ F (t+
e ) − F (te ) for any function F (t). Note that this implies continuity
of φ and H across the kink. There is, however, a discontinuity in φ̇: the kinetic energy
of the field is reduced by the height of the kink.
Therefore, to ensure that the transition carries the ekpyrotic solution at xbef ore =
p p
ek /3 to the matter-like solution at xaf ter = md /3, the height, ∆V , of the kink
must be chosen to satisfy
s
1 ! φ̇ ∆V
√ = xaf ter = √af terf = x2bef ore − f 2
. (2.21)
2 6Hek 3(Hek )
19
f 2
) = Vekf /(3(x2bef ore − 1)) and xbef ore =
p
Since (Hek ek /3, this gives ∆V = (1 +
3
2(ek −3)
)Vekf or
3
i
Vmd = Vekf , (2.22)
2(ek − 3)
or equivalently, that |H| must grow during the matter-like phase by a factor
exp (3Nmd ). Although we have modeled the pressureless matter as a scalar field, it
is clear from Eq. (2.3) that this growth in Hubble is independent of the nature of
the pressureless matter (so long as it can support density fluctuations, e.g., a scalar
field with the same equation of state as matter). During this period, recall from
Eq. (2.5) that the fractional energy density in anisotropy will have grown by a factor
of exp (6Nmd ). Therefore, the preceding ekpyrotic phase must suppress anisotropy
by at least this much. This requires
!
6Nmd
Vekf /Veki > exp 3 , (2.24)
1 − ek
20
or equivalently that |H| must grow by a factor of at least exp (3Nmd ) during the
ekpyrotic phase. Therefore, combining Eqs. (2.23) and (2.24) with Eq. (2.22), we find
that from the beginning of the ekpyrotic phase to the end of the matter-like phase,
the potential must grow by many orders of magnitude such that
f 3 18
Vmd /Veki > exp 12 + Nmd , (2.25)
2(ek − 3) ek − 3
or equivalently, that |H| must grow by a factor exp (6Nmd ). This is independent of
the nature of the degrees of freedom driving ekpyrosis. Thus, we have shown that
resolving the anisotropy problem requires an extremely fine-tuned potential.
2.6 Discussion
In this chapter, we have emphasized some of the difficulties imposed by the anisotropy
problem on the matter bounce scenario. Collecting the model-independent observa-
tions of the previous section, these are:
21
3. If ekpyrosis is that suppression mechanism, any implementation will require an-
other phase during which H changes by another 2.6Nmd orders of magnitude.
It will also require that soft, pressureless matter somehow overtake stiff, ekpy-
rotic matter. This is impossible unless the degrees of freedom responsible for
ekpyrotic contraction are coupled somehow to the pressureless matter. (In the
toy model presented here, in which both the stiff and the soft phases result from
the same scalar field, this requires fine-tuning a potential over 5.2Nmd orders
of magnitude with a kink of just the right height in between. The only other
possibility is to arrange for direct decay of an ekpyrotic field into pressureless
matter, which introduces the additional problems discussed in Sec. 2.3.)
23
Chapter 3
Scale-invariant perturbations in
ekpyrotic cosmologies without
fine-tuning of initial conditions
Parts of this chapter were published in Ref. [66] in collaboration with Dr. Anna Ijjas
and Professor Paul Steinhardt.
3.1 Introduction
25
that generating a scale-invariant spectrum of adiabatic or entropic fluctuations is only
possible if the ekpyrotic phase begins with finely tuned initial conditions. Their argu-
ment hinged on a phase plane analysis which revealed that the background solutions
are not attracted to a fixed point. From this, they concluded that the scenario suffers
from a sensitivity to initial conditions. If true, this argument represents a serious
blow to bouncing cosmologies because it implies that ekpyrosis cannot resolve the
standard puzzles without introducing another fine-tuning problem of its own.
In this chapter, we present simple ekpyrotic models that generate a scale-invariant,
nearly Gaussian spectrum of density perturbations but do not require fine-tuning of
initial conditions. Although these models belong in the class considered by Tolley
and Wesley, we show that the background solutions are attracted to a fixed curve
along which scale-invariant fluctuations are generated. The existence of such a fixed
curve is sufficient to ensure that the observational predictions are insensitive to the
choice of initial conditions. In other words, being attracted to a fixed point is not
necessary to avoid fine-tuning.
In Section 3.2, we summarize the general argument that suggests the need for
finely tuned initial conditions. In Section 3.3, we review a simple ekpyrotic model for
which we find a fixed-curve but no fixed-point attractor. In Section 3.4, we describe
how to construct more general examples that also avoid the need for fine-tuning of
initial conditions. Finally, we discuss the implications for cosmology.
Scaling solutions are solutions to the equations of motion for which there exists a set
of field variables such that all contributions (treating the kinetic and potential energy
densities as distinct) to the total energy density scale identically with time, keeping
their fractional contributions constant. Scaling background solutions are particularly
26
important because they are exactly solvable and can yield a scale-invariant spectrum
of perturbations.
In Ref. [106], Tolley and Wesley presented an instability argument that applies to
contracting, ekpyrotic, scaling solutions derived from two-derivative, two-field actions
2
√
1 1 ab
Z X
4 µν
S= d x −g R − G (Φ)g ∂µ Φa ∂ν Φb − V (Φ) (3.1)
2 a,b=1
2
dΦa
= ξa (Φ), gµν → eκ gµν , S → eκ S. (3.2)
dκ
Here gµν is the spacetime metric with (− + ++) signature convention, R is the Ricci
scalar, ξa (Φ) is a function of two scalar fields Φ = {Φa } where a = 1, 2, Gab is the
metric on field space, and V (Φ) is the potential energy density; reduced Planck units
(8πGN = 1 where GN is Newton’s gravitational constant) are used throughout. This
symmetry guarantees the existence of a set of field variables (Φ1 , Φ2 ) 7→ (φ, σ) such
that the Lagrangian density can be rewritten as
1 1 1
L = R − (∂σ)2 − f (σ)(∂φ)2 − V0 e−cφ h(σ), (3.3)
2 2 2
where c is a real constant, and V0 < 0 (that is, the ekpyrotic potential V (φ, σ) is
negative). Along the background solution, f (0) = h(0) = 1 and σ = 0.
It proves useful to introduce the dynamical variables
p p c
!
f (σ)φ0 σ 0 a −V0 h(σ)e− 2 φ
(w, x, y, z) ≡ √ , √ ,− √ ,σ (3.4)
6H 6H 3H
where τ is conformal time running from large negative to small negative values, a
prime denoting a derivative with respect to conformal time, a is the scale factor, and
27
H ≡ a0 /a is the conformal Hubble parameter. The four variables are dimensionless
using reduced Planck units. With these variables, the Friedmann-Robertson-Walker
(FRW) equations of motion become
! r
c
3 f,z
w,N = 3(w2 + x2 − 1) w − p − xw, (3.5)
6f (z) 2 f (z)
r
2 2 1 h,z 3 f,z 2
x,N = 3(w + x − 1) x + √ + w , (3.6)
6 h(z) 2 f (z)
√
z,N = 6x. (3.7)
Here we introduced the dimensionless time variable N ≡ ln a that denotes the num-
ber of e-folds of ekpyrotic contraction and runs from large positive to small positive
values. We eliminated y using the Friedmann constraint
w2 + x2 − y 2 = 1. (3.8)
H0
≡1− = 3(w2 + x2 ). (3.9)
H2
From the Friedmann constraint, we also see that the square of each variable w, x, and
y is a fractional contribution to the total energy density: w2 is the φ-kinetic energy;
x2 is the σ-kinetic energy; and y 2 is the potential energy. The equation-of-state
parameter is the sum of kinetic energies.
As can be verified by direct substitution, the background Eqs. (3.5)-(3.7) admit a
fixed-point solution,
r !
c c2
(w, x, y, z) = √ , 0, − 1, 0 , (3.10)
6 6
28
Φ2
Φ1
δσ
δφ
29
Linearized around the fixed point in Eq. (3.10), the background Eqs. (3.5)-(3.7) reduce
to a matrix equation, where the parameters in Eq. (3.11) determine the eigenvalues
of the matrix. Tolley and Wesley’s analysis showed that any combination of the
parameters in Eq. (3.11) that results in a scale-invariant spectrum of perturbations
(entropic or adiabatic) renders the background solution dynamically unstable to per-
turbations in the sense that the matrix associated with the linearized system has at
least one negative eigenvalue. In a contracting universe, a negative eigenvalue means
a dynamically unstable direction in (w − x − z) space.
An example is given by the Lagrangian density
1 1 1
L = R − (∂Φ1 )2 + (∂Φ2 )2 − Ṽ0 e−c1 ·Φ1 − Ṽ0 e−c2 ·Φ2 , (3.12)
2 2 2
where c1 , c2 are positive-definite constants and Ṽ0 < 0. The corresponding FRW
equations of motion admit a scaling solution with Φi = Ai ln |τ | + Bi and c1 A1 =
c2 A2 that has been shown [98, 99] to generate a scale-invariant spectrum of entropic
perturbations. According to the instability argument, it should have an unstable
background which in this simple case (Gab = δab ) can be depicted as in Fig. 3.2.
For the purpose of illustration, we briefly outline how the instability emerges from
a negative eigenvalue of the linearized system. The change of variables,
c2 Φ1 + c1 Φ2
φ = p , (3.13)
c21 + c22
c1 Φ1 − c2 Φ2
σ = p + σ0 (3.14)
c21 + c22
with σ0 = 2 ln(c2 /c1 )/(c21 + c22 ), brings the Lagrangian density in Eq. (3.12) to the
form of Eq. (3.3). The coupling function to the kinetic energy of φ, f (σ), and the
30
Figure 3.2: The instability of the scaling solution corresponding to the ac-
tion in Eq. (3.12): The background cosmological solution corresponds to a trajectory
along the ridge of the potential as indicated by the red arrow and red curve. Quantum
fluctuations along the trajectory produce a blue spectrum of adiabatic perturbations
and fluctuations normal to the trajectory produce a scale-invariant spectrum of en-
tropic perturbations. After the ekpyrotic phase, the entropic perturbations convert
into adiabatic perturbations due to the bending of the trajectory curve (not shown
here). The fact that the background trajectory is a ridge means that it is unstable if
the initial conditions are sufficiently far from the ridge.
f (σ) = 1, (3.15)
2
c 2
h(σ) = 1 + σ + O(σ 3 ), (3.16)
2
1 1 1
2
= 2 + 2, (3.17)
c c c2
1
22c21 2 22c22 2
c2 c1 +c2 c1 c1 +c2
V0 = + Ṽ0 . (3.18)
c1 c2
31
Linearizing the background Eqs. (3.5)-(3.7) about the fixed point in Eq. (3.10), the
perturbations (δw, δx, δz) ≡ (w − √c , x, z) satisfy
6
δw,N δw
δx, =M · (3.19)
δx
N
δz,N δz
with M defined as
1 2
2
(c − 6) 0 0
M ≡ 1 c2 (c√2 −6)
. (3.20)
0 2
(c2 − 6) 2 6
√
0 6 0
√
The eigenvalues of M are (c2 − 6)/2 and (c2 − 6) ± 9c4 − 60c2 + 36/4. Note that the
smallest eigenvalue is negative for ekpyrosis, as is clear from substituting Eq. (3.10)
√
into the equation of state, Eq. (3.9): > 3 requires c > 6. Therefore, as the universe
contracts, N decreases, and perturbations along the eigenvector corresponding to the
negative eigenvalue grow so that the system is carried away from the fixed-point
solution in Eq. (3.10). In this case, the negative eigenvalue means that the initial
conditions for the fields must be fine-tuned to lie close to the trajectory or else the
fields will evolve far-off course as illustrated in Fig. 3.2.
ing
In this section, we describe the case where the negative eigenvalue exists but is phys-
ically irrelevant. As we will show, the occurrence of the negative eigenvalue only
means that the attractor is a fixed curve rather than a fixed point.
32
We consider the Lagrangian density first discussed by Li in Ref. [107],
1 1 1
L = R − (∂ψ)2 − e−λψ (∂χ)2 − Ṽ0 e−λψ , (3.21)
2 2 2
where λ is positive and Ṽ0 < 0. The model involves an ekpyrotic field, ψ, with a
negative potential, similar to ordinary, single-field ekpyrosis. The novel feature is
the non-canonical, exponential coupling to the massless spectator field, χ. We begin
with the simple case where V (ψ) = 0. This corresponds to the borderline ekpyrotic
equation of state = 3. Then, we generalize to V (ψ) 6= 0 ( > 3) and provide a full,
analytic treatment.
3.3.1 V (ψ) = 0
The basic idea is captured in Fig. 3.3(a), which illustrates background trajectories
0
corresponding to different initial conditions in the space (w̃, x̃, ỹ,
z̃) ≡
*0
p
ψ 0 e− λ2 ψ χ0 a −Ṽ0 λ
e− 2 ψ − λ ψ
√ , √ , − √
, e 2 (χ − χ ) .
0 (3.22)
6H 6H 3H
33
√
Figure 3.3: The Lagrangian density in Eq. (3.21) with V = 0 and λ = 6:
Green curves and red arrows map onto each other according to the transformations
in Eqs. (3.38)-(3.41). Solutions follow the blue arrows. (a) The trajectories in
√ λ √ λ
(ψ, χ) variables: For ỹ = 0, (w̃, x̃, z̃) ≡ ψ /( 6H), e− 2 ψ χ0 /) 6H), e− 2 ψ (χ − χ0 ) .
0
The green curve is the fixed-curve attractor (w̃, x̃, z̃) = (1, 0, z̃). The red arrow
points to the fixed point (w̃, x̃, z̃) = (1, 0, 0). (b) The trajectories in (φ, σ) vari-
p √ √
ables: For y = 0, (w, x, z) ≡ f (σ)φ0 /( 6H), σ 0 /( 6H), σ . The green curve,
(w, x, z) = sech( λ2 z), − tanh( λ2 z), z , is the fixed-curve attractor. The red arrow
34
For any other initial conditions, (i.e., any other point on the cylinder), the blue
arrows carry the background solution toward the green curve. The green curve is
therefore a strong attractor for generic initial conditions. Henceforth, we call such
an attractor curve a “fixed-curve attractor.” Solutions evolving along this curve are
scaling solutions since w̃, x̃, and ỹ are constant. They describe a universe dominated
by the kinetic energy of the adiabatic field, ψ; the entropic field, χ, is constant. For
√
the case (λ = 6) depicted in Fig. 3.3, these solutions can be shown to generate a
scale-invariant spectrum of perturbations in the entropic field, χ.
As for the negative eigenvalue associated with the linearized equations of motion,
in the cases discussed in this chapter, it only indicates the existence of a fixed-curve
attractor instead of a fixed-point attractor. The existence of a fixed-curve attractor
means there is no need for fine-tuning of initial conditions.
One can describe the same dynamics in the special (φ, σ) variables (see Fig.
3.3(b)). Since the Lagrangian density in Eq. (3.21) has the shift symmetry of Eq. (3.2)
assumed by the instability argument, it can be put into the form of Eq. (3.3) through
a variable transformation (ψ, χ) 7→ (φ, σ),
2 λσ
ψ = φ + ln sech + ψ0 , (3.23)
λ 2
2 λ (φ+ψ0 ) λσ
χ = e2 tanh + χ0 , (3.24)
λ 2
where ψ0 is a real constant (see Appendix A). Substituting this transformation into
Eq. (3.21) yields the Lagrangian density in Eq. (3.3) with f (σ) = h(σ) = cosh2 ( λσ
2
)
and c = λ. The two Lagrangian densities describe the same theory in different field
variables: the cylinder on the right is a twisted version of the one on the left. It is
clear from the blue arrows that the green curve in Fig. 3.3(b) is an attractor, just like
the green curve in Fig. 3.3(a). Solutions along the green-curve attractor generate a
scale-invariant spectrum of entropic perturbations.
35
3.3.2 V (ψ) 6= 0
We extend our analysis to the more general form of the Lagrangian density in
Eq. (3.21) that applies both to V (ψ) = 0 and to V (ψ) 6= 0.
Extremizing the action with respect to variations of the fields (ψ, χ) yields the
field equations
λ
ψ 00 + 2Hψ 0 − λṼ0 e−λψ a2 + e−λψ χ02 = 0, (3.25)
2
χ00 + 2Hχ0 − λψ 0 χ0 = 0. (3.26)
1 02
H2 = ψ + e−λψ χ02 + 2a2 Ṽ0 e−λψ . (3.27)
6
Using the variables defined in Eq. (3.22), these equations can be recast as the au-
tonomous, dynamical system
r
λ 3 2
w̃,N = 3(w̃2 + x̃2 − 1) w̃ − √ − λx̃ , (3.28)
6 2
r
2 2 3
x̃,N = 3x̃(w̃ + x̃ − 1) + λw̃x̃, (3.29)
2
√
r
3
z̃,N = − λw̃z̃ + 6x̃. (3.30)
2
If V 6= 0, this system admits three fixed-point solutions at (w̃, x̃, ỹ, z̃) =
(−1, 0, 0, 0) , (3.31)
(+1, 0, 0, 0) , (3.32)
r !
λ λ2
√ , 0, − 1, 0 , (3.33)
6 6
36
all of which are unstable (i.e., associated with a negative eigenvalue). The third
fixed-point solution given by Eq. (3.33) bisects two fixed-curve solutions
r !
λ λ2
(w̃, x̃, ỹ, z̃) = √ , 0, − 1, ±Z̃ (3.34)
6 6
λ2
with Z̃ ∝ e− 2
N
, that generate a scale-invariant spectrum of entropic perturbations,
as shown in Ref. [107].
If V = ỹ = 0, λ2 must be 6 in order for the fixed point in Eq. (3.33) to be a solution.
√
For λ = 6, this coincides with the fixed point in Eq. (3.32) and corresponds to the
red arrow shown in Fig. 3.3(a); Eq. (3.34) parameterizes the vertical, green, fixed-
curve attractor.
Changing variables (ψ, χ) 7→ (φ, σ) as defined in Eqs. (3.23) and (3.24), the
Lagrangian density in Eq. (3.21) takes the form of Eq. (3.3) with f (σ) = h(σ) =
cosh2 λ2 σ , c = λ, and V0 = Ṽ0 e−λψ0 . Repeating the same analysis in the new vari-
√
2 2 λ
w,N = 3(w + x − 1) w − √ sech (λz) − 6c tanh(λz)xw, (3.35)
6
!
√
r
2
x,N = 3(w2 + x2 − 1) x + λ tanh(λz) + 6c tanh(λz)w2 , (3.36)
3
√
z,N = 6x. (3.37)
37
The variable transformations as defined in Eqs. (3.23) and (3.24) imply the fol-
lowing relationship between the variables (w, x, y, z) 7→ (w̃, x̃, ỹ, z̃):
λ λ
w̃ = sech z w − tanh z x, (3.38)
2 2
λ λ
x̃ = tanh z w + sech z x, (3.39)
2 2
ỹ = y, (3.40)
2 λ
z̃ = sinh z . (3.41)
λ 2
(−1, 0, 0, 0) , (3.42)
(+1, 0, 0, 0) , (3.43)
r !
λ λ2
√ , 0, − 1, 0 , (3.44)
6 6
r 2 !
λ λ λ λ λ
√ sech z , − √ tanh z , − 1, ±Z (3.45)
6 2 6 2 6
with Z = (2/λ) sinh−1 λZ̃/2 . These fixed curves lie on the surface of the cylinder
2 √
w2 + x2 = λ6 . For V = 0 and λ = 6, Eq. (3.45) corresponds to the twisted green
curve that is confined to the surface of the unit cylinder in Fig. 3.3(b).
Direct substitution verifies that the curves in Eq. (3.34) and Eq. (3.45) are solu-
tions to the background equations given in Eqs. (3.28)-(3.30) and Eqs. (3.35)-(3.37)
√
for both V = 0 with λ = 6 and for V 6= 0.
38
To show the existence of a negative eigenvalue, we linearize the equations of motion
about the fixed points, in Eq. (3.33) and (3.44), respectively, for the two sets of
variables. Linearizing Eqs. (3.28)-(3.30) about Eq. (3.33) yields a matrix equation
√
like that given in Eq. (3.19) with (δ w̃, δx̃, δz̃) ≡ w̃ − λ/ 6, x̃, z̃ and
λ2
2
−3 0 0
M̃ ≡
0 λ2 − 3 0 .
(3.46)
√ 2
0 6 − λ2
Similarly, linearizing Eqs. (3.36)-(3.37) about Eq. (3.44) yields Eq. (3.19) with
√
(δw, δx, δz) ≡ w − λ/ 6, x, z and
λ2
2
−3 0 0
M ≡ λ2 λ2 (λ√2 −3)
. (3.47)
0 2
−3 2 6
√
0 6 0
Both M̃ and M have eigenvalues {−λ2 /2, λ2 /2 − 3, λ2 − 3}, the first of which is neg-
ative. In the first set of variables, the eigenvector corresponding to the eigenvalue
−λ2 /2 is parallel to the unit vector in the z̃-direction, which is tangent to the fixed-
curve solution in Eq. (3.34). In the second set of variables, the eigenvector associated
with eigenvalue −λ2 /2 is parallel to a linear combination of unit vectors x̂, ẑ, namely
√
ẑ − λ2 /(2 6)x̂, which is tangent to the fixed-curve solution in Eq. (3.45).
√
For the case, V = 0 and λ = 6, these eigenvectors are tangent to the green fixed
curves in Fig. 3.3 at the red arrows. The existence of a negative eigenvalue in this
model is harmless, since it only means that the system is attracted to a fixed-curve
solution (instead of a fixed-point solution) that generates a scale-invariant spectrum
of entropic perturbations.
39
3.3.3 Further generalizations
Although the remainder of this chapter will consider actions with the shift symmetry
in Eq. (3.2), our results can be generalized to cases without shift symmetry. For
example, the field contribution to the shift-symmetric Lagrangian density in Eq. (3.21)
is a special case of the more general Lagrangian density
1 1 1
L = R − (∂ψ)2 − e−λψ (∂χ)2 − V0 e−µψ [1 + r(χ)] + q(χ) (3.48)
2 2 2
with µ = λ and r(χ) = q(χ) = 0. The addition of q(χ) and r(χ) breaks the shift
symmetry since
6= e−κ V. (3.49)
If µ = λ, the ekpyrotic Lagrangian density in Eq. (3.48) admits a scaling solution that
is a fixed-curve attractor with χ0 = 0 and that generates a scale-invariant spectrum
of entropic perturbations. This is due to the fact that as χ0 → 0, r(χ) and q(χ)
approach constants r(χ0 ) and q(χ0 ). The first, r(χ0 ), can be reabsorbed into V0 , and
the second, q(χ0 ), is negligible along the fixed-curved attractor.
In this section, we derive the most general ekpyrotic, two-field Lagrangian density
with shift symmetry that admits scaling solutions which are either fixed-point or fixed-
curve attractors and generate a scale-invariant spectrum of entropy perturbations.
40
First, we consider the Lagrangian density in Eq. (3.3) with arbitrary parameters
and couplings,
{V0 < 0, c ∈ R, h(σ) > 0, f (σ) > 0}. (3.50)
Property P1 says that the coupling f (σ) must grow without bound. Property P2
constrains the form of the potential energy density to be exponential at late times.
Property P3 defines the scaling solution. It implies w = 0 so that the φ field is fixed;
furthermore, since the background equations in Eqs. (3.5)-(3.7) depend explicitly on
σ, which will vary, it also implies that the scaling solution is a fixed curve (rather
than a fixed point) in (w −x−z) space. Property P4 is necessary for ekpyrosis ( > 3)
as follows from substituting this solution into the equation of state, Eq. (3.9).
The example from the last section has these four properties. At late times, the
fixed-curve attractor in Eq. (3.45) goes to z ≡ σ → ±∞. In this limit, f (σ) = h(σ) =
cosh2 (λσ/2) is dominated by the single exponential e|λσ| /4. For example, if λ > 0
and σ → −∞, choosing µ = λ in the solution in property P3 reproduces Eq. (3.45)
at late times. Similar arguments apply for the different combinations of sgn(λ) and
sgn(σ).
Assuming properties P1 through P4, the only remaining degrees of freedom are
the parameters, c, V0 , and the late-time behavior of f (σ), modulo property P1. We
41
show now that, given these four properties, it is possible to obtain a scale-invariant
spectrum for the entropic modes but not for the adiabatic modes.
We perturb Einstein’s equations about the fixed-curve solution specified by Prop-
erty P3, working in the longitudinal gauge [108, 23] where the metric takes the form
θσ
α β 0
u00σ 2
+ k − uσ = us + u, (3.52)
(−τ )2 (−τ ) 2 (−τ ) s
θs
00 2 γ δ 0
us + k − us = u σ + u , (3.53)
(−τ )2 (−τ )2 (−τ ) σ
42
where the background-dependent quantities can be derived, for example, from the
expressions in Ref. [111]:
2
σ 2 (µ2 − 4) (µ2 − 6) c2
θ = − 2 − 2 2 , (3.54)
(µ − 2)2 µ (µ − 2)2 f
2(µ2 − 4) µ2 − 6 c2 µ(µ2 − 6) f,σ
θs = − 2 + 3 − 2
(µ − 2)2 µ2 (µ2 − 2) f (µ − 2)2 f
2
µ2 2µ2 f,σσ
f,σ
− + , (3.55)
(µ2 − 2)2 f (µ2 − 2)2 f
µ2 − 6 c µ2 − 6 c f,σ
α = 4 √ + 2 √ , (3.56)
µ(µ2 − 2)2 f (µ2 − 2)2 f f
µ2 − 6 c
β = 2 2 2
√ , (3.57)
µ(µ − 2) f
µ2 − 6 c
γ = −4 2
√ , (3.58)
µ (µ2 − 2) f
δ = −β. (3.59)
f0 2
0µ − 2
f,σ = 0 = −f (−τ ). (3.60)
σ 2µ
2
µ 2 − 4 c2 2 f0 1 f0 1 f 00
s
θ = −2 2 − (−τ ) − (−τ ) + (−τ )2 , (3.61)
(µ − 2)2 f µ2 − 2 f 4 f 2 f
µ2 − 6 c µ2 − 6 c f 0
α = 4 √ + √ (−τ ). (3.62)
µ(µ2 − 2) f µ(µ2 − 2) f f
Eqs. (3.52) and (3.53) are a coupled, linear system of differential equations which
must be solved as τ → 0− to find the spectra. Depending on the growth rate of the
coupling function f (σ), different terms in Eqs. (3.54), (3.61), (3.62), (3.57), (3.58),
and (3.59) come to dominate in this regime. For example, the first term in θσ always
43
dominates over the second since f → ∞ as τ → 0− . By contrast, the relative sizes of
the different terms in θs depend on the magnitude of f 0 /f .
For clarity, we define the symbol “&” to mean
d ln |A(τ )| d ln |B(τ )|
|A(τ )| & |B(τ )| if < as τ → 0− . (3.63)
d ln |τ | d ln |τ |
d ln |A(τ )| d ln |B(τ )|
A(τ ) ∼ B(τ ) if = as τ → 0. (3.64)
d ln |τ | d ln |τ |
γ δ 0 θs
uσ , uσ us (3.65)
(−τ )2 (−τ ) (−τ 2 )
00
0
2 !
1 f 1 f
u00s − − us = 0. (3.66)
2 f 2 f
τ
dτ̄
p Z
us (τ ) = f (τ ) c1 (k) + c2 (k) , (3.67)
−1/k f (τ̄ )
where c1 (k) and c2 (k) are constants of integration. Choosing c1 (k) and c2 (k) so
√
that us and u0s match the Bunch-Davies solution, (1/ 2k)e−ikτ , at horizon crossing
44
(−τ = 1/k), yields
where
ei (−f 0 (−1/k) + (2 − 2i)kf (−1/k))
Σ(k) ≡ √ . (3.71)
2 2k 3/2 f (−1/k)3/2
Substituting this result into the right side of the adiabatic equation, Eq. (3.52), we
find that uσ satisfies
µ2 − 4 1 µ2 − 6 cΣ(k)
u00σ + 2 u σ = 4 (3.72)
(µ2 − 2)2 (−τ )2 µ(µ2 − 2)2 (−τ )2
with solution
µ2 −4 2 µ2 − 6
uσ = c3 (k)(−τ ) µ2 −2 + c4 (k)(τ ) µ2 −2 + 2 cΣ(k) (3.73)
µ(µ2 − 4)
where c3 (k) and c4 (k) are constants. The first two terms vanish as τ → 0 by prop-
erty P4. This shows that our assumption in Eq. (3.65) is justified.
From the solutions for the mode functions in Eqs. (3.67) and (3.73), it is clear
that both the adiabatic and the entropic spectra are proportional to k 3 |Σ(k)|2 . Scale
invariance is obtained if and only if
|Σ(k)|2 = ξk −3 , (3.74)
45
for some constant ξ that is independent of k. Eq. (3.74) is a first order differential
equation for the coupling function f , namely
2
f0 4 f0
8
− = 8ξf − ; (3.75)
f (−τ ) f (−τ )2
(µ2 − 4) 1
00
u +2 2 u = 0, (3.77)
(µ − 2)2 (−τ )2
as is clear from Eqs. (3.52), (3.53) and (3.54)-(3.62). In this case, both the adiabatic
and the entropic spectra are given by
2
µ − 6
ns = 4 − 2 , (3.78)
µ − 2
which is blue by property P4. Hence, we conclude scale invariance is impossible for
slow-growing f .
If |f,σ /f | ∼ 1, f (σ) = e−λσ for some λ ∈ R such that sgn(λ) = sgn(µ). The coupling
√
functions α, β, γ, δ are all proportional to 1/ f so the right sides of Eqs. (3.52) and
46
(3.53) can be neglected. The mode functions effectively decouple and the adiabatic
spectral index is again given by Eq. (3.78); the entropic spectral index is
λµ − 2
ns = 4 − 2 2 + 1 , (3.79)
µ −2
√
which is scale invariant when λ = µ. For a given µ > 6, any ns < (3µ2 − 2)/(µ2 − 2)
can be achieved by choosing λ = (ns − 1)/µ − µ(ns − 3)/2.
Note that since “just-so” growth implies f (σ) = e−λσ (and property P2 requires
h(σ) = e−µσ ), the Lagrangian density is given by
1 1 1
L = R − (∂σ)2 − e−λσ (∂φ)2 − V0 e−cφ e−µσ , (3.80)
2 2 2
which is equivalent to the Lagrangian density in Eq. (3.48) with the identifications
σ ←→ ψ,
φ ←→ χ. (3.81)
3.5 Discussion
47
those that generate the observed fluctuation spectrum– suffer from a different kind
of fine-tuning problem that manifests as a sensitivity to initial conditions.
In this chapter, we presented explicit ekpyrotic models, with and without shift
symmetry, that generate a scale-invariant spectrum of entropic fluctuations and do
not require finely tuned initial conditions. Rather, we showed that although the
negative eigenvalue identified by Tolley and Wesley can indicate a true instability in
some models, in others it simply implies the existence of a fixed-curve solution rather
than a fixed point. For the actions discussed here, the fixed curve is a stable attractor;
fine-tuning of initial conditions is thereby avoided.
It is worthwhile to pause to examine these beneficial features in a broader context.
First, as with all ekpyrotic models, those discussed here avoid the mutiverse problem
of inflation and therefore make falsifiable predictions that are the same in all parts of
the universe without resorting to additional inputs like anthropic arguments or prob-
ability measures. From an aesthetic perspective, it is remarkable that the avoidance
of finely tuned initial conditions alleviates, rather than exacerbates, concomitant re-
quirements on the equation of state, , during the ekpyrotic phase. These models no
longer require ultra-stiff pressures (i.e., 3) to obtain a scale-invariant spectrum of
density perturbations; the observed spectral index can be achieved– and fine-tuning
problems averted– if is only marginally greater than 3. Moreover, as in all ekpyrotic
models, these naturally generate an undetectable spectrum of primordial gravitational
waves (the ratio of the tensor-perturbation amplitude to the scalar-perturbation am-
plitude, r ≈ 0), consistent with current bounds [47, 112]. They also generate zero
non-Gaussianity during the ekpyrotic contraction phase. A small amount of local non-
Gaussianity may be generated during the bounce, but at a level well within current
observational bounds on fNL [113, 114].
Interestingly, as observational constraints impose tighter bounds on r, inflation
and ekpyrosis are moving in diametrically opposed directions in terms of parameters,
48
degrees of freedom, and initial conditions. That is, inflationary models are becoming
more complicated [28] while ekpyrotic models are becoming simpler.
mation
In this appendix, we will derive the coordinate transformation in Eqs. (3.23) and
(3.24). We will work in three steps. First, we will show quite generally that the
symmetry transformation in Eq. (3.2) guarantees that the existence of a Killing field,
namely ξ a , on field space. Next, we will specialize to the model in Eq. (3.21), showing
that it possesses the symmetry and then constructing the corresponding Killing field.
Finally, we will choose coordinates (φ, σ) on field space such that the Killing direction
is φ and the orthogonal direction is σ, thereby producing Eqs. (3.23) and (3.24).
In order to conform to the (standard) notation that contravariant vector indices ap-
pear as superscripts and that covariant covector indices appear as subscripts, we will
depart from the (nonstandard) convention taken in Eq. (3.1), in which Φa with lower
index denotes field-space coordinates and in which Gab with upper indices denotes
field-space metric components. Rather, henceforth, Φa with upper indices will de-
note field-space coordinates, and Gab with lower indices will denote field-space metric
components. This should present no ambiguity.
The instability argument in Ref. [106] applies to actions of the form in Eq. (3.1),
which in this new notation reads
√
1 1
Z
4 µν a b
S= d x −g R − Gab g ∂µ Φ ∂ν Φ − V (Φ) ≡ SR + SK + SV . (3.82)
2 2
49
Perturbing Eq. (3.2) about an arbitrary point κ = κ0 , we find that under
∂Φa (x; κ0 + )
δ Φ a ≡ =0
= ξa (3.86)
∂
∂gµν (x; κ0 + )
δ gµν ≡ =0
= gµν . (3.87)
∂
Then the demand that the action transforms as in Eq. (3.85) implies
dS(κ0 + )
δ S ≡ =0
= S(κ0 ). (3.88)
d
or equivalently that
0 = δ S − S (3.89)
50
√
The first term in parentheses vanishes trivially since in 3+1 dimensions, δ ( −gg µν ) =
√
−gg µν . Expanding the other two terms gives
√
Z
4 µν a b
0 = δ d x −g g Gab ∂µ Φ ∂ν Φ + V (Φ) − SK − SV (3.91)
√ √
Z
= d x δ ( −gg µν )Gab ∂µ Φa ∂ν Φb + −gg µν δ Gab ∂µ Φa ∂ν Φb
4
√ √ √
µν a b µν a b
+ −gg Gab ∂µ δ Φ ∂ν Φ + −gg Gab ∂µ Φ ∂ν δ Φ + δ ( −gV )
−SK − SV (3.92)
√
Z
d4 x −g g µν δ Gab ∂µ Φa ∂ν Φb + 2Gab ∂µ δ Φa ∂ν Φb + δ V + V
= (3.93)
Since Gab and ξ a depend only on Φa , we can use the chain rule to write
∂Gab
δ Gab = δ Φc (3.94)
∂Φc
and
∂δ Φa
∂µ δ Φa = ∂µ (Φc ) . (3.95)
∂Φc
Using these and relabeling indices, we can rewrite the term in the square brackets in
Eq. (3.93) as
∂δ Φc ∂δ Φc
∂Gab c µν a b dV
δ Φ + Gac + Gcb g ∂µ Φ ∂ν Φ + δ φ a + V . (3.96)
∂Φc ∂Φb ∂Φa dΦ a
µν a b dV a
(Lξ Gab )g ∂µ Φ ∂ν Φ + ξ +V . (3.97)
dΦa
Since this transformation is to hold for all fields and all metrics, both terms in paren-
theses must vanish separately, i.e., (Lξ G)ab = 0 and δ V = −V . Thus, we have shown
51
that if the symmetry transformation in Eq. (3.2) holds, then ξ a is a Killing field on
field space (and moreover that V → e−κ V ).
The model in Eq. (3.21) possesses the symmetry in Eq. (3.2). Note that under
κ
ψ → ψ+ , (3.98)
λ
χ → χ0 + eκ/2 (χ − χ0 ), (3.99)
X = χ − χ0 (3.101)
2 λψ
Y = e2 . (3.102)
λ
52
It is also useful to rewrite Killing’s equation for a vector field K a in terms of its
covariant components by substituting the identity
where Γνρµ ≡ Gνσ Γσρµ and Γσρµ are the components of the Levi-Civita connection for
the field-space metric. This yields
For the field-space metric in Eq. (3.103), the only nontrivial connection components
−1
are ΓYXX = −ΓX Y
XY = −ΓY Y = Y . Thus, Eq. (3.109) yields from the (µ, ν) =
(X, X), (X, Y ), (Y, Y ) components, respectively, the following differential equations,
for KX and KY
KX,X = Y −1 KY , (3.110)
KY,Y = −Y −1 KY . (3.112)
53
Equation (3.112) integrates immediately to
where X (X) is an arbitrary function of X. Putting this into Eq. (3.110) yields
where Y(Y ) is an arbitrary function of Y . Substituting Eqs. (3.113) and (3.114) into
Eq. (3.111) and rearranging yields the separable equation
from which it follows that both sides are equal to the same constant, which we set to
A. The remaining integrals are trivial and give
A 2
X (X) = X + BX + C, (3.116)
2
A D
Y(Y ) = − + 2 , (3.117)
2 Y
where B, C, and D are integration constants. Putting these into Eqs. (3.113) and
(3.114) and raising the indices yields
2
X λ A 2 2
K = (X − Y ) + BX + (C + D) , (3.118)
2 2
2
λ
KY = (AXY + BY ) . (3.119)
2
Thus, we have derived the most general Killing field of the field-space metric. It is a
linear combination of the independent Killing fields, K1 ≡ (X 2 − Y 2 )∂X + 2XY ∂Y ,
K2 ≡ X∂X + Y ∂Y , and K3 ≡ ∂X .
54
Selecting the coordinates
We will now choose coordinates (φ, σ) such that φ increases along K2a , namely ∂φa =
ωK2a , where ω is a constant. This implies
∂X
= ωX =⇒ X = α(σ)eωφ , (3.120)
∂φ
∂Y
= ωY =⇒ Y = β(σ)eωφ , (3.121)
∂φ
where α and β are functions of σ that will be chosen such that the Lagrangian density
takes the form in Eq. (3.3). Comparing to the field-space metric,
2
dX 2 + dY 2
2 1 2 2
= (α,σ + β,σ )dσ 2 + ω 2 (α2 + β 2 )dφ2
λ Y2 (ωβ)2
−ω(αα,σ + ββ,σ )(dφdσ + dσdφ) , (3.122)
this requires
2 2
α,σ + β,σ = ωβ 2 (3.123)
1 d
α2 + β 2 = 0
αα,σ + ββ,σ = (3.124)
2 dσ
for some function F (σ), we can substitute into Eq. (3.123). This yields F,σ2 = ω 2 ,
which has a solution
F (σ) = ωσ + c2 . (3.127)
55
Substituting into Eqs. (3.120) and (3.121) gives
λ
Using Eqs. (3.101) and (3.102), we find (setting c2 = 0, c1 = λ2 e 2 ψ0 , ω = λ/2),
2 λσ
ψ = φ + ln sech + ψ0 , (3.130)
λ 2
2 λ (φ+ψ0 ) λσ
χ = e2 tanh + χ0 , (3.131)
λ 2
In this appendix, we derive the four properties listed at the beginning of Sec. 3.4
using a series of lemmas:
σ̄ ≡ σ − z0 , (3.132)
p 1
φ̄ ≡ f (z0 ) φ − ln h(z0 ) , (3.133)
c
c
c̄ ≡ p , (3.134)
f (z0 )
f (σ̄ + z0 )
F (σ̄) ≡ , (3.135)
f (z0 )
h(σ̄ + z0 )
H(σ̄) ≡ , (3.136)
h(z0 )
56
the Lagrangian density can be recast as
1 1 1
L = R − (∂ σ̄)2 − F (σ̄)(∂ φ̄)2 − V0 H(σ̄)e−c̄φ̄ . (3.137)
2 2 2
Eq. (3.137) has the shift-symmetric form of Eq. (3.3) with the solution (x, z) = (0, z0 )
at (x, z̄) = (0, 0) (with z̄ ≡ σ̄) for which Tolley and Wesley proved that any solutions
of interest are unstable. For fixed-point solutions like this (as opposed to fixed-curve
solutions) instability implies the need for finely tuned initial conditions. Thus, we
conclude |σ| → ∞; we are forced to fixed-curve solutions.
1 1 1
L = R − (∂σ)2 − f0 (∂φ)2 − V0 e−cφ e−dσ (3.138)
2 2 2
√ √
linearized around (w0 , x0 ) = (d/ 6, c/ 6f0 ) yield the same equation for both the
adiabatic and the entropic mode functions
2
c + (d2 − 6) f0
ns = 4 − 2
. (3.140)
c + (d2 − 2) f0
57
Avoiding fine tuning, (i.e., negative eigenvalues in the linearized equations of motion)
requires c2 /f0 + d2 > 6. In this regime, the spectral indices are blue for both the
adiabatic and entropic spectra.
For scaling solutions, the background equation for w, Eq. (3.5), can be recast as
df p p
=A f f −B , (3.141)
dz
√ √
with A ≡ 6 (w2 + x2 − 1) /x and B ≡ c/( 6w). There are various cases to consider,
depending on the sign of A and B. For example, if A > 0 and B > 0, then x > 0
so that z 0 < 0. Then sgn(f 0 ) = −sgn(df /dz), so that f decreases whenever the
√
right side of Eq. (3.141) is positive ( f > B) and increases whenever it is negative
√
( f < B). It follows that at late times f → B 2 = const. The other cases can be
analyzed similarly, with the result that f → (B 2 or 0). With lemma 2, we only have
to consider the case w = 0.
Lemma 5 The scaling solution must correspond to the limit given in property P3.
With h(σ) = e−µσ , the only scaling solutions of Eqs. (3.5) and (3.6) are (w, x) =
√
(0, ±1), (0, µ/ 6) if f,σ /f 6= const. For the first two solutions, = 3. Since
ekpyrosis corresponds to > 3, we only consider the third solution (property P3).
58
Note, if f,σ /f = const ≡ −λ, there is another scaling solution at
p √ !
µ(λ + µ) − 6 6
(w, x) = ± , . (3.142)
|λ + µ| λ+µ
At late times, this solution is equivalent to one in Ref. [100] that was shown never to
be an attractor.
2
µ µf,z
(δw),N = 3 −1 − δx, (3.143)
6 2f
2
µ
(δx),N =3 − 1 δx, (3.144)
6
√
where δw ≡ w and δx ≡ x − µ/ 6. The eigenvalues of this system are
{(µ2 − 6)/2, (µ2 − 6 − µf,z /f ) /2}. Since µf,z /f < 0, this reduced 2 × 2 sys-
√
tem is stable i.e., has positive eigenvalues, if |µ| > 6 (property P4).
59
Chapter 4
Warm ekpyrosis
Most of this chapter was published in Ref. [67] in collaboration with a fellow graduate
student Gustavo Turiaci, who has granted permission for its reproduction here.
4.1 Introduction
61
amplitude, r ≈ 0) and zero non-Gaussianity during the ekpyrotic contraction phase;
only a small amount of local non-Gaussianity (fNL = O(1)) is generated by the con-
version process [112, 113, 114]. These models also impose less stringent constraints
on the equation of state parameter of the universe and hence require less fine-tuning
of parameters than actions with canonical kinetic terms.
In all of these models, during the slow contracting phase, the ekpyrotic fields are
assumed to have no direct interaction with any other fields. They simply traverse
their potential energy surface in a supercooled universe, and only after the ekpyrotic
phase is the universe assumed to reheat, either through some coupling to Standard
Model particles or through stringy, higher dimensional effects [58]. In this chapter, we
consider a single, ekpyrotic field coupled to a perfect fluid of ultra-relativistic matter
(e.g., radiation) in thermal equilibrium. As the field falls down its steep, negative
potential, it decays continuously into lighter fields which are thermally excited, thus
generating a dissipative friction term in its equation of motion. We assume that
the dissipation occurs under adiabatic conditions in which the microscopic dynamical
processes operate much faster than the macroscopic evolution of the field and of the
universe [122]. In this way, we approximate the dissipation to be local in time, a
condition which may be difficult to achieve in a microphysical model. We return to
this point in Sec. 4.4. As in the non-canonical, two-field models discussed above,
this friction term allows a scale-invariant spectrum to be produced. In contrast to
the non-canonical models, the scale-invariant spectrum is immediately adiabatic; no
conversion is necessary.
To describe the interaction between the fluid and scalar field, we strive for gen-
erality, leaving the details of specific microphysical model-building for future work.
Therefore, we work at the level of the equations of motion, adding generic dissipa-
tive and noise terms. As we will show, if the dissipation is too strong, the radiation
fluid dominates the energy density of the universe; if it is too weak, the scalar field
62
dominates. Hence, our results require that the dissipation coefficient evolve in fixed
proportion to the Hubble parameter. This is the main source of fine-tuning (although
this tuning can be relaxed somewhat by changing the details of the interaction).
The idea of particle production during a cosmological smoothing and flattening
phase has been investigated previously in models referred to as warm inflation, where
thermal fluctuations sourced by radiation-induced noise were shown to dominate over
vacuum fluctuations [123, 124, 125, 126, 122, 127]. Similar effects appear in mod-
els such as trapped inflation [128]. In contracting universes, however, the thermal
fluctuations are suppressed on the largest scales, and the density perturbations are
dominated by vacuum fluctuations. The reason is that contracting universes grow
hotter with time, so that longer modes cross the horizon at lower temperature with
correspondingly smaller thermal fluctuations.
This chapter is organized as follows. In Sec. 4.2, we solve and analyze the back-
ground evolution, showing the appearance of a new family of attractors introduced
by the interaction between the ekpyrotic field and the radiation fluid. In Sec. 4.3, we
compute the power spectrum for the comoving curvature perturbation by studying
scalar perturbations to linear order. This results in a Langevin-like equation that we
solve using Green’s function techniques. We find that the thermal contribution to the
power spectrum is subleading to the vacuum contribution over the observable modes,
and we show how to fix the parameters of the model to obtain scale invariance. In
Sec. 4.4, we discuss implications of our results and directions for future work.
4.2 Background
In this section, we derive an explicit solution for the background cosmology. The
main results of this section are Eqs. (4.17) and Fig. 4.1.
63
We employ reduced Planck units in which 8πGN = kB = ~ = cL = 1 where
GN is Newton’s gravitational constant, kB is Boltzmann’s constant, ~ is the re-
duced Planck’s constant, and cL is the speed of light. We use the metric signature
(−, +, +, +). Commas denote ordinary derivatives, and semicolons denote covariant
derivatives.
We consider a contracting universe populated with a radiation fluid and a mini-
mally coupled scalar field obeying Einstein’s equations,
(r) (φ)
Here, Gab is the Einstein tensor, and Tab = Tab + Tab is the total energy-momentum
tensor which has been decomposed into a term describing the radiation fluid, de-
noted by the superscript (r), and a term describing the scalar field, denoted by the
superscript (φ). The radiation fluid is characterized by a four-velocity, ua , an energy
density, ρr , and a pressure, pr , so that its energy-momentum tensor is given by
(r)
Tab = (ρr + pr )ua ub + pr gab . (4.2)
For simplicity, we take pr = ρr /3, although this is not central to our results. The
scalar field is characterized by a potential energy density, V (φ), so that its energy-
momentum tensor is given by
(φ) 1 ;c
Tab = φ,a φ,b − φ φ;c + V (φ) gab . (4.3)
2
For convenience, we take the negative, exponential form, V (φ) = V0 e−cφ , where V0 < 0
and c > 0. The interaction between the radiation fluid and the scalar field is described
64
by a flux term, Qa ≡ −Γub φ,b φ,a , satisfying
(r)b (φ)b
T a;b = −T a;b = Qa . (4.4)
where a is the scale factor, t < 0 is cosmic time, and τ < 0 is conformal time defined
by dτ ≡ dt/a. At background, Eqs. (4.4) gives two equations (from the t-component)
where overdots represent derivatives with respect to cosmic time and H ≡ ȧ/a < 0
is the Hubble parameter. The flux term, proportional to Γ, describes the decay of
the φ-field into the particles comprising the radiation fluid. It appears in two places:
on the right side of Eq. (4.7), it sources the energy density of the radiation, ensuring
that ρr is not rapidly outstripped by the energy density in the ekpyrotic field, φ; but
most important, in Eq. (4.6) it manifests as a dissipative friction term for φ. As we
will see, this friction term is critical for the production of a scale-invariant spectrum.
As it stands, Eq. (4.6) for the scalar field is incomplete. It is well understood in
the context of classical and quantum theory that whenever a process generates an
effective dissipative interaction, it also generates fluctuations that can be described
by a stochastic noise source, Ξ, with zero mean [129]. Thus, Eq. (4.6) should read
65
If the microphysical process responsible for this noise, Ξ, is in thermal equilibrium at
some temperature, T , then the Fluctuation-Dissipation Theorem relates the dissipa-
tion it induces, Γ, to its correlation function via
where angular brackets denote ensemble averaging. If this process is not in thermal
equilibrium, then its correlation, hΞΞi, can depend more generally on (x, τ ) and
(x0 , τ 0 ). As discussed in Sec. 4.1, this noise term is critical in warm inflation because
it significantly enhances the power spectrum of scalar perturbations relative to the
vacuum result. In the contracting models considered here, the opposite is true: as we
will show, the noise, Ξ, is completely irrelevant to the power spectrum of the comoving
curvature perturbation. Moreover, since it has zero mean, hΞi, it is irrelevant to the
background dynamics as well and will be omitted in the reminder of this section.
To find the background dynamics, Eqs. (4.6) and (4.7) must be solved subject to
the Friedmann constraint (from the t-t component of Eq. (4.1)),
21 1 2
H = φ̇ + V + ρr . (4.10)
3 2
To this end, it proves useful to introduce the dimensionless “Ω-variables,” (or more
properly their square roots)
p √ !
φ̇ |V | ρr
(x, y, z) ≡ √ ,−√ , −√ , (4.11)
6H 3H 3H
characterizing respectively the fractional kinetic energy density in the scalar field, the
fractional potential energy density in the scalar field, and the fractional energy density
in the radiation fluid. In terms of these variables, Eq. (4.10) can be rewritten as
√
y = x2 + z 2 − 1, where we have taken the positive root since H < 0 in a contracting
66
universe. The equation of state of the universe takes the simple form
as can be obtained by differentiating Eq. (4.10) and substituting Eqs. (4.6) and (4.7).
Therefore, ekpyrosis occurs whenever 3x2 + 2z 2 > 3. Meanwhile, Eqs. (4.6) and (4.7)
can be rewritten as
dx 2 2 c 2 Γ
= 3(x + z − 1) x − √ −x z + , (4.13)
d ln a 6 H
dz Γ x2
= (3x2 + 2z 2 − 2)z + . (4.14)
d ln a H z
Note that Γ appears only in the ratio γ ≡ Γ/H < 0. It is at this point that the fine-
tuning enters: we assume in this chapter that γ is a constant, independent of time.
To motivate this assumption, note that if |γ| grows rapidly, the universe becomes
dominated by radiation only, and if it shrinks rapidly, the universe becomes dominated
by the ekpyrotic field only. It is only when γ is roughly constant that these two
components coexist. Therefore, we assume it in what follows, and merely observe
√
in passing that if, e.g., Γ ∝ T 2 or ρφ , where ρφ ≡ 21 φ̇2 + V , then γ is constant
along the solution of interest, and our assumption is justified. With this assumption,
Eqs. (4.13) and (4.14) admit a fixed-point, scaling solution at (x0 , z0 ) with
2 1/2
24 c2 + 4 γ + 9 c2 − 4 + 16γ 2
x0 ≡
√
+3c2 + 4γ + 12 (6 6c)−1 , (4.15)
p
z0 ≡ [(γ + 3)( 9c4 + 24(γ − 3)c2 + 16(γ + 3)2
√
+4γ + 12) − 3c2 (γ − 3)]1/2 (3 2c)−1 . (4.16)
67
As a consistency check, note that in the absence of radiation, i.e., when γ = 0,
√
this solution reduces to (x, z) = (c/ 6, 0), which reproduces ordinary, single-field
√
ekpyrosis (when c > 6).
To summarize, Eqs. (4.15) and (4.16) describe a cosmological background whose
evolution is given by
a = (t/te )
1
a = (τ /τe ) −1
1 1
H ≡ ȧ/a = H ≡ a0 /a =
( − 1)τ
√t √ (4.17)
6x0 6x0
φ = φe + ln (t/te ) φ = φe + ln (τ /τe )
−1
3z 2 3z 2 (τ /τe )−2/(−1)
ρr = 2 02 ρr = 0 2 ,
t τe ( − 1)2
where 0 ≡ d/dτ and we have normalized the scale factor to unity when ekpyrosis ends
at some time te < 0. For convenience, we have included the results in conformal time
and defined τe ≡ ( − 1)−1 te and
V0 τe2 ( − 1)2
1
φe ≡ ln − . (4.18)
c 3 (x20 + z02 − 1)
These dynamics are pictured in Fig. 4.1, which shows that this solution is an attractor
for a wide range of initial conditions.
We close this section by noting that if the flux term is changed to Qa =
Γ(ub φ,b )n φ,a , for n > 1, it can be shown that the updated equations of motion
admit a similar attractor so long as Γ ∝ H 2−n . This alleviates the finely-tuned time
dependence of Γ required for the stability of the background solution.
68
�
��
�
�
�
�
� � � � �
�
Figure 4.1: This streamplot shows that the warm ekpyrotic background solution is
an attractor for a wide range of initial conditions. For illustration, we have chosen
parameter values (c, γ) = (15, −56.8). Any set of initial conditions
√ for φ, φ̇, ρ̇, anda
√ √
corresponds to a particular point in this plane (x, z) ≡ φ̇/( 6H), − ρr /( 3H) .
The background solution follows the blue arrows originating at this point. The red
and green curves are included simply to guide the eye: they are nullclines, where
dx/d ln a = 0 (red) and dz/d ln a = 0 (green). The intersections of the nullclines
are shown as red dots. These so called “fixed-point, scaling solutions” are special
because the blue streamlines vanish here. If the background solution starts at one of
these points, it stays there. The rightmost such point corresponds to (x0 , z0 ) defined in
Eqs. (4.15) and (4.16). Clearly, it is an attractor for a wide range of initial conditions.
The analysis below shows that the comoving curvature perturbation generated by this
solution acquires a scale-invariant spectrum on large scales.
69
4.3 Perturbations
In this section, we study scalar perturbations to linear order about the background
solution described in Eqs. (4.17). We show that there exists a wide range in parameter
space (i.e., choices of c and γ) for which the comoving curvature perturbation acquires
a scale-invariant power spectrum. This result is displayed in Fig. 4.2; it is the main
result of this chapter.
A full derivation of the scalar perturbation equations in spatially-flat gauge is
presented in Appendix 4.5. In this gauge, all perturbed quantities can be expressed
in terms of the scalar potentials of the four-velocities of the radiation fluid, δur , and
of the scalar field, δuφ . In particular, the comoving curvature perturbation is
H
R≡− (6x20 δuφ + 4z02 δur ). (4.19)
2
C1 0 C2
δu00φ+ 2
δuφ + k + 2 δuφ = Jr (δur , k, τ ) + ξ(k, τ ), (4.20)
τ τ
2
00 C5 0 k C6
δur + δur + + 2 δur = Jφ (δuφ , k, τ ), (4.21)
τ 3 τ
where
C3 C4 0
Jr (δur , k, τ ) ≡ 2
δur + δur , (4.22)
τ τ
C7 C8 0
Jφ (δuφ , k, τ ) ≡ 2 δuφ + δuφ , (4.23)
τ τ
ξ ≡ Ξ/φ̇2 , and the Ci are constants that depend on c and γ, whose explicit definitions
are given in Eqs. (4.71) and (4.62)-(4.69).
70
The purpose of the rest of this section is to solve Eqs. (4.20) and (4.21) for general
c and γ so that we can find R via Eq. (4.19). Before solving this system in general
(Sec. 4.3.2), we first review the solution in the cold case when γ = 0 (Sec. 4.3.1).
C1 0 C2
δu00φ + 2
δuφ + k + 2 δuφ = 0, (4.24)
τ τ
−2
with C1 = −2 and C2 = 2c2 (c2 − 3) (c2 − 2) . The selection of Bunch-Davies vac-
uum fixes
r
−1 π 1−C1
δuφ (k, τ ) = √ (−τ ) 2 Hν(1) (−kτ ), (4.25)
2 4 φ
1p
ν(X, Y ) ≡ (X − 1)2 − 4Y , (4.26)
2
we have defined νφ ≡ ν(C1 , C2 ). In the next subsection, we will use this same nor-
malization for the perturbation, δuφ , since at early times, the temperature, T , and
dissipation, Γ, are small. In the super-horizon limit, −kτ → 0, Eq. (4.25) approaches
δuφ ∝ k −νφ , so the spectral index is given by
4
ns = 4 − 2νφ = 3 + , (4.27)
c2 − 2
71
√
which is clearly blue (in particular > 3) for ekpyrosis (which requires c > 6). Thus,
scale-invariance is impossible in the single-field model.
In this subsection, we consider the “warm” case when γ 6= 0. The presence of the
radiation fluid introduces into Eqs. (4.20) and (4.21) two crucial differences: the first
is that C1 and C2 depend not only on the steepness, c, of the potential, but also on
the dissipation rate, γ; the second is that δur is no longer negligible.
To solve the system, we decompose the scalar potential for the four-velocity of the
radiation fluid as δur = δuhr + δupr , where the first term is a homogeneous solution to
Eq. (4.21) with Jφ set to 0, i.e.,
1−C5
δuhr (k, τ ) = (−kτ ) 2 [a1 (k)Jνr (−kτ ) + a2 (k)Yνr (−kτ )], (4.28)
and the second term is the particular solution given by integrating over the retarded
Green’s function, i.e.,
Z τ
−1
δupr (k, τ ) =k Gr (−kτ, −kτ̄ )Jφ (δuφ , k, τ̄ )dτ̄ . (4.29)
−∞
π 1−C5
√ √ √ √
Gr (z, y) ≡ y (z/y) 2 [Jνr z/ 3 Yνr y/ 3 − Yνr z/ 3 Jνr y/ 3 ] (4.30)
2
with νr ≡ ν(C5 , C6 ). In Appendix 4.6, we show that the integral in Eq. (4.29) can be
approximated by
δupr ≈ (C7 /C6 )θ(1 − k|τ |)δuφ , (4.31)
72
where θ(x) is the Heaviside step function. That is, δupr is negligible before horizon
crossing and is a constant multiple of δuφ after horizon crossing (see Fig. 4.4).
Armed with these solutions, we now turn to Eq. (4.20). The right side is a sum of
three terms, Jr (δupr , k, τ ) + Jr (δuhr , k, τ ) + ξ. The last term is negligible as discussed
in detail in Appendix 4.7. The second term is a rapidly decreasing function that
depends on the initial state of δuhr . We restrict attention to models where this term
begins sufficiently small that it can be neglected. Therefore, we need only consider
Jr (δupr ). Inside the horizon, it has no effect, but outside the horizon, it renormalizes
the “dissipation” and “frequency” terms on the left side of Eq. (4.20)
as is clear from substituting Eq. (4.31) into Eq. (4.22) and putting the result into
Eq. (4.20).
Therefore, the subhorizon solution is given by
r
−1 π 1−C1
δusub
φ = √ (−τ ) 2 Hν(1) (−kτ ), (4.34)
2 4 φ
1−C̃1
δusup
φ = (−τ ) κ1 Jν̃φ (−kτ ) + κ2 Yν̃φ (−kτ ) , (4.35)
2
where ν̃φ ≡ ν(C̃1 , C̃2 ), and κ1 and κ2 are approximated by the following matching
conditions at horizon crossing (−kτ = 1)
δusub
φ = δusup
φ , (4.36)
(δusup 0 sub 0
φ ) − (δuφ ) = −C4 C7 C6−1 k δusub
φ , (4.37)
73
i.e.,
π 3/2 ( − 1)
κ1 = − √ √ Hν(1) φ
(1)(Yν̃φ (1)(C̃1 − C1
8 2C6
+2(C4 C7 C6−1 + ν̃φ − νφ )) − 2Yν̃φ −1 (1))
C1 −C̃1
(1)
+2Hνφ −1 (1)Yν̃φ (1) × k 2 , (4.38)
π 3/2 ( − 1)
κ2 = √ √ Hν(1) φ
(1)(Jν̃φ (1)(C̃1 − C1
8 2C6
+2(C4 C7 C6−1 + ν̃φ − νφ )) − 2Jν̃φ −1 (1))
C1 −C̃1
(1)
+2Hνφ −1 (1)Jν̃φ (1) × k 2 , (4.39)
Substituting the solution in Eq. (4.35), together with Eqs. (4.38) and (4.39), into
Eq. (4.19), we find that the primordial power spectrum of the comoving curvature
perturbation on superhorizon scales is given by
k3 −4
1+C̃1 +2ν̃φ
where Vend ≡ |V0 |e−cφe is the magnitude of the potential energy density when ekpyrosis
ends, and the spectral index is given by
74
Figure 4.2: This shows ns as a function of c and γ for the background solution in
Eqs. (4.17).
ing the dissipation rate, |γ|, reddens the spectrum. In particular, the thick, black
curve has ns = 1. Any choice of c and γ along this curve corresponds to an exactly
scale-invariant spectrum. For such a choice, the exponent of Vend in Eq. (4.40) can
be computed and is roughly .61, so that to match the observed amplitude, Vend must
1/4
be made of order Vend ∼ 1016 GeV, which is high enough to recover the successful
predictions of hot big bang nucleosynthesis.
In this section, we have ignored isocurvature perturbations. The reason is that
even though they are not negligible at the end of ekpyrosis, reheating will render
them thus, provided the universe reheats in local thermal and chemical equilibrium
with no nonzero conserved quantities (see Ref. [130]). Therefore, our assumption that
isocurvature perturbations can be ignored amounts to this rather mild constraint on
reheating. This can be achieved, e.g., if the φ-field completely decays into thermalized
radiation at sufficiently early time.
75
4.4 Discussion
In this chapter, we have presented a scenario for ekpyrosis that continuously generates
a scale-invariant spectrum of adiabatic perturbations. The key is the continuous
decay of the ekpyrotic field; this decay introduces a friction term that allows a scale-
invariant spectrum to be achieved. More generally, as can be seen by following a
curve of constant c along the surface in Fig. 4.2, we showed that the effect of particle
production is to redden the power spectrum of the supercooled theory.
We view the elimination of the second scalar field and hence any subsequent con-
version mechanism as a major simplification, and a return to the spirit of the original
formulation of ekpyrosis, since the hydrodynamical behavior at finite temperature
is universal regardless of the details of its microscopic origin. While we have not
attempted to embed this phase into a complete cosmological history, the decay into
radiation presents the tantalizing possibility of evading the need for additional reheat-
ing (provided the radiation so generated does not present complications for realizing
a bounce).
There are two key assumptions that merit attention. The first is that the dissipa-
tion rate must scale with the Hubble parameter. This scaling represents the greatest
source of fine-tuning (although see the last paragraph of Sec. 4.2 for a possible al-
ternative). The second is that the initial fluctuations of the fluid, δuhr , are small
enough to be neglected. If these conditions are met, it is always possible to choose
the parameters c and γ such that a scale-invariant spectrum is achieved.
There are many directions for future work. One possibility is to consider gener-
alizations of the radiation fluid within the framework presented here, as was done in
warm inflation [131]. For example, one could analyze a fluid with a non-relativistic
equation of state or that is out of thermal equilibrium. One could also include vis-
(r)
cosity by adding corrections to its energy-momentum tensor, Tab .
76
Another possibility is to devise a microphysical theory– to identify the micro-
scopic degrees of freedom comprising the fluid that realizes the effective dynamics
described here. For example, one could try to reproduce the trapped inflation sce-
nario in a contracting universe [128]. In warm inflation, this is difficult, though not
impossible [132, 133, 134], to achieve because, as argued in Ref. [135], the dissipation
coefficient, Γ, appears as the result of a small correction to a sub-leading thermal
correction to the potential energy density of the inflaton, which must be extremely
flat to support inflation. Thus, non-negligible Γ requires large thermal corrections,
which spoil the extreme flatness of the potential. In ekpyrosis, such extreme flatness
is neither required nor permitted. Nevertheless, constructing a microphysical model
may also prove difficult for warm ekpyrosis. We assumed in this chapter that the
degrees of freedom in the thermal bath induce a dissipation coefficient that is local
in time. However, as indicated above, interactions between a scalar field and the
degrees of freedom in an ambient thermal bath generically lead to non-local terms in
the equation of motion for the scalar field. These terms can be approximated with a
local friction term only when the microphysical dynamics of these degrees of freedom
operate much faster than the evolution of the ekpyrotic field and the contraction of
the universe. Since the background solution in Eq. (4.17) involves only a logarithmic
dependence of the ekpyrotic field on time, it may be possible to find a model where
these non-local interactions are well approximated by a local term, but we leave this
for future work. In addition, the scaling Γ ∝ H may require nontrivial microphysics.
For example, if the dissipation coefficient depends only on temperature, this scaling
requires Γ ∝ T 2 , which may be difficult to achieve [136, 137, 138, 135, 139, 140].
It is important to understand to what extent dissipation affects the tensor-to-
scalar ratio. One does not expect dissipation to alter the tensor spectrum from the
predictions of ordinary, single-field ekpyrosis, in which it is strongly blue-tilted and
exponentially suppressed on large scales [112]. Since we have shown that dissipation
77
flattens the scalar spectrum, we expect the tensor-to-scalar ratio to be exponentially
suppressed on observable scales.
It is also important to understand the effects of dissipation on the non-Gaussian
signatures. A reason to be optimistic is that, as we have shown, neither the steepness,
c, of the potential nor the equation of state parameter of the universe, , needs to be
tuned particularly large. As discussed above, one microphysical realization for which
the machinery already exists to compute non-Gaussianities would be an ekpyrotic
analog of the trapped inflation scenario [128]. In such a realization, the ekpyrotic
field falls down its potential, transferring its kinetic energy into the production of
other particles via couplings of the form
1X 2
gi (φ − φi )χ2i . (4.42)
2
As φ passes each φi , the particles corresponding to the field χi become light and
are copiously produced with an energy density controlled by gi . As a consequence,
the ekpyrotic field receives an effective friction term in its equation of motion– also
controlled by the {gi }– which can be chosen to approximate the dissipative dynamics
studied here. In trapped inflation, these effects were used to support slow roll on
potentials that were otherwise too steep. The techniques developed in Ref. [128] for
the computation of non-Gaussianities would apply as well to ekpyrotic potentials.
For comparison, the models discussed in Refs. [107, 120, 114, 66] generate no non-
Gaussianity during the ekpyrotic phase.
In this appendix, we will derive the linearized perturbation equations. We will follow
the notation of Ref. [141] (see also Ref. [142]). To simplify the derivation, we will
78
find the linearized equations for the ensemble expectation values of the fields. This
implies that any stochastic contribution to these equations will vanish.
A metric with the most general scalar-type perturbation in a flat Friedman-
Robertson-Walker background is
ds2 = −a2 (1 + 2α)dτ 2 − 2a2 β,i dτ dxi + a2 [δij (1 + ϕ) + 2ψ,ij ] dxi dxj (4.43)
Ignoring anisotropic stress, the energy-momentum tensor for the fluid can be decom-
posed as
Thus, perturbations in the fluid are parameterized by δρr , δpr and δur . For simplicity,
we will assume δpr = δρr /3, though this is not central to our results. In writing the
perturbation equations, it is useful to define the shear, χ ≡ a(β + aψ̇), and the
k2
perturbed expansion of the normal-frame vector field κ ≡ 3(−ϕ̇ + Hα) + a2
χ. In
Fourier space, the perturbation equations are
k2 1
− 2
ϕ + Hκ = − δρ, (4.47)
a 2
k2 3 X
κ− χ + (ρi + pi )δui = 0, (4.48)
a2 2
i=r,φ
χ̇ + Hχ − α − ϕ = 0, (4.49)
k2
1
κ̇ + 2Hκ + 3Ḣ − 2 α = (δρ + 3δp), (4.50)
a 2
k 2
δ ρ̇r + 3H(δρr + δpr ) + 2 (ρr + pr )δur = +δqr + ρ̇r α + (ρr + pr )κ, (4.51)
a
−1 d 3 δpr jr
a (ρr + p r )δur = +α− , (4.52)
a3 (ρr + pr ) dt ρr + pr ρr + pr
2
k
δ φ̈ + 3Hδ φ̇ + + V,φφ δφ = φ̇(κ + α̇) − δqφ + (2φ̈ + 3H φ̇)α, (4.53)
a2
79
where
jr ≡ −Γφ̇δφ. (4.59)
Eqs. (4.47)-(4.52) are, respectively, the Gtt component of the field equations, the Gti
(r)b
component, the Gij − 13 δji Gkk component, the Gii − Gtt component, the T i;b = Qi
(r)b (φ)b
component, the T t;b = Qt component, and the T t;b = −Qt component.
Henceforth, we work in spatially flat gauge (G = ϕ = 0). Then Eqs. (4.47) and
(4.48) can be solved algebraically for the metric variables α and β in terms of the
matter variables δφ, δur , and δρr . Eq. (4.52) can then be solved algebraically for δρr
in terms of δur and δφ. Substituting these results into Eqs. (4.51) and (4.53) leaves
two closed equations for the variables δur and δuφ . Specializing to the background
solution in Eqs. (4.17), these are
k2
δüφ + c1 Hδ u̇φ + + c2 H δuφ = c3 H 2 δur + c4 Hδ u̇r
2
(4.60)
a2
2
k
δür + c5 Hδ u̇r + 2
+ c6 H δur = c7 H 2 δuφ + c8 Hδ u̇φ , (4.61)
3a2
80
with the constants ci defined by
√ √
6cz02 √ 6c
c1 ≡ − − 6cx0 + − γ − 3, (4.62)
x0 x0
√ √ √
c2 ≡ 6 6cx30 + 6 6cx0 z02 − 6 6cx0 − 18x40
√
2 2 2 2 2 6
−24x0 z0 + 27x0 + 2γz0 + 6z0 + cγx0 , (4.63)
√ 4 √ 2 2
2 6cz0 √ 2 6cz0
c3 ≡ − − 2 6cx0 z02 +
x0 x0
−8γx20 + 12x20 z02 + 16z04 − 2γz02 − 8z02 , (4.64)
c4 ≡ −4z02 , (4.65)
4γx20
c5 ≡ − 1, (4.66)
z02
√ √
2 6cγx30 √ 2 6cγx0
c6 ≡ − − 2 6cγx0 +
z02 z02
6γx4 4γ 2 x20 6γx20
− 2 0 − 7γx20 − − 2
z0 z02 z0
−8x20 z02 + 3x20 − 8z04 + 10z02 , (4.67)
√ √
2 6cγx30 √ 3 √ 2 6cγx0
c7 ≡ − − 3 6cx0 − 2 6cγx0 +
z02 z02
√ √ 4γ 2 x20
−3 6cx0 z02 + 3 6cx0 + 12x40 − 5γx20 −
z02
6γx2
− 2 0 + 12x20 z02 − 18x20 , (4.68)
z0
5γx20
c8 ≡ 2
+ 2x20 , . (4.69)
2z0
p
For concreteness, we have assumed Γ ∝ V (φ), independent of ρr and φ̇, i.e.,
p
Γ = −γ −V (φ)/(3(x20 + z02 − 1)). For this choice, δΓ = γcHδφ/2. Of course, for
a realistic model, Γ should depend on the temperature of the radiation fluid [136];
the field dependence is not required [137, 138, 135]. One possibility, which pre-
serves the Γ ∝ H scaling necessary for the warm ekpyrotic background solution, is
Γ ∝ T 2 /M, where M is some mass scale fixed by the background solution. More
generally, Γ can be any function of T and φ, provided the scaling Γ ∝ H is preserved.
81
p
We have performed the analysis for the two extreme cases, where Γ ∝ V (φ) and
where Γ ∝ T 2 and found that both cases give qualitatively similar results, namely, a
scale-invariant scalar spectrum is possible. There are only minor differences, e.g., in
the ci in Eqs. (4.62)-(4.69). This is important because in warm inflation, a tempera-
ture dependence in the dissipation coefficient can have large effects on the primordial
scalar spectrum as shown in Refs. [136, 139, 140].
As we explained above, in deriving these equations, we have averaged out the
stochastic fluctuations. However, at background level, we know that whatever mi-
crophysical process generates the dissipation, Γ, in the equation of motion for the
scalar field, must also be accompanied by a stochastic source, Ξ, whose correlation
satisfies the Fluctuation-Dissipation Theorem (see Appendix 4.7 for details). That
is, Eq. (4.60) must be replaced with
k2
δüφ + c1 Hδ u̇φ + 2
+ c2 H δuφ = c3 H 2 δur + c4 Hδ u̇r + ξ(k, t), (4.70)
a2
where ξ ≡ Ξ/φ̇2 with the extra factors of φ̇ in the denominator coming from the
change of variables from δφ to δuφ . In conformal time, Eqs. (4.70) and (4.61) become
Eqs. (4.20) and (4.21) , with
ci −1
−1
if i = 1, 5
Ci = ci
if i = 2, 3, 6, 7 (4.71)
(−1)2
ci
−1
if i = 4, 8
82
4.6 Appendix B: Locality of fluid response
Since, in general, fluids behave non-locally we will show in this appendix how locality
can be recovered in the late time limit. This is because the sources for the fluid are
τ −2 δuφ (k, τ ), τ −1 δu0φ (k, τ ) and not δuφ itself. Using the expression for the source, Jφ ,
in Eq. (4.23) and integrating the derivative term by parts, the particular solution for
the radiation fluid in Eq. (4.29) can be rewritten as
Z ∞
δupr (z) = dyKr (z, y)δuφ (y), (4.72)
z
C7 + C8 C8
Kr (z, y) ≡ Gr (z, y) − Gr,y (z, y). (4.73)
y2 y
Now, we will show that this kernel behaves locally in the small z (superhorizon)
approximation. It follows from the explicit expression of the Green’s functions that
Kr (z, z) → ∞ and Kr (z, y) → 0 for z 6= y as z → 0. These properties are illustrated
in Fig. 4.3. To compute the particular solution for the radiation fluid, we can therefore
make the local approximation
Z ∞ Z ∞
δupr (z) = dyKr (z, y)δuφ (y) ≈ δuφ (z) dyKr (z, y). (4.74)
z z
In the small z limit, this integral can be done exactly and gives
∞
C7
Z
1−C5 +2νr
dyKr (z, y) = + O(z 2 ). (4.75)
z C6
83
���
���
�� (���)
��� �� (������)
��� �� (���)
��� �� (���)
���� ��� ��� �
�
(a)
�� (���)
�� (���)
���
��� � ⩵ ����
��� � ⩵ ���
��� � ⩵ ���
���
Figure 4.3: (a) Comparison of the locality of the Green’s function, Gr , (blue) and
the kernel, Kr , (black) for z = 10−2 . Both are normalized such that their maximum
value is 1: as should be clear, ymax is the argument for which Gr (z, y) is maximized.
Note the logarithmic scale on the horizontal axis. (b) Comparison of the fluid kernel,
Kr at different final times, z. As modes are stretched beyond the horizon z 1, this
kernel becomes increasingly local. For both plots, we used (c, γ) = (15, −56.8).
84
���
���
���
��(δ��� /δ�ϕ )
��� ��(δ��� /δ�ϕ )
1−C5 +2νr
δupr (z) ≈ (C7 /C6 )δuφ (z) + O(z 2 ). (4.76)
spectrum
In this appendix, we will show that the thermal contribution to the power spectrum
of the comoving curvature perturbation, ∆2R , is negligible for the observable modes
in comparison to the vacuum, scale-invariant contribution.
85
To prove this, we write the particular solution for the scalar field perturbation in
terms of the stochastic noise
Z τe
−1
δuφ (k, τe ) = k dτ Gφ (−kτe , −kτ )ξ(k, τ ), (4.77)
−∞
where Gφ is the retarded Green’s function for Eq. (4.20), and again, τe is the time
at which ekpyrosis ends. The two-point function of the noise follows from the
Fluctuation-Dissipation Theorem in Eq. (4.9)
with the extra factors of φ̇ in the denominator coming from the change of variables
from δφ to δuφ . Substituting Eq. (4.78) into Eq. (4.77) and changing the integration
variable to y = −kτ , the thermal power spectrum for δuφ is given by
To compute this integral, we separate out the time dependence of the noise kernel,
1 1
i.e., NFD = N0 (−τ ) 2 + 2(−1) with
1/4
−γ( − 1)1/2
45 2 1
N0 ≡ z (−τe )− 2(−1) . (4.81)
3x20 π2 0
86
For the Green’s function, Gφ , we use the same approximation we used to compute
δuφ in Eqs. (4.34) and (4.35), namely finding the solutions for y > 1 and y < 1
and matching at horizon crossing, taking account of the effect of Jr according to the
change in dissipation and frequency in Eqs. (4.32) and (4.33).
�(�)
��-�
��-�
��-�
Then, dropping the arguments on the left side of Eq. (4.80), the power spectrum
is
1 1
k 3 hδu2φ i = N0 (−τe ) 2 + 2(−1) A(−kτe ) ×
87
where the function
Z ∞
− 12 − 2(−1)
1 1 1
A(z) ≡ z dy[Gφ (z, y)]2 y 2 + 2(−1) (4.83)
z
is plotted in Fig. 4.5. Using Eqs. (4.81), (4.82), and (4.19) in Eq. (4.40), we find
that this contribution to ∆2R is suppressed relative to the vacuum result by a factor
.14
that is weakly dependent on c and γ and is of order A(−kτe )Vend , e.g., for the choice
(c, γ) = (15, −56.8). Thus, the ratio of this thermal contribution to the vacuum
contribution is of order O(10−9 ) for the largest observable modes.
88
Chapter 5
Early Universe theories were introduced to address fine-tuning problems with stan-
dard big bang cosmology. To date, inflation remains the most popular proposal, but
open questions remain– both phenomenological and conceptual. In this thesis, we
studied bouncing alternatives in which the large-scale structure of the present-day
Universe originates from the contracting phase before the bounce. These scenarios
have the attractive features that they avoid a multiverse and make falsifiable predic-
tions for cosmological observables.
Throughout the duration of my PhD (2012-2017), early Universe cosmology has
undergone a paradigm shift. In 2013, the BICEP2 collaboration reported a detection
of B-modes in the polarization of the CMB at large angular scales (around ` ≈ 80)– a
detection they concluded was “well-fit by a lensed-ΛCDM + tensor theoretical model
with tensor-to-scalar ratio r = 0.20+0.07
−0.05 ” [6]. At the time, this detection was hailed as
“smoking gun” evidence for cosmic inflation. After all, when combined with previous
bounds from, e.g., PLANCK2013 on the spectral index of the scalar perturbation
spectrum, ns = 0.9603 ± 0.0073 [143], the simplest, single-field inflationary models
predict a primordial gravitational wave spectrum with a tensor-to-scalar ratio con-
89
sistent with the claimed detection. Bets were settled and champagne was poured in
celebration of this putative verification of inflation.
As time passed, more detailed studies revealed that the B-mode polarization de-
tected by BICEP2 was caused not by primordial gravitational waves but by a local
contaminant that obscures our view of the CMB– the polarized emission of dust
within our galaxy. Taking proper account of the contribution of dust, a joint anal-
ysis of data from the BICEP2/Keck Array and the Planck collaboration revised the
bounds on the tensor-to-scalar ratio, r0.05 < 0.12 at 95% confidence [144]. With that,
the simple models of inflation– those that precipitated the uncorking of champagne
bottles– were suddenly disfavored by the data. One would think that if a detection of
primordial gravitational waves proves inflation, then a non-detection should disprove
it. But this is not so for a theory as versatile as inflation because it can accommodate
nearly any cosmological observables (see Ref. [145] for a recent inflationary model
consistent with the revised bounds). Some view unfalsifiability as a feature. It cer-
tainly affords theorists a justification for permanent employment. But it also means
that no observation can rule it out, a big problem for any scientific theory.
Along with these recent phenomenological obstacles, inflation comes with the more
serious philosophical hurdle of the multiverse– a nearly unavoidable consequence of
quantum fluctuations coupled with accelerated, exponential expansion. With the
multiverse, any and every cosmological outcome is produced an infinite number of
times. The multiverse predicts everything, and so it predicts nothing. In light of
these shortcomings, early Universe cosmologists have two options: we can either seek
to redress these shortcomings within the inflationary framework, or we can search for
alternative mechanisms for resolving the standard problems. In this thesis, we have
taken the second approach, choosing to focus our efforts on two popular bouncing
paradigms.
90
Chapter 2 was based on work published in Ref. [65]. In it, we assessed the mat-
ter bounce scenario. We showed that its instability to anisotropic stresses renders
it fundamentally unable to address the standard fine-tuning problems without in-
troducing a more extreme fine-tuning requirement of its own. In particular, since
the growth of anisotropies is exponentially sensitive to the number of scale-invariant
modes that leave the Hubble horizon during the matter-like phase, any implementa-
tion must involve a powerful mechanism for suppressing anisotropy beforehand. We
argued that orchestrating such a suppression requires that the degrees of freedom
responsible for the matter-like phase be coupled to those driving the isotropizing
phase. As we pointed out, there are two possibilities: either the isotropizing degrees
of freedom must decay directly into matter or they must drive the matter-like contrac-
tion themselves. We showed that both possibilities impose extraordinary, exponential
fine-tuning requirements on any realization. As a result of these considerations, we
arrived at two robust conclusions: 1) new insights are required for the viability of
the matter bounce scenario and 2) ekpyrosis is a seemingly indispensable ingredient
in any bouncing cosmology that hopes to explain the standard puzzles within the
framework of Einstein gravity.
Chapter 3 was based on work published in Ref. [66] in collaboration with Dr.
Anna Ijjas and Professor Paul Steinhardt. In it, we examined recent models of two-
field ekpyrosis. For these models, we found no such fine-tuning challenges. On the
contrary, we illustrated that the simplest realizations are strong attractors for a wide
range of initial conditions. The attractor behavior is produced by a non-canonical
kinetic coupling between the adiabatic and entropic fields. The coupling introduces
friction into the equation of motion for the entropic field, stabilizing it against per-
turbations, thereby avoiding the need for finely tuned initial conditions. The entropic
field acquires (nearly) scale-invariant perturbations, which convert into curvature per-
turbations via the entropic mechanism. These models generate zero non-Gaussianity
91
during the ekpyrotic phase and an undetectable spectrum of primordial gravitational
waves, consistent with current bounds. Therefore, detecting large-amplitude primor-
dial gravitational waves would represent an unequivocal falsification of this scenario.
Finally, chapter 4 was based on work published in Ref. [67] in collaboration with
my fellow graduate student, Gustavo Turiaci. In it, we introduced a new model
called “warm ekpyrosis,” which is also an attractor and has the aesthetic advantage
of requiring only a single scalar field. We showed that an ekpyrotic field decaying
into hot radiation receives dissipative corrections to its equation of motion, which
can be chosen such that its fluctuations acquire a (nearly) scale-invariant spectrum.
The spectrum is immediately adiabatic, and thus no conversion is necessary– another
significant advantage. It is notable that the same ingredient responsible for the scale-
invariant spectrum– the dissipation induced by the decay– also provides a framework
for explaining reheating, a largely unexplored topic in the context of ekpyrotic cos-
mologies and one that merits further study. The general hydrodynamical analysis we
applied provides many directions for future work, chief among them to devise a micro-
physical realization that reproduces the effective dynamics described here. As with
all ekpyrotic scenarios, this too would be falsified by a detection of large-amplitude
primordial gravitational waves.
These results suggest that ekpyrotic bouncing cosmologies hold promise for de-
scribing the early Universe. In the near future, high-sensitivity measurements of
B-mode polarization in the CMB will yield tighter constraints on the tensor-to-scalar
ratio, providing a crucial discriminant between these different early Universe scenar-
ios. Within the framework of Einstein gravity, it is difficult to imagine an ekpyrotic
universe that can generate large-scale primordial gravitational waves. Both mech-
anisms studied in this thesis for producing scale-invariant curvature perturbations
(noncanonical kinetic couplings in chapter 3 and direct decay in chapter 4) altered only
the scalar (compressional) modes; they left the tensor (radiative) perturbations intact
92
with respect to single-field ekpyrosis [112]. All models to date generate no detectable
spectrum of primordial gravitational waves. To alter this spectrum, one might imag-
ine tensorial corrections to the anisotropic inertia (see e.g., πijT in Eq. (5.1.43) of
Ref. [146]). Such terms vanish for perfect fluid matter or for matter in local thermal
equilibrium, but in general they can act as sources for the production of primordial
gravitational waves. We leave the investigation of such effects for future work.
93
Bibliography
[3] P. S. Wesson, “Olbers’s paradox and the spectral intensity of the extragalactic
background light,” The Astrophysical Journal, vol. 367, pp. 399–406, 1991.
[9] S. Masi et al., “The BOOMERanG experiment and the curvature of the uni-
verse,” Prog. Part. Nucl. Phys., vol. 48, pp. 243–261, 2002. [,243(2002)].
[13] P. A. R. Ade et al., “Planck 2013 results. XVI. Cosmological parameters,” As-
tron. Astrophys., vol. 571, p. A16, 2014.
[14] C. W. Misner, “Mixmaster universe,” Phys. Rev. Lett., vol. 22, pp. 1071–1074,
May 1969.
[15] R. Dicke, “Gravitation and the universe, volume 78 of memoirs of the american
philosophical society, jayne lecture for 1969,” American Philosophical Society,
Philadelphia, USA, vol. 82, 1970.
95
[17] A. Albrecht and P. J. Steinhardt, “Cosmology for grand unified theories with
radiatively induced symmetry breaking,” Phys.Rev.Lett., vol. 48, pp. 1220–1223,
1982.
[21] A. H. Guth and S.-Y. Pi, “Fluctuations in the new inflationary universe,” Phys-
ical Review Letters, vol. 49, no. 15, p. 1110, 1982.
[26] L. Berezhiani and M. Trodden, “How Likely are Constituent Quanta to Initiate
Inflation?,” Phys. Lett., vol. B749, pp. 425–430, 2015.
96
[27] S. Ferrara, R. Kallosh, A. Linde, and M. Porrati, “Minimal Supergravity Models
of Inflation,” Phys. Rev., vol. D88, no. 8, p. 085038, 2013.
[31] A. Vilenkin, “The Birth of Inflationary Universes,” Phys.Rev., vol. D27, p. 2848,
1983.
[32] A. H. Guth, “Inflation and eternal inflation,” Phys.Rept., vol. 333, pp. 555–574,
2000.
[35] A. Linde, “Inflationary Cosmology after Planck 2013,” in Proceedings, 100th Les
Houches Summer School: Post-Planck Cosmology: Les Houches, France, July
8 - August 2, 2013, pp. 231–316, 2015.
[36] A. H. Guth, “Eternal inflation and its implications,” J. Phys., vol. A40,
pp. 6811–6826, 2007.
97
[38] I. Bars, P. J. Steinhardt, and N. Turok, “Cyclic Cosmology, Conformal Sym-
metry and the Metastability of the Higgs,” Phys. Lett., vol. B726, pp. 50–55,
2013.
[39] A. Friedmann, “Über die krümmung des raumes,” Zeitschrift für Physik, vol. 10,
no. 1, pp. 377–386, 1922.
[40] A. Einstein and W. De Sitter, “On the relation between the expansion and the
mean density of the universe,” Proceedings of the National Academy of Sciences,
vol. 18, no. 3, pp. 213–214, 1932.
[43] R. C. Tolman, “On the theoretical requirements for a periodic behaviour of the
universe,” Physical Review, vol. 38, no. 9, p. 1758, 1931.
[46] P. J. Steinhardt and N. Turok, “The cyclic model simplified,” New Astronomy
Reviews, vol. 49, no. 2âĂŽÄì6, pp. 43 – 57, 2005. Sources and Detection of Dark
98
Matter and Dark Energy in the Universe 6th {UCLA} Symposium on Sources
and Detection of Dark Matter and Dark Energy in the Universe.
[49] J. Acacio de Barros, M. A. Sagioro Leal, and M. Pinto-Neto, “The Causal Inter-
pretation of Dust and Radiation Fluids Non-Singular Quantum Cosmologies,”
in Recent Developments in Theoretical and Experimental General Relativity,
Gravitation, and Relativistic Field Theories (T. Piran and R. Ruffini, eds.),
p. 1403, 1999.
[52] L. Smolin, “An Invitation to loop quantum gravity,” in Proceedings, 3rd Inter-
national Symposium on Quantum theory and symmetries (QTS3): Cincinnati,
USA, September 10-14, 2003, pp. 655–682, 2004. [Submitted to: Rev. Mod.
Phys.(2004)].
[53] N. Turok, B. Craps, and T. Hertog, “From big crunch to big bang with ads/cft,”
arXiv preprint arXiv:0711.1824, 2007.
99
[54] S. Gielen and N. Turok, “Perfect Quantum Cosmological Bounce,” Phys. Rev.
Lett., vol. 117, no. 2, p. 021301, 2016.
[57] Y.-F. Cai, T. Qiu, X. Zhang, Y.-S. Piao, and M. Li, “Bouncing universe with
quintom matter,” Journal of High Energy Physics, vol. 2007, no. 10, p. 071,
2007.
[59] P. Creminelli and L. Senatore, “A smooth bouncing cosmology with scale in-
variant spectrum,” Journal of Cosmology and Astroparticle Physics, vol. 2007,
no. 11, p. 010, 2007.
[60] T. Qiu, J. Evslin, Y.-F. Cai, M. Li, and X. Zhang, “Bouncing galileon cosmolo-
gies,” Journal of Cosmology and Astroparticle Physics, vol. 2011, no. 10, p. 036,
2011.
100
[64] A. Ijjas and P. J. Steinhardt, “Fully stable cosmological solutions with a non-
singular classical bounce,” Phys. Lett., vol. B764, pp. 289–294, 2017.
[65] A. M. Levy, “Fine-tuning challenges for the matter bounce scenario,” Phys. Rev.
D, vol. 95, p. 023522, Jan 2017.
[67] A. M. Levy and G. J. Turiaci, “Warm Ekpyrosis,” Phys. Rev., vol. D94, no. 8,
p. 083514, 2016.
[70] P. A. R. Ade et al., “Planck 2015 results. XVII. Constraints on primordial non-
Gaussianity,” 2015.
[73] F. Finelli and R. Brandenberger, “On the generation of a scale invariant spec-
trum of adiabatic fluctuations in cosmological models with a contracting phase,”
Phys. Rev., vol. D65, p. 103522, 2002.
[77] G. W. Gibbons, S. W. Hawking, and S. Siklos, The Very Early Universe: Pro-
ceedings of the Nuffield Workshop, Cambridge 21 June to 9 July, 1982. CUP
Archive, 1985.
[80] Y.-B. Li, J. Quintin, D.-G. Wang, and Y.-F. Cai, “Matter bounce cosmology
with a generalized single field: non-Gaussianity and an extended no-go theo-
rem,” 2016.
[81] Y.-F. Cai, R. Brandenberger, and X. Zhang, “The Matter Bounce Curvaton
Scenario,” JCAP, vol. 1103, p. 003, 2011.
[83] Y.-F. Cai and E. Wilson-Ewing, “A ΛCDM bounce scenario,” JCAP, vol. 1503,
no. 03, p. 006, 2015.
102
[84] A. Fertig, J.-L. Lehners, E. Mallwitz, and E. Wilson-Ewing, “Converting en-
tropy to curvature perturbations after a cosmic bounce,” JCAP, vol. 1610,
no. 10, p. 005, 2016.
[85] Y.-F. Cai, A. Marciano, D.-G. Wang, and E. Wilson-Ewing, “Bouncing cos-
mologies with dark matter and dark energy,” Universe, vol. 3, p. 1, 2017.
[91] Y.-F. Cai, S.-H. Chen, J. B. Dent, S. Dutta, and E. N. Saridakis, “Mat-
ter Bounce Cosmology with the f(T) Gravity,” Class. Quant. Grav., vol. 28,
p. 215011, 2011.
[92] Y.-F. Cai and E. Wilson-Ewing, “Non-singular bounce scenarios in loop quan-
tum cosmology and the effective field description,” JCAP, vol. 1403, p. 026,
2014.
103
[93] R. Brandenberger and F. Finelli, “On the spectrum of fluctuations in an effective
field theory of the Ekpyrotic universe,” JHEP, vol. 11, p. 056, 2001.
[95] J.-c. Hwang, “Cosmological structure problem in the ekpyrotic scenario,” Phys.
Rev., vol. D65, p. 063514, 2002.
[98] J.-L. Lehners, “Ekpyrotic and Cyclic Cosmology,” Phys.Rept., vol. 465, pp. 223–
263, 2008.
[101] F. Finelli, “A contracting universe driven by two scalar fields,” Physics Letters
B, vol. 545, no. 1–2, pp. 1 – 7, 2002.
104
[103] K. Koyama and D. Wands, “Ekpyrotic collapse with multiple fields,” JCAP,
vol. 0704, p. 008, 2007.
[104] E. I. Buchbinder, J. Khoury, and B. A. Ovrut, “On the initial conditions in new
ekpyrotic cosmology,” JHEP, vol. 0711, p. 076, 2007.
[113] A. Fertig, J.-L. Lehners, and E. Mallwitz, “Ekpyrotic Perturbations With Small
Non-Gaussian Corrections,” 2013.
105
[114] A. Ijjas, J.-L. Lehners, and P. J. Steinhardt, “General mechanism for producing
scale-invariant perturbations and small non-Gaussianity in ekpyrotic models,”
Phys.Rev., vol. D89, no. 12, p. 123520, 2014.
[116] F. Finelli, “Assisted contraction,” Phys. Lett., vol. B545, pp. 1–7, 2002.
[119] D. H. Lyth and D. Wands, “Cold dark matter isocurvature perturbation in the
curvaton scenario,” Phys. Rev. D, vol. 68, p. 103516, Nov 2003.
[120] M. Li, “Entropic mechanisms with generalized scalar fields in the Ekpyrotic
universe,” Phys.Lett., vol. B741, pp. 320–326, 2015.
[122] A. Berera, I. G. Moss, and R. O. Ramos, “Warm Inflation and its Microphysical
Basis,” Rept. Prog. Phys., vol. 72, p. 026901, 2009.
[123] A. Berera and L.-Z. Fang, “Thermally induced density perturbations in the
inflation era,” Phys. Rev. Lett., vol. 74, pp. 1912–1915, 1995.
[124] A. Berera, “Warm inflation,” Phys. Rev. Lett., vol. 75, pp. 3218–3221, 1995.
106
[126] A. Berera, “The warm inflationary universe,” Contemp. Phys., vol. 47, pp. 33–
49, 2006.
[127] M. Bastero-Gil and A. Berera, “Warm inflation model building,” Int. J. Mod.
Phys., vol. A24, pp. 2207–2240, 2009.
[135] J. Yokoyama and A. D. Linde, “Is warm inflation possible?,” Phys. Rev.,
vol. D60, p. 083509, 1999.
[141] J.-c. Hwang and H. Noh, “Cosmological perturbations with multiple fluids and
fields,” Class. Quant. Grav., vol. 19, pp. 527–550, 2002.
[142] K. A. Malik and D. Wands, “Adiabatic and entropy perturbations with inter-
acting fluids and fields,” JCAP, vol. 0502, p. 007, 2005.
[143] P. Ade et al., “Planck 2013 results. XXII. Constraints on inflation,” As-
tron.Astrophys., vol. 571, p. A22, 2014.
[145] A. Linde, “Does the first chaotic inflation model in supergravity provide the
best fit to the planck data?,” Journal of Cosmology and Astroparticle Physics,
vol. 2015, no. 02, p. 030, 2015.
108