Inflation and String Cosmology∗

arXiv:hep-th/0503195v1 24 Mar 2005

Andrei Linde

Department of Physics, Stanford University, Stanford, CA 94305, USA

Abstract
After 25 years of its existence, inflationary theory gradually becomes the standard cosmological
paradigm. However, we still do not know which of the many versions of inflationary cosmology
will be favored by the future observational data. Moreover, it may be quite nontrivial to obtain a
natural realization of inflationary theory in the context of the ever changing theory of all funda-
mental interactions. In this paper I will describe the history and the present status of inflationary
cosmology, including recent attempts to implement inflation in the context of string theory.

∗
This is an extended version of my talks at the SLAC Summer School “Nature’s Greatest Puzzles,” at
the conference Cosmo04 in Toronto, at the VI Mexican School on Gravitation, and at the XXII Texas
Symposium on Relativistic Astrophysics in 2004.

1
 Contents

I. Brief history of inflation 3

II. Chaotic Inflation 4

III. Hybrid inflation 8

IV. Quantum fluctuations and density perturbations 8

V. Eternal inflation 12

VI. Creation of matter after inflation: reheating and preheating 15

VII. Inflation and observations 16

VIII. Alternatives to inflation? 17

IX. Shift symmetry and chaotic inflation in supergravity 19

X. Towards Inflation in String Theory 22

A. de Sitter vacua in string theory 22
B. Inflation in string theory 23
1. Modular inflation 23
2. Brane inflation and shift symmetry 24

XI. Scale of inflation, scale of SUSY breaking and the gravitino mass 27

XII. Initial conditions for the low-scale inflation and topology of the universe29

XIII. Eternal inflation and the string theory landscape 31

Acknowledgments 33

References 33

2
I. BRIEF HISTORY OF INFLATION

The first model of inflationary type was proposed by Alexei Starobinsky [1]. It was based
on investigation of conformal anomaly in quantum gravity. This model was rather compli-
cated, and its goal was somewhat different from the goals of inflationary cosmology. Instead
of attempting to solve the homogeneity and isotropy problems, Starobinsky considered the
model of the universe which was homogeneous and isotropic from the very beginning, and
emphasized that his scenario was “the extreme opposite of Misner’s initial “chaos”.”
On the other hand, Starobinsky model did not suffer from the graceful exit problem, and it
was the first model predicting gravitational waves with flat spectrum [1]. More importantly,
the first mechanism of production of adiabatic perturbations of metric with a flat spectrum,
which are responsible for galaxy production, and which were found by the observations of
the CMB anisotropy, was proposed by Mukhanov and Chibisov [2] in the context of this
model.
A much simpler inflationary model with a very clear physical motivation was proposed
by Alan Guth [3]. His model, which is now called “old inflation,” was based on the theory
of supercooling during the cosmological phase transitions [4]. Even though this scenario did
not work, it played a profound role in the development of inflationary cosmology since it
contained a very clear explanation how inflation may solve the major cosmological problems.
According to this scenario, inflation is as exponential expansion of the universe in a
supercooled false vacuum state. False vacuum is a metastable state without any fields or
particles but with large energy density. Imagine a universe filled with such “heavy nothing.”
When the universe expands, empty space remains empty, so its energy density does not
change. The universe with a constant energy density expands exponentially, thus we have
inflation in the false vacuum. This expansion makes the universe very big and very flat. Then
the false vacuum decays, the bubbles of the new phase collide, and our universe becomes
hot.
Unfortunately, this simple and intuitive picture of inflation in the false vacuum state,
which is often presented in the popular literature, is somewhat misleading. If the bubbles
of the new phase are formed near each other, their collisions make the universe extremely
inhomogeneous. If they are formed far away from each other, each of them represents a
separate open universe with a vanishingly small Ω. Both options are unacceptable, which
has lead to the conclusion that this scenario cannot be improved (graceful exit problem)
[3, 5, 6].
This problem was resolved with the invention of the new inflationary theory [7, 8]. In this
theory, inflation may begin either in the false vacuum, or in an unstable state at the top of the
effective potential. Then the inflaton field φ slowly rolls down to the minimum of its effective
potential. The motion of the field away from the false vacuum is of crucial importance:
density perturbations produced during the slow-roll inflation are inversely proportional to φ̇
[2, 9, 10]. Thus the key difference between the new inflationary scenario and the old one is
that the useful part of inflation in the new scenario, which is responsible for the homogeneity
of our universe, does not occur in the false vacuum state, where φ̇ = 0.
Some authors recently started using a generalized notion of the false vacuum, defining it as
a vacuum-like state with a slowly changing energy density. While the difference between this

3
definition and the standard one is very subtle, it is exactly this difference that is responsible
for solving all problems of the old inflation scenario, and it is exactly this subtlety that made
it so difficult to find this solution.
The new inflationary scenario became so popular in the beginning of the 80’s that even
now most textbooks on astrophysics incorrectly describe inflation as an exponential expan-
sion during high temperature phase transitions in grand unified theories. Unfortunately,
new inflation was plagued by its own problems. It works only if the effective potential of
the field φ has a very a flat plateau near φ = 0, which is somewhat artificial. In most ver-
sions of this scenario the inflaton field has an extremely small coupling constant, so it could
not be in thermal equilibrium with other matter fields. The theory of cosmological phase
transitions, which was the basis for old and new inflation, did not work in such a situation.
Moreover, thermal equilibrium requires many particles interacting with each other. This
means that new inflation could explain why our universe was so large only if it was very
large and contained many particles from the very beginning. Finally, inflation in this theory
begins very late. During the preceding epoch the universe can easily collapse or become so
inhomogeneous that inflation may never happen [11].
Old and new inflation represented a substantial but incomplete modification of the big
bang theory. It was still assumed that the universe was in a state of thermal equilibrium from
the very beginning, that it was relatively homogeneous and large enough to survive until the
beginning of inflation, and that the stage of inflation was just an intermediate stage of the
evolution of the universe. In the beginning of the 80’s these assumptions seemed most natural
and practically unavoidable. On the basis of all available observations (CMB, abundance of
light elements) everybody believed that the universe was created in a hot big bang. That is
why it was so difficult to overcome a certain psychological barrier and abandon all of these
assumptions. This was done with the invention of the chaotic inflation scenario [12]. This
scenario resolved all problems of old and new inflation. According to this scenario, inflation
may occur even in the theories with simplest potentials such as V (φ) ∼ φn . Inflation may
begin even if there was no thermal equilibrium in the early universe, and it may start even
at the Planckian density, in which case the problem of initial conditions for inflation can be
easily resolved [11].

II. CHAOTIC INFLATION

Consider the simplest model of a scalar field φ with a mass m and with the potential
2
energy density V (φ) = m2 φ2 . Since this function has a minimum at φ = 0, one may expect
that the scalar field φ should oscillate near this minimum. This is indeed the case if the
universe does not expand, in which case equation of motion for the scalar field coincides
with equation for harmonic oscillator, φ̈ = −m2 φ.
However, because of the expansion of the universe with Hubble constant H = ȧ/a, an
additional term 3H φ̇ appears in the harmonic oscillator equation:

φ̈ + 3H φ̇ = −m2 φ . (1)

The term 3H φ̇ can be interpreted as a friction term. The Einstein equation for a homoge-

4
neous universe containing scalar field φ looks as follows:
k 1  2 
H2 + = φ̇ + m2 2
φ ) . (2)
a2 6
Here k = −1, 0, 1 for an open, flat or closed universe respectively. We work in units Mp−2 =
8πG = 1.
If the scalar field φ initially was large, the Hubble parameter H was large too, according
to the second equation. This means that the friction term 3H φ̇ was very large, and therefore
the scalar field was moving very slowly, as a ball in a viscous liquid. Therefore at this stage
the energy density of the scalar field, unlike the density of ordinary matter, remained almost
constant, and expansion of the universe continued with a much greater speed than in the
old cosmological theory. Due to the rapid growth of the scale of the universe and a slow
motion of the field φ, soon after the beginning of this regime one has φ̈ ≪ 3H φ̇, H 2 ≫ ak2 ,
φ̇2 ≪ m2 φ2 , so the system of equations can be simplified:
s
ȧ mφ 2
H= = √ , φ̇ = −m . (3)
a 6 3
The first equation shows that if the field φ changes slowly, the size of the universe in this
regime grows approximately as eHt , where H = mφ √ . This is the stage of inflation, which
6
ends when the field φ becomes much smaller than Mp = 1. Solution of these equations shows
that after a long stage of inflation the universe initially filled with the field φ ≫ 1 grows
exponentially [11],
2
a = a0 eφ /4 . (4)

This is as simple as it could be. Inflation does not require supercooling and tunneling
from the false vacuum [3], or rolling from an artificially flat top of the effective potential
[7, 8]. It appears in the theories that can be as simple as a theory of a harmonic oscillator
[12]. Only when it was realized, I started to really believe that inflation is not just a trick
necessary to fix problems of the old big bang theory, but a generic cosmological regime.
But what’s about the initial conditions required for chaotic inflation? Let us consider first
a closed Universe of initial size l ∼ 1 (in Planck units), which emerges from the space-time
foam, or from singularity, or from ‘nothing’ in a state with the Planck density ρ ∼ 1. Only
starting from this moment, i.e. at ρ ∼< 1, can we describe this domain as a classical Universe.
Thus, at this initial moment the sum of the kinetic energy density, gradient energy density,
and the potential energy density is of the order unity: 12 φ̇2 + 12 (∂i φ)2 + V (φ) ∼ 1.
We wish to emphasize, that there are no a priori constraints on the initial value of the
scalar field in this domain, except for the constraint 21 φ̇2 + 21 (∂i φ)2 + V (φ) ∼ 1. Let us
consider for a moment a theory with V (φ) = const. This theory is invariant under the shift
symmetry φ → φ + c. Therefore, in such a theory all initial values of the homogeneous
component of the scalar field φ are equally probable.
The only constraint on the amplitude of the field appears if the effective potential is
not constant, but grows and becomes greater than the Planck density at φ > φp , where
V (φp ) = 1. This constraint implies that φ ∼< φp , but there is no reason to expect that
initially φ must be much smaller than φp . This suggests that the typical initial value of the
field φ in such a theory is φ ∼ φp .

5
 V
Planck density

C

2
FIG. 1: Motion of the scalar field in the theory with V (φ) = m2 φ2 . Several different regimes
are possible, depending on the value of the field φ. If the potential energy density of the field is
greater than the Planck density Mp4 , quantum fluctuations of space-time are so strong that one
cannot describe it in usual terms. Such a state is called space-time foam. At a somewhat smaller
energy density (region A: mMp3 < V (φ) < Mp4 ) quantum fluctuations of space-time are small,
but quantum fluctuations of the scalar field φ may be large. Jumps of the scalar field due to
quantum fluctuations lead to a process of eternal self-reproduction of inflationary universe which
we are going to discuss later. At even smaller values of V (φ) (region B: m2 Mp2 < V (φ) < mMp3 )
fluctuations of the field φ are small; it slowly moves down as a ball in a viscous liquid. Inflation
occurs both in the region A and region B. Finally, near the minimum of V (φ) (region C) the scalar
field rapidly oscillates, creates pairs of elementary particles, and the universe becomes hot.

Thus, we expect that typical initial conditions correspond to 12 φ̇2 ∼ 21 (∂i φ)2 ∼ V (φ) =
< V (φ) in the domain under consideration, then inflation begins, and
O(1). If 21 φ̇2 + 21 (∂i φ)2 ∼
then within the Planck time the terms 12 φ̇2 and 21 (∂i φ)2 become much smaller than V (φ),
which ensures continuation of inflation. It seems therefore that chaotic inflation occurs under
rather natural initial conditions, if it can begin at V (φ) ∼ 1.
As we will see shortly, the realistic value of the mass m is about 3 × 10−6 , in Planck
units. Therefore, according to Eq. (4), the total amount of inflation achieved starting from
10
V (φ) ∼ 1 is of the order 1010 . The total duration of inflation in this model is about
10−30 seconds. When inflation ends, the scalar field φ begins to oscillate near the minimum
of V (φ). As any rapidly oscillating classical field, it looses its energy by creating pairs of
elementary particles. These particles interact with each other and come to a state of thermal
equilibrium with some temperature Tr [13, 14, 15, 16, 17, 18, 19]. From this time on, the
universe can be described by the usual big bang theory.
The main difference between inflationary theory and the old cosmology becomes clear
when one calculates the size of a typical inflationary domain at the end of inflation. Inves-
tigation of this question shows that even if the initial size of inflationary universe was as
small as the Planck size lP ∼ 10−33 cm, after 10−30 seconds of inflation the universe acquires
10
a huge size of l ∼ 1010 cm! This number is model-dependent, but in all realistic models

6
the size of the universe after inflation appears to be many orders of magnitude greater than
the size of the part of the universe which we can see now, l ∼ 1028 cm. This immediately
solves most of the problems of the old cosmological theory [11, 12].
Our universe is almost exactly homogeneous on large scale because all inhomogeneities
were exponentially stretched during inflation. The density of primordial monopoles and other
undesirable “defects” becomes exponentially diluted by inflation. The universe becomes
enormously large. Even if it was a closed universe of a size ∼ 10−33 cm, after inflation the
distance between its “South” and “North” poles becomes many orders of magnitude greater
than 1028 cm. We see only a tiny part of the huge cosmic balloon. That is why nobody has
ever seen how parallel lines cross. That is why the universe looks so flat.
If our universe initially consisted of many domains with chaotically distributed scalar field
φ (or if one considers different universes with different values of the field), then domains in
which the scalar field was too small never inflated. The main contribution to the total volume
of the universe will be given by those domains which originally contained large scalar field
φ. Inflation of such domains creates huge homogeneous islands out of initial chaos. (That is
why I called this scenario “chaotic inflation.”) Each homogeneous domain in this scenario
is much greater than the size of the observable part of the universe.
The first models of chaotic inflation were based on the theories with polynomial potentials,
2
such as V (φ) = ± m2 φ2 + λ4 φ4 . But, as was emphasized in [12], the main idea of this scenario
is quite generic. One should consider any particular potential V (φ), polynomial or not, with
or without spontaneous symmetry breaking, and study all possible initial conditions without
assuming that the universe was in a state of thermal equilibrium, and that the field φ was
in the minimum of its effective potential from the very beginning.
This scenario strongly deviated from the standard lore of the hot big bang theory and
was psychologically difficult to accept. Therefore during the first few years after invention
of chaotic inflation many authors claimed that the idea of chaotic initial conditions is un-
natural, and made attempts to realize the new inflation scenario based on the theory of
high-temperature phase transitions, despite numerous problems associated with it. Some
authors believed that the theory must satisfy so-called ‘thermal constraints’ which were
necessary to ensure that the minimum of the effective potential at large T should be at
φ = 0 [20], even though the scalar field in the models they considered was not in a state of
thermal equilibrium with other particles. It took several years until it finally became clear
that the idea of chaotic initial conditions is most general, and it is much easier to construct
a consistent cosmological theory without making unnecessary assumptions about thermal
equilibrium and high temperature phase transitions in the early universe.
The issue of thermal initial conditions played the central role in the long debate about
new inflation versus chaotic inflation in the 80’s [11], but now the debate is over: no realistic
versions of new inflation based on the theory of thermal phase transitions and supercooling
have been proposed so far. As a result, the corresponding terminology gradually changed.
Chaotic inflation, as defined in [12], occurs in all models with sufficiently flat potentials,
including the potentials with a flat maximum, originally used in new inflation [21]. Now
the versions of inflationary scenario with such potentials for simplicity are often called ‘new
inflation’, even though inflation begins there not as in the original new inflation scenario,
but as in the chaotic inflation scenario. A new twist in terminology was suggested very
recently, when this version of chaotic inflation was called ‘hilltop inflation’ [22].

7
III. HYBRID INFLATION

In the previous section we considered the simplest chaotic inflation theory based on the
theory of a single scalar field φ. The models of chaotic inflation based on the theory of two
scalar fields may have some qualitatively new features. One of the most interesting models
of this kind is the hybrid inflation scenario [23]. The simplest version of this scenario is
based on chaotic inflation in the theory of two scalar fields with the effective potential

1 m2 2 g 2 2 2
V (σ, φ) = (M 2 − λσ 2 )2 + φ + φ σ . (5)
4λ 2 2
The effective mass squared of the field σ is equal to −M 2 +g 2φ2 . Therefore for φ > φc = M/g
the only minimum of the effective potential V (σ, φ) is at σ = 0. The curvature of the effective
potential in the σ-direction is much greater than in the φ-direction. Thus at the first stages
of expansion of the universe the field σ rolled down to σ = 0, whereas the field φ could
remain large for a much longer time.
At the moment when the inflaton field φ becomes smaller than φc = M/g, the phase tran-
sition with the symmetry breaking occurs. The fields rapidly fall to the absolute minimum
of the potential at φ = 0, σ 2 = M 2 /λ. If m2 φ2c = m2 M 2 /g 2 ≪ M 4 /λ, the Hubble constant
M4 2
at the time of the phase transition is given by H 2 = 12λ (in units Mp = 1). If M 2 ≫ λm g2
and m2 ≪ H 2 , then the universe at φ > φc undergoes a stage of inflation, which abruptly
ends at φ = φc .
One of the advantages of this scenario is the possibility to obtain small density pertur-
bations even if coupling constants are large, λ, g = O(1), and if the inflaton field φ is much
smaller than Mp . The last condition is absolutely unnecessary in the usual theory of scalar
fields coupled to gravity, but it may be important if the scalar field has a certain geomet-
ric meaning, which is often the case in supergravity and string theory. This makes hybrid
inflation an attractive playground for those who wants to achieve inflation in supergravity
and string theory. We will return to this question later.

IV. QUANTUM FLUCTUATIONS AND DENSITY PERTURBATIONS

The vacuum structure in the exponentially expanding universe is much more compli-
cated than in ordinary Minkowski space. The wavelengths of all vacuum fluctuations of the
scalar field φ grow exponentially during inflation. When the wavelength of any particular
fluctuation becomes greater than H −1 , this fluctuation stops oscillating, and its amplitude
freezes at some nonzero value δφ(x) because of the large friction term 3H φ̇ in the equation
of motion of the field φ. The amplitude of this fluctuation then remains almost unchanged
for a very long time, whereas its wavelength grows exponentially. Therefore, the appearance
of such a frozen fluctuation is equivalent to the appearance of a classical field δφ(x) that
does not vanish after averaging over macroscopic intervals of space and time.
Because the vacuum contains fluctuations of all wavelengths, inflation leads to the cre-
ation of more and more new perturbations of the classical field with wavelengths greater
than H −1 . The average amplitude of such perturbations generated during a typical time

8
interval H −1 is given by [24, 25]
H
|δφ(x)| ≈ . (6)
2π
These fluctuations lead to density perturbations that later produce galaxies. The theory
of this effect is very complicated [2, 9], and it was fully understood only in the second part
of the 80’s [10]. The main idea can be described as follows:
Fluctuations of the field φ lead to a local delay of the time of the end of inflation,
δt = δφφ̇
∼ 2πHφ̇ . Once the usual post-inflationary stage begins, the density of the universe
starts to decrease as ρ = 3H 2 , where H ∼ t−1 . Therefore a local delay of expansion leads to
a local density increase δH such that δH ∼ δρ/ρ ∼ δt/t. Combining these estimates together
yields the famous result [2, 9, 10]

δρ H2
δH ∼ ∼ . (7)
ρ 2π φ̇
This derivation is oversimplified; it does not tell, in particular, whether H should be calcu-
lated during inflation or after it. This issue was not very important for new inflation where
H was nearly constant, but it is of crucial importance for chaotic inflation.
The result of a more detailed investigation [10] shows that H and φ̇ should be calculated
during inflation, at different times for perturbations with different momenta k. For each of
these perturbations the value of H should be taken at the time when the wavelength of the
perturbation becomes of the order of H −1 . However, the field φ during inflation changes
H2
very slowly, so the quantity 2π φ̇
remains almost constant over exponentially large range of
wavelengths. This means that the spectrum of perturbations of metric is flat.
A detailed calculation in our simplest chaotic inflation model the amplitude of perturba-
tions gives
mφ2
δH ∼ √ . (8)
5π 6
The perturbations on scale of the horizon were produced at φH ∼ 15 [11]. This, together
with COBE normalization δH ∼ 2 × 10−5 gives m ∼ 3 × 10−6 , in Planck units, which is
approximately equivalent to 7 × 1012 GeV. An exact value of m depends on φH , which in its
turn depends slightly on the subsequent thermal history of the universe.
When the fluctuations of the scalar field φ are first produced (frozen), their wavelength
2
is given by H(φ)−1 . At the end of inflation, the wavelength grows by the factor of eφ /4 , see
Eq. (4). In other words, the logarithm of the wavelength l of the perturbations of metric is
proportional to the value of φ2 at the moment when these perturbations were produced. As a
result, according to (8), the amplitude of perturbations of metric depends on the wavelength
logarithmically: δH ∼ m ln l. A similar logarithmic dependence (with different powers of
the logarithm) appears in other versions of chaotic inflation with V ∼ φn and in the simplest
versions of new inflation.
Since the observations provide us with information about a rather limited range of l, it
is often possible to parametrize the scale dependence of density perturbations by a simple
power law, δH ∼ l(1−ns )/2 . An exactly flat spectrum, called Harrison-Zeldovich spectrum,
would correspond to ns = 1.

9
 Flatness of the spectrum of density perturbations, together with flatness of the universe
(Ω = 1), constitute the two most robust predictions of inflationary cosmology. It is possible
to construct models where δH changes in a very peculiar way, and it is also possible to
construct theories where Ω 6= 1, but it is difficult to do so.
However, there is a subtle but important difference between the prediction of flatness
of the universe and the flatness of the spectrum of perturbations of metric. It is very
difficult (though possible) to construct an inflationary model deviating from the simple
prediction Ω = 1. Meanwhile, the situation with the flatness of the spectrum is opposite:
it is very difficult (though possible) to construct a model with an exactly flat spectrum of
perturbations of metric. In this sense, existence of a small deviation of the spectrum of
inflationary perturbations from the flat Harrison-Zeldovich spectrum (i.e. the breaking of
the scale invariance of the spectrum) represents an additional robust prediction of inflation.
The small deviation from the scale invariance discussed above is fundamentally important
for understanding of the global structure of inflationary universe. Here an analogy with a
different branch of physics may be rather helpful.
This year we celebrated a discovery of asymptotic freedom [26, 27], which is based, in
the simplest cases, on the logarithmic growth of the coupling constants with distance. If
one ignores this logarithmic dependence, one could incorrectly say that the theory is scale-
invariant.
Similarly, in the simplest models of inflation the amplitude of density perturbations loga-
rithmically increases with distance. If one ignores this logarithmic dependence, one obtains
the flat Harrison-Zeldovich spectrum.
In QCD, the logarithmic dependence leads to the asymptotic freedom on small scale and
to the growth of the coupling on large scale, which results in a nonperturbative regime and
quark confinement. In inflationary cosmology, a similar logarithmic dependence leads to the
growth of density perturbations on large scales, which results in a nonperturbative regime,
fractal structure of the universe, and eternal inflation, which we are going to discuss in the
next section. In both cases, the deviation of the scale invariance appears to be profoundly
important. That is why it would be very interesting to find a possible deviation of the index
ns from 1. Hopefully, this can be achieved in future observations; in particular, Planck
satellite may be capable of measuring the deviation ns − 1 with an accuracy about 0.5%.
For future reference, we will give here a list of equations which are often used for compar-
ison of predictions of inflationary theories with observations in the slow roll approximation.
The amplitude of scalar perturbations of metric can be characterized either by δH , or by
a closely related quantity ∆R [28]. Similarly, the amplitude of tensor perturbations is given
by ∆h . Following [28, 29, 30, 31], one can represent these quantities as
!n −1
k s
∆2R (k) = ∆2R (k0 ) , (9)
k0
!n
2 2 k t
∆h (k) = ∆h (k0 ) , (10)
k0

where ∆2 (k0 ) is a normalization constant, and k0 is a normalization point. Here we ignored

running of the indexes ns and nt since there is no observational evidence that it is significant.

10
 One can also introduce the tensor/scalar ratio r, the relative amplitude of the tensor to
scalar modes,
∆2 (k0 )
r ≡ 2h . (11)
∆R (k0 )

There are three slow-roll parameters [28]

!2
1 V′ V ′′ V ′ V ′′′
ǫ= , η= , ξ= , (12)
2 V V V2
where prime denotes derivatives with respect to the field φ. All parameters must be smaller
than one for the slow-roll approximation to be valid.
A standard slow roll analysis gives observable quantities in terms of the slow roll param-
eters to first order as
V V3
∆2R = = , (13)
24π 2 ǫ 12π 2 (V ′ )2
ns − 1 = −6ǫ + 2η, (14)
r = 16ǫ, (15)
r
nt = −2ǫ = − . (16)
8
The equation nt = −r/8 is known as the consistency relation for single-field inflation models;
it becomes an inequality for multi-field inflation models. If V during inflation is sufficiently
large, as in the simplest models of chaotic inflation, one may have a chance to find the tensor
contribution to the CMB anisotropy. The possibility to determine nt is less certain. It may
be rather difficult, though maybe not impossible [32], to find the tensor contribution in new
inflation and in hybrid inflation. The most important information which can be obtained
now from the cosmological observations at present is related to Eqs. (13) and (14).
Following notational conventions in [29], we use A(k0 ) for the scalar power spectrum
amplitude, where A(k0 ) and ∆2R (k0 ) are related through
∆2R (k0 ) ≃ 3 × 10−9 A(k0 ). (17)
The parameter A is often normalized at k0 ∼ 0.05/Mpc; its observational value is about 0.8
[29, 30]. This leads to the observational constraint on V (φ) following from the normalization
of the spectrum of the large-scale density perturbations:
V 3/2
≃ 5 × 10−4 . (18)
V′
Here V (φ) should be evaluated for the value of the field φ which is determined by the con-
dition that the perturbations produced at the moment when the field was equal φ evolve
into the present time perturbations with momentum k0 ∼ 0.05/Mpc. In the first approxi-
mation, one can find the corresponding moment by assuming that it happened 60 e-foldings
before the end of inflation. The number of e-foldings can be calculated in the slow roll
approximation using the relation
φ V
Z
N≃ dφ . (19)
φend V′

11
Finally, recent observational data imply [31] that
!2
V′ V ′′
ns = 1 − 3 +2 = 0.98 ± 0.03 . (20)
V V

These relations are very useful for comparing inflationary models with observations. In
particular, the simplest versions of chaotic and new inflation predict ns < 1, whereas in
hybrid inflation one may have either ns < 1 or ns > 1, depending on the model.
Here we concentrated on the simplest and most general mechanism of production of adi-
abatic perturbations of metric in inflationary cosmology. However, one should keep in mind
that quantum fluctuations produced during inflation may also lead to isocurvature fluctu-
ations [33], which may later convert into adiabatic perturbations (the so-called curvaton
mechanism [34, 35, 36, 37]). One may also produce adiabatic perturbations by perturbing
effective coupling constants, which modulates the process of reheating [38, 39].

V. ETERNAL INFLATION

A significant step in the development of inflationary theory was the discovery of the
process of self-reproduction of inflationary universe. This process was known to exist in old
inflationary theory [3] and in the new one [40, 41], but its significance was fully realized only
after the discovery of the regime of eternal inflation in the simplest versions of the chaotic
inflation scenario [42, 43]. It appears that in many models large quantum fluctuations
produced during inflation which may locally increase the value of the energy density in some
parts of the universe. These regions expand at a greater rate than their parent domains,
and quantum fluctuations inside them lead to production of new inflationary domains which
expand even faster. This leads to an eternal process of self-reproduction of the universe.
To understand the mechanism of self-reproduction one should remember that the pro-
cesses separated by distances l greater than H −1 proceed independently of one another. This
is so because during exponential expansion the distance between any two objects separated
by more than H −1 is growing with a speed exceeding the speed of light. As a result, an
observer in the inflationary universe can see only the processes occurring inside the horizon
of the radius H −1 . An important consequence of this general result is that the process of
inflation in any spatial domain of radius H −1 occurs independently of any events outside it.
In this sense any inflationary domain of initial radius exceeding H −1 can be considered as a
separate mini-universe.
To investigate the behavior of such a mini-universe, with an account taken of quantum
fluctuations, let us consider an inflationary domain of initial radius H −1 containing suffi-
ciently homogeneous field with initial value φ ≫ Mp . Equation (3) implies that during a
typical time interval ∆t = H −1 the field inside this domain will be reduced by ∆φ = φ2 . By
comparison this expression with |δφ(x)| ≈ 2π H
= 2πmφ
√ one can easily see that if φ is much
6
less than φ∗ ∼ √5m , then the decrease of the field φ due to its classical motion is much
greater than the average amplitude of the quantum fluctuations δφ generated during the
same time. But for φ ≫ φ∗ one has δφ(x) ≫ ∆φ. Because the typical wavelength of the
fluctuations δφ(x) generated during the time is H −1 , the whole domain after ∆t = H −1

12
effectively becomes divided into e3 ∼ 20 separate domains (mini-universes) of radius H −1 ,
each containing almost homogeneous field φ − ∆φ + δφ. In almost a half of these domains
the field φ grows by |δφ(x)| − ∆φ ≈ |δφ(x)| = H/2π, rather than decreases. This means
that the total volume of the universe containing growing field φ increases 10 times. During
the next time interval ∆t = H −1 this process repeats. Thus, after the two time intervals
H −1 the total volume of the universe containing the growing scalar field increases 100 times,
etc. The universe enters eternal process of self-reproduction.
The existence of this process implies that the universe will never disappear as a whole.
Some of its parts may collapse, the life in our part of the universe may perish, but there
always will be some other parts of the universe where life will appear again and again, in all
of its possible forms.
One should be careful, however, with the interpretation of these results. There is still an
ongoing debate of whether eternal inflation is eternal only in the future or also in the past.
In order to understand what is going on, let us consider any particular time-like geodesic line
at the stage of inflation. One can show that for any given observer following this geodesic,
the duration ti of the stage of inflation on this geodesic will be finite. One the other hand,
eternal inflation implies that if one takes all such geodesics and calculate the time ti for each
of them, then there will be no upper bound for ti , i.e. for each time T there will be such
geodesic which experience inflation for the time ti > T . Even though the relative number of
long geodesics can be very small, exponential expansion of space surrounding them will lead
to an eternal exponential growth of the total volume of inflationary parts of the universe.
Similarly, if one concentrates on any particular geodesic in the past time direction, one
can prove that it has finite length [44], i.e. inflation in any particular point of the universe
should have a beginning at some time τi . However, there is no reason to expect that there is
an upper bound for all τi on all geodesics. If this upper bound does not exist, then eternal
inflation is eternal not only in the future but also in the past.
In other words, there was a beginning for each part of the universe, and there will be
an end for inflation at any particular point. But there will be no end for the evolution of
the universe as a whole in the eternal inflation scenario, and at present we do not have any
reason to believe that there was a single beginning of the evolution of the whole universe at
some moment t = 0, which was traditionally associated with the Big Bang.
To illustrate the process of eternal inflation, we present here the results of computer
simulations of evolution of a system of two scalar fields during inflation. The field φ is
the inflaton field driving inflation; it is shown by the height of the distribution of the field
φ(x, y) in a two-dimensional slice of the universe. The second field, Φ, determines the type of
spontaneous symmetry breaking which may occur in the theory. We paint the surface black
if this field is in a state corresponding to one of the two minima of its effective potential; we
paint it white if it is in the second minimum corresponding to a different type of symmetry
breaking, and therefore to a different set of laws of low-energy physics.
In the beginning of the process the whole inflationary domain is black, and the distribution
of both fields is very homogeneous. Then the domain became exponentially large (but it has
the same size in comoving coordinates, as shown in Fig. 2). Each peak of the mountains
corresponds to nearly Planckian density and can be interpreted as a beginning of a new
“Big Bang.” The laws of physics are rapidly changing there, but they become fixed in the

13
FIG. 2: Evolution of scalar fields φ and Φ during the process of self-reproduction of the universe.
The height of the distribution shows the value of the field φ which drives inflation. The surface
is painted black in those parts of the universe where the scalar field Φ is in the first minimum of
its effective potential, and white where it is in the second minimum. Laws of low-energy physics
are different in the regions of different color. The peaks of the “mountains” correspond to places
where quantum fluctuations bring the scalar fields back to the Planck density. Each of such places
in a certain sense can be considered as a beginning of a new Big Bang.

parts of the universe where the field φ becomes small. These parts correspond to valleys in
Fig. 2. Thus quantum fluctuations of the scalar fields divide the universe into exponentially
large domains with different laws of low-energy physics, and with different values of energy
density.
If this scenario is correct, then physics alone cannot provide a complete explanation for
all properties of our part of the universe. The same physical theory may yield large parts
of the universe that have diverse properties. According to this scenario, we find ourselves
inside a four-dimensional domain with our kind of physical laws not because domains with
different dimensionality and with alternate properties are impossible or improbable, but
simply because our kind of life cannot exist in other domains.

14
 The fact that inflation may happen in a different manner in different parts of the universe
was recognized at the very early stages of development of inflationary theory, which allowed
us to justify the use of anthropic principle in inflationary cosmology [45]. Eternal inflation
makes it possible to go even further: It implies that even if the universe started in a single
domain with well defined initial conditions, the process of eternal inflation will divide it into
infinitely many exponentially large domains that have different laws of low-energy physics
[42, 43]. Among all of these domains, we can live and make observations only in those that
are compatible with our existence.
Eternal inflation scenario was extensively studied during the last 20 years. I should
mention, in particular, the discovery of the topological eternal inflation [46], calculation of
the fractal dimension of the universe [43, 47], and development of various methods describing
statistical/probabilistic aspects of the self-reproducing universe, see [43, 48, 49, 50] and
references therein. This scenario may have especially interesting implications in the context
of string theory, which allows exponentially large number of different de Sitter vacuum states
[11, 51, 52, 53], see Sect. XIII.

VI. CREATION OF MATTER AFTER INFLATION: REHEATING AND PRE-

HEATING

The theory of reheating of the universe after inflation is the most important application
of the quantum theory of particle creation, since almost all matter constituting the universe
was created during this process.
At the stage of inflation all energy is concentrated in a classical slowly moving inflaton
field φ. Soon after the end of inflation this field begins to oscillate near the minimum of
its effective potential. Eventually it produces many elementary particles, they interact with
each other and come to a state of thermal equilibrium with some temperature Tr .
Early discussions of reheating of the universe after inflation [13] were based on the idea
that the homogeneous inflaton field can be represented as a collection of the particles of the
field φ. Each of these particles decayed independently. This process can be studied by the
usual perturbative approach to particle decay. Typically, it takes thousands of oscillations
of the inflaton field until it decays into usual elementary particles by this mechanism. More
recently, however, it was discovered that coherent field effects such as parametric resonance
can lead to the decay of the homogeneous field much faster than would have been predicted
by perturbative methods, within few dozen oscillations [14]. These coherent effects produce
high energy, nonthermal fluctuations that could have significance for understanding develop-
ments in the early universe, such as baryogenesis. This early stage of rapid nonperturbative
decay was called ‘preheating.’ In [15] it was found that another effect known as tachyonic
preheating can lead to even faster decay than parametric resonance. This effect occurs
whenever the homogeneous field rolls down a tachyonic (V ′′ < 0) region of its potential.
When that occurs, a tachyonic, or spinodal instability leads to exponentially rapid growth
of all long wavelength modes with k 2 < |V ′′ |. This growth can often drain all of the energy
from the homogeneous field within a single oscillation.
We are now in a position to classify the dominant mechanisms by which the homogeneous
inflaton field decays in different classes of inflationary models. Even though all of these

15
models, strictly speaking, belong to the general class of chaotic inflation (none of them is
based on the theory of thermal initial conditions), one can break them into three classes:
small field, or new inflation models [7], large field, or chaotic inflation models of the type
of the model m2 φ2 /2 [12], and multi-field, or hybrid models [23]. This classification is
incomplete, but still rather helpful.
In the simplest versions of chaotic inflation, the stage of preheating is generally dominated
by parametric resonance, although there are parameter ranges where this can not occur
[14]. In [15] it was shown that tachyonic preheating dominates the preheating phase in
hybrid models of inflation. New inflation in this respect occupies an intermediate position
between chaotic inflation and hybrid inflation: If spontaneous symmetry breaking in this
scenario is very large, reheating occurs due to parametric resonance and perturbative decay.
However, for the models with spontaneous symmetry breaking at or below the GUT scale,
φ ≪ 10−2 Mp , preheating occurs due to a combination of tachyonic preheating and parametric
resonance. The resulting effect is very strong, so that the homogeneous mode of the inflaton
field typically decays within few oscillations [16].
A detailed investigation of preheating usually requires lattice simulations, which can be
achieved following [17, 18]. Note that preheating is not the last stage of reheating; it is
followed by a period of turbulence [19], by a much slower perturbative decay described by
the methods developed in [13], and by eventual thermalization.

VII. INFLATION AND OBSERVATIONS

Inflation is not just an interesting theory that can resolve many difficult problems of the
standard Big Bang cosmology. This theory made several predictions which can be tested by
cosmological observations. Here are the most important predictions:
1) The universe must be flat. In most models Ωtotal = 1 ± 10−4 .
2) Perturbations of metric produced during inflation are adiabatic.
3) Inflationary perturbations have nearly flat spectrum. In most inflationary models the
spectral index ns = 1 ± 0.2 (ns = 1 means totally flat).
4) The spectrum of inflationary perturbations should be slightly non-flat. (It is very
difficult to construct a model with ns = 1.)
5) These perturbations are gaussian.
6) Perturbations of metric could be scalar, vector or tensor. Inflation mostly produces
scalar perturbations, but it also produces tensor perturbations with nearly flat spectrum, and
it does not produce vector perturbations. There are certain relations between the properties
of scalar and tensor perturbations produced by inflation.
7) Inflationary perturbations produce specific peaks in the spectrum of CMB radiation.
(For a simple pedagogical interpretation of this effect see e.g. [54]; a detailed theoretical
description can be found in [55].)
It is possible to violate each of these predictions if one makes this theory sufficiently
complicated. For example, it is possible to produce vector perturbations of metric in the

16
FIG. 3: CMB data versus the predictions of one of the simplest inflationary models (thick yellow
line), according to [30].

models where cosmic strings are produced at the end of inflation, which is the case in some
versions of hybrid inflation. It is possible to have an open or closed inflationary universe, or
even a small periodic inflationary universe, it is possible to have models with nongaussian
isocurvature fluctuations with a non-flat spectrum. However, it is very difficult to do so,
and most of the inflationary models obey the simple rules given above.
It is not easy to test all of these predictions. The major breakthrough in this direction was
achieved due to the recent measurements of the CMB anisotropy. The latest results based
on the WMAP experiment, in combination with the Sloan Digital Sky Survey, are consistent
with predictions of the simplest inflationary models with adiabatic gaussian perturbations,
with Ω = 1.01 ± 0.02, and ns = 0.98 ± 0.03 [29, 30, 31], see Fig. 3.
There are still some question marks to be examined, such as an unexpectedly small
anisotropy of CMB at large angles [29] and possible correlations between low multipoles.
However, the observational status and interpretation of these effects is rather uncertain [56],
and it is quite significant that all proposed explanations of these observations are based on
inflationary cosmology, see e.g. [57].

VIII. ALTERNATIVES TO INFLATION?

Inflationary scenario is very versatile, and now, after 20 years of persistent attempts of
many physicists to propose an alternative to inflation, we still do not know any other way

17
to construct a consistent cosmological theory. Indeed, in order to compete with inflation a
new theory should make similar predictions and should offer an alternative solution to many
difficult cosmological problems. Let us look at these problems before starting a discussion.
1) Homogeneity problem. Before even starting investigation of density perturbations and
structure formation, one should explain why the universe is nearly homogeneous on the
horizon scale.
2) Isotropy problem. We need to understand why all directions in the universe are similar
to each other, why there is no overall rotation of the universe. etc.
3) Horizon problem. This one is closely related to the homogeneity problem. If different
parts of the universe have not been in a causal contact when the universe was born, why do
they look so similar?
4) Flatness problem. Why Ω ≈ 1? Why parallel lines do not intersect?
5) Total entropy problem. The total entropy of the observable part of the universe is
greater than 1087 . Where did this huge number come from? Note that the lifetime of a
closed universe filled with hot gas with total entropy S is S 2/3 × 10−43 seconds [11]. Thus S
must be huge. Why?
6) Total mass problem. The total mass of the observable part of the universe has mass
∼ 1060 Mp . Note also that the lifetime of a closed universe filled with nonrelativistic particles
of total mass M is MMP × 10−43 seconds. Thus M must be huge. But why?
7) Structure formation problem. If we manage to explain the homogeneity of the universe,
how can we explain the origin of inhomogeneities required for the large scale structure
formation?
8) Monopole problem, gravitino problem, etc.
This list is very long. That is why it was not easy to propose any alternative to inflation
even before we learned that Ω ≈ 1, ns ≈ 1, and that the perturbations responsible for
galaxy formation are mostly adiabatic, in agreement with the predictions of the simplest
inflationary models.
Despite this difficulty, there was always a tendency to announce that we have eventually
found a good alternative to inflation. This was the ideology of the models of structure
formation due to topological defects, or textures, which were often described as competitors
to inflation, see e.g. [58]. However, it was clear from the very beginning that these theories at
best could solve only one problem (structure formation) out of 8 problems mentioned above.
Therefore the true question was not whether one can replace inflation by the theory of cosmic
strings/textures, but whether inflation with cosmic strings/textures is better than inflation
without cosmic strings/textures. Recent observational data favor the simplest version of
inflationary theory, without topological defects, or with an extremely small (few percent)
admixture of the effects due to cosmic strings.
A similar situation emerged with the introduction of the ekpyrotic scenario [59]. In the
original version of this theory it was claimed that this scenario can solve all cosmological
problems without using the stage of inflation, i.e. without a prolonged stage of an accelerated
expansion of the universe, which was called in [59] “superluminal expansion.” However, this
original idea did not work [60, 61], and the idea to avoid “superluminal expansion” was

18
abandoned by the authors of [59]. A more recent version of this scenario, the cyclic scenario
[62], uses an infinite number of periods of “superluminal expansion”, i.e. inflation, in order
to solve the major cosmological problems. In this sense, the cyclic scenario is not a true
alternative to inflationary scenario, but its rather peculiar version. The main difference
between the usual inflation and the cyclic inflation, just as in the case of topological defects
and textures, is the mechanism of generation of density perturbations. However, since the
theory of density perturbations in cyclic inflation requires a solution of the cosmological
singularity problem [63, 64], it is difficult to say anything definite about it. The latest
attempts to do so (despite our inability to address the singularity problem) indicate that
the spectrum of metric perturbations produced in the cyclic scenario is incompatible with
observations [65, 66].
Thus at the moment it is hard to see any real alternative to inflationary cosmology; instead
of a competition between inflation and other ideas, we witness a competition between many
different models of inflationary theory.
This competition goes in several different directions. First of all, we must try to im-
plement inflation in realistic theories of fundamental interactions. But what do we mean
by ‘realistic?’ Here we have an interesting and even somewhat paradoxical situation. In
the absence of a direct confirmation of M/string theory and supergravity by high energy
physics experiments (which may change when we start receiving data from the LHC), the
definition of what is realistic becomes increasingly dependent on cosmology and the results
of the cosmological observations. In particular, one may argue that those versions of the
theory of all fundamental interactions that cannot describe inflation and the present stage
of acceleration of the universe are disfavored by observations.
On the other hand, not every theory which can lead to inflation does it in an equally good
way. Many inflationary models have been already ruled out be observations. This happened
long ago with such models as extended inflation [67] and the simplest versions of “natural
inflation” [68]. Recent data from WMAP and SDSS almost ruled out a particular version
of chaotic inflation with V (φ) ∼ φ4 [29, 30].
However, observations test only the last stages of inflation. In particular, they do not say
anything about the properties of the inflaton potential at V (φ) ∼> 10−10 M 4 . Thus there may
p
exist many different models which describe all observational data equally well. In order to
compare such models, one should not only compare their predictions with the results of the
cosmological observations, but also carefully examine whether these models are internally
consistent and whether it is possible to solve the problem of initial conditions for inflation
in these models. One should also try to find out the best way to implement inflationary
scenario in the context of realistic models of all fundamental interactions, including the
models based on supergravity and string theory.

IX. SHIFT SYMMETRY AND CHAOTIC INFLATION IN SUPERGRAVITY

Most of the existing inflationary models are based on the idea of chaotic initial conditions,
which is the trademark of the chaotic inflation scenario. In the simplest versions of chaotic
inflation scenario with the potentials V ∼ φn , the process of inflation occurs at φ > 1, in
Planck units. Meanwhile, there are many other models where inflation may occur at φ ≪ 1.

19
 There are several reasons why this difference may be important. First of all, some authors
argue that the generic expression for the effective potential can be cast in the form

m2 2 β 3 λ 4 X φ4+n
V (φ) = V0 + αφ + φ + φ + φ + λn , (21)
2 3 4 n Mp n

and then they assume that generically λn = O(1), see e.g. Eq. (128) in [72]. If this
assumption were correct, one would have little control over the behavior of V (φ) at φ > Mp .
Here we have written Mp explicitly, to expose the implicit assumption made in [72]. Why
do we write Mp in the denominator, instead of 1000Mp ? An intuitive reason is that quantum
gravity is non-renormalizable, so one should introduce a cut-off at momenta k ∼ Mp . This
is a reasonable assumption, but it does not imply validity of Eq. (21). Indeed, the constant
part of the scalar field appears in the gravitational diagrams not directly, but only via its
effective potential V (φ) and the masses of particles interacting with the scalar field φ. As a
n
result, the terms induced by quantum gravity effects are suppressed not by factors Mφp n , but
2
by factors MVp 4 and mMp(φ)
2 [11]. Consequently, quantum gravity corrections to V (φ) become
large not at φ > Mp , as one could infer from (21), but only at super-Planckian energy density,
or for super-Planckian masses. This justifies our use of the simplest chaotic inflation models.
The simplest way to understand this argument is to consider again the case where the
potential of the field φ is a constant, V = V0 . Then the theory has a shift symmetry,
φ → φ + c. This symmetry is not broken by perturbative quantum gravity corrections, so no
4+n
such terms as n λn φMp n are generated. This symmetry may be broken by nonperturbative
P

quantum gravity effects (wormholes? virtual black holes?), but such effects, even if they
exist, can be made exponentially small [69].
The idea of shift symmetry appears to be very fruitful in application to inflation; we will
return to it many times in this paper. However, in some cases the scalar field φ itself may
have physical (geometric) meaning, which may constrain the possible values of the fields
during inflation. The most important example is given by N = 1 supergravity.
The effective potential of the complex scalar field Φ in supergravity is given by the well-
known expression (in units Mp = 1):
h i
V = eK KΦ−1Φ̄ |DΦ W |2 − 3|W |2 . (22)

Here W (Φ) is the superpotential, Φ denotes the scalar component of the superfield Φ;
DΦ W = ∂W ∂Φ
+ ∂K
∂Φ
W . The kinetic term of the scalar field is given by KΦΦ̄ ∂µ Φ∂µ Φ̄. The stan-
dard textbook choice of the Kähler potential corresponding to the canonically normalized
fields Φ and Φ̄ is K = ΦΦ̄, so that KΦΦ̄ = 1.
This immediately reveals a problem: At Φ > 1 the potential is extremely steep. It
2
blows up√as e|Φ| , which makes it very difficult to realize chaotic inflation in supergravity
at φ ≡ 2|Φ| > 1. Moreover, the problem persists even at small φ. If, for example, one
considers the simplest case when there are many other scalar fields in the theory and the
superpotential does not depend on the inflaton field φ, then Eq. (22) implies that at φ ≪ 1
the effective mass of the inflaton field is m2φ = 3H 2. This violates the condition m2φ ≪ H 2
required for successful slow-roll inflation (so-called η-problem).

20
 The major progress in SUGRA inflation during the last decade was achieved in the context
of the models of the hybrid inflation type, where inflation may occur at φ ≪ 1. Among the
best models are the F-term inflation, where different contributions to the effective mass
term m2φ cancel [70], and D-term inflation [71], where the dangerous term eK does not affect
the potential in the inflaton direction. A detailed discussion of various versions of hybrid
inflation in supersymmetric theories can be found in [72]. A recent version of this scenario,
P-term inflation, which unifies F-term and D-term models, was proposed in [73].
However, hybrid inflation occurs only on a relatively small energy scale, and many of its
versions do not lead to eternal inflation. Therefore it would be nice to obtain inflation in a
context of a more general class of supergravity models.
This goal seemed very difficult to achieve; it took almost 20 years to find a natural
realization of chaotic inflation model in supergravity. Kawasaki, Yamaguchi and Yanagida
suggested to take the Kähler potential
1
K = (Φ + Φ̄)2 + X X̄ (23)
2
of the fields Φ and X, with the superpotential mΦX [74].
At the first glance, this Kähler potential may seem somewhat unusual. However, it can be
obtained from the standard Kähler potential K = ΦΦ̄ + X X̄ by adding terms Φ2 /2 + Φ̄2 /2,
which do not give any contribution to the kinetic term of the scalar fields KΦΦ̄ ∂µ Φ∂µ Φ̄. In
other words, the new Kähler potential, just as the old one, leads to canonical kinetic terms
for the fields Φ and X, so it is as simple and legitimate as the standard textbook Kähler
potential. However, instead of the U(1) symmetry with respect to rotation of the field Φ in
the complex plane, the new Kähler potential has a shift symmetry; it does not depend on
the imaginary part of the field Φ. The shift symmetry is broken only by the superpotential.
This leads to a profound change of the potential (22): the dangerous term eK continues
growing exponentially in the direction (Φ + Φ̄), but it remains constant in the direction
(Φ − Φ̄). Decomposing the complex field Φ into two real scalar fields, Φ = √12 (η + iφ), one
can find the resulting potential V (φ, η, X) for η, |X| ≪ 1:

m2 2
V = φ (1 + η 2 ) + m2 |X|2. (24)
2
This potential has a deep valley, with a minimum at η = X = 0. Therefore the fields η and
X rapidly fall down towards η = X = 0, after which the potential for the field φ becomes
2
V = m2 φ2 . This provides a very simple realization of eternal chaotic inflation scenario in
supergravity [74]. This model can be extended to include theories with different power-law
potentials, or models where inflation begins as in the simplest versions of chaotic inflation
scenario, but ends as in new or hybrid inflation, see e.g. [75, 76].
It is amazing that for almost 20 years nothing but inertia was keeping us from using
the version of the supergravity which was free from the famous η problem. As we will see
shortly, the situation with inflation in string theory is very similar, and may have a similar
resolution.

21
X. TOWARDS INFLATION IN STRING THEORY

A. de Sitter vacua in string theory

For a long time, it seemed rather difficult to obtain inflation in M/string theory. The
main problem here was the stability of compactification of internal dimensions. For exam-
ple, ignoring non-perturbative effects to be discussed below, a typical effective potential of
the effective 4d theory obtained by compactification in string theory of type IIB can be
represented in the following form:
√ √
2ϕ− 6ρ
V (ϕ, ρ, φ) ∼ e Ṽ (φ) (25)
Here ϕ and ρ are canonically normalized fields representing the dilaton field and the volume
of the compactified space; φ stays for all other fields, including the inflaton field.
If ϕ and ρ were constant, then the potential Ṽ√
(φ)√could drive inflation. However, this
2ϕ− 6ρ
does not happen because of the steep exponent e , which rapidly pushes the dilaton
field ϕ to −∞, and the volume modulus ρ to +∞. As a result, the radius of compactification
becomes infinite; instead of inflating, 4d space decompactifies and becomes 10d.
Thus in order to describe inflation one should first learn how to stabilize the dilaton and
the volume modulus. The dilaton stabilization was achieved in [77]. The most difficult
problem was to stabilize the volume. The solution of this problem was found in [78] (KKLT
construction). It consists of two steps.
First of all, due to a combination of effects related to warped geometry of the compactified
space and nonperturbative effects calculated directly in 4d (instead of being obtained by
compactification), it was possible to obtain a supersymmetric AdS minimum of the effective
potential for ρ. In the original version of the KKLT scenario, it was done in the theory with
the Kähler potential
K = −3 log(ρ + ρ̄), (26)
and with the nonperturbative superpotential of the form
W = W0 + Ae−aρ , (27)
with a = 2π/N. The corresponding effective potential for the complex field ρ = σ + iα had
a minimum at finite, moderately large values of the volume modulus field σ0 , which fixed
the volume modulus in a state with a negative vacuum energy. Then an anti-D3 brane with
the positive energy ∼ σ −2 was added. This addition uplifted the minimum of the potential
to the state with a positive vacuum energy, see Fig. 4.
Instead of adding an anti-D3 brane, which explicitly breaks supersymmetry, one can add
a D7 brane with fluxes. This results in the appearance of a D-term which has a similar
dependence on ρ, but leads to spontaneous supersymmetry breaking [79]. In either case, one
ends up with a metastable dS state which can decay by tunneling and formation of bubbles
of 10d space with vanishing vacuum energy density. The decay rate is extremely small [78],
so for all practical purposes, one obtains an exponentially expanding de Sitter space with
the stabilized volume of the internal space1 .

1
It is also possible to find de Sitter solutions in noncritical string theory [80].

22
 V
2

100 150 200 250 Σ

-1

-2

FIG. 4: KKLT potential as a function of σ = Re ρ. Thin green line corresponds to AdS stabilized
potential for W0 = −10−4 , A = 1, a = 0.1. Dashed line shows the additional term, which appears
either due to the contribution of a D3 brane or of a D7 brane. Thick black line shows the resulting
potential with a very small but positive value of V in the minimum. All potentials are shown
multiplied by 1015 .

B. Inflation in string theory

There are two different versions of string inflation. In the first version, which we will call
modular inflation, the inflaton field is associated with one of the moduli, the scalar fields
which are already present in the KKLT construction. In the second version, the inflaton
is related to the distance between branes moving in the compactified space. (This scenario
should not be confused with inflation in the brane world scenario [81, 82]. This is a separate
interesting subject, which we are not going to discuss in this paper.)

1. Modular inflation

An example of the KKLT-based modular inflation is provided by the racetrack inflation

model of Ref. [83]. It uses a slightly more complicated superpotential

W = W0 + Ae−aρ + Be−bρ . (28)

The potential of this theory has a saddle point as a function of the real and the complex part
of the volume modulus: It has a local minimum in the direction Re ρ, which is simultaneously
a very flat maximum with respect to Im ρ. Inflation occurs during a slow rolling of the field
Im ρ away from this maximum (i.e. from the saddle point). The existence of this regime
requires a significant fine-tuning of parameters of the superpotential. However, in the context
of the string landscape scenario describing from 10100 to 101000 different vacua (see below),
this may not be such a big issue. A nice feature of this model is that it does not require
adding any new branes to the original KKLT scenario, i.e. it is rather economical. Another
attractive feature of this model is the existence of the regime of eternal inflation near the

23
saddle point.

160
X 140

120

100

1.5

V 0.5

0
20

Y -20

FIG. 5: Plot for the potential in the racetrack model (rescaled by 1016 ). Here X stays for σ = Re ρ
and Y stays for α = Im ρ. Inflation begins in a vicinity of the saddle point at Xsaddle = 123.22,
Ysaddle = 0. Units are Mp = 1.

2. Brane inflation and shift symmetry

During the last few years there were many suggestions how to obtain hybrid inflation in
string theory by considering motion of branes in the compactified space, see [84, 85] and
references therein. The main problem of all of these models was the absence of stabilization
of the compactified space. Once this problem was solved for dS space [78], one could try
to revisit these models and develop models of brane inflation compatible with the volume
stabilization.
The first idea [86] was to consider a pair of D3 and anti-D3 branes in the warped geometry
studied in [78]. The role of the inflaton field φ could be played by the interbrane separation.
A description of this situation in terms of the effective 4d supergravity involved Kähler
potential
K = −3 log(ρ + ρ̄ − k(φ, φ̄)), (29)
where the function k(φ, φ̄) for the inflaton field φ, at small φ, was taken in the simplest form
k(φ, φ̄) = φφ̄. If one makes the simplest assumption that the superpotential does not depend
on φ, then the φ dependence of the potential (22) comes from the term eK = (ρ + ρ̄ − φφ̄)−3 .
Expanding this term near the stabilization point ρ = ρ0 , one finds that the inflaton field
has a mass m2φ = 2H 2 . Just like the similar relation m2φ = 3H 2 in the simplest models of
supergravity, this is not what we want for inflation.

24
 One way to solve this problem is to consider φ-dependent superpotentials. By doing so,
one may fine-tune m2φ to be O(10−2)H 2 in a vicinity of the point where inflation occurs [86].
Whereas fine-tuning is certainly undesirable, in the context of string cosmology it may not
be a serious drawback. Indeed, if there exist many realizations of string theory [53], then one
might argue that all realizations not leading to inflation can be discarded, because they do
not describe a universe in which we could live. Meanwhile, those non-generic realizations,
which lead to eternal inflation, describe inflationary universes with an indefinitely large
and ever-growing volume of inflationary domains. This makes the issue of fine-tuning less
problematic. A more detailed investigation of this issue can be found in [87].
Can we avoid fine-tuning altogether? One of the possible ideas is to find theories with
some kind of shift symmetry. Another possibility is to construct something like D-term
inflation, where the flatness of the potential is not spoiled by the term eK . Both of these
ideas were explored in Ref. [88] based on the model of D3/D7 inflation in string theory [89].
In this model the Kähler potential is given by
1
K = −3 log(ρ + ρ̄) − (φ − φ̄)2 , (30)
2
and superpotential depends only on ρ. The role of the inflaton field is played by the field
s = Re φ, which represents the distance between the D3 and D7 branes. The shift symmetry
s → s + c in this model is related to the requirement of unbroken supersymmetry of branes
in a BPS state.
The effective potential with respect to the field ρ in this model coincides with the KKLT
potential [78, 79]. The potential is exactly flat in the direction of the inflaton field s, see
Fig. 6, until one adds a hypermultiplet of other fields φ± , which break this flatness due
to quantum corrections and produce a logarithmic potential for the field s. The resulting
potential with respect to the fields s and φ± is very similar to the potential of D-term hybrid
inflation [71].

10
5
0 s
-5

3 -10

2
V
1

0 150 200 250 300

100
σ
FIG. 6: Inflationary potential as a function of the compactification modulus σ and the inflaton field
s. This potential is exactly flat in the inflaton direction due to shift symmetry, which is violated
only by radiative corrections.

25
 During inflation, φ± = 0, and the field s slowly rolls down to its smaller values. When
it becomes sufficiently small, the theory becomes unstable with respect to generation of the
field φ+ , see Fig. 7. The fields s and φ+ roll down to the KKLT minimum, and inflation
ends.

0.6

0.4

V 0.2 2

0
-2 1.5

-1

0
1
s
Re Φ+ 1
0.5

2 0

FIG. 7: Inflationary potential as a function of the inflaton field s and Re φ+ . In the beginning, the
field s rolls along the valley φ+ = 0, and then it falls down to the KKLT minimum.

One may wonder whether the shift symmetry is just a condition which we want to impose
on the theory in order to get inflation [90], or an unavoidable property of the theory, which
remains valid even after the KKLT volume stabilization. The answer to this question appears
to be model-dependent. It was shown in [91, 92] that in a certain class of models, including
D3/D7 models [88, 89, 93], the shift symmetry of the effective 4d theory is not an assumption
but an unambiguous consequence of the underlying mathematical structure of the theory.
This may allow us to obtain a natural realization of inflation in string theory2 . For the latest
developments in D3/D7 inflation see [95, 96, 97].
All inflationary models discussed above were formulated in the context of Type IIB string
theory with the KKLT stabilization. A discussion of the possibility to obtain inflation in
the heterotic string theory with stable compactification can be found in [98, 99].
Finally, we should mention that making the effective potential flat is not the only way to
achieve inflation. There are some models with nontrivial kinetic terms where inflation may
occur even without any potential [100]. One may also consider models with steep potentials
but with anomalously large kinetic terms for the scalar fields see e.g. [101]. In application
to string theory, such models, called ‘DBI inflation,’ were developed in [102].

2
This issue was recently debated in [94], but the brane configuration in the model discussed there (one
D3 brane interacting with a stack of many coincident D7 branes) is quite different from the configuration
considered in D3/D7 scenario of [88, 89, 92].

26
XI. SCALE OF INFLATION, SCALE OF SUSY BREAKING AND THE GRAV-
ITINO MASS

So far, we did not discuss relation of the new class of models with particle phenomenology.
This relation is rather unexpected and may impose strong constraints either on particle
phenomenology or on inflationary models: Recently it was shown that the Hubble constant
and the inflaton mass in the simplest models based on the KKLT mechanism with the
superpotential (27) are always smaller than the gravitino mass [103],
< m3/2 .
H∼ (31)
The reason for this constraint is that adding a large vacuum energy density to the KKLT
potential may destabilize it, see Fig. 8.

V
4

100 150 200 250 Σ

FIG. 8: The lowest curve with dS minimum is the one from the KKLT model. The height of
the barrier in this potential is of the order m23/2 . The second line shows the σ-dependence of
the D3/D7 inflationary potential with the term Vinfl = V (s,φσ3
±)
added to the KKLT potential; it
originates from fluxes on D7 brane. The top curve shows that when the inflationary potential
becomes too large, the barrier disappears, and the internal space decompactifies. This explains the
origin of the constraint H <
∼ m3/2 . This constraint appears in a broad class of inflationary models
based on the simplest KKLT potential.

Since in the slow-roll models the inflaton mass must be much smaller than H, its mass
must be much smaller than m3/2 . Therefore if one insists on the standard SUSY phenomenol-
ogy assuming that the gravitino mass is smaller than O(1) TeV, one will need to find realistic
particle physics model where the nonperturbative string theory dynamics occurs at the LHC
scale (!!!), and inflation occurs at least 30 orders of magnitude below the Planck energy den-
sity. Such models are possible, but their parameters should be substantially different from
the parameters used in all presently existing models of string theory inflation.
There are several different ways to address this problem. First of all, one may try to
construct realistic particle physics models with superheavy gravitinos [104, 105]. Another
possibility is to consider models with the racetrack superpotential (28) and find such param-
eters that the minimum of the potential even before the uplifting will occur at vanishingly
small energy density. This goal was achieved in [103].

27
 V
-9
8·10

-9
6·10

-9
4·10

-9
2·10

20 40 60 80 100 Σ
-9
-2·10

FIG. 9: The potential in the theory (28) for A = 1, B = −5, a = 2π/100, b = 2π/50, W0 = −0.05.
A Minkowski minimum at V = 0 stabilizes the volume at σ0 ≈ 37. AdS vacuum at V < 0 occurs
at σ0 ∼ 66. There is a barrier protecting the Minkowski minimum, with the height VB ∼ 5 × 10−9 ,
in units of the Planck energy density. The height of the barrier is not correlated with the gravitino
mass, which vanishes if the system is trapped in Minkowski vacuum.

An intriguing property of the new version of the KKLT construction is that the minimum
of the potential prior to the uplifting corresponds to a supersymmetric Minkowski vacuum.
The gravitino mass in this minimum (and the magnitude of SUSY breaking there) can be
<
vanishingly small as compared to all other parameters of the model, and the constraint H ∼
m3/2 disappears. Anywhere outside this minimum the gravitino mass and the magnitude
of SUSY breaking is extremely large. This means that the Minkowski minimum shown in
Fig. 9 is a point of enhanced symmetry, which is a trapping point for the motion of the
moduli fields, in accordance with [106]. This fact may increase the probability that among
all possible minima in the string theory landscape, the minimum with a low-scale SUSY
breaking is dynamically preferred.
The original KKLT model with the potential shown by the black line in Fig. 4, and its
new version with the potential shown in Fig. 9, represent the two presently available models
of dark energy based on string theory. In both cases the universe experiences constant ac-
celeration determined by the small positive vacuum energy in the minimum of the potential.
In both cases, these vacuum states are expected to be metastable, with an extremely long
lifetime. These states may decay by forming expanding bubble of a new phase. It would be
very hard to detect such a bubble and even harder to report the results of the observations,
since the bubble walls move with the speed approaching the speed of light, so an observer
would see the bubble wall at the moment when it hits him.
The difference between these two versions is that the interior of the bubble in the simplest
version of the KKLT scenario represents a 10D Minkowski space, whereas an interior of the
bubble in its modified version shown in Fig. 9 will correspond to a very rapidly collapsing
open universe with a negative energy density.
The only good thing about this grim picture is that in an eternally existing self-
reproducing universe there will always remain some parts of space where somebody else
will enjoy life.

28
XII. INITIAL CONDITIONS FOR THE LOW-SCALE INFLATION AND TOPOL-
OGY OF THE UNIVERSE

One of the advantages of the simplest chaotic inflation scenario is that inflation may
begin in the universe immediately after its creation at the largest possible energy density
Mp4 , of a smallest possible size (Planck length), with the smallest possible mass M ∼ Mp and
with the smallest possible entropy S = O(1). This provides a true solution to the flatness,
horizon, homogeneity, mass and entropy problems [11].
Meanwhile, in the new inflation scenario, inflation occurs on the mass scale 3 orders of
magnitude below Mp , when the total size of the universe was very large. If, for example,
the universe is closed, its total mass at the beginning of new inflation must be greater than
106 Mp , and its total entropy must be greater than 109 . In other words, in order to explain
why the entropy of the universe at present is greater than 1087 one should assume that it
was extremely large from the very beginning. This does not look like a real solution of the
entropy problem. A similar problem exists for many of the models advocated in [22, 107].
Finally, in cyclic inflation, the process of exponential expansion of the universe occurs only
if the total mass of the universe is greater than its present mass M ∼ 1060 Mp and its total
entropy is greater than 1087 . This scenario does not solve the flatness, mass and entropy
problems.
Is it at all possible to solve the problem of initial conditions for the low scale inflation?
The answer to this question appears to be positive though perhaps somewhat unexpected:
One should consider a compact flat or open universe with nontrivial topology (usual flat or
open universes are infinite). The universe may initially look like a nearly homogeneous torus
of a Planckian size containing just one or two photons or gravitons. It can be shown that
such a universe continues expanding and remains homogeneous until the onset of inflation,
even if inflation occurs only at a very low scale [108, 109, 110, 111, 112].
Consider for simplicity the flat compact universe having the topology of a torus, S13 ,

ds2 = dt2 − a2i (t) dx2i (32)

with identification xi + 1 = xi for each of the three dimensions. Consider for simplicity
the case a1 = a2 = a3 = a(t). In this case the curvature of the universe and the Einstein
equations written in terms of a(t) will be the same as in the infinite flat Friedmann universe
with metric ds2 = dt2 − a2 (t) dx2 . In our notation, the scale factor a(t) is equal to the size
of the universe in Planck units Mp−1 = 1.
Let us assume, that at the Planck time tp ∼ Mp−1 = 1 the universe was radiation domi-
nated, V ≪ T 4 = O(1). Let us also assume that at the Planck time the total size of the box
was Planckian, a(tp ) = O(1). In such case the whole universe initially contained only O(1)
relativistic particles such as photons or gravitons, so that the total entropy of the whole
universe was O(1).
√
The size of the universe dominated by relativistic particles was growing as a(t) ∼ t,
whereas the mean free path of the gravitons was growing as H −1 ∼ t. If the initial size of
the universe was O(1), then at the time t ≫ 1 each particle (or a gravitational perturbation
of metric) within one cosmological time would run all over the torus many times, appearing
in all of its parts with nearly equal probability. This effect, called “chaotic mixing,” should

29
lead to a rapid homogenization of the universe [109, 110]. Note, that to achieve a modest
degree of homogeneity required for inflation to start when the density of ordinary matter
drops down, we do not even need chaotic mixing. Indeed, density perturbations do not grow
in a universe dominated by ultrarelativistic particles if the size of the universe is smaller
than H −1 . This is exactly what happens in our model. Therefore the universe should remain
relatively homogeneous until the thermal energy density drops below V and inflation begins.
Thus we see that in this scenario, just as in the simplest chaotic inflation scenario, inflation
begins if we had a sufficiently homogeneous domain of a smallest possible size (Planck scale),
with the smallest possible mass (Planck mass), and with the total entropy O(1). The only
additional requirement is that this domain should have identified sides, in order to make a
flat or open universe compact. We see no reason to expect that the probability of formation
of such domains is strongly suppressed.
One can come to a similar conclusion from a completely different point of view. Investi-
gation of the quantum creation of a closed or an infinite open inflationary universe with a
small value of the vacuum energy shows that this process is forbidden at the classical level,
and therefore it occurs only due to tunneling. As a result, the probability of this process is
exponentially suppressed [114, 115, 116]. Meanwhile, creation of the flat or open universe
is possible without any need for the tunneling, and therefore there is no exponential sup-
pression for the probability of quantum creation of a topologically nontrivial compact flat
or open inflationary universe [108, 111, 112, 113].
These results suggest that if inflation can occur only much below the Planck density, then
the topologically nontrivial flat or open universes should be much more probable than the
standard Friedmann universes described in every textbook on cosmology.3
Until now, we discussed creation of compact universes which have approximately equal
size in all directions. If at the Planck time our universe was of a Planck size in some
directions, but it was much larger in some other directions, then it consisted of many
causally disconnected Planck size regions. This results in a particular version of the hori-
zon/homogeneity problem: The probability that the universe was homogeneous in all of
these causally disconnected regions should be exponentially small [113].
In application to the initial conditions in the 10D universe described by string theory,
this suggests that it is more natural to start with the universe which would have similar
initial size in all 9 spatial directions. In terms of the KKLT potential, this implies that the
initial value of the volume modulus σ should be very small, so that V (σ) = O(1). But then
the field σ will rapidly fall down. It can easily roll over the very law KKLT barrier, and
continue moving to infinitely large σ.
One of the recent attempts to solve this problem was based on the dynamical slow-down
of the field σ in the universe filled with radiation [117]. But this mechanism typically works
only if the initial value of the effective potential is several orders of magnitude smaller than
O(1).

3
Note, however, that unless one fine-tunes the parameters of the theory and the shape of the inflationary
potential, inflation makes the observable part of the universe so large that its topology should not affect
observational data.

30
 It is not quite clear whether this is a real problem because those parts of the universe
which start at large V (σ), become ten-dimensional, so we cannot live there. We will live
in the parts of the universe which started at smaller V (σ), even if it may seem improbable
from the point of view of initial conditions. This is similar to the fact that we live at the 2D
surface of a relatively small planet instead of living in the vast but empty interstellar space.
One can also argue [43] that eternal inflation may alleviate some of the problems of the
problem of initial conditions for the low-scale inflation. Note, however, that eternal inflation,
which naturally occurs in the simplest versions of chaotic inflation, in new inflation, and in
the racetrack scenario [83], may not exist in many versions of string theory inflation of the
hybrid type, such as the models of [86, 88, 97]. Of course, models of low-scale non-eternal
inflation are still much better than the models with no inflation at all, but I do not think
that we should settle for the second-best. An indeed we should not, if we are able to combine
the slow-roll inflation, which ends in our vacuum, with the eternal inflation in a collection
of different metastable dS states in the string theory landscape.

XIII. ETERNAL INFLATION AND THE STRING THEORY LANDSCAPE

Even though we are still at the very first stages of implementing inflation in string theory,
it is very tempting to speculate about possible generic features and consequences of such a
construction.
First of all, KKLT construction shows that the vacuum energy after the volume stabiliza-
tion is a function of many different parameters in the theory. One may wonder how many
different choices do we actually have. There were many attempts to investigate this issue.
For example, many years ago it was argued [118] that there are infinitely many AdS4 × X7
vacua of D=11 supergravity. An early estimate of the total number of different 4d string
vacua gave the number 101500 [119]. At present we are more interested in counting different
flux vacua [52, 53], where the possible numbers, depending on specific assumptions, may
vary in the range from 1020 to 101000 . Some of these vacuum states with positive vacuum
energy can be stabilized using the KKLT approach. Each of such states will correspond to
a metastable vacuum state. It decays within a cosmologically large time, which is, however,
2
smaller than the ‘recurrence time’ eS(φ) , where S(φ) = V24π
(φ)
is the entropy of dS space with
the vacuum energy density V (φ) [78].
But this is not the whole story; old inflation does not describe our world. In addition to
these metastable vacuum states, there should exist various slow-roll inflationary solutions,
where the properties of the system practically do not change during the cosmological time
H −1 . It might happen that such states, corresponding to flat directions in the string the-
ory landscape, exist not only during inflation in the very early universe, but also at the
present stage of the accelerated expansion of the universe. This would simplify obtaining an
anthropic solution of the cosmological constant problem along the lines of [52, 120].
If the slow-roll condition V ′′ ≪ V is satisfied all the way from one dS minimum of the
effective potential to another, then one can show, using stochastic approach to inflation, that
the probability to find the field φ at any of these minima, or at any given point between them,
is proportional to eS(φ) . In other words, the relative probability to find the field taking some
value φ1 as compared to some other value φ0 , is proportional to e∆S = eS(φ1 )−S(φ0 ) [78, 116].

31
One may argue, using Euclidean approach, that this simple thermodynamic relation should
remain valid for the relative probability to find a given point in any of the metastable dS
vacua, even if the trajectory between them does not satisfy the slow-roll condition m2 ≪ H 2
[121, 122, 123, 124].
The resulting picture resembles eternal inflation in the old inflation scenario. However,
now we have an incredibly large number of false vacuum states, plus some states which
may allow slow-roll inflation. Once inflation begins, different parts of the universe start
jumping from one of these vacuum states to another, so that the universe becomes divided
into indefinitely many regions with all possible laws of low-energy physics corresponding to
different 4D vacua of string theory [11].
The best inflationary scenario would describe a slow-roll eternal inflation starting at the
maximal possible energy density (minimal dS entropy). It would be almost as good to have
a low-energy slow-roll eternal inflation. Under certain conditions, such regimes may exist in
string theory [86]. Here we are going to discuss a different but equally interesting possibility.
Multi-level eternal inflation of the old inflation type, which appears in string theory in the
context of the KKLT construction, may be very useful being combined with the slow-roll
inflation, even if the slow-roll inflation by itself is not eternal. We will give a particular
example, which is based on the ideas developed in Ref. [125].
Suppose we have two noninteracting scalar fields: field φ with the potential of the old
inflation type, and field s with the potential which may lead to a slow-roll inflation. Let us
consider the worst case scenario: the slow-roll inflation occurs only on low energy scale, and
it is not eternal.
Let us assume that the Hubble constant at the stage of old inflation is much greater
than the curvature of the potential which drives the slow-roll inflation. This is a natural
assumption, considering huge number of possible dS states, and the presumed smallness
of energy scale of the slow-roll inflation. The validity of this assumption is ensured if the
shift symmetry with respect to the slow-rolling inflaton field s is preserved in many different
string theory vacua. This may be the case in the D3/D7 scenario, where the shift symmetry
is related to the isometry of the compactified space but not to the particular values of the
fluxes [88, 92, 97]. In this case large inflationary fluctuations of the field s will be generated
during eternal old inflation. These fluctuations will push the field s from the minimum of
its effective potential and will give it different values in different exponentially large parts of
the universe. When old inflation ends, there will be many practically homogeneous parts of
the universe where the field s will take values corresponding to good initial conditions for a
slow-roll inflation.
Thus, the existence of many different dS vacua in string theory leads to the regime of
eternal inflation. This regime may help us to solve the problem of initial conditions for the
slow-roll inflation even in the models where the slow-roll inflation by itself is not eternal and
would occur only on a small energy scale.

32
 Acknowledgments

I am grateful to the organizers of the SLAC Summer School “Nature’s Greatest Puzzles,”
of the conference Cosmo04 in Toronto, of the VI Mexican School on Gravitation, and of the
XXII Texas Symposium on Relativistic Astrophysics. I would like to thank Lev Kofman,
Renata Kallosh and Slava Mukhanov for valuable comments. This work was supported by
NSF grant PHY-0244728.

[1] A. A. Starobinsky, “Spectrum Of Relict Gravitational Radiation And The Early State Of
The Universe,” JETP Lett. 30, 682 (1979) [Pisma Zh. Eksp. Teor. Fiz. 30, 719 (1979)];
A. A. Starobinsky, “A New Type Of Isotropic Cosmological Models Without Singularity,”
Phys. Lett. B 91, 99 (1980).
[2] V. F. Mukhanov and G. V. Chibisov, “Quantum Fluctuation And ‘Nonsingular’ Universe,”
JETP Lett. 33, 532 (1981) [Pisma Zh. Eksp. Teor. Fiz. 33, 549 (1981)].
[3] A. H. Guth, “The Inflationary Universe: A Possible Solution To The Horizon And Flatness
Problems,” Phys. Rev. D 23, 347 (1981).
[4] D. A. Kirzhnits and A. D. Linde, “Symmetry Behavior In Gauge Theories,” Annals Phys.
101, 195 (1976).
[5] S. W. Hawking, I. G. Moss and J. M. Stewart, “Bubble Collisions In The Very Early Uni-
verse,” Phys. Rev. D 26, 2681 (1982).
[6] A. H. Guth and E. J. Weinberg, “Could The Universe Have Recovered From A Slow First
Order Phase Transition?,” Nucl. Phys. B 212, 321 (1983).
[7] A. D. Linde, “A New Inflationary Universe Scenario: A Possible Solution Of The Horizon,
Flatness, Homogeneity, Isotropy And Primordial Monopole Problems,” Phys. Lett. B 108,
389 (1982); A. D. Linde, “Coleman-Weinberg Theory And A New Inflationary Universe Sce-
nario,” Phys. Lett. B 114, 431 (1982); A. D. Linde, “Temperature Dependence Of Coupling
Constants And The Phase Transition In The Coleman-Weinberg Theory,” Phys. Lett. B 116,
340 (1982); A. D. Linde, “Scalar Field Fluctuations In Expanding Universe And The New
Inflationary Universe Scenario,” Phys. Lett. B 116, 335 (1982).
[8] A. Albrecht and P. J. Steinhardt, “Cosmology For Grand Unified Theories With Radiatively
Induced Symmetry Breaking,” Phys. Rev. Lett. 48, 1220 (1982).
[9] S. W. Hawking, “The Development Of Irregularities In A Single Bubble Inflationary Uni-
verse,” Phys. Lett. B 115, 295 (1982); A. A. Starobinsky, “Dynamics Of Phase Transition In
The New Inflationary Universe Scenario And Generation Of Perturbations,” Phys. Lett. B
117, 175 (1982); A. H. Guth and S. Y. Pi, “Fluctuations In The New Inflationary Universe,”
Phys. Rev. Lett. 49, 1110 (1982); J. M. Bardeen, P. J. Steinhardt and M. S. Turner, “Spon-
taneous Creation Of Almost Scale - Free Density Perturbations In An Inflationary Universe,”
Phys. Rev. D 28, 679 (1983).
[10] V. F. Mukhanov, “Gravitational Instability Of The Universe Filled With A Scalar Field,”
JETP Lett. 41, 493 (1985) [Pisma Zh. Eksp. Teor. Fiz. 41, 402 (1985)]; V. F. Mukhanov,
H. A. Feldman and R. H. Brandenberger, “Theory Of Cosmological Perturbations,” Phys.
Rept. 215, 203 (1992); V. F. Mukhanov, Physical Foundations of Cosmology, Cambridge
University Press, to be published.

33
[11] A.D. Linde, Particle Physics and Inflationary Cosmology (Harwood, Chur, Switzerland,
1990).
[12] A. D. Linde, “Chaotic Inflation,” Phys. Lett. B 129, 177 (1983).
[13] A. D. Dolgov and A. D. Linde, “Baryon Asymmetry In Inflationary Universe,” Phys. Lett. B
116, 329 (1982); L. F. Abbott, E. Farhi and M. B. Wise, “Particle Production In The New
Inflationary Cosmology,” Phys. Lett. B 117, 29 (1982).
[14] L. Kofman, A. D. Linde and A. A. Starobinsky, “Reheating after inflation,” Phys. Rev.
Lett. 73, 3195 (1994) [arXiv:hep-th/9405187]; L. Kofman, A. D. Linde and A. A. Starobin-
sky, “Towards the theory of reheating after inflation,” Phys. Rev. D 56, 3258 (1997)
[arXiv:hep-ph/9704452].
[15] G. N. Felder, J. Garcia-Bellido, P. B. Greene, L. Kofman, A. D. Linde and I. Tkachev,
“Dynamics of symmetry breaking and tachyonic preheating,” Phys. Rev. Lett. 87, 011601
(2001) [arXiv:hep-ph/0012142]; G. N. Felder, L. Kofman and A. D. Linde, “Tachyonic insta-
bility and dynamics of spontaneous symmetry breaking,” Phys. Rev. D 64, 123517 (2001)
[arXiv:hep-th/0106179].
[16] M. Desroche, G. N. Felder, J. M. Kratochvil and A. Linde, “Preheating in new inflation,”
arXiv:hep-th/0501080.
[17] S. Y. Khlebnikov and I. I. Tkachev, “Classical decay of inflaton,” Phys. Rev. Lett. 77, 219
(1996) [arXiv:hep-ph/9603378]; S. Y. Khlebnikov and I. I. Tkachev, “Resonant decay of Bose
condensates,” Phys. Rev. Lett. 79, 1607 (1997) [arXiv:hep-ph/9610477].
[18] G. N. Felder and I. Tkachev, “LATTICEEASY: A program for lattice simulations of scalar
fields in an expanding universe,” arXiv:hep-ph/0011159.
[19] G. N. Felder and L. Kofman, “The development of equilibrium after preheating,” Phys.
Rev. D 63, 103503 (2001) [arXiv:hep-ph/0011160]; R. Micha and I. I. Tkachev, “Turbulent
thermalization,” Phys. Rev. D 70, 043538 (2004) [arXiv:hep-ph/0403101].
[20] B. A. Ovrut and P. J. Steinhardt, “Supersymmetry And Inflation: A New Approach,” Phys.
Lett. B 133, 161 (1983); B. A. Ovrut and P. J. Steinhardt, “Inflationary Cosmology And
The Mass Hierarchy In Locally Supersymmetric Theories,” Phys. Rev. Lett. 53, 732 (1984);
B. A. Ovrut and P. J. Steinhardt, “Locally Supersymmetric Cosmology And The Gauge Hi-
erarchy,” Phys. Rev. D 30, 2061 (1984); B. A. Ovrut and P. J. Steinhardt, “Supersymmetric
Inflation, Baryon Asymmetry And The Gravitino Problem,” Phys. Lett. B 147, 263 (1984).
[21] A. D. Linde, “Primordial Inflation Without Primordial Monopoles,” Phys. Lett. B 132 (1983)
317.
[22] L. Boubekeur and D. H. Lyth, “Hilltop inflation,” arXiv:hep-ph/0502047.
[23] A. D. Linde, “Axions in inflationary cosmology,” Phys. Lett. B 259, 38 (1991); A. D. Linde,
“Hybrid inflation,” Phys. Rev. D 49, 748 (1994) [astro-ph/9307002].
[24] A. Vilenkin and L. H. Ford, “Gravitational Effects Upon Cosmological Phase Transitions,”
Phys. Rev. D 26, 1231 (1982).
[25] A. D. Linde, “Scalar Field Fluctuations In Expanding Universe And The New Inflationary
Universe Scenario,” Phys. Lett. B 116, 335 (1982).
[26] D. J. Gross and F. Wilczek, “Ultraviolet Behavior Of Non-Abelian Gauge Theories,” Phys.
Rev. Lett. 30, 1343 (1973).
[27] H. D. Politzer, “Reliable Perturbative Results For Strong Interactions?,” Phys. Rev. Lett.
30, 1346 (1973).
[28] A.R. Liddle and D. H. Lyth, Cosmological Inflation and Large-Scale Structure, (Cambridge
University Press, Cambridge 2000).

34
[29] H. V. Peiris et al., “First year Wilkinson Microwave Anisotropy Probe (WMAP) observations:
Implications for inflation,” Astrophys. J. Suppl. 148, 213 (2003) [arXiv:astro-ph/0302225].
[30] M. Tegmark et al. [SDSS Collaboration], “Cosmological parameters from SDSS and WMAP,”
Phys. Rev. D 69, 103501 (2004) [arXiv:astro-ph/0310723].
[31] U. Seljak et al., “Cosmological parameter analysis including SDSS Ly-alpha forest and galaxy
bias: Constraints on the primordial spectrum of fluctuations, neutrino mass, and dark en-
ergy,” arXiv:astro-ph/0407372.
[32] K. Sigurdson and A. Cooray, “Cosmic 21-cm Delensing of Microwave Background Polariza-
tion and the Minimum Detectable Energy Scale of Inflation,” arXiv:astro-ph/0502549.
[33] A. D. Linde, “Generation Of Isothermal Density Perturbations In The Inflationary Universe,”
Phys. Lett. B 158, 375 (1985); L. A. Kofman and A. D. Linde, “Generation Of Density
Perturbations In The Inflationary Cosmology,” Nucl. Phys. B 282 (1987) 555.
[34] D. H. Lyth and D. Wands, “Generating the curvature perturbation without an inflaton,”
Phys. Lett. B 524, 5 (2002) [arXiv:hep-ph/0110002].
[35] T. Moroi and T. Takahashi, “Effects of cosmological moduli fields on cosmic mi-
crowave background,” Phys. Lett. B 522, 215 (2001) [Erratum-ibid. B 539, 303 (2002)]
[arXiv:hep-ph/0110096].
[36] S. Mollerach, “Isocurvature Baryon Perturbations And Inflation,” Phys. Rev. D 42, 313
(1990).
[37] A. D. Linde and V. Mukhanov, “Nongaussian isocurvature perturbations from inflation,”
Phys. Rev. D 56, 535 (1997) [arXiv:astro-ph/9610219].
[38] G. Dvali, A. Gruzinov and M. Zaldarriaga, “A new mechanism for generating density per-
turbations from inflation,” Phys. Rev. D 69, 023505 (2004) [arXiv:astro-ph/0303591].
[39] L. Kofman, “Probing string theory with modulated cosmological fluctuations,”
arXiv:astro-ph/0303614; F. Bernardeau, L. Kofman and J. P. Uzan, “Modulated fluctua-
tions from hybrid inflation,” Phys. Rev. D 70, 083004 (2004) [arXiv:astro-ph/0403315].
[40] P. J. Steinhardt, “Natural Inflation,” In: The Very Early Universe, ed. G.W. Gibbons,
S.W. Hawking and S.Siklos, Cambridge University Press, (1983); A. D. Linde, “Nonsingular
Regenerating Inflationary Universe,” Cambridge University preprint Print-82-0554 (1982).
[41] A. Vilenkin, “The Birth Of Inflationary Universes,” Phys. Rev. D 27, 2848 (1983).
[42] A. D. Linde, “Eternally Existing Self-reproducing Chaotic Inflationary Universe,” Phys. Lett.
B 175, 395 (1986); A. S. Goncharov, A. D. Linde and V. F. Mukhanov, “The Global Structure
Of The Inflationary Universe,” Int. J. Mod. Phys. A 2, 561 (1987).
[43] A. D. Linde, D. A. Linde and A. Mezhlumian, “From the Big Bang theory to the theory of
a stationary universe,” Phys. Rev. D 49, 1783 (1994) [arXiv:gr-qc/9306035].
[44] A. Borde, A. H. Guth and A. Vilenkin, “Inflationary space-times are incomplete in past
directions,” Phys. Rev. Lett. 90, 151301 (2003) [arXiv:gr-qc/0110012].
[45] A.D. Linde, “The New Inflationary Universe Scenario,” In: The Very Early Universe, ed.
G.W. Gibbons, S.W. Hawking and S.Siklos, pp. 205-249 (Cambridge University Press,1983);
A.D. Linde, “Inflation Can Break Symmetry In SUSY,” Phys. Lett. B 131, 330 (1983); A.D.
Linde, “The Inflationary Universe,” Rept. Prog. Phys. 47, 925 (1984).
[46] A. D. Linde, “Monopoles as big as a universe,” Phys. Lett. B 327, 208 (1994)
[arXiv:astro-ph/9402031]; A. Vilenkin, “Topological inflation,” Phys. Rev. Lett. 72, 3137
(1994) [arXiv:hep-th/9402085]; A. D. Linde and D. A. Linde, “Topological defects as seeds
for eternal inflation,” Phys. Rev. D 50, 2456 (1994) [arXiv:hep-th/9402115]; I. Cho and
A. Vilenkin, “Spacetime structure of an inflating global monopole,” Phys. Rev. D 56, 7621

35
 (1997) [arXiv:gr-qc/9708005].
[47] M. Aryal and A. Vilenkin, “The Fractal Dimension Of Inflationary Universe,” Phys. Lett. B
199, 351 (1987).
[48] J. Garcia-Bellido, A. D. Linde and D. A. Linde, “Fluctuations of the gravitational
constant in the inflationary Brans-Dicke cosmology,” Phys. Rev. D 50, 730 (1994)
[arXiv:astro-ph/9312039]; J. Garcia-Bellido and A. D. Linde, “Stationarity of inflation and
predictions of quantum cosmology,” Phys. Rev. D 51, 429 (1995) [arXiv:hep-th/9408023];
J. Garcia-Bellido and A. D. Linde, “Stationary solutions in Brans-Dicke stochastic infla-
tionary cosmology,” Phys. Rev. D 52, 6730 (1995) [arXiv:gr-qc/9504022]; A. D. Linde and
A. Mezhlumian, “On Regularization Scheme Dependence of Predictions in Inflationary Cos-
mology,” Phys. Rev. D 53, 4267 (1996) [arXiv:gr-qc/9511058].
[49] A. Vilenkin, “Making predictions in eternally inflating universe,” Phys. Rev. D 52, 3365
(1995) [arXiv:gr-qc/9505031]; S. Winitzki and A. Vilenkin, “Uncertainties of predictions in
models of eternal inflation,” Phys. Rev. D 53, 4298 (1996) [arXiv:gr-qc/9510054]; A. Vilenkin,
“Unambiguous probabilities in an eternally inflating universe,” Phys. Rev. Lett. 81, 5501
(1998) [arXiv:hep-th/9806185]; V. Vanchurin, A. Vilenkin and S. Winitzki, “Predictabil-
ity crisis in inflationary cosmology and its resolution,” Phys. Rev. D 61, 083507 (2000)
[arXiv:gr-qc/9905097]; J. Garriga and A. Vilenkin, “A prescription for probabilities in eter-
nal inflation,” Phys. Rev. D 64, 023507 (2001) [arXiv:gr-qc/0102090].
[50] M. Tegmark, “What does inflation really predict?,” arXiv:astro-ph/0410281.
[51] L. Susskind, “The anthropic landscape of string theory,” arXiv:hep-th/0302219.
[52] R. Bousso and J. Polchinski, “Quantization of four-form fluxes and dynamical neutralization
of the cosmological constant,” JHEP 0006, 006 (2000) [arXiv:hep-th/0004134].
[53] M. R. Douglas, “The statistics of string / M theory vacua,” JHEP 0305, 046 (2003)
[arXiv:hep-th/0303194] M. R. Douglas, B. Shiffman and S. Zelditch, “Critical points and
supersymmetric vacua,” arXiv:math.cv/0402326. F. Denef and M. R. Douglas, “Distri-
butions of flux vacua,” JHEP 0405, 072 (2004) [arXiv:hep-th/0404116]; A. Giryavets,
S. Kachru and P. K. Tripathy, “On the taxonomy of flux vacua,” JHEP 0408, 002 (2004)
[arXiv:hep-th/0404243]; F. Denef, M. R. Douglas and B. Florea, “Building a better race-
track,” JHEP 0406, 034 (2004) [arXiv:hep-th/0404257].
[54] S. Dodelson, “Coherent phase argument for inflation,” AIP Conf. Proc. 689, 184 (2003)
[arXiv:hep-ph/0309057].
[55] V. Mukhanov, “CMB-slow, or How to Estimate Cosmological Parameters by Hand,”
arXiv:astro-ph/0303072.
[56] G. Efstathiou, “The Statistical Significance of the Low CMB Multipoles,”
arXiv:astro-ph/0306431; G. Efstathiou, “A Maximum Likelihood Analysis of the
Low CMB Multipoles from WMAP,” Mon. Not. Roy. Astron. Soc. 348, 885 (2004)
[arXiv:astro-ph/0310207]. A. Slosar and U. Seljak, “Assessing the effects of foregrounds and
sky removal in WMAP,” Phys. Rev. D 70, 083002 (2004) [arXiv:astro-ph/0404567].
[57] C. R. Contaldi, M. Peloso, L. Kofman and A. Linde, “Suppressing the lower Multipoles in
the CMB Anisotropies,” JCAP 0307, 002 (2003) [arXiv:astro-ph/0303636].
[58] D. Spergel and N. Turok, “Textures and cosmic structure,” Scientific American 266, 52
(1992).
[59] J. Khoury, B. A. Ovrut, P. J. Steinhardt and N. Turok, “The ekpyrotic universe: Col-
liding branes and the origin of the hot big bang,” Phys. Rev. D 64, 123522 (2001)
[arXiv:hep-th/0103239].

36
[60] R. Kallosh, L. Kofman and A. D. Linde, “Pyrotechnic universe,” Phys. Rev. D 64, 123523
(2001) [arXiv:hep-th/0104073].
[61] R. Kallosh, L. Kofman, A. D. Linde and A. A. Tseytlin, “BPS branes in cosmology,” Phys.
Rev. D 64, 123524 (2001) [arXiv:hep-th/0106241].
[62] P. J. Steinhardt and N. Turok, “Cosmic evolution in a cyclic universe,” Phys. Rev. D 65,
126003 (2002) [arXiv:hep-th/0111098].
[63] H. Liu, G. Moore and N. Seiberg, “Strings in a time-dependent orbifold,” JHEP 0206, 045
(2002) [arXiv:hep-th/0204168].
[64] G. T. Horowitz and J. Polchinski, “Instability of spacelike and null orbifold singularities,”
Phys. Rev. D 66, 103512 (2002) [arXiv:hep-th/0206228].
[65] P. Creminelli, A. Nicolis and M. Zaldarriaga, “Perturbations in bouncing cosmologies: Dy-
namical attractor vs scale invariance,” arXiv:hep-th/0411270.
[66] V. Bozza and G. Veneziano, “Scalar perturbations in regular two-component bouncing cos-
mologies,” arXiv:hep-th/0502047.
[67] D. La and P. J. Steinhardt, “Extended Inflationary Cosmology,” Phys. Rev. Lett. 62, 376
(1989) [Erratum-ibid. 62, 1066 (1989)].
[68] K. Freese, J. A. Frieman and A. V. Olinto, “Natural Inflation With Pseudo - Nambu-
Goldstone Bosons,” Phys. Rev. Lett. 65, 3233 (1990).
[69] R. Kallosh, A. Linde, D. Linde and L. Susskind, “Gravity and global symmetries,” Phys.
Rev. D52, 912 (1995) [hep-th/9502069].
[70] E. J. Copeland, A. R. Liddle, D. H. Lyth, E. D. Stewart and D. Wands, “False vacuum infla-
tion with Einstein gravity,” Phys. Rev. D 49, 6410 (1994) [astro-ph/9401011]; G. R. Dvali,
Q. Shafi and R. Schaefer, “Large scale structure and supersymmetric inflation without fine
tuning,” Phys. Rev. Lett. 73, 1886 (1994) [hep-ph/9406319]; A. D. Linde and A. Riotto,
“Hybrid inflation in supergravity,” Phys. Rev. D 56, 1841 (1997) [arXiv:hep-ph/9703209].
[71] P. Binetruy and G. Dvali, “D-term inflation,” Phys. Lett. B 388, 241 (1996)
[hep-ph/9606342]; E. Halyo, “Hybrid inflation from supergravity D-terms,” Phys. Lett. B
387, 43 (1996) [hep-ph/9606423].
[72] D. H. Lyth and A. Riotto, “Particle physics models of inflation and the cosmological density
perturbation,” Phys. Rept. 314, 1 (1999) [hep-ph/9807278].
[73] R. Kallosh and A. Linde, “P-term, D-term and F-term inflation,” JCAP 0310, 008 (2003)
[arXiv:hep-th/0306058].
[74] M. Kawasaki, M. Yamaguchi and T. Yanagida, “Natural chaotic inflation in supergravity,”
Phys. Rev. Lett. 85, 3572 (2000) [arXiv:hep-ph/0004243].
[75] M. Yamaguchi and J. Yokoyama, “New inflation in supergravity with a chaotic initial con-
dition,” Phys. Rev. D 63, 043506 (2001) [arXiv:hep-ph/0007021]; M. Yamaguchi, “Natural
double inflation in supergravity,” Phys. Rev. D 64, 063502 (2001) [arXiv:hep-ph/0103045].
[76] M. Yamaguchi and J. Yokoyama, “Chaotic hybrid new inflation in supergravity with a run-
ning spectral arXiv:hep-ph/0307373.
[77] S. B. Giddings, S. Kachru and J. Polchinski, “Hierarchies from fluxes in string compactifica-
tions,” Phys. Rev. D66, 106006 (2002) [arXiv:hep-th/0105097].
[78] S. Kachru, R. Kallosh, A. Linde and S. P. Trivedi, “De Sitter vacua in string theory,” Phys.
Rev. D 68, 046005 (2003) [arXiv:hep-th/0301240].
[79] C. P. Burgess, R. Kallosh and F. Quevedo, “de Sitter string vacua from supersymmetric
D-terms,” JHEP 0310, 056 (2003) [arXiv:hep-th/0309187].
[80] E. Silverstein, “(A)dS backgrounds from asymmetric orientifolds,” arXiv:hep-th/0106209;

37
 A. Maloney, E. Silverstein and A. Strominger, “De Sitter space in noncritical string theory,”
arXiv:hep-th/0205316.
[81] N. Arkani-Hamed, S. Dimopoulos and G. R. Dvali, “The hierarchy problem and new dimen-
sions at a millimeter,” Phys. Lett. B 429, 263 (1998) [arXiv:hep-ph/9803315]; I. Antoniadis,
N. Arkani-Hamed, S. Dimopoulos and G. R. Dvali, “New dimensions at a millimeter to a
Fermi and superstrings at a TeV,” Phys. Lett. B 436, 257 (1998) [arXiv:hep-ph/9804398].
[82] L. Randall and R. Sundrum, “A large mass hierarchy from a small extra dimension,” Phys.
Rev. Lett. 83, 3370 (1999) [arXiv:hep-ph/9905221].
[83] J. J. Blanco-Pillado et al., “Racetrack inflation,” JHEP 0411, 063 (2004)
[arXiv:hep-th/0406230].
[84] G. R. Dvali and S. H. H. Tye, “Brane inflation,” Phys. Lett. B 450, 72 (1999)
[arXiv:hep-ph/9812483].
[85] F. Quevedo, “Lectures on String/Brane Cosmology,” hep-th/0210292.
[86] S. Kachru, R. Kallosh, A. Linde, J. Maldacena, L. McAllister and S. P. Trivedi, “Towards
inflation in string theory,” JCAP 0310, 013 (2003) [arXiv:hep-th/0308055].
[87] A. Buchel and R. Roiban, “Inflation in warped geometries,” Phys. Lett. B 590, 284 (2004)
[arXiv:hep-th/0311154]; M. Berg, M. Haack and B. Kors, “Loop corrections to volume mod-
uli and inflation in string theory,” Phys. Rev. D 71, 026005 (2005) [arXiv:hep-th/0404087];
C. P. Burgess, J. M. Cline, H. Stoica and F. Quevedo, “Inflation in realistic D-brane mod-
els,” JHEP 0409, 033 (2004) [arXiv:hep-th/0403119]; O. DeWolfe, S. Kachru and H. Ver-
linde, “The giant inflaton,” [arXiv:hep-th/0403123]; N. Iizuka and S. P. Trivedi, “An infla-
tionary model in string theory,” Phys. Rev. D 70, 043519 (2004) [arXiv:hep-th/0403203];
A. Buchel and A. Ghodsi, “Braneworld inflation,” Phys. Rev. D 70, 126008 (2004)
[arXiv:hep-th/0404151]; H. Firouzjahi and S. H. Tye, “Brane inflation and cosmic string
tension in superstring theory,” arXiv:hep-th/0501099.
[88] J. P. Hsu, R. Kallosh and S. Prokushkin, “On brane inflation with volume stabilization,”
JCAP 0312, 009 (2003) [arXiv:hep-th/0311077].
[89] R. Kallosh, “N = 2 supersymmetry and de Sitter space,” arXiv:hep-th/0109168; C. Herdeiro,
S. Hirano and R. Kallosh, “String theory and hybrid inflation/acceleration,” JHEP
0112 (2001) 027 [arXiv:hep-th/0110271]; K. Dasgupta, C. Herdeiro, S. Hirano and
R. Kallosh, “D3/D7 inflationary model and M-theory,” Phys. Rev. D 65, 126002 (2002)
[arXiv:hep-th/0203019].
[90] H. Firouzjahi and S. H. H. Tye, “Closer towards inflation in string theory,” Phys. Lett. B
584, 147 (2004) [arXiv:hep-th/0312020].
[91] C. Angelantonj, R. D’Auria, S. Ferrara and M. Trigiante, “K3 x T**2/Z(2) orientifolds
with fluxes, open string moduli and critical points,” Phys. Lett. B 583, 331 (2004)
[arXiv:hep-th/0312019].
[92] J. P. Hsu and R. Kallosh, “Volume Stabilization and the Origin of the Inflaton Shift Sym-
metry in String Theory,” hep-th/0402047.
[93] F. Koyama, Y. Tachikawa and T. Watari, “Supergravity analysis of hybrid inflation model
from D3-D7 system,” Phys. Rev. D 69, 106001 (2004) [Erratum-ibid. D 70, 129907 (2004)]
[arXiv:hep-th/0311191].
[94] L. McAllister, “An inflaton mass problem in string inflation from threshold corrections to
volume stabilization,” arXiv:hep-th/0502001.
[95] K. Dasgupta, J. P. Hsu, R. Kallosh, A. Linde and M. Zagermann, “D3/D7 brane inflation
and semilocal strings,” JHEP 0408, 030 (2004) [arXiv:hep-th/0405247].

38
 [96] P. Chen, K. Dasgupta, K. Narayan, M. Shmakova and M. Zagermann, “Brane inflation,
solitons and cosmological solutions: I,” arXiv:hep-th/0501185.
[97] R. Kallosh and A. Linde, in preparation.
[98] E. I. Buchbinder, “Five-brane dynamics and inflation in heterotic M-theory,” Nucl. Phys. B
711, 314 (2005) [arXiv:hep-th/0411062].
[99] K. Becker, M. Becker and A. Krause, “M-theory inflation from multi M5-brane dynamics,”
arXiv:hep-th/0501130.
[100] C. Armendariz-Picon, T. Damour and V. Mukhanov, “k-inflation,” Phys. Lett. B 458, 209
(1999) [arXiv:hep-th/9904075].
[101] S. Dimopoulos and S. Thomas, “Discretuum versus continuum dark energy,” Phys. Lett. B
573, 13 (2003) [arXiv:hep-th/0307004].
[102] E. Silverstein and D. Tong, “Scalar speed limits and cosmology: Acceleration from D-
cceleration,” Phys. Rev. D 70, 103505 (2004) [arXiv:hep-th/0310221]; M. Alishahiha,
E. Silverstein and D. Tong, “DBI in the sky,” Phys. Rev. D 70, 123505 (2004)
[arXiv:hep-th/0404084].
[103] R. Kallosh and A. Linde, “Landscape, the scale of SUSY breaking, and inflation,” JHEP
0412, 004 (2004) [arXiv:hep-th/0411011].
[104] O. DeWolfe and S. B. Giddings, “Scales and hierarchies in warped compactifications and
brane worlds,” Phys. Rev. D 67, 066008 (2003) [arXiv:hep-th/0208123].
[105] N. Arkani-Hamed and S. Dimopoulos, “Supersymmetric unification without low energy su-
persymmetry and signatures for fine-tuning at the LHC,” arXiv:hep-th/0405159; N. Arkani-
Hamed, S. Dimopoulos, G. F. Giudice and A. Romanino, “Aspects of split supersymmetry,”
Nucl. Phys. B 709, 3 (2005) [arXiv:hep-ph/0409232].
[106] L. Kofman, A. Linde, X. Liu, A. Maloney, L. McAllister and E. Silverstein, “Beauty
is attractive: Moduli trapping at enhanced symmetry points,” JHEP 0405, 030 (2004)
[arXiv:hep-th/0403001].
[107] D. H. Lyth, “Which is the best inflation model?,” arXiv:hep-th/0311040.
[108] Y. B. Zeldovich and A. A. Starobinsky, “Quantum Creation Of A Universe In A Nontrivial
Topology,” Sov. Astron. Lett. 10, 135 (1984).
[109] O. Heckmann & E. Schucking, in Handbuch der Physik, ed. S. Flugge (Springer, Berlin, 1959),
Vol. 53, p.515; G.F. Ellis, Gen. Rel. Grav. 2, 7 (1971); J.R. Gott, “Chaotic Cosmologies,”
Mon. Not. R. Astron. Soc. 193, 153 (1980); C.N. Lockhart, B.Misra and I. Prigogine, Phys.
Rev. D25, 921 (1982); H.V. Fagundes, Phys. Rev. Lett. 51, 517 (1983).
[110] N. J. Cornish, D. N. Spergel and G. D. Starkman, “Does chaotic mixing facilitate Ω < 1
inflation?” Phys. Rev. Lett. 77, 215 (1996) [arXiv:astro-ph/9601034].
[111] D. H. Coule and J. Martin, “Quantum cosmology and open universes,” Phys. Rev. D 61,
063501 (2000) [arXiv:gr-qc/9905056].
[112] A. Linde, “Creation of a compact topologically nontrivial inflationary universe,” JCAP 0410,
004 (2004) [arXiv:hep-th/0408164].
[113] L.Kofman, A. Linde, and A. Starobinsky, in preparation.
[114] A. D. Linde, “Quantum Creation Of The Inflationary Universe,” Lett. Nuovo Cim. 39, 401
(1984).
[115] A. Vilenkin, “Quantum Creation Of Universes,” Phys. Rev. D 30, 509 (1984); A. Vilenkin,
“Quantum Cosmology And The Initial State Of The Universe,” Phys. Rev. D 37, 888 (1988);
A. Vilenkin, “The Interpretation Of The Wave Function Of The Universe,” Phys. Rev. D
39, 1116 (1989).

39
[116] A. D. Linde, “Quantum creation of an open inflationary universe,” Phys. Rev. D 58, 083514
(1998) [arXiv:gr-qc/9802038].
[117] R. Brustein, S. P. de Alwis and P. Martens, “Cosmological stabilization of moduli with steep
potentials,” Phys. Rev. D 70, 126012 (2004) [arXiv:hep-th/0408160].
[118] M. J. Duff, B. E. W. Nilsson and C. N. Pope, “Kaluza-Klein Supergravity,” Phys. Rept. 130,
1 (1986).
[119] W. Lerche, D. Lust and A. N. Schellekens, “Chiral Four-Dimensional Heterotic Strings From
Selfdual Lattices,” Nucl. Phys. B 287, 477 (1987); M. J. Duff, “Not The Standard Superstring
Review,” CERN-TH-4749-87 Proc. of Int. School of Subnuclear Physics, The Super-World
II, Erice, Italy, Aug 6-14, 1987.
[120] A. D. Linde, “The Inflationary Universe,” Rept. Prog. Phys. 47, 925 (1984); A. D. Sakharov,
“Cosmological Transitions With A Change In Metric Signature,” Sov. Phys. JETP 60, 214
(1984) [Zh. Eksp. Teor. Fiz. 87, 375 (1984)]; A.D. Linde, “Inflation And Quantum Cos-
mology,” Print-86-0888 (June 1986) in 300 Years of Gravitation, ed. by S.W. Hawking and
W. Israel, Cambridge University Press, Cambridge (1987); S. Weinberg, “Anthropic Bound
On The Cosmological Constant,” Phys. Rev. Lett. 59, 2607 (1987); A. Vilenkin, “Predic-
tions From Quantum Cosmology,” Phys. Rev. Lett. 74, 846 (1995) [arXiv:gr-qc/9406010]; G.
Efstathiou, MNRAS 274, L73 (1995); H. Martel, P. R. Shapiro and S. Weinberg, “Likely Val-
ues of the Cosmological Constant,” Astrophys. J. 492, 29 (1998) [arXiv:astro-ph/9701099];
J. Garriga and A. Vilenkin, “On likely values of the cosmological constant,” Phys. Rev.
D 61, 083502 (2000) [arXiv:astro-ph/9908115]; J. L. Feng, J. March-Russell, S. Sethi and
F. Wilczek, “Saltatory relaxation of the cosmological constant,” Nucl. Phys. B 602, 307
(2001) [arXiv:hep-th/0005276].
[121] S. W. Hawking and I. G. Moss, “Supercooled Phase Transitions In The Very Early Universe,”
Phys. Lett. B 110, 35 (1982).
[122] K. M. Lee and E. J. Weinberg, “Decay Of The True Vacuum In Curved Space-Time,” Phys.
Rev. D 36, 1088 (1987).
[123] J. Garriga and A. Vilenkin, “Recycling universe,” Phys. Rev. D 57, 2230 (1998)
[arXiv:astro-ph/9707292].
[124] L. Dyson, M. Kleban and L. Susskind, “Disturbing implications of a cosmological constant,”
JHEP 0210, 011 (2002) [arXiv:hep-th/0208013]; N. Goheer, M. Kleban and L. Susskind,
“The trouble with de Sitter space,” JHEP 0307, 056 (2003) [arXiv:hep-th/0212209];
B. Freivogel and L. Susskind, “A framework for the landscape,” Phys. Rev. D 70, 126007
(2004) [arXiv:hep-th/0408133].
[125] A. D. Linde, “Chaotic Inflation With Constrained Fields,” Phys. Lett. B 202, 194 (1988);
A. D. Linde and A. Mezhlumian, “Inflation with Ω 6= 1,” Phys. Rev. D 52, 6789 (1995)
[arXiv:astro-ph/9506017].

40