Dynamics of The Inflationary Era

Download as pdf or txt
Download as pdf or txt
You are on page 1of 40

Dynamics of the Inationary Era

Edward W. Kolb NASA/Fermilab Astrophysics Center Fermi National Accelerator Laboratory, Batavia, Illinois 60510-0500 and Department of Astronomy and Astrophysics, Enrico Fermi Institute The University of Chicago, Chicago, Illinois 60637-1433

arXiv:hep-ph/9910311v2 21 Oct 1999

There is very strong circumstantial evidence that there was an inationary epoch very early in the history of the universe. In this lecture I will describe how we might be able to piece together some understanding of the dynamics during and immediately after the inationary epoch.

Introduction

We live in a very large, very old, and nearly (perhaps exactly) spatially at universe. The universe we observe is, at least on large scales, remarkably homogeneous and isotropic. These attributes of our universe can be ascribed to a period of very rapid expansion, or ination, at some time during the very early history of the universe [1]. A suciently long epoch of primordial ination leads to a homogeneous/isotropic universe that is old and at. The really good news is that very many reasonable models have been proposed for ination.1 In some ways, ination is generic. That is also the really bad news, since we would like to use the early universe to learn something about physics at energy scales we cant produce in terrestrial laboratories. We want to know more than just that ination occurred, we want to learn something of the dynamics of the expansion of the universe during ination. That may tell us something about physics at very high energies. It would also allow us to restrict the number of ination models. If we can dierentiate between various ination models, then ination can be used as a phenomenological guide for understanding physics at very high energies. Oldness, atness, and homogeneity/isotropy only tell us the minimum length of the inationary era. They are not very useful instruments to probe the dynamics of ination. If that is our goal, we must nd another tool. In this lecture I will discuss two things associated with ination that allow us to probe the dynamics of ination: perturbations and preheating/reheating. While the universe is homogeneous and isotropic on large scales, it is inhomogeneous on small scales. The inhomogeneity in the distribution of galaxies, clusters, and other
Perhaps a more accurate statement is that there are many models that seemed reasonable to the people who proposed them at the time they were proposed.
1

luminous objects is believed to result from small seed primordial perturbations in the density eld produced during ination. Density perturbations produced in ination also lead to the observed anisotropy in the temperature of the cosmic microwave background radiation. A background of primordial gravitational waves is also produced during ination. While the background gravitational waves do not provide the seeds or inuence the development of structure, gravitational waves do lead to temperature variations in the cosmic microwave background radiation. If we can extract the primordial density perturbations from observations of large-scale structure and cosmic microwave background radiation temperature uctuations, then we can learn something about the dynamics of ination. If we can discover evidence of a gravitational wave background, then we will know even more about the dynamics of the expansion rate during ination. Ination was wonderful, but all good things must end. The early universe somehow made the transition from an inationary phase to a radiation-dominated phase. Perhaps there was a brief matter-dominated phase between the ination and radiation eras. In the past few years we have come to appreciate that some interesting phenomena like phase transitions, baryogenesis, and dark matter production can occur at the end of ination. Perhaps by studying how ination ended, we can learn something of the dynamics of the universe during ination.

Perturbations Produced During Ination

One of the striking features of the cosmic background radiation (CBR) temperature uctuations is the growing evidence that the uctuations are acausal.2 The CBR uctuations were largely imprinted at the time of last-scattering, about 300,000 years after the bang. However, there seems to be uctuations on length scales much larger than 300,000 light years!3 How could a causal process imprint correlations on scales larger than the light-travel distance since the time of the bang? The answer is ination. In order to see how ination solves this problem, rst consider the evolution of the Hubble radius with the scale factor a(t):4 RH H 1 = a a
1

1/2

a2 (RD) 3/2 a (MD).

(1)

In a k = 0 matter-dominated universe the age is related to H by t = (2/3)H 1, so RH = (3/2)t. In the early radiation-dominated universe t = (1/2)H 1, so RH = 2t. On length scales smaller than RH it is possible to move material around and make an imprint upon the universe. Scales larger than RH are beyond the Hubble radius, and
Exactly what is meant by acausal will be explained shortly. Acausal is in fact somewhat of a misnomer since, as we shall see, ination produces acausal perturbations by completely causal physics. 3 Although denitive data is not yet in hand, the issue of the existence of acausal perturbations will be settled very soon. 4 Here and throughout the paper RD is short for radiation dominated, and MD implies matter dominated.
2

=H

/ a1

=H

/ a1

H
H

/ a2 (RD) 1 / a3=2 (MD)


1

H
H

/ a2 (RD) 1 / a3=2 (MD)


1

/ an(n < 1)

log a

end inflation

log a

Figure 1: Physical sizes increase as a(t) in the expanding universe. The Hubble radius evolves as RH = H 1 = (8G(a)/3)1/2 . In a radiation-dominated or matter-dominated universe (illustrated by the left panel) any physical length scale starts larger than RH , then crosses the Hubble radius ( = H 1 ) only once. However, if there was a period of early ination when RH increased more slowly than a (as illustrated in the right panel), it is possible for a physical length scale to start smaller than RH , become larger than RH , and after ination ends become once again smaller than RH . Periods during which the scale is larger than the Hubble radius are indicated by the dotted line. the expansion of the universe prevents the establishment of any perturbation on scales larger than RH . Next consider the evolution of some physical length scale . Clearly, any physical length scale changes in expansion in proportion to a(t). Correlations on physical length scales larger than RH (t) are often called acausal. Now let us form the dimensionless ratio L /RH . If L is smaller than unity, the length scale is smaller than the Hubble radius and it is possible to imagine some microphysical process establishing perturbations on that scale, while if L is larger than unity, no microphysical process can account for perturbations on that scale. scales as a Since RH = a/a , and a, the ratio L is proportional to a , and L , which in turn is proportional to ( + 3p). There are two possible scenarios for L depending

upon the sign of + 3p: L < 0 RH grows faster than , happens for + 3p > 0 > 0 RH grows more slowly than , happens for + 3p < 0. (2)

In the standard MD or RD universe, + 3p > 0, and RH grows faster than . Perturbations that appear to be noncausal at last scattering can be produced if sometime during the early evolution of the universe the expansion was such that a > 0. If a > 0, then the Hubble radius will increase more slowly than any physical scale increases during ination. Physical length scales larger than the Hubble radius at the time of last scattering would have been smaller than the Hubble radius during the accelerated era. Therefore it is possible to imprint correlations during ination as a scale passes out of the horizon and have it appear as an acausal perturbation at last scattering. From Einsteins equations, the acceleration is related to the energy density and the pressure as a ( + 3p). Therefore, an acceleration (positive a ) requires an unusual equation of state with + 3p < 0. This is the condition for accelerated expansion or ination. We do not yet know when ination occurred, but the best guess for the dierent epochs in the history of the universe is given in Table 1. It is useful to spend a few minutes discussing the movements of Table 1. The rst movement of the Cosmic Symphony may be dominated by the string section if on the smallest scales there is a fundamental stringiness to elementary particles. If this is true, then the rst movement in the cosmic symphony would have been a pizzicato movement of vibrating strings about 1043 s after the bang. There is basically nothing known about the stringy phase, if indeed there was one. We do not yet know enough about this era to predict relics, or even the equation of state. The earliest phase we have information about is the inationary phase. The inationary movement probably followed the string movement, lasting approximately 1036 seconds. During ination the energy density of the universe was dominated by vacuum energy, with equation of state pV V . As we shall see, the best information we have of the inationary phase is from the quantum uctuations during ination, which were imprinted upon the metric, and can be observed as CBR uctuations and the departures from homogeneity and isotropy in the matter distribution, e.g., the power spectrum. Ination also produces a background of gravitational radiation, which can be detected by its eect on the CBR, or if ination was suciently exotic, by direct detection of the relic background by experiments such as LIGO or LISA. Ination was wonderful, but all good things must end. A lot of eort has gone into studying the end of ination (For a review, see Kofman et al., [2].) It was likely that there was a brief period during which the energy density of the universe was dominated by coherent oscillations of the inaton eld. During coherent oscillations the inaton energy density scales as a3 where a is the scale factor, so the expansion rate of the universe decreased as in a matter-dominated universe with p 0. Very little is known about this period immediately after ination, but there is hope that one day we will 4

tempo pizzicato

epoch string dominated vacuum dominated (ination)


<

age 1043 s

p ?

+ 3p ?

relic ?

prestissimo

1038 s

density perturbations gravitational waves dark matter? phase transitions? dark matter? baryogenesis? dark matter? baryogenesis? neutrino decoupling nucleosynthesis recombination radiation decoupling growth of structure

presto

matter dominated

10

36

allegro

radiation dominated

<

104 yr

T4

T 4 /3

andante

matter dominated vacuum dominated (ination) da capo?

>

10 yr

largo

recent

acceleration of the universe

Table 1: Dierent epochs in the history of the universe and the associated tempos of the ever decreasing expansion rate H , along with the equation of state and some of the relics produced during the various eras.

discover a relic. Noteworthy events that might have occurred during this phase include baryogenesis, phase transitions, and generation of dark matter. We do know that the universe was radiation dominated for almost all of the rst 10,000 years. The best preserved relics of the radiation-dominated era are the light elements. The light elements were produced in the radiation-dominated universe one second to three minutes after the bang.5 If the baryon asymmetry is associated with the electroweak transition, then the asymmetry was generated in the radiation-dominated era. The radiation era is also a likely source of dark matter such as WIMPS or axions. If one day we can detect the 1.9 K neutrino background, it would be a direct relic of the radiation era. The equation of state during the radiation era is pR = R /3. The earliest picture of the matter-dominated era is the CBR. Recombination and matterradiation decoupling occurred while the universe was matter dominated. Structure developed from small primordial seeds during the matter-dominated era. The pressure is negligible during the matter-dominated era. Finally, if recent determinations of the Hubble diagram from observations of distant high-redshift Type-I supernovae are correctly interpreted, the expansion of the universe is increasing today ( a > 0). This would mean that the universe has recently embarked on another inationary era, but with the Hubble expansion rate much less than the rate during the rst inationary era.

2.1

Simple Dynamics of Ination: The Inaton

In building ination models it is necessary to nd a mechanism by which a universe dominated by vacuum energy can make a transition from the inationary universe to a matter-dominated or radiation-dominated universe. There is some unknown dynamics causing the expansion rate to change with time. There may be several degrees of freedom involved in determining the expansion rate during ination, but the simplest assumption is that there is only one dynamical degree of freedom responsible for the evolution of the expansion rate. If there is a single degree of freedom at work during ination, then the evolution from the inationary phase may be modeled by the action of a scalar eld evolving under the inuence of a potential V (). Lets imagine the scalar eld is displaced from the minimum of its potential as illustrated in Fig. 2. If the energy density of the universe is dominated by the potential energy of the scalar eld , known as the inaton, then +3p will be negative. The vacuum energy disappears when the scalar eld evolves to its minimum. The amount of time required for the scalar eld to evolve to its minimum and ination to end (or even more useful, the number of e-folds of growth of the scale factor) can be found by solving the classical eld equation for the evolution of the inaton eld,
Although I may speak of time after the bang, I will not address the issue of whether the universe had a beginning or not, which in the modern context is the question of whether ination is eternal. For the purpose of this discussion, time zero of the bang can be taken as some time before the end of ination in the region of the universe we observe.
5

+ 3H + dV /d = 0. which is simply given by

2.2

Quantum uctuations

In addition to the classical motion of the inaton eld, during ination there are quantum uctuations.6 Since the total energy density of the universe is dominated by the inaton potential energy density, uctuations in the inaton eld lead to uctuations in the energy density. Because of the rapid expansion of the universe during ination, these uctuations in the energy density are frozen into super-Hubble-radius-size perturbations. Later, in the radiation or matter-dominated era they will come within the Hubble radius as if they were noncausal perturbations. The spectrum and amplitude of perturbations depend upon the nature of the inaton potential. Mukhanov [3] has developed a very nice formalism for the calculation of density perturbations. One starts with the action for gravity (the EinsteinHilbert action) plus a minimally-coupled scalar inaton eld : S= d4x g 1 m2 Pl R g + V () . 16 2 (3)

Here R is the Ricci curvature scalar. Quantum uctuations result in perturbations in the metric tensor and the inaton eld
F RW g g + g ;

0 + ,

(4)

F RW where g is the FriedmannRobertsonWalker metric, and 0 (t) is the classical solution for the homogeneous, isotropic evolution of the inaton. The action describing the dynamics of the small perturbations can be written as

2 S =

1 2

d4x u u + z 1

d2 z 2 u d 2

z = a/H ,

(5)

i.e., the action in conformal time (d 2 = a2 (t)dt2 ) for a scalar eld in Minkowski 1 2 space, with mass-squared m2 d z/d 2 . Here, the scalar eld u is a combination u = z of metric uctuations g and scalar eld uctuations . This scalar eld is related to the amplitude of the density perturbation. The simple matter of calculating the perturbation spectrum for a noninteracting scalar eld in Minkowski space will give the amplitude and spectrum of the density perturbations. The problem is that the solution to the eld equations depends upon the background eld evolution through the dependence of the mass of the eld upon z . Dierent choices for the inaton potential V () results in dierent background eld evolutions, and hence, dierent spectra and amplitudes for the density perturbations. Before proceeding, now is a useful time to remark that in addition to scalar density perturbations, there are also uctuations in the transverse, traceless component of the
6

Here I will continue to assume there is only one dynamical degree of freedom.

spatial part of the metric. These uctuations (known as tensor uctuations) can be thought of as a background of gravitons. Although the scalar and tensor spectra depend upon V (), for most potentials they S can be characterized by QP RM S (the amplitude of the scalar and tensor spectra on large length scales added in quadrature), n (the scalar spectral index describing the best power-law t of the primordial scalar spectrum), r (the ratio of the tensor-to-scalar contribution to C2 in the angular power spectrum), and nT ( the tensor spectral index describing the best power-law t of the primordial tensor spectrum). For single-eld, slow-roll ination models, there is a relationship between nT and r , so in fact there are only three independent variables. Furthermore, the amplitude of the uctuations often S depends upon a free parameter in the potential, and the spectra are normalized by QP RM S . This leads to a characterization of a wide-range of inaton potentials in terms of two numbers, n and r .

2.3

Models of ination

A quick perusal of the literature will reveal many models of ination. Some of the familiar names to be found include: old, new, pre-owned, chaotic, quixotic, ergodic, exotic, heterotic, autoerotic, natural, supernatural, au natural, power-law, powerless, power-mad, one-eld, two-eld, home-eld, modulus, modulo, moduli, self-reproducing, self-promoting, hybrid, low-bred, white-bread, rst-order, second-order, new-world order, pre-big-bang, no-big-bang, post-big-bang, D-term, F-term, winter-term, supersymmetric, superstring, superstitious, extended, hyperextended, overextended, D-brane, p-brane, No-brain, dilaton, dilettante, . . . Probably the rst step in sorting out dierent models is a classication scheme. One proposed classication scheme has two main types. Type-I ination models are models based on a single inaton eld, slowly rolling under the inuence of an inaton potential V (). This may seem like a restrictive class, but in fact many more complicated models can be expressed in terms of an equivalent Type-I model. For instance extended ination, which is a JordanBransDicke model with an inaton eld and a JBD scalar eld, can be recast as an eective Type-I model. Anything that is not Type I is denoted as a Type-II model. There are subclasses within Type I. Type-Ia models are large-eld models, where the inaton eld starts large and evolves toward its minimum. Examples of large-eld models are chaotic ination and power-law ination. Type-Ib models are small-eld models, where the inaton eld starts small and evolves to its minimum at larger values of the inaton eld. Examples of small-eld models are new ination and natural ination. Finally, hybrid-ination models are classied as Type-Ic models. In Type-Ia and Type-Ib models the vacuum energy is approximately zero at the end of ination. Hybrid models have a signicant vacuum energy at the end of ination (see Fig 2). Hybrid ination is usually terminated by a rst-order phase transition or by the action of a second scalar eld. A more accurate description of large-eld and small-eld potential is the sign of 8

Figure 2: Schematic illustrations of the inaton potential energy. The rst potential is a large-eld model (Type-Ia). The second gure illustrates a small-eld model (Type-Ib). The nal gure illustrates an example of hybrid ination (Type-Ic). Notice that the minimum of the potential is nonzero in the hybrid model. the second derivative of the potential: large-eld models have V > 0 while small-eld models have V < 0. Of course a classication scheme is only reasonable if there are some observable quantities that can dierentiate between dierent schemes. It turns out that the dierent Type-I models ll in dierent regions of the nr plane, as shown in Fig. 3 (from [4]). For a given spectral index n, small-eld models have a smaller value of r . Shown as an ellipse is a very conservative estimate of the uncertainties in n and r that are expected after the next round of satellite observations. Although we dont know where the error ellipse will fall on the graph, an error ellipse at the indicated size will restrict models. So we expect a true ination phenomenology, where models of ination are confronted by precision observations.

2.4

Ination Models in the Era of Precision Cosmology

It was once said that the words precision and cosmology could not both appear in a sentence containing an even number of negatives. However that statement is now out of date, or at the very least, very soon will be out of date. A number of new instruments will come on line in the next few years and revolutionize cosmology. There is now a world-wide campaign to pin down the microwave anisotropies. In the near future, long-duration balloon ights, as well as observations from the cold, dry observatory in Antarctica will completely change the situation. Finally, in the next decade two satellites, a NASA mission the Microwave Anisotropy Probe (MAP) 9

Figure 3: Large-eld (Type-Ia), small-eld (Type-Ib), and hybrid (Type-Ic) models seem to populate dierent regions of the nr plane. Observational determination of n and r could rule out entire classes of models. The dark ellipse indicated the anticipated size of the errors in the post-Planck era. Of course the location is yet to be determined. This gure is from Dodelson, Kinney and Kolb [4]. and an ESA missionPLANCK, will culminate in a determination of the spectrum with errors much smaller than the present errors. Of course we dont know what the shape of the spectrum will turn out to be, but we can anticipate errors as small as shown in the gure. With errors of this magnitude, tting the spectrum will allow determination of n and r , useful for ination, as well as determination of the cosmological parameters (H0 , 0 , B , , etc.) to a few percent.

2.5

Reconstruction

In addition to restricting the class of ination models (Type Ia, Type Ib, etc.), it may be possible to use the data from precision microwave experiments to reconstruct a fragment of the inationary potential. Reconstruction of the inaton potential (see Lidsey, et al., [5] for a review) refers to the process of using observational data, especially microwave background anisotropies, to determine the inaton potential capable of generating the perturbation spectra inferred from observations [6]. Of course there is no way to prove that the reconstructed inaton potential was the agent responsible for generating the perturbations. What can be hoped for is that one can determine a unique (within observational errors) inaton potential 10

lowest-order V ( ) V ( ) V ( ) V ( ) H ( ) H ( ), ( ) H ( ), ( ), ( ) H , ( ), ( ), ()

next-order (exact) H (), ( ) H (), ( ), ( ) H (), ( ), ( ), ( )

Table 2: A summary of the slow-roll parameters, H , , , and , needed to reconstruct a given derivative of the potential to a certain order. Note that the next-order result is exact. capable of producing the observed perturbation spectra. The reconstructed inaton potential may well be the rst concrete piece of information to be obtained about physics at scales close to the Planck scale. As is well known, ination produces both scalar and tensor perturbations, and each generate microwave anisotropies (see Liddle and Lyth [7] for a review). If V () is known, the perturbation spectra can be computed exactly in linear perturbation theory through integration of the relevant mode equations [8]. If the scalar eld is rolling suciently slowly, the solutions to the mode equations may be approximated using something known as the slow-roll expansion [9, 10, 11]. The standard reconstruction program makes use of the slow-roll expansion, taking advantage of a calculation of the perturbation spectra by Stewart and Lyth [12], which gives the next-order correction to the usual lowest-order slow-roll results. Two crucial ingredients for reconstruction are the primordial scalar and tensor perturbation spectra AS (k ) and AT (k ) which are dened as in Lidsey et al., [5]. The scalar and tensor perturbations depend upon the behavior of the expansion rate during ination, which in turn depends on the value of the scaler inaton eld . In order to track the change in the expansion rate, we dene slow-roll parameters , , and as ()
m2 Pl H () 4 H ( ) 2

( )

m2 Pl H () 4 H ( )

( )

m2 Pl 4

H ()H () H 2 ( )

1/2

, (6)

where here the prime superscript implies d/d. If the slow-roll parameters are small, , , and can be expressed in terms of derivatives of the inaton potential: m2 () = Pl 16 V V
2

m2 V () = Pl 8 V 11

( ) =

m2 Pl 8

V V V2

(7)

parameter H

lowest-order A2 T
2 A2 T , AS 2 A2 T , AS , n 2 A2 T , AS , n, dn/d ln k

next-order
2 A2 T , AS 2 A2 T , AS , n 2 A2 T , AS , n, dn/d ln k

Table 3: The ination parameters may be expressed in terms of the primordial scalar 2 and tensor perturbation spectra A2 S , AT , and the scalar and tensor spectral indices n, and dn/d ln k . As long as the slow-roll parameters are small compared to unity, the scalar and tensor perturbation amplitudes AS (k ) and AT (k ) are given by (see Stewart and Lyth [12] for the normalization) 4 25 4 A2 T (k ) 25 A2 S (k ) H mPl H mPl
2

1 [1 (2C + 1) + C ]2 [1 (C + 1)]2 ,

(8) (9)

where H , , and are to be determined at the value of when k = aH during ination, and where C = 2+ln 2+ 0.73 is a numerical constant, being the Euler constant. These equations are the basis of the reconstruction process. 2 From the expressions for A2 S and AT one can express the scalar and tensor spectral indices, dened as d ln A2 d ln A2 S T n(k ) 1 nT (k ) = (10) d ln k d ln k in terms of slow-roll parameters. Perturbative reconstruction requires that one ts an expansion, usually a Taylor series of the form
2 ln A2 S (k ) = ln AS (k ) + (n 1) ln

k 1 d2 n k 1 dn 3 k ln2 + ln + + , (11) k 2 d ln k k 6 d(ln k )2 k

where n = n(k ), to the observed spectrum in order to extract the coecients, where stars indicate the value at k . The scale k is most wisely chosen to be at the (logarithmic) center of the data, about k = 0.01 h Mpc1 . For reconstruction, one takes a HamiltonJacobi approach where the expansion rate is considered fundamental, and the expansion rate is parameterized by a value of the 12

lowest-order V V V V
S R , QP RM S S R , QP RM S S R , QP RM S , n S R , QP RM S , n , n

next-order
S R , QP RM S S R , QP RM S , n S R , QP RM S , n , n

next-to-next-order
S R , QP RM S , n S R , QP RM S , n , n

Table 4: A summary of the observables needed to reconstruct a given derivative of the potential to a certain order. R is the ratio of the tensor to scalar contributions to the S CBR anisotropies at any conveniently chosen scale. QP RM S is the normalization of the total (scalar plus tensor) contributions at small l, n and n refer to the scalar spectral S index and its derivative at some scale. Knowledge of R and QP RM S is equivalent to 2 2 determination of AT (k ) and AS (k ). inaton eld. The Friedmann equation may be expressed in terms of the potential V (), the expansion rate H (), and the derivative of H as dH d
2

32 2 12 2 H = V ( ) 2 MP m4 l Pl

(12)

or equivalently in terms of the slow-roll parameter , as


2 m2 P lH (3 ) . (13) 8 Subsequent derivatives of V may be expressed in terms of additional slow-roll parameters:

V =

m2 V = P l H 2 1/2 (3 ) 4

V = H 2 3 + 3 2 + 2

(14)

and so on. So if one can determine the slow-roll parameters, one has information about the potential. The slow-roll parameters needed to reconstruct a given derivative of the potential is given in Table 2. Of course, the problem is that the slow-roll parameters are not directly observable! But one can construct an iterative scheme to express the slow-roll parameters in terms of observables. The result is shown in Table 3. 2 Using the notation n to indicate dn/d ln k | and R = A2 T (k )/AS (k ), the reconstruction equations are
4 m Pl V ( )

75 2 A (k )R 1 + [0.21R ] 32 S 13

75 2 3/2 1 0.85R 0.53(1 n ) AS (k )R V ( ) 8 25 2 2 2 m A (k )R 9R 1.5(1 n ) 24.3R + 0.25(1 n )2 Pl V ( ) 4 S


3 m Pl

14.8R (1 n ) 1.6 n

(15)

The observables needed to reconstruct a given derivation of the potential are listed in Table 4. The biggest hurdle for successful reconstruction is that many ination models predict a tensor perturbation amplitude (hence, R ) well below the expected threshold for detection, and even above detection threshold the errors can be considerable. If the tensor modes cannot be identied a unique reconstruction is impossible, as the scalar perturbations are governed not only by V (), but by the rst derivative of V () as well. Knowledge of only the scalar perturbations leaves an undetermined integration constant in the non-linear system of reconstruction equations. Another problem is that simple potentials usually lead to a nearly exact power-law scalar spectrum with the spectral index close to unity. In such a scenario only a very limited amount of information could be obtained about high energy physics from astrophysical observations. However if a tensor mode can be determined, then one may follow the following reconstruction procedure. R , n , n , . . . are to be determined from observations. Fortunately, parameter estimation from the microwave background has been explored in some detail [13, 14]. We shall use error estimates for Planck assuming polarized detectors are available, following the analysis of Zaldarriaga et al., [14]. Most analyses have assumed that R and n are the only parameters needed to describe the spectra. In a recent paper [15], Copeland, Grivell, and Liddle have generalized their treatment to allow the power spectrum to deviate from scale-invariance. Including extra parameters leads to a deterioration in the determination of all the parameters, as it introduces extra parameter degeneracies. Fortunately, for most parameters the uncertainty is not much increased by including the rst few derivatives of n [15], but the parameter n itself has a greatly increased error bar. If a power-law is assumed it can be determined to around n 0.004 [14, 15], but including scale dependence increases this error bar by a factor of ten or more. Notice that unless one assumes a perfect power-law behavior, this increase in uncertainty is applicable even if the deviation from power-law behavior cannot be detected within the uncertainty. From Grivell and Liddle [8], an estimate of the relevant uncertainties is (R ) 0.004 n 0.15 ; ; (dn/d ln k ) 0.04 [d2 n/d(ln k )2 ] 0.005 . (16)

Once R , A2 S , n , and n are determined, then it is possible to nd V ( ), V ( ), and V ( ) from Eq. 15. Then one can express V () as a Taylor series about V ( ):

1 V () = V ( ) + V ( ) + V ( )2 + . 2 14

(17)

is found from an exact expression connecting changes in with changes in k [5], d m2 H 1 = Pl d ln k 4 H 1 R (1 + R ) . 4 (18)

Let me illustrate reconstruction by considering two sample potentials. The rst potential, discussed by Lidsey et al., [5] is a power-law potential, V () = V0 exp(/mP l ), with 1.6. The potential is shown in the upper left-hand-side of Fig. 4. This potential generates a spectral index of n = 0.9 and n = 0. It also results in a value of R = 0.1. So one might guess that precision CBR measurements will determine R , n , and n of these values, with uncertainties of Eq. 16. In order to see what sort of reconstructed potential results, one can imagine a model universe with R , n , and n generated as Gaussian random variables with mean determined by the underlying potential and variance determined by the expected observational uncertainties. The result of ten such reconstructions of the potential are shown in the upper left-hand gure of Fig. 4. The range of is determined by the range of wavenumber over which one expects to have accurate determinations of the parameters from the CBR. Here k was chosen to be 0.01h Mpc1 and the range of k taken to be three decades. Also shown in Fig. 4 in the upper right-hand panel is information about the reconstruction of the rst derivative of the potential. The second example of reconstruction was considered in Copeland et al., [16] The potential, rst considered by Wang et al., [17] is a potential of the form V ( ) = 4 1 2 5 tan1 mPl , (19)

which is shown in the lower left-hand side of Fig. 4. A t to a Taylor expansion to the exact spectrum, using k as above, yields the following results: ln A2 R = 0.014 ; n = 0.579 ; S (k ) = 3.84 ; 2 dn dn = 0.134 = 0.052 . d ln k d(ln k )2

(20)

From the estimated observational uncertainties of Eq. (16), we see that all these coecients should be successfully determined at high signicance and a simple chibyeye demonstrates that the spectrum reconstructed from these data is an adequate t to the observed spectrum. This stresses the point that the more unusual a potential is, the more information one is likely to be able to extract about it, though the uncertainties on the individual pieces of information may be greater. We reconstruct the potential in a region about = 0.22 (the reconstruction program does not determine ), the width of the region given via Eq. (18). The result for the reconstructed potential including the eect of observational errors and cosmic variance are shown in Fig. 4, where it can immediately be seen that the reconstruction has 15

Figure 4: Sample reconstructions of two inationary potentials. The upper two gures are the reconstruction of a power-law potential of the form V () = V0 exp(/mP l ). The light curves are from the true potential, while the heavy curves are ten reconstructions. The lower gures are the actual potential and ten reconstructions of a potential of the form V () = 4 [1 (2/ ) tan1 (5/mPl )]. been very successful in reproducing the main features of the potential while perturbations on interesting scales are being developed. The uncertainty is dominated by that of A2 T (k ); although the gravitational waves are detectable in this model, it is only about a three-sigma detection and the error bar is thus large. Since the overall magnitude of the potential is proportional to A2 T , the visual impression is of a large uncertainty. Fortunately, information in combinations of the higher derivatives is more accurately determined. Fig. 4 also shows the reconstruction of V /V 3/2 with observational errors; this combination is chosen as it is independent of the tensors to lowest order. Not only is it reconstructed well at the central point, but both the gradient and curvature are well t too, conrming useful information has been obtained about not just V but V as well, which is only possible because of the extra information contained in the scaledependence of the power spectrum. So rather accurate information is being obtained about the potential. 16

Preheating, Reheating, and Dark Matter

If the inaton is completely decoupled, then once ination ends it will oscillate about the minimum of the potential, with the cycle-average of the energy density decreasing as a3 , i.e., as a matter-dominated universe. But at the end of ination the universe is cold and frozen in a low-entropy state: the only degree of freedom is the zero-momentum mode of the inaton eld. It is necessary to defrost the universe and turn it into a hot high-entropy universe with many degrees of freedom in the radiation. Exactly how this is accomplished is still unclear. It probably requires the inaton eld to be coupled to other degrees of freedom, and as it oscillates, its energy is converted to radiation either through incoherent decay, or through a coherent process involving very complicated dynamics of coupled oscillators with time-varying masses. In either case, it is necessary to extract the energy from the inaton and convert it into radiation. I will now turn to a discussion of how defrosting might occur. It may be a complicated several-step process. I will refer to nonlinear eects in defrosting as preheating [2] and refer to linear processes as reheating [18]. The possible role of nonlinear dynamics leading to explosive particle production has recently received a lot of attention. This process, known as preheating [2] may convert a fair fraction of the inaton energy density into other degrees of freedom, with extremely interesting cosmological eects such as symmetry restoration, baryogenesis, or production of dark matter. But the eciency of preheating is very sensitive to the model and the model parameters. Perhaps some relic of defrosting, such as symmetry restoration, baryogenesis, or dark matter may provide a clue of the exact mechanism, and even shed light on ination.

3.1
3.1.1

Defrosting the Universe After Ination


Reheating

In one extreme is the assumption that the vacuum energy of ination is immediately converted to radiation resulting in a reheat temperature TRH . A second (and more plausible) scenario is that reheating is not instantaneous, but is the result of the slow decay of the inaton eld. The simplest way to envision this process is if the comoving energy density in the zero mode of the inaton decays into normal particles, which then scatter and thermalize to form a thermal background. It is usually assumed that the decay width of this process is the same as the decay width of a free inaton eld. There are two reasons to suspect that the inaton decay width might be small. The requisite atness of the inaton potential suggests a weak coupling of the inaton eld to other elds since the potential is renormalized by the inaton coupling to other elds [19]. However, this restriction may be evaded in supersymmetric theories where the nonrenormalization theorem ensures a cancellation between elds and their superpartners. A second reason to suspect weak coupling is that in local supersymmetric theories 17

gravitinos are produced during reheating. Unless reheating is delayed, gravitinos will be overproduced, leading to a large undesired entropy production when they decay after big-bang nucleosynthesis [20]. With the above assumptions, the Boltzmann equations describing the redshift and interchange in the energy density among the dierent components is + 3H + = 0 R + 4HR = 0 ,

(21)

where dot denotes time derivative. The dynamics of reheating will be discussed in detail in Section 3.2.3. The reheat temperature is calculated quite easily [18]. After ination the inaton eld executes coherent oscillations about the minimum of the potential. Averaged over several oscillations, the coherent oscillation energy density redshifts as matter: a3 , where a is the RobertsonWalker scale factor. If we denote as I and aI the total inaton energy density and the scale factor at the initiation of coherent oscillations, then the Hubble expansion rate as a function of a is (MP l is the Planck mass) H (a) = 8 I 2 3 MP l aI a
3

(22)

Equating H (a) and leads to an expression for aI /a. Now if we assume that all available coherent energy density is instantaneously converted into radiation at this value of aI /a, we can dene the reheat temperature by setting the coherent energy density, 4 = I (aI /a)3 , equal to the radiation energy density, R = ( 2 /30)g TRH , where g is the eective number of relativistic degrees of freedom at temperature TRH . The result is TRH = 90 8 3 g
1/4

MP l

200 = 0.2 g
<

1/4

MP l .

(23)

The limit from gravitino overproduction is TRH 3.1.2 Preheating

109 to 1010 GeV.

The main ingredient of the preheating scenario introduced in the early 1990s is the nonperturbative resonant transfer of energy to particles induced by the coherently oscillating inaton elds. It was realized that this nonperturbative mechanism can be much more ecient than the usual perturbative mechanism for certain parameter ranges of the theory [21]. The basic picture can be seen as follows. Assume there is an inaton eld oscillating about the minimum of a potential V () = m2 2 /2. It is convenient to parameterize the motion of the inaton eld as (t) = (t) sin mt. (24) 18

In an expanding universe, even in the absence if interactions the amplitude of the oscillations of the inaton eld, , decreases slowly due to the redshift of the momentum. In Minkowski space, would be a constant in the absence of interactions.. Suppose there is a scalar eld X with a coupling to the inaton of g 2 2 X 2 /2. The mode equation for the X eld can be written in terms of a redened variable k Xk a3/2 as a k2 k + 3 k + 2 + g 22 (t) sin2 (mt) k = 0 . (25) a a In Minkowski space a/a = 0 and (t) is constant, and the mode equation becomes (prime denotes d/dz where z = mt) k + [Ak 2q cos(2z )]k (t) = 0 . (26)

The parameter q depends on the inaton eld oscillation amplitude, and Ak depends on the energy of the particle and q : q = g 2 2 4m2 k2 . m2 (27)

Ak = 2q +

When Ak and q are constants, the equation is the Mathieu equation, which exhibits resonant mode instabilities for certain values of Ak and q . If Ak and q are constant, then there are instability regions where there is explosive particle production. The instability regions are shown in Fig. 5 from the paper of Chung [22]. Integration of the eld equations for the number of particles created in a particular k mode is shown in Fig. 6. Explosive growth occurs every time the inaton passes through the origin. In an expanding universe, Ak and q will vary in time, but if they vary slowly compared to the frequency of oscillations, the eects of resonance will remain. An example of particle production where Ak and q vary due to expansion is shown in Fig. 7. Details of the parameters and the calculation can be found in Kofman, Linde, and Starobinski [2]. If the mode occupation number for the X particles is large, the number density per mode of the X particles will be proportional to |k |2 . If Ak and q have the appropriate values for resonance, k will grow exponentially in time, and hence the number density will attain an exponential enhancement above the usual perturbative decay. This period of enhanced rate of energy transfer has been called preheating primarily because the particles that are produced during this period have yet to achieve thermal equilibrium. This resonant amplication leads to an ecient transfer of energy from the inaton to other particles which may have stronger coupling to other particles than the inaton, thereby speeding up the reheating process and leading to a higher reheating temperature than in the usual scenario. One possible result of preheating is a phase transition caused by the large number of soft particles created in preheating [23]. 19

20 15 10 5 0 -5 2 4 q 6 8 10 A

Figure 5: Shaded areas are regions of explosive particle production. The line is Ak = 2q (recall Ak = 2q + k 2 /m2 ). This gure is from Chung [22].

Figure 6: Explosive particle production if Ak and q are constant. This gure is from Kofman, Linde, and Starobinski [2].

20

Figure 7: Explosive particle production if Ak and q are not constant. The lower gure shows integration for a longer time. These gures are from Kofman, Linde, and Starobinski [2].

21

Another interesting feature is that particles of mass larger than the inaton mass can be produced through this coherent resonant eect. This has been exploited to construct a baryogenesis scenario [24] in which the baryon number violating bosons with masses larger than the inaton mass are created through the resonance mechanism.

3.2

Dark Matter

There are many reasons to believe the present mass density of the universe is dominated by a weakly interacting massive particle (wimp), a fossil relic of the early universe. Theoretical ideas and experimental eorts have focused mostly on production and detection of thermal relics, with mass typically in the range a few GeV to a hundred GeV. Here, I will review scenarios for production of nonthermal dark matter. Since the masses of the nonthermal wimps are in the range 1012 to 1019 GeV, much larger than the mass of thermal wimpy wimps, they may be referred to as wimpzillas. In searches for dark matter it may be well to remember that size does matter. 3.2.1 Thermal RelicsWimpy WIMPS

It is usually assumed that the dark matter consists of a species of a new, yet undiscovered, massive particle, traditionally denoted by X . It is also often assumed that the dark matter is a thermal relic, i.e., it was in chemical equilibrium in the early universe. A thermal relic is assumed to be in local thermodynamic equilibrium (lte) at early times. The equilibrium abundance of a particle, say relative to the entropy density, depends upon the ratio of the mass of the particle to the temperature. Dene the variable Y nX /s, where nX is the number density of WIMP X with mass MX , and s T 3 is the entropy density. The equilibrium value of Y , YEQ , is proportional to exp(x) for x 1, while YEQ constant for x 1, where x = MX /T . A particle will track its equilibrium abundance as long as reactions which keep the particle in chemical equilibrium can proceed rapidly enough. Here, rapidly enough means on a timescale more rapid than the expansion rate of the universe, H . When the reaction rate becomes smaller than the expansion rate, then the particle can no longer track its equilibrium value, and thereafter Y is constant. When this occurs the particle is said to be frozen out. A schematic illustration of this is given in Fig. 8. The more strongly interacting the particle, the longer it stays in lte, and the smaller its eventual freeze-out abundance. Conversely, the more weakly interacting the particle, the larger its present abundance. The freeze-out value of Y is related to the mass of the particle and its annihilation cross section (here characterized by 0 ) by [18] Y 1 . MX mP l 0 (28)

Since the contribution to is proportional to MX nX , which in turn is proportional to MX Y , the present contribution to from a thermal relic roughly is independent of its 22

Figure 8: A thermal relic starts in lte at T MX . When the rates keeping the relic in chemical equilibrium become smaller than the expansion rate, the density of the relic relative to the entropy density freezes out. mass,7 and depends only upon the annihilation cross section. The cross section that results in X h2 1 is of order 1037 cm2 , i.e., of the order of the weak scale. This is one of the attractions of thermal relics. The scale of the annihilation cross section is related to a known mass scale. The simple assumption that dark matter is a thermal relic is surprisingly restric2 tive. The largest possible annihilation cross section is roughly MX . This implies that large-mass wimps would have such a small annihilation cross section that their present abundance would be too large. Thus one expects a maximum mass for a thermal WIMP, which turns out to be a few hundred TeV [25]. The standard lore is that the hunt for dark matter should concentrate on particles with mass of the order of the weak scale and with interaction with ordinary matter on the scale of the weak force. This has been the driving force behind the vast eort in dark matter direct detection. In view of the unitarity argument, in order to consider thermal wimpzillas, one must invoke, for example, late-time entropy production to dilute the abundance of these supermassive particles [26], rendering the scenario unattractive. 3.2.2 Nonthermal RelicsWIMPZILLAS

There are two necessary conditions for the wimpzilla scenario. First, the wimpzilla must be stable, or at least have a lifetime much greater than the age of the universe.
To rst approximation the relic dependence depends upon the mass only indirectly through the dependence of the annihilation cross section on the mass.
7

23

This may result from, for instance, supersymmetric theories where the breaking of supersymmetry is communicated to ordinary sparticles via the usual gauge forces [27]. In particular, the secluded and the messenger sectors often have accidental symmetries analogous to baryon number. This means that the lightest particle in those sectors might be stable and very massive if supersymmetry is broken at a large scale [28]. Other natural candidates arise in theories with discrete gauge symmetries [29] and in string theory and M theory [30, 31]. It is useful here to note that wimpzilla decay might be able to account for ultra-high energy cosmic rays above the GreisenZatzepinKuzmin cuto [32, 33]. A wimpy little thermal relic would be too light to do the job, a wimpzilla is needed. The second condition for a wimpzilla is that it must not have been in equilibrium when it froze out (i.e., it is not a thermal relic), otherwise X h2 would be much larger than one. A sucient condition for nonequilibrium is that the annihilation rate (per particle) must be smaller than the expansion rate: nX |v | < H , where |v | is the annihilation rate times the Mller ux factor, and H is the expansion rate. Conversely, if the dark matter was created at some temperature T and X h2 < 1, then it is easy to show that it could not have attained equilibrium. To see this, assume X s were created in a radiation-dominated universe at temperature T . Then X h2 is given by X h2 = h2 (T /T0 )mX nX (T )/ (T ) , (29)

2 where T0 is the present temperature. Using the fact that (T ) = H (T )MP l T , 2 nX (T )/H (T ) = (X / )T0 MP l T /MX . One may safely take the limit |v | < MX , 3 so nX (T ) |v |/H (T) must be less than (X / )T0 MP l T /MX . Thus, the requirement for nonequilibrium is 200 TeV 2 T <1. (30) MX MX > 200 TeV was created at T < M This implies that if a nonrelativistic particle with MX X < with a density low enough to result in X 1, then its abundance must have been so small that it never attained equilibrium. Therefore, if there is some way to create wimpzillas in the correct abundance to give X 1, nonequilibrium is automatic. Examples of wimpzilla evolution and freezeout are shown in Fig. 9. Any wimpzilla production scenario must meet these two criteria. Before turning to several wimpzilla production scenarios, it is useful to estimate the fraction of the total energy density of the universe in wimpzillas at the time of their production that will eventually result in 1 today. The most likely time for wimpzilla production is just after ination. The rst step in estimating the fraction of the energy density in wimpzillas is to estimate the total energy density when the universe is reheated after ination. Consider the calculation of the reheat temperature, denoted as TRH . The reheat temperature is calculated by assuming an instantaneous conversion of the energy density in the inaton eld into radiation when the decay width of the inaton energy, , is

24

Figure 9: A nonthermal relic typically has abundance much less than lte at T MX . Here are two examples of the evolution of a nonthermal relic. In the left panel Y YEQ at freezeout, while in the right panel Y YEQ at freezeout. Again, the solid curve is the lte abundance while the dashed curve is the actual abundance. equal to H , the expansion rate of the universe. The derivation of the reheat temperature was given in Section 3.1.1. Now consider the wimpzilla density at reheating. Suppose the wimpzilla never attained lte and was nonrelativistic at the time of production. The usual quantity X h2 associated with the dark matter density today can be related to the dark matter density when it was produced. First write X (tRH ) X (t0 ) = R (t0 ) R (tRH ) TRH , T0 (31)

where R denotes the energy density in radiation, X denotes the energy density in the dark matter, TRH is the reheat temperature, T0 is the temperature today, t0 denotes the time today, and tRH denotes the approximate time of reheating.8 To obtain X (tRH )/R (tRH ), one must determine when X particles are produced with respect to the completion of reheating and the eective equation of state between X production and the completion of reheating. At the end of ination the universe may have a brief period of matter domination resulting either from the coherent oscillations phase of the inaton condensate or from the preheating phase [21]. If the X particles are produced at time t = te when the de Sitter phase ends and the coherent oscillation period just begins, then both the X particle energy density and the inaton energy density will redshift at approximately the same rate until reheating is completed and radiation domination begins. Hence, the
More specically, this is approximately the time at which the universe becomes radiation dominated after ination.
8

25

ratio of energy densities preserved in this way until the time of radiation domination is 8 X (te ) X (tRH ) , 2 2 R (tRH ) 3 MP l H (te ) (32)

where MP l 1019 GeV is the Planck mass and most of the energy density in the universe just before time tRH is presumed to turn into radiation. Thus, using Eq. 31, one may 2 2 obtain an expression for the quantity X X (t0 )/C (t0 ), where C (t0 ) = 3H0 MP l /8 1 1 and H0 = 100 h km sec Mpc : X h2 R h2 TRH T0 8 3 MX MP l nX (te ) . MP l H 2 (te ) (33)

Here R h2 4.31 105 is the fraction of critical energy density in radiation today and nX is the density of X particles at the time when they were produced. Note that because the reheating temperature must be much greater than the temper> 4.2 1014 ), in order to satisfy the cosmological bound h2 < 1, ature today (TRH /T0 X the fraction of total wimpzilla energy density at the time when they were produced must be extremely small. One sees from Eq. 33 that X h2 1017 (TRH /109GeV)(X (te )/(te )). It is indeed a very small fraction of the total energy density extracted in wimpzillas. This means that if the wimpzilla is extremely massive, the challenge lies in creating very few of them. Gravitational production discussed in Section 3.2.3 naturally gives the needed suppression. Note that if reheating occurs abruptly at the end of ination, then the matter domination phase may be negligibly short and the radiation domination phase may follow immediately after the end of ination. However, this does not change Eq. 33. 3.2.3 Wimpzilla Production

3.2.3.a Gravitational Production First consider the possibility that wimpzillas are produced in the transition between an inationary and a matter-dominated (or radiation-dominated) universe due to the nonadiabatic expansion of the background spacetime acting on the vacuum quantum uctuations [34]. The distinguishing feature of this mechanism is the capability of generating particles with mass of the order of the inaton mass (usually much larger than the reheating temperature) even when the particles only interact extremely weakly (or not at all) with other particles and do not couple to the inaton. They may still be produced in sucient abundance to achieve critical density today due to the classical gravitational eect on < M /H < 2, where the vacuum state at the end of ination. More specically, if 0.04 X I 13 HI m 10 GeV is the Hubble constant at the end of ination (m is the mass of the inaton), wimpzillas produced gravitationally can have a density today of the order of the critical density. This result is quite robust with respect to the ne details 26

Figure 10: The contribution of gravitationally produced wimpzillas to X h2 as a function of MX /HI . The shaded area is where thermalization may occur if the annihilation cross section is its maximum value. Also shown is the contribution assuming that the wimpzilla is present at the end of ination with a temperature T = HI /2 . of the transition between the inationary phase and the matter-dominated phase, and independent of the coupling of the wimpzilla to any other particle. Conceptually, gravitational wimpzilla production is similar to the inationary generation of gravitational perturbations that seed the formation of large scale structures. In the usual scenarios, however, the quantum generation of energy density uctuations from ination is associated with the inaton eld that dominated the mass density of the universe, and not a generic, sub-dominant scalar eld. Another dierence is that the usual density uctuations become larger than the Hubble radius, while most of the wimpzilla perturbations remain smaller than the Hubble radius. There are various inequivalent ways of calculating the particle production due to interaction of a classical gravitational eld with the vacuum (see for example [35, 36, 37]). Here, I use the method of nding the Bogoliubov coecient for the transformation be< 1 the tween positive frequency modes dened at two dierent times. For MX /HI results are quite insensitive to the dierentiability or the ne details of the time depen< M /H < 10, all the dark matter needed for closure dence of the scale factor. For 0.03 X I of the universe can be made gravitationally, quite independently of the details of the transition between the inationary phase and the matter dominated phase. Start with the canonical quantization of the X eld in an action of the form (with metric ds2 = dt2 a2 (t)dx2 = a2 ( ) [d 2 dx2 ] where is conformal time) S= dt d3x a3 2 (X )2 2 X MX X 2 RX 2 2 a2 27 (34)

where R is the Ricci scalar. After transforming to conformal time coordinate, use the mode expansion X (x) = d3k ikx , ak hk ( )eikx + a k hk ( )e 3 / 2 (2 ) a( ) (35)

where because the creation and annihilation operators obey the commutator [ak1 , a k2 ] = (3) (k1 k2 ), the hk s obey a normalization condition hk hk hk hk = i to satisfy the canonical eld commutators (henceforth, all primes on functions of refer to derivatives with respect to ). The resulting mode equation is
2 h k ( ) + wk ( )hk ( ) = 0,

(36) (37)

where The parameter is 1/6 for conformal coupling and 0 for minimal coupling. From now on, = 1/6 for simplicity but without much loss of generality. By a change in variable k/a, one can rewrite the dierential equation such that it depends only on H ( ), H ( )/k , k/a( ), and MX . Hence, the parameters HI and aI correspond to the Hubble parameter and the scale factor evaluated at an arbitrary conformal time I , which can be taken to be the approximate time at which X s are produced (i.e., I is the conformal time at the end of ination). One may then rewrite Eq. 36 as
2 MX 2 h ( ) + k + ) = 0 , a 2 hk ( k 2 HI 2 2 2 wk = k 2 + MX a + (6 1)a /a .

(38)

= k/(aI HI ). For simplicity of notation, drop all where = aI HI , a = a/aI , and k the tildes. This dierential equation can be solved once the boundary conditions are supplied. The number density of the wimpzillas is found by a Bogoliubov transformation from the vacuum mode solution with the boundary condition at = 0 (the initial time at which the vacuum of the universe is determined) into the one with the boundary condition at = 1 (any later time at which the particles are no longer being created). 0 will be taken to be while 1 will be taken to be at +. Dening the Bogoliubov 0 0 1 transformation as h k ( ) = k hk ( ) + k hk ( ) (the superscripts denote where the boundary condition is set), the energy density of produced particles is X (1 ) = MX nX (1 ) =
3 MX HI

1 a (1 )

3 0

dk 2 | |2 , k k 2 2

(39)

where one should note that the number operator is dened at 1 while the quantum state (approximated to be the vacuum state) dened at 0 does not change in time in the Heisenberg representation. 28

Figure 11: The evolution of the Bogoliubov coecient with conformal time for several wavenumbers. = I corresponds to the end of the inationary era. As one can see from Eq. 38, the input parameter is MX /HI . One must also specify the behavior of a( ) near the end of ination. In Fig. 10 (from [34]), I show the resulting values of X h2 as a function of MX /HI assuming the evolution of the scale factor smoothly interpolates between exponential expansion during ination and either a matter-dominated universe or radiation-dominated universe. The peak at MX /HI 1 is similar to the case presented in Ref. [38]. As expected, for large MX /HI , the number density falls o faster than any inverse power of MX /HI . Now most of the action occurs around the transition from ination to the matterdominated or radiation-dominated universe. This is shown in Fig. 11. Also from Fig. 11 one can see that most of the particles are created with wavenumber of order HI . To conclude, there is a signicant mass range (0.03HI to 10HI , where HI 1013 GeV) for which wimpzillas will have critical density today regardless of the ne details of the transition out of ination. Because this production mechanism is inherent in the dynamics between the classical gravitational eld and a quantum eld, it needs no ne tuning of eld couplings or any coupling to the inaton eld. However, only if the particles are stable (or suciently long lived) will these particles give contribution of the order of critical density. 3.2.3.b Production During Reheating Another attractive origin for wimpzillas is during the defrosting phase after ination. It is important to recall that it is not necessary to convert a signicant fraction of the available energy into massive particles; in fact, it must be an innitesimal amount. I now will discuss how particles of mass much greater than TRH may be created in the correct amount after ination in reheating [39]. 29

In one extreme is the assumption that the vacuum energy of ination is immediately converted to radiation resulting in a reheat temperature TRH . In this case X can be calculated by integrating the Boltzmann equation with initial condition NX = 0 at T = TRH . One expects the X density to be suppressed by exp(2MX /TRH ); indeed, one nds X 1 for MX /TRH 25+0.5 ln(m2 X |v | ), in agreement with previous estimates [32] that for TRH 109 GeV, the wimpzilla mass would be about 2.5 1010 GeV. It is simple to calculate the wimpzilla abundance in the slow reheating scenario. It will be important to keep in mind that what is commonly called the reheat temperature, TRH , is not the maximum temperature obtained after ination. The maximum temperature is, in fact, much larger than TRH . The reheat temperature is best regarded as the temperature below which the universe expands as a radiation-dominated universe, 1/3 with the scale factor decreasing as g T 1 . In this regard it has a limited meaning [18, 40]. One implication of this is that it is incorrect to assume that the maximum abundance of a massive particle species produced after ination is suppressed by a factor of exp(M/TRH ). To estimate wimpzilla production in reheating, consider a model universe with three components: inaton eld energy, , radiation energy density, R , and wimpzilla energy density, X . Assume that the decay rate of the inaton eld energy density is . Also assume the wimpzilla lifetime is longer than any timescale in the problem (in fact it must be longer than the present age of the universe). Finally, assume that the light degrees of freedom are in local thermodynamic equilibrium. With the above assumptions, the Boltzmann equations describing the redshift and interchange in the energy density among the dierent components is [cf. Eq. (21)] + 3H + = 0 R + 4HR X + 3HX + |v | MX |v | MX
EQ 2 X X 2 2

=0 (40)

EQ 2 X X

=0,

where dot denotes time derivative. As already mentioned, |v | is the thermal average of the X annihilation cross section times the Mller ux factor. The equilibrium energy density for the X particles, EQ X , is determined by the radiation temperature, T = 2 1/4 (30R / g ) . It is useful to introduce two dimensionless constants, and X , dened in terms of and |v | as 2 = M |v | = X MX . (41) For a reheat temperature much smaller than M , must be small. From Eq. (23), the 1/2 M MP l . For M = 1013 GeV, reheat temperature in terms of X and MX is TRH must be smaller than of order 1013 . On the other hand, X may be as large as of order unity, or it may be small also. 30

It is also convenient to work with dimensionless quantities that can absorb the eect of the expansion of the universe. This may be accomplished with the denitions
1 3 M a ;

R R a4 ;

1 3 X X MX a .

(42)

It is also convenient to use the scale factor, rather than time, for the independent variable, so one may dene a variable x = aM . With this choice the system of equations can be written as (prime denotes d/dx) = c1 R X x x + R x2 x1 = c1 + c2 x + R x + R 2 x 2 . X 2 XEQ = c3 x + R

2 X 2 XEQ

(43)

The constants c1 , c2 , and c3 are given by c1 = 3 MP l 8 M c2 = c1 M X MX c3 = c2 M . MX (44)

XEQ is the equilibrium value of X , given in terms of the temperature T as (assuming a single degree of freedom for the X species) XEQ =
3 MX 3 M

1 2

3/2

x3

T MX

3/2

exp(MX /T ) .

(45)

The temperature depends upon R and g , the eective number of degrees of freedom in the radiation: 1/4 M R1/4 30 T = . (46) MX g 2 MX x It is straightforward to solve the system of equations in Eq. (43) with initial conditions at x = xI of R(xI ) = X (xI ) = 0 and (xI ) = I . It is convenient to express (x = xI ) in terms of the expansion rate at xI , which leads to I =
2 2 3 MP l HI x3 . 2 2 8 M M I

(47)

The numerical value of xI is irrelevant. Before solving numerically the system of equations, it is useful to consider the earlytime solution for R. Here, early times means H , i.e., before a signicant fraction of the comoving coherent energy density is converted to radiation. At early times I , 31

Figure 12: The evolution of energy densities and T /MX as a function of the scale factor. Also shown is X/XEQ. and R X 0, so the equation for R becomes R = c1 x3/2 I . Thus, the early time solution for R is simple to obtain: 2 1/2 5/2 R c1 x5/2 xI I 5 ( H ) . (48)
1/2

Now express T in terms of R to yield the early-time solution for T : T M 12 2 g


1/4

c1

1/4

I x3 I

1/8

x xI

3/2

x xI

4 1/4

( H ) .

(49)

Thus, T has a maximum value of TM AX M 12 = 0.77 2 g =


1/4 0.77 1/4 1/4 c1 1/4

I x3 I

1/8

9 2 3 g

2 MP l HI 3 M

1/4

(50)

which is obtained at x/xI = (8/3)2/5 = 1.48. It is also possible to express in terms of TRH and obtain 1/4 1/8 HI MP l 9 TM AX = 0.77 . (51) 2 TRH 5 3 g TRH For an illustration, in the simplest model of chaotic ination HI M with M 10 GeV, which leads to TM AX /TRH 103 (200/g )1/8 for TRH = 109 GeV.
13

32

We can see from Eq. (48) that for x/xI > 1, in the early-time regime T scales as a , which implies that entropy is created in the early-time regime [40]. So if one is producing a massive particle during reheating it is necessary to take into account the fact that the maximum temperature is greater than TRH , and that during the early-time evolution, T a3/8 . An example of a numerical evaluation of the complete system in Eq. (43) is shown in Fig. 12 (from [39]). The model parameters chosen were M = 1013 GeV, = 2 1013 , MX = 1.15 1012 GeV, X = 102 , and g = 200. The expansion rate at the beginning of the coherent oscillation period was chosen to be HI = M . These parameters result in TRH = 109 GeV and X h2 = 0.3. Figure 12 serves to illustrate several aspects of the problem. Just as expected, the comoving energy density of (i.e., a3 ) remains roughly constant until H , which for the chosen model parameters occurs around a/aI 5 108 . But of course, that does not mean that the temperature is zero. Notice that the temperature peaks well before reheating. The maximum temperature, TM AX = 1012 GeV, is reached at a/aI slightly larger than unity (in fact at a/aI = 1.48 as expected), while the reheat temperature, TRH = 109 GeV, occurs much later, around a/aI 108 . Note that TM AX 103 TRH in agreement with Eq. (51). From the gure it is clear that X XEQ at the epoch of freeze out of the comoving X number density, which occurs around a/aI 102 . The rapid rise of the ratio after freeze out is simply a reection of the fact that X is constant while XEQ decreases exponentially. A close examination of the behavior of T shows that after the sharp initial rise of the temperature, the temperature decreases as a3/8 [as follows from Eq. (49)] until H , and thereafter T a1 as expected for the radiation-dominated era. For the choices of M , , g , and X used for the model illustrated in Fig. 12, X h2 = 0.3 for MX = 1.15 1012 GeV, in excellent agreement with the mass predicted by using an analytic estimate for the result [39]
3/8 2 X h2 = MX |v |

g 200

3/2

2000TRH MX

(52)

Here again, the results have also important implications for the conjecture that ultrahigh cosmic rays, above the Greisen-Zatsepin-Kuzmin cut-o of the cosmic ray spectrum, may be produced in decays of superheavy long-living particles [31, 32, 33, 41]. In order to produce cosmic rays of energies larger than about 1013 GeV, the mass of the X -particles > 1013 GeV and their lifetime must be very large, MX X cannot be smaller than the 10 > age of the Universe, X 10 yr. With the smallest value of the lifetime, the observed ux of ultra-high energy cosmic rays will be reproduced with a rather low density of X -particles, X 1012 . It has been suggested that X -particles can be produced in the right amount by usual collisions and decay processes taking place during the reheating stage after ination if the reheat temperature never exceeded MX [41]. Again, assuming naively that that the maximum number density of a massive particle species X produced 33

after ination is suppressed by a factor of (MX /TRH )3/2 exp(MX /TRH ) with respect to the photon number density, one concludes that the reheat temperature TRH should be in the range 1011 to 1015 GeV [32]. This is a rather high value and leads to the gravitino problem in generic supersymmetric models. This is one reason alternative production mechanisms of these superheavy X -particles have been proposed [34, 42, 43]. However, our analysis show that the situation is much more promising. Making use of Eq. (52), the right amount of X -particles to explain the observed ultra-high energy cosmic rays is produced for TRH g 3/14 MX , (53) 10 15 10 GeV 200 10 GeV
2 where it has been assumed that |v | MX . Therefore, particles as massive as 1015 GeV may be generated during the reheating stage in abundances large enough to explain the ultra-high energy cosmic rays even if the reheat temperature satises the gravitino bound.

3.2.3.c Production During Preheating Another way to produce wimpzillas after ination is in a preliminary stage of reheating called preheating [21], where nonlinear quantum eects may lead to an extremely eective dissipational dynamics and explosive particle production. As discussed in Section 3.1.2, particles can be created in a broad parametric resonance with a fraction of the energy stored in the form of coherent inaton oscillations at the end of ination released after only a dozen oscillation periods. A crucial observation for our discussion is that particles with mass up to 1015 GeV may be created during preheating [42, 24, 44], and that their distribution is nonthermal. If these particles are stable, they may be good candidates for wimpzillas [22]. Interestingly enough, what was found [22] is that in the context of a slow-roll ination 2 2 2 2 with the potential V () = m2 /2 with the inaton coupling of g X /2, the resonance phenomenon is mostly irrelevant to wimpzilla production because too many particles would be produced if the resonance is eective. For the tiny amount of energy conversion needed for wimpzilla production, the coupling g 2 must be small enough (for a xed MX ) such that the motion of the inaton eld at the transition out of the inationary phase generates just enough nonadiabaticity in the mode frequency to produce wimpzillas. The rest of the oscillations, damped by the expansion of the universe, will not contribute signicantly to wimpzilla production as in the resonant case. In other words, the quasiperiodicity necessary for a true resonance phenomenon is not present in the case when only an extremely tiny fraction of the energy density is converted into wimpzillas. Of course, if the energy scales are lowered such that a fair fraction of the energy density can be converted to wimpzillas without overclosing the universe, this argument may not apply. The main nding of a detailed treatment [22] is that wimpzillas with a mass as large as 103 HI , where HI is the value of the Hubble expansion rate at the end of ination, can be produced in sucient abundance to be cosmologically signicant today. 34

Figure 13: A graph of X h2 versus MX /HI for gMP l /HI = 106 . The solid curve is a numerical result, while the dashed and dotted curves are analytic approximations [22]. If the wimpzilla is coupled to the inaton by a term g 2 2 X 2 /2, then the mode equation in Eq. 37 is now changed to
2 2 + g 2 2 a2 , k + k 2 + MX

(54)

again taking = 1/6. The procedure to calculate the wimpzilla density is the same as in Section 3.2.3. Now, in addition to the parameter MX /HI , there is another parameter gMP l /HI . Now in large-eld models HI 1013 GeV, so MP l /HI might be as large as 106 . The choice of g = 103 would yield gMP l /HI = 103 . Fig. 13 (from [22]) shows the dependence of the wimpzilla density upon MX /HI for the particular choice gMP l /HI = 106 . This would correspond to g 1 in largeeld ination models where MP l /HI = 106 , about the largest possible value. Note that X 1 obtains for MX /HI 103 . The dashed and dotted curves are two analytic approximations discussed in [22], while the solid curve is the numerical result. The approximations are in very good agreement with the numerical results. Fig. 14 (from [22]) shows the dependence of the wimpzilla density upon gMP l /HI . For this graph MX /HI was chosen to be unity. This gure illustrates the fact that the dependence of X h2 on gMP l /HI is not monotonic. For a detailed explanation of this curious eect, see the paper of Chung [22]. 3.2.3.d Production in Bubble Collisions wimpzillas may also be produced [43] if ination is completed by a rst-order phase transition [45], in which the universe exits from a false-vacuum state by bubble nucleation [46]. When bubbles of true vacuum form, the energy of the false vacuum is entirely 35

Figure 14: An illustration of the nonmonotonic behavior of the particle density produced with the variation of the coupling constant. The value of MX /HI is set to unity [22]. transformed into potential energy in the bubble walls. As the bubbles expand, more and more of their energy becomes kinetic as the walls become highly relativistic. In bubble collisions the walls oscillate through each other [47] and their kinetic energy is dispersed into low-energy scalar waves [47, 48]. We are interested in the potential energy of the walls, MP = 4R2 , where is the energy per unit area of a bubble wall of radius R. The bubble walls can be visualized as a coherent state of inaton particles, so the typical energy E of the products of their decays is simply the inverse thickness of the wall, E 1 . If the bubble walls are highly relativistic when they collide, there is the possibility of quantum production of nonthermal particles with mass well above the mass of the inaton eld, up to energy 1 = M , with the relativistic Lorentz factor. Suppose for illustration that the wimpzilla is a fermion coupled to the inaton eld by a Yukawa coupling gXX . One can treat (the bubbles or walls) as a classical, external eld and the wimpzilla as a quantum eld in the presence of this source. The number of wimpzillas created in the collisions from the wall potential energy is NX fX MP /MX , where fX parametrizes the fraction of the primary decay products in wimpzillas. The fraction fX will depend in general on the masses and the couplings of a particular theory in question. For the Yukawa coupling g , it is fX g 2ln (M /2MX ) [48, 49]. wimpzillas may be produced in bubble collisions out of equilibrium and never attain chemical equilibrium. Even with TRH as low as 100 GeV, the present wimpzilla abundance would be X 1 if g 105 1/2 . Here 1 1 is the fraction of the bubble energy at nucleation in the form of potential energy at the time of collision. This simple analysis indicates that the correct magnitude for the abundance of wimpzillas may be naturally obtained in the process of reheating in theories where ination is terminated 36

by bubble nucleation. 3.2.4 Wimpzilla Conclusions

In this talk I have pointed out several ways to generate nonthermal dark matter. All of the methods can result in dark matter much more massive than the feeble little weakscale mass thermal relics. The nonthermal dark matter may be as massive as the GUT scale, truly in the wimpzilla range. The mass scale of the wimpzillas is determined by the mass scale of ination, more exactly, the expansion rate of the universe at the end of ination. For large-eld ination models, that mass scale is of order 1013 GeV. For small-eld ination models, it may be less, perhaps much less. The mass scale of ination may one day be measured! In addition to scalar density perturbations, tensor perturbations are produced in ination. The tensor perturbations are directly proportional to the expansion rate during ination, so determination of a tensor contribution to cosmic background radiation temperature uctuations would give the value of the expansion rate of the universe during ination and set the scale for the mass of the wimpzilla. Undoubtedly, other methods for wimpzilla production will be developed. But perhaps even with the present scenarios one should start to investigate methods for wimpzilla detection. While wimpy wimps must be color singlets and electrically neutral, wimpzillas may be endowed with color and electric charge. This should open new avenues for detection and exclusion of wimpzillas. The lesson here is that wimpzillas may surprise and be the dark matter, and we may learn that size does matter!

Acknowledgements
This work was supported by the DOE and NASA under Grant NAG5-7092. This is an expanded version of my paper Early-Universe Issues: Seeds of Structure and Dark Matter, which was part of the proceedings of the Nobel Symposium Particles and the Universe.

References
[1] A. H. Guth, Phys. Rev. D23, 347 (1981). [2] L. Kofman, A. D. Linde, and A. A. Starobinsky, Phys. Rev. D 56, 3258 (1997). [3] For a review, see V. F. Mukhanov, H. A. Feldman, and R. H. Brandenberger. Phys. Rep. 215, 203 (1992). [4] S. Dodelson, W. H. Kinney, and E. W. Kolb, Phys. Rev. D56, 3207 (1997) . 37

[5] J. E. Lidsey, A. R. Liddle, E. W. Kolb, E. J. Copeland, T. Barreiro and M. Abney, Rev. Mod. Phys. 69, 373 (1997). [6] E. J. Copeland, E. W. Kolb, A. R. Liddle and J. E. Lidsey, Phys. Rev. Lett. 71, 219 (1993), Phys. Rev. D 48, 2529 (1993), 49, 1840 (1994); M. S. Turner, Phys. Rev. D 48, 5539 (1993). [7] A. R. Liddle and D. H. Lyth, Phys. Rep. 231, 1 (1993). [8] I. J. Grivell and A. R. Liddle, Phys. Rev. D 54, 7191 (1996). [9] D. S. Salopek and J. R. Bond, Phys. Rev. D 42, 3936 (1990). [10] A. R. Liddle and D. H. Lyth, Phys. Lett. B 291, 391 (1992). [11] A. R. Liddle, P. Parsons and J. D. Barrow, Phys. Rev. D 50, 7222 (1994). [12] E. D. Stewart and D. H. Lyth, Phys. Lett. B 302, 171 (1993). [13] G. Jungman, M. Kamionkowski, A. Kosowsky and D. N. Spergel, Phys. Rev. D 54, 1332 (1996); J. R. Bond, G. Efstathiou and M. Tegmark, Mon. Not. R. Astron. Soc. 291, L33 (1997). [14] M. Zaldarriaga, D. Spergel and U. Seljak, Astrophys. J. 488, 1 (1997). [15] E. J. Copeland, I. J. Grivell and A. R. Liddle, Sussex preprint astro-ph/9712028 (1997). [16] E. J. Copeland, I. J. Grivell, E. W. Kolb, and A. R. Liddle, Phys. Rev. D58 043002 (1998) . [17] L. Wang, V. F. Mukhanov and P. J. Steinhardt, Phys. Lett. B 414, 18 (1997). [18] E. W. Kolb and M. S. Turner, The Early Universe, (Addison-Wesley, Menlo Park, Ca., 1990). [19] D. H. Lyth and A. Riotto, hep-ph/9807278. [20] J. Ellis, J. Kim and D. V. Nanopoulos, Phys. Lett. B145, 181 (1984); L. M. Krauss, Nucl. Phys. B227, 556 (1983); M. Yu. Khlopov and A. D. Linde, Phys. Lett. 138B, 265 (1984); J. Ellis, D. V. Nanopoulos, and S. Sarkar, Nucl. Phys. B461, 597 (1996). [21] L. A. Kofman, A. D. Linde and A. A. Starobinsky, Phys. Rev. Lett. 73, 3195 (1994); S. Yu. Khlebnikov and I. I. Tkachev, Phys. Rev. Lett. 77, 219 (1996); Phys. Lett. B390, 80 (1997); Phys. Rev. Lett. 79, 1607 (1997); Phys. Rev. D 56, 653 (1997); G. W. Anderson, A. Linde and A. Riotto, Phys. Rev. Lett. 77, 3716 (1996); see

38

L. Kofman, The origin of matter in the Universe: reheating after ination, astroph/9605155, UH-IFA-96-28 preprint, 16pp., to appear in Relativistic Astrophysics: A Conference in Honor of Igor Novikovs 60th Birthday, eds. B. Jones and D. Markovic for a more recent review and a collection of references; see also L. Kofman, A. D. Linde and A. A. Starobinsky, Phys. Rev. D 56, 3258 (1997); J. Traschen and R. Brandenberger, Phys. Rev. D 42, 2491 (1990); Y. Shtanov, J. Traschen, and R. Brandenberger, Phys. Rev. D 51, 5438 (1995). [22] D. J. Chung, hep-ph/9809489. [23] L. Kofman, A. D. Linde, and A. A. Starobinski, Phys. Rev. Lett. 76, 1011 (1996); I. I. Tkachev, Phys. Lett. B376, 35 (1996). [24] E.W. Kolb, A. D. Linde and A. Riotto, Phys. Rev. Lett. 77, 4290 (1996). [25] K. Griest and M. Kamionkowski, Phys. Rev. Lett. 64, 615 (1990). [26] J. Ellis, J.L. Lopez and D.V. Nanopoulos, Phys. Lett. B247, 257 (1990). [27] See, for instance, G. F. Giudice and R. Rattazzi, hep-ph/9801271. [28] S. Raby, Phys. Rev. D 56, (1997). [29] K. Hamaguchi, Y. Nomura and T. Yanagida, hep-ph/9805346. [30] K. Benakli, J. Ellis and D.V. Nanopoulos, hep-ph/9803333. [31] D. V. Nanopoulos, Proceedings of DARK98, (1998) L. Baudis and H. KlapdorKleingrothaus, eds. [32] V. A. Kuzmin and V. A. Rubakov, Phys. Atom. Nucl. 61, 1028 (1998). [33] M. Birkel and S. Sarkar, hep-ph/9804285. [34] D. J. H. Chung, Edward W. Kolb, and A. Riotto, hep-ph/9802238. [35] S. Fulling, Gen. Rel. and Grav. 10, 807 (1979). [36] N. D. Birrell and P. C. W. Davies, Quantum Fields in Curved Space (Cambridge University Press, Cambridge, 1982). [37] D. M. Chitre and J. B. Hartle, Phys. Rev. D 16, 251 (1977); D. J. Raine and C. P. Winlove, Phys. Rev. D 12, 946 (1975); G. Schaefer and H. Dehnen, Astron. Astrophys. 54, 823 (1977). [38] N. D. Birrell and P. C. W. Davies, J. Phys. A:Math. Gen. 13, 2109 (1980). [39] D. J. H. Chung, E. W. Kolb, and A. Riotto, hep-ph/9809453. 39

[40] R. J. Scherrer and M. S. Turner, Phys. Rev. D31, 681 (1985). [41] V. Berezinsky, M. Kachelriess and A. Vilenkin, Phys. Rev. Lett. 79, 4302 (1997). [42] V. Kuzmin and I. I. Tkachev, hep-ph/9802304. [43] D. J. H. Chung, E. W. Kolb and A. Riotto, hep-ph/9805473. [44] E. W. Kolb, A. Riotto and I. I. Tkachev, Phys. Lett. B423, 348 (1998). [45] D. La and P. J. Steinhardt, Phys. Rev. Lett. 62, 376 (1989). [46] A. H. Guth, Phys. Rev. D 23, 347 (1981). [47] S. W. Hawking, I. G. Moss and J. M. Stewart, Phys. Rev. D 26, 2681 (1982). [48] R. Watkins and L. Widrow, Nucl. Phys. B374, 446 (1992). [49] A. Masiero and A. Riotto, Phys. Lett. B289, 73 (1992).

40

You might also like