Reheating and Preheating after Inflation : an Introduction

Kai Schmitz and Gilles Vertongen

(Dated: November 23, 2010)

These notes are intended to give a global picture of the general mechanism responsible for the
creation of the hot Big Bang Universe as we are observing it. They are mainly based on Mukhanov’s
book [1], as well as on the Kofman et al. articles [2, 3], from which the plots have also been taken.

I. REHEATING where

MP
Φ(t) = √ . (8)
3π mt
A. Evolution of the inflaton field
This solution has a simple interpretation which is illus-
trated in Fig. 1: at the end of the inflationary era, the
The dynamics of the inflaton field is described by a friction term 3H φ̇ becomes subdominant and Eq. (1) de-
Klein-Gordon equation coupled to a Friedmann equation scribes an oscillator, the amplitude of which gets damped
due to the universe expansion.
0 = φ̈ + 3H φ̇ + m2 φ , (1) The behavior of the scale factor can be extracted from
 
8π 1 2 Eq. (6) :
H2 = φ̇ + V (φ) . (2)
3MP2 2
a(t) ∝ t2/3 . (9)
For chaotic inflation V (φ) = 21 mφ2 , Eq. (2) can be conve-
niently parametrized using the Hubble parameter H and In consequence, the energy density of the field φ decreases
the angular variable θ defined via in the same way as the energy density of non relativistic
r r particles of mass m:
3 3
φ̇ = HMP sin θ , mφ = HMP cos θ . (3)
4π 4π 1 2 1 2 2
ρφ = φ̇ + m φ ∼ a−3 . (10)
2 2
Using Eq. (1), the dynamics of these two independent
variables is described by The inflaton oscillations can be interpreted as a collection
of scalar particles, independent from each other, oscillat-
Ḣ = −3H 2 sin2 θ , (4) ing coherently at the same frequency m.
3
θ̇ = −m − 2H sin 2θ . (5) For time intervals larger than the oscillating period,
the energy and number densities are related to the am-
For mt  1, the second term in the r.h.s of Eq. (5) can plitude Φ in a simple way
be neglected ; the scalar field φ thus oscillates with a
frequency ω ' m. The Hubble rate can then be extracted 1 2 2 1
ρφ = m Φ , nφ = mΦ2 . (11)
from Eq. (4) 2 2

   −1
ȧ 2 sin(2mt)
H(t) ≡ = 1− , (6)
a 3t 2mt B. Decay of the scalar field

where the oscillating term is small compared to unity.

The behavior of φ(t) is obtained by solving Eq. (3) using The reheating process occurs when the inflaton energy
an expansion in (mt)−1 density is transferred to the energy density of other fields,
leading to a decrease of the oscillating amplitude much
 
sin(2mt) faster than the one described in Eq. (7).
φ(t) ' Φ(t) cos(mt) 1 + , (7)
2mt
Consider the interaction of the inflaton field φ with a
 2

scalar field χ and a fermion field ψ due to the universe expansion as well as due to particle
production. The field Φ(t) obeys the equation
,i
L = 1
2 φ,i φ− V (φ) + 12 χ,i χ,i − 21 m2χ (0)χ2
1 d 3 2
i
+ ψ̄ iγ ∂i − mψ (0) ψ − 12 g 2 φ2 χ2 − hψ̄ψφ . (12) (a Φ ) = −ΓΦ2 .


(21)
a3 dt

To be general, we suppose that the effective potential Multiplying the latter by m and using Eq. (11), one ob-
possess a minimum at φ = σ, and is quadratic in φ tains the following equation for the number density

2
V (φ) ∼ 21 m2 (φ − σ) . (13) d 3
(a nφ ) = −a3 nφ Γ . (22)
dt
Doing the usual shift φ−σ → φ after symmetry breaking,
This is nothing else than a so-called Boltzmann equation,
the potential gets interaction terms linear in φ
showing that the comoving number density of φ particles
∆L = −g 2 σφχ2 − hψ̄ψφ , (14) exponentially decreases with the decay rate Γ.

which eventually lead to inflaton tree-body decays

C. The reheating temperature
4 2 2
g σ h m
Γ(φ → χχ) = , Γ(φ → ψ̄ψ) = . (15)
8πm 8π During the oscillating phase the Universe behaves in
the same way as if it was dominated by non-relativistic
particles of mass m : H(t) ∼ 2/(3t). The inflaton energy
In order to consider the effects related to particle pro- density is then transferred to the relativistic decay prod-
duction, let us consider the following Klein Gordon equa- ucts, the energy density of which decreasing much faster
tion describing the homogeneous scalar field than the energy of the oscillating field φ. The reheating
process eventually ends when H < Γφ . The reheating
φ̈ + 3H(t) φ̇ + m2 + Π(ω) φ = 0 .


(16)
time is defined as the time at which the transition be-
tween these two regimes occurs : treh ' 2/(3Γφ ). By
where Π(ω) is the inflaton polarisation operator with 4-
equating the inflaton energy density
momentum k = (ω, 0, 0, 0). Neglecting for simplicity the
time dependance of H and Im Π, Eq. (16) has the follow- MP2
ing solution ρφ = , (23)
6π t2reh

φ(t) ≈ Φ(t) sin(mt) , (17) with the one of a thermal bath

π2
where ρrad = g? TR4 , (24)
30
  
1 Im Π(m)
Φ(t) ≈ Φ0 exp − 3H + t . (18) containing g? DOFs, one gets the reheating temperature
2 m
1/4 p
For m  min(2 mχ , 2 mψ ), Π(ω) has an imaginary part TR ' 0.2 (200/g? ) Γ φ MP (25)
which can be identify with the decay width thanks to the
optical theorem through unitarity: It is remarquable that TR does only depend on the par-
ticle theory parameters and not on the initial value of
Im Π(ω) = m Γφ . (19) φ.
For h ∼ 10−2 , Γφ ∼ 108 GeV and TR ∼ 108 GeV. For
The amplitude of the oscillations of the field φ decreases g 2 ∼ 10−2 and v ∼ 1011 GeV, Γ ∼ 103 GeV and TR ∼
as 1010 GeV.
  Imposing Γφ  m ∼ 10−6 MP and g? ∼ O(100), TR 
1 15
Φ(t) = Φ0 exp − (3H + Γ)t (20) 10 GeV, thus rendering GUT baryogenesis impossible
2
in such kind of scenarios. Also, TR ≤ 109 − 1010 GeV in
 3

order to avoid the overproduction of gravitinos, an upper where we assumed m2  m2χ + 2gφ and where
bound which turns out to be in tension with thermal
leptogenesis scenarios [5]. m 2gΦ
k0 = , ∆k = . (29)
2 m

The occupation number nk can then be expressed

D. Remarks

nχ π 2 Φ nχ
1. If one would have considered the interaction ∆L = nk=m/2 = ' . (30)
2
(4πk02 ∆k)/(2π 3 ) g nφ
− g2 φ2 χ2 , then the rate in that case is Γ(φφ →
2 2
χχ) ∼ g8πmφ
, which is a function of φ. Since φ ∼ t−1 In consequence, we have Bose enhancement as soon as
and H ∼ t−1 , this interaction rate never overcomes
the Hubble expansion rate, thus rendering reheat- π2 Φ
nχ > nφ , (31)
ing through such an interaction impossible. g

2. We have perturbative reheating if and only if Γφ i.e. as soon as a fraction ∼ g of φ particles has been con-
can decrease more slowly than t−1 . verted into χ ones. The elementary theory of reheating
fails quickly after the beginning of reheating.
3. Typically, for reheating to complete, one either has
The Boltzmann equation Eq. (15) can then be rewrit-
to have symmetry breaking or coupling to fermions.
ten
Reheating thus implies constraints on the structure of
1 d(a3 nχ ) g2 2π 2 Φ nχ
 
the theory and on the couplings of φ to the other fields. = 1 + nφ , (32)
a3 dN 2m2 g nφ

where N ≡ mt/2π is the number of inflation oscillations

II. PREHEATING
up to t. Neglecting the expansion, and assuming Φ to be
A. Intuitive Picture constant disregarding the particle production, one gets

nχ ∝ exp (4π µ N ) , (33)

Up to know, particle production has been addressed
disregarding the previously created χ particles. It turns
where µ ≡ πgΦMP /(4m2 ) is the parameter of instability.
out that Γeff , the effective φ decay width, can be much
larger than Γφ due to Bose enhancement effects.
If m  mχ , most of the χ particles produced in φ
B. EOM for quantum fluctuations
decays carry momentum k ' m/2, and lead to a Bose
enhancement of the φ decay. This can be seen at first if
Collective effects such as Bose enhancement limit the
we consider the matrix element of such a process:
range of applicability of the elementary theory of reheat-
|hnφ − 1, nk + 1, n−k + 1|â+ + − 2 ing. Therefore, a rigorous description of the first stage
k â−k âφ |nφ , nk , n−k i|
of reheating, commonly referred to as preheating, can
= (nk + 1)(n−k + 1)nφ (26)
only be based on non-perturbative techniques. In par-
ticular, we will show in this second part of the talk how
where n±k are the occupation numbers of the χ particles.
the phenomenon of parametric resonance may result in
One can thus define an effective decay width
explosive particle production. The physical picture one
should keep in mind is the following: Due to their cou-
Γeff ' Γφ→χχ (1 + 2 nk ) (27)
pling to the coherently oscillating classical inflaton field
The φ particles decay at rest into two χ particles, and if φ quantum fluctuations of the scalar field χ experience a
gφ  m2 /8, each of them carries a momentum k located resonant amplication. This causes an exponential growth
in the thin shell ∆k around k0 of the corresponding occupation numbers.
The details of reheating in combination with sponta-
k = k0 ± ∆k , (28) neous symmetry breaking will be the subject of the next
 4

seminar talk. For now, we thus restrict ourselves to a Eq. (37) now turns into:
toy model-like scenario of chaotic inflation in which the
scalar potential is given as: χ̈k + ωk2 (t)χk = 0 ; ωk2 (t) = k 2 + g 2 Φ2 sin2 (mt) (39)

1 2 2 1 2 1 Substituting z = mt and using sin2 (z) = 21 (1 − cos(2z))

V (φ, χ) = m φ + mχ (0)χ2 + g 2 φ2 χ2 (34)
2 2 2 we may rewrite Eq. (39) as Mathieu’s differential equa-
The time evolution of the quantum fluctuations of the tion:
field χ is ruled by the classical equation of motion – the
Klein-Gordan equation in an expanding flat Friedmann- χ00k + (Ak − 2q cos(2z)) χk = 0 (40)
Lemaı̂tre-Robertson-Walker Universe:
with the parameters Ak and q being defined as:
1
χ̈ + 3H χ̇ − 2 ∇2x χ + V,χ = 0 (35)
a k2 g 2 Φ2
Ak = 2
+ 2q ; q = (41)
Writing χ in the Heisenberg representation allows us to m 4m2
proceed to Fourier space. With ~x and ~k denoting the
In solid state physics when describing charge carri-
comoving position and momentum vectors:
ers in certain periodic crystalline solids the stationary
Z
d3 k h
~ ~
i Schrödinger equation may be cast as an Mathieu equa-
χ (t, ~x) = 3/2
ak χk (t)e−ik·~x + a†k χ∗k (t)eik·~x tion as well. That the unique solution of Eq. (40) for
(2π)
(36) vacuum initial conditions closely resembles Bloch waves:

~
The temporal part of the momentum eigenfunction with ψn~k (~r) = eik·~r P (~k, n, ~r) (42)
momentum k = |~k| (field mode) χk hence satisfies:
where P is invariant under translations by a lattice vec-
k2
 
2 2 2 tor. Standard references now tell us [4]:
χ̈k + 3H χ̇k + + mχ (0) + g φ (t) χk = 0 (37)
a2
χk (z) = emk z P (Ak , q, z) (43)
where φ(t) = Φ(t) sin (mt). Eq. (37) describes a har-
monic oscillator with variable frequency (parametric os- where P is periodic in z with period π. The real part µk
cillator) that is damped by the expansion of the Universe, of the Mathieu exponent mk is always non-negative:
viz. the friction term 3H χ̇k . As is well known from clas-
sical mechanics a concerted choice of parameters may µk (q) = 0 ⇒ |χk | is stable. (44)
cause parametric oscillators to resonantly excite them- µk (q) > 0 ⇒ |χk | grows exponentially. (45)
selves, a feature which goes by the name of parametric
resonance. In our context this means that depending on A primitive contour plot of µk as function of q and Ak ,
k we expect some of the modes χk to get parametrically the so-called stability-instability chart of the Mathieu
excited. equation, is shown in Fig. 2. According to Eq. (41) we
identify the different ranges of physical momenta that ex-
To alleviate our further calculations we neglect the ex- perience parametric resonance as the white regions above
pansion of the Universe, set the bare χ mass mχ (0) to the line A0 = 2q. The width of the resonance bands
(l)
zero and assume a slow variation in Φ(t) compared to the ∆Ak , l ∈ N, and the preheating effeciency are solely
oscillation frequencies of the fields φ and χ: controlled by the parameter q which is related to the in-
flaton amplitude Φ:
H = 0 ; a = 1 ; mχ (0) = 0 ; Φ(t) ≈ const. (38)
q <1 ⇒ Narrow resonance; 2πµk  1 (46)
This leaves us with an effective χ mass meff.
χ = gφ. The q >1 ⇒ Broad resonance; 2πµk ∼ O(1) (47)
scenario mχ (0)  m would allow for the production of
particles above the scale of inflation which could rescue
GUT baryogenesis models. The occupation numbers nk count by how many
 5

quanta the respective modes χk are populated. Matching Combining Eqs. (50) and (53) we find ΓPR = mq. Given
the quantum with the classical expression for the energy k ∼ lm in the lth instability band preheating has a natu-
eigenvalue of a harmonic oscillator we find: ral interpretation in the particle picture as the collective
  process 2l × φ → χχ.
εk 1 1 1 2 1 2 1
nk = − = |χ̇k | + ωk2 |χk | − (48) There are three main reasons why preheating is so
ωk 2 ωk 2 2 2
much more efficient than ordinary inflaton decays: First,
Instable modes χk ∼ exp (µk z) hence entail exponentially in preheating the growth rate of nk is proportional to nk
growing occupation numbers nk : itself, not to the number density of inflatons nφ as in de-
cays. Second, the perturbative decay rate of inflatons Γφ
nk ∼ exp (2µk z) = exp (2µk mt) (49) is suppressed by g 4 /m whereas ΓPR can be quite sizable
if q is not too small (see below). Third, preheating can
That is, the growth rate of those nk is proportional to
also produce χ particles off the mass shell which opens
the present occupation number:
new channels for the energy transfer from the φ to the χ
field.
ṅk ∼ ΓPR nk ; ΓPR = 2µk m (50)
Decays: ωk2 − k 2 = 0 (54)
whereby we have rediscovered the effect of Bose enhance-
ment in our non-perturbative calculation. Finally notice Preheating: ωk2 − k 2 = g 2 Φ2 sin2 (mt) (55)
that Eq. (49) implies that also the number density of χ
particles nk increases explosively: A necessary condition for successful preheating is that
it proceeds at a faster rate than decays:
d3 p
Z
nχ (t) = 3 nk (t) ; p = k/a (51)
(2π) g 2 Φ2
ΓPR = qm = & Γφ (56)
4m
Having outlined the generic features of preheating we
will now turn to a more detailed discussion of the two Once decays take over, Γφ > ΓPR , the inflaton amplitude
regimes of narrow and broad resonance. decreases exponentially and any resonance disappears. In
more visual terms a violation of condition (56) also means
that the intrinsic width of the inflaton mass eigenstate
has become broader than the first resonance band around
C. Narrow resonance regime (q < 1)
k ' m.
A realistic treatment of preheating beyond our toy
Fig. 3 displays numerical solutions for χk (t) and nk (t)
model faces further complications. The expansion of
with k ' m in the narrow resonance regime (q = 0.1). As
the Universe, for instance, augments the inflaton de-
ωk ' m for k ' m and q  1 the mode χk oscillates ap-
cay rate with the friction term 3H and, within a time
proximately with the same frequency as the inflaton field
∆t ∼ qH −1 , redshifts the χ modes out of the reso-
φ. The occupation number nk increases ally in agreement
nance layers. This provides us with two conditions for
with Eq. (49).
ΓPR = qm the stronger of which is the second:
The structure of the instability bands is dictated by
the theory of Mathieu’s equation: qm & Γφ + 3H ; qm & ∆t−1 ⇒ q 2 m & H (57)
(l) (l)
Ak ' l2 ; ∆Ak ' q l ⇒ k 2 ' m2 l2 − 2q ± q l


(52) The χ bosons may also be removed from the resonance
bands as they change their momenta or decay into other
We recognize the first band as the widest and most im-
particles due to secondary interactions (rescatterings).
portant one. It is centered around k ' m, has a width of
(1) Finally, one must not neglect the backreactions of the
mq and exhibits an instability parameter µk of:
χ particles on the inflaton field. Not only does the in-
(1)
q
2 2
flaton amplitude decrease due to χ production, also its
µk ' (q/2) − (k/m − 1) ' q/2 (53) effective mass receives ever-growing and eventually dom-
 6

inating contributions from the χ fluctuations: adiabatic invariant:

2 Φ2 1 2 2 −1/2
meff. = m 2 + g 2 χ2 ⇒ q →


(58) nk ∼ ω |χk | /ωk ' const. ⇒ |χk | ∼ ωk (63)
φ
4 hχ2 i 2 k

It only varies when the adiabaticity condition imposed

Preheating typically ends at time tPR after NPR in- on the change of ωk is violated:
flaton oscillations when condition (57) becomes violated.
This usually happens for: Particle production when: ω̇ & ω 2 (64)

q ∼ O 10−1


⇒ Φ ∼ m/g (59) which, as we will show, only happens around the time
gMp gMp when the inflaton passes through zero, |φ| . φ∗  Φ. In
⇒ tPR ∼ ; NPR ∼ ∼ few × 10 (60)
3m2 6πm these acts of creation the effective χ mass vanishes, ωk
becomes very small and the amplitude of the χk mode
Subsequently, reheating is described by the elementary blows up, cf. Eq. (64).
theory of perturbative inflaton decays. As TR is mainly
sensitive to the last stages of reheating it should be cal- For small φ we may approximate φ̇ = mΦ cos (mt) '
culated within the elementary theory and not at the end mΦ. With ωk2 = k 2 + g 2 φ2 and ω̇k ' ω −1 g 2 φmΦ we then
of preheating. deduce the range of excited momenta k from Eq. (64):

2/3
0 ≤ k 2 . g 2 φmΦ − g 2 φ2 (65)

D. Broad resonance regime (q > 1) which forces φ to be smaller than (mΦ/g)

1/2
and becomes
maximal for:
Although we now know everything about preheating,
1 1/2 1
there is still more to say about it, viz. about its sensitiv- φ = φ∗ ' (mΦ/g) ' Φq −1/4 (66)
2 3
ity to initial conditions. In particular in chaotic inflation
the initial amplitude of inflaton oscillations may be very The typical momenta k of particles that are produced in
large, Φ0 & Mp , resulting in a very broad (q  1) and the broad resonance regime may then be estimated as:
extremly efficient parametric resonance. It can be shown 1/2
√
that typically µk ' 0.18. From Eq. (49) we then deduce: 0 ≤ 2k . k∗ = (gmΦ) = 2mq 1/4 (67)

nk ∼ exp (2µk mt) = exp (4πµk N ) ∼ 10N (61) where k∗ is a measure for the maximal momentum scale
that can be reached during preheating. Eq.(67) tells us
that typically k  m which indicates the collective inter-
Fig. 4 displays numerical solutions for χk (t) and nk (t)
action of many φ quanta in the production of χ particles.
with k ' m in the broad resonance regime (q ' 200).
Finally, we compute the duration of each act of creation:
As we see, the χk mode oscillates much faster than the
inflaton field and particle production only occurs for very 2φ∗ (mΦ/g)
1/2

small values of φ(t). The former observation is due to the ∆t∗ ' ' = k∗−1 ∼ ω∗−1 (68)
φ̇ mΦ
mostly large effective χ mass:
Particle production occurs within one period of oscilla-

−1/2
meff.
χ = gφ ' gΦ  m ; at most times (62) tion ∼ ω∗−1 = k∗2 + g 2 φ2∗ of the mode χk which is
in agreement with the uncertainty principle:
explaining a ratio of order q 1/2 between the χ and φ os-
cillation frequencies. Turning to our second observation ∆t∗ ω∗ ∼ 1 (69)
we note that in view of the comparably slow φ oscillation
the frequency ωk mostly only experiences an adiabatic
variation. Hence the quantum number nk of the para- Due to the decrease in the inflaton amplitude the broad
metric oscillator with variable frequency ωk is almost an resonance eventually becomes narrow. Ref. [2] estimates
 7

that this will happen after a time tBR : tion effects. On the other hand, as a consequence of the
  momentum redshift and the rapid decrease in the infla-
1 m
tBR ∼ ln ⇒ NBR ∼ O (10) (70) ton amplitude, the resonance turns into what has been
m g 5 Mp
named a stochastic resonance.
At this time we approximately have: k2
ωk2 = + g 2 Φ2 (t) sin2 (mt) (72)
a2
q ' 1 ; Φ ' χ2 ; ρφ ' ρχ ; p ' ρ/3
2


(71)
Due to the non-periodical variation of the frequency ωk
That is, preheating facilitates an almost instantaneous the phases of the mode χk at successive zero-crossings of
transition from the epoch with vacuum-like equation of the inflaton field, φ (ti ) = 0, are completely uncorrelated
state to the epoch of radiation domination. Contrary thereby leading to only stochastic changes in the occupa-
to earlier expectations a prolonged intermediate stage of tion number nk . The number density of χ particles then
matter domination does not exist. only grows on average exponentially. Intermediately it
Bringing back into play the expansion of the Universe, may either increase or decrease. Such a process could
the complexity in describing preheating in the broad res- never be explained in the classical particle picture. We
onance regime increases significantly. As a virtue of the thus realize that the creation of almost all particles that
expansion we note that all excited modes, 0 ≤ k . k∗ /2, populate our present Universe was a purely quantum me-
are redshifted away from k∗ . This stabilizes preheating chanical effect.
and makes it less sensitive to rescattering and backreac-

[1] V. Mukhanov, Cambridge, UK: Univ. Pr. (2005) 421 p [5] An elegant solution to circumvent the gravitino problem
[2] L. Kofman, A. D. Linde, A. A. Starobinsky, Phys. Rev. is to consider right handed neutrinos as being responsible
Lett. 73, 3195-3198 (1994). [hep-th/9405187]. for the reheating of the Universe. This scenario present
[3] L. Kofman, A. D. Linde, A. A. Starobinsky, Phys. Rev. the peculiar property of being able to produce matter and
D56, 3258-3295 (1997). [hep-ph/9704452]. dark matter from a single mechanism (see Kai’s talk this
[4] N.W. Mac Lachlan, “Theory and Application of Mathieu afternoon !)
functions,” (Dover, New York, 1961).
 8

0.08

0.06

0.04

0.02

20 40 60 80

-0.02

-0.04

-0.06

FIG. 1: Oscillations of the field φ after inflation in the theory 21 m2 φ2 . The value of the scalar field is measured in units of Mp ,
time is measured in units of m−1 . Figure taken from Ref. [3].

FIG. 2: Sketch of the stability-instability chart of the canonical Mathieu equation (40) taken from Ref. [2]. Gray bands indicate
regions of stability (µk = 0), white bands regions of instability (µk > 0). The line A0 = 2q shows the values of Ak and q for
k = 0. For k 6= 0 the corresponding function graphs Ak (q) are obtained by parallely shifting upwards the line A0 by k2 /m2 .
 9

FIG. 3: Narrow parametric resonance for the momentum k corresponding to the maximal speed of growth, k ' m. Time is
shown in units of m/2π, which is equal to the number of oscillations of the inflaton field φ. For each oscillation of the field φ
the mode of the field χ oscillates one time. Left: growth of the mode χk , Right: logarithm of the occupation number nk of
particles in this mode. In this particular example q = 0.1 and hence µk ' q/2 = 0.05. Figures taken from Ref. [3].

FIG. 4: Broad parametric resonance for k ' m. For each oscillation of the field φ the mode of the field χ oscillates many times.
The peaks in the χk oscillations corresponds to the time intervals when |φ| . φ∗ . In this particular example q ' 200 and the
average rate of growth of nk is close to maximal in the context of our model, µk ' 0.3. Figures taken from Ref. [3].