Vous êtes sur la page 1sur 9

Quenching preheating by light fields

Olga Czerwinska,1 Seishi Enomoto,1, 2 and Zygmunt Lalak1

Institute of Theoretical Physics, Faculty of Physics,
University of Warsaw ul. Pasteura 5, 02-093 Warsaw, Poland
University of Florida, Department of Physics, P.O. Box 118440, Gainesville, FL 32611-8440
(Dated: June 22, 2017)
In this paper we investigate the role of additional light fields not directly coupled to the background
during preheating. We extend our previous study that proved that the production of particles
associated with such fields can be abundant due to quantum corrections, even for the massless states.
We also obtain the expression for the occupation number operator in terms of interacting fields which
includes the non-linear effects important for non-perturbative particle production. We show that
adding too many light degrees of freedom without direct interactions with the background might
attenuate or even quench preheating as the result of back-reaction effects and quantum corrections.
arXiv:1701.00015v3 [hep-ph] 20 Jun 2017

PACS numbers: 98.80.-k, 98.80.Cq

I. INTRODUCTION role of different parameters in the theory, summarize the

paper and conclude.
Post-inflationary particle production is a very complex
stage in the evolution of the universe that mixes pertur-
bative and non-perturbative processes [15]. Usually it II. PARTICLE PRODUCTION IN TERMS OF
is divided into two main stages: INTERACTING FIELDS

a) preheating - when exponentially and non-

perturbatively produced states typically corre- Usually the occupation number operator of produced
spond to the fields directly interacting with the particles is defined in terms of the creation and annihi-
inflaton, they do affect the mass term of the lation operators as Nk ak ak . This definition assumes
inflaton through back-reaction effects that produced states can be treated as free fields which
means that their equations of motion are linear. How-
b) reheating (thermalization) - when the inflaton de- ever, in general fields associated with the produced par-
cays perturbatively and produced particles end up ticles interact with other fields which spoils linearity and
in thermal equilibrium with a well-defined temper- results in the non-perturbative production. In that case
ature. it is not clear how to define the number operator properly.
In this section we address this issue and describe parti-
Interesting is the question about the impact of the
cle number using the theory of interacting fields which
additional fields, especially light ones, on preheating.
takes into account the non-linear effects. To compare
Their presence in the theory during [68] and after in-
these results with a simpler theory of a free field with
flation is important for multi-field inflation models and
time-dependent mass term see Appendix A.
for curvaton scenarios [911]. For recent reviews of post-
inflationary particle production see [12] or [13]. For simplicity let us consider a real scalar field with
In our previous study [14] we showed that light fields the Lagrangian of the form:
which are not coupled directly to the background can be
1 1
produced due to quantum corrections and their abun- L= ()2 m20 2 V [, (other fields)], (1)
dance can be sizeable, even for the massless case. In 2 2
this paper we want to develop these results addressing where m0 is a bare mass of and V is a general potential.
the problem of additional light degrees of freedom and Then equation of motion reads:
avoiding at the same time the infinite growth resulting
from the approximation used previously. The crucial dif- V
ference between present considerations at that of [14] lies 0 = ( 2 + m20 ) + = ( 2 + M 2 ) + J, (2)

in the fact that presently the inflaton is massive.
The outline of the paper is as follows. In Section II 1
where M is a physical mass that can depend on time
we develop the formalism necessary to describe the cre-
ation of particles in the presence of interactions, with and
without time-varying vacuum expectation value (vev) of
the considered field. In Section III we apply our formal- 1 In general physical mass can depend not only on time but also
ism to a number of well-motivated cosmological scenarios, on space coordinates. For simplicity we consider only the time-
including a sector of very light fields. In Section IV we dependent case as it is more common in cosmological considera-
compare our results with the earlier ones [14], discuss the tions.

and should be a c-number, and These relations are equivalent to the Bogoliubov trans-
formation for the wave function
J (m20 M 2 ) + (3) (t)
k = k as as
k k k , (12)
is a source term that can be an operator. where as
k denotes the asymptotic value of the mode k ,
Formal solution of (2) can be presented in a form of as in out
k = k , k .
the Yang-Feldman equation
Lagrangian (1) corresponds to the Hamiltonian
Z x0  
(x) = (t) (x) d4 y i[(t) (x), (t) (y)]J(y),
(4) 1 2 1 1
H= d3 x + ()2 + m20 2 + V . (13)
2 2 2
where the first term describes an asymptotic field defined
at x0 = t which satisfies the free field equation of motion Substituting (8) into the above results in a quite compli-
cated expression for the Hamiltonian which can be sim-
0 = ( 2 + M 2 )(t) . (5) plified by choosing the Bogoliubov coefficients of the form

In case that does not have a vev, the (t) can be de- in 1 in 1
|k |2 = k
+ , |k |2 = k , Arg(k k ) = Arg in
composed into modes 2k 2 2k 2
d3 k ikx  (t) (t) 3

(t) (t) (t)
(x) = e k ak + k ak (6)
in in 2 2 in 2
k |k | + k |k | , in in 2 2 in 2
k (k ) + k (k ) . (15)
fulfilling harmonic oscillator equation
(t) (t)
Then the Hamiltonian with diagonalized kinetic terms
0 = k + k2 k (7) reads
R d3 k  
with k |k|, k k 2 + M 2 and obeys the inner prod- H = (2) 3 k
(t) (t)
ak ak + 12 (2)3 3 (k = 0) + (16)
(t) (t)
uct 2 relation: k , k = 1.
+ d3 x 12 (m20 M 2 )2 + V ,
From (4) also the relation between two asymptotic
fields defined at different times x0 = t and x0 = tin can which indicates that the operator
be derived
(t) (t)
(t) in
Z t Nk (t) ak ak (17)
(x) = (x) d4 y i[in (x), in (y)]J(y), (8)
really plays the role of the occupation number. This is
where we denoted (x) (tin )
(x). Evaluating the in- because in the system which has a potential energy par-

(t) (t)
 ticle number N would be described classically as
ner product of the above equation with k : , k ,
we can obtain the Bogoliubov transformation for annihi- H Veff V0
lation operators [14] N= , (18)
where H is the total Hamiltonian, Veff an effective po-
(t) in in
ak = k ain
k +k ak d4 y i[k ain
k +k ak , (y)]J(y), tential, V0 a zero-point energy and E is an one-particle
(9) energy. Therefore particle number is just the kinetic en-
where ergy of the system divided by the one-particle energy.
Substituting (9) into (17) and using (14), (15), we can
(t) (t)
k = k (t, tin ) (k , ink = k (t, tin ) (k , in
k ), k ) finally obtain the expression for the occupation operator
(10) in terms of the interacting fields as
and the normalization condition reads
1 +
Nk (t) + Nk (t)

|k |2 |k |2 = 1. (11) Nk (t) = (19)

3 in in
k and k are constrained by the relation:
2 We use the following definition of the inner product:
(A, B) i(A B A B). |in 2 in 2 2
k | |k | = k .

with4 number with a general Lagrangian (1). In order to de-

scribe the time evolution of distributions nk = hNk i/V

Nk+ (t) = 1
kk k + k2 k k (2)3 3 (k = 0),(23) for each type of produced states we need to determine
the time evolution of bilinear products of field operators

Nk (t) = i k k k k + (2)3 3 (k = 0), (24)
hk k i, hk k i and hk k i. Equations of motion for
these operators can be derived by calculating their time
where we defined the Fourier transformation as derivatives and using (2)as
k (t) d3 xeikx (t, x).
(25) hk k i = hk k i + hk k i (27)

(t) (t) (t) (t)
hk k i = hk k i + hk k i
Since Nk (t) = ak ak ak ak , Nk+ denotes a total
and Nk a net number of particles with momentum be- = hk k i k2 hk k i hk Jk i (28)
tween k and -k. In the case of a complex scalar this
(t) (t) (t) (t)
hk k i = hk k i + hk k i
expression changes to Nk (t) = ak ak bk bk , where
(t) = k2 (hk k i + hk k i) hk Jk i hJk k i
bk is an annihilation operator for anti-state, but (23) and
R d3 k (29)
(24) still hold. In such a case (2) 3 Nk corresponds to

the U (1) Noether charge. where

In the
case where has a non-vanishing vev hi Z
Jk d3 xekx J(t, x).

0 0in , we just have to replace hi in (25) to (30)
obtain the proper expression for the occupation number.
Physical mass of is determined by the relation:

III. NUMERICAL RESULTS FOR 0 = hk Jk i = (m2 M 2 )hk k i (31)

+ d3 xeikx k dVd(x)

To obtain numerical results for some specific models to remove the infinite part of the mass correction.
we follow the procedure described in Section II. We are
especially interested in time-evolution of particle number
density for each considered species: A. Two scalar system
d3 k hNk i
n(t) = , (26) At first we apply our formalism to the simple theory
(2)3 V
consisting of two scalar fields
where V is the volume of the system, and in time- 1 1 1 1 1
dependence of the background (inflaton). We consider L= ()2 + ()2 m2 2 m2 2 g 2 2 2 .
2 2 2 2 4
a time range and starting from the initial state we solve (32)
equations of motion for all the species and calculate their We assume that it is the field
that has time-varying vev
number density. Then we move to a slightly later time

and plays the role of inflaton, 0in 0in = h(t)i, while
and repeat the procedure taking into account the back- is another scalar field with vanishing vev that can be,
reaction of previously produced states on the evolution of for instance. a mediator field between the inflaton and
the background (given by the induced potential coming the Standard Model. We also assume m  m . The
from non-zero energy density) and all the species. details of the calculation in this system can be found in
Before we present numerical results for specific mod- Appendix B.
els, let us focus on a subtlety in calculation of the particle Asymptotically, when quantum effects can be ne-
glected, we can choose a vacuum solution for (32) of the
4 The zero point term can be regarded as the volume of the system hi = 0 cos(m (t t0 )), (33)
Z Z where 0 denotes the initial amplitude of the oscillations,
(2)3 3 (k = 0) = d3 xeikx |k=0 = d3 x = V. (20) hi(t = t0 ) = 0 . When this trajectory crosses the non-
Therefore, we can also find distribution operators:
adiabatic area for : |hi| < m |0 |/g, the mass of
becomes very small and kinetic energy of the background
Nk 1
k = V = k
+ k2 V1 k k 1,
V k k
(21) field is transferred to the field . This results in the
creation of particles with the distribution [5]
Nk 1 1
k = V = i V k k V k k + 1. (22)
gmk |
nk = e 0| , (34)

where k is a momentum of a particle. Once particles scalar field domination phase this means that gv > 3H,
are produced and trajectory of hi goes away from the
for matter domination: gv > 32 H, while for radiation
non-adiabatic region, the energy density of particles
domination: gv > 2H.
can be represented as Following [15] and the analytical method of estimating
the number density of producing particles in the expand-
d3 k
g|hi| nk , (35) ing universe presented there for which
which corresponds to the linear potential acting on hi 5 1
n(j) (1)
n 3
, (38)
describing the backreaction effects. Then trajectory of 2 j 5/2
hi goes back towards the origin and particles can be
produced again both due to the oscillatory behaviour of where j denotes the number of oscillations, we can see
hi and backreaction. the agreement with our results. If we take j 10 as in
In the Figure 1 we show an example of the numerical the Figure 1 and n 1 106 , we can see that the
results for the Lagrangian (32). According to [5] the first oscillation phase indeed finishes when 12 m2 hj i2
production of particles results in the number density (j)
ghj in .
(gm h(0)i)3/2
4 109 , (36)
(2)3 Im t

which is consistent with our numerical results. On the

other hand, it is difficult to obtain the analytic results
for indirect production products, like hi. But
one can see in the Figure 1 that for the considered La-  pgv 1
w H0
3(1+ )

grangian energy transfer from the background to and Re t

the production of particles associated with the inflaton is
in t0
small for generic choices of parameters. Therefore in this non-adiabati
system it is a good approximation to neglect the quantum
part of and the production of its fluctuation.



10-13 FIG. 2: Time spent by the trajectory in the adiabatic region

in comparison with the Hubble time.

0. 2.104 4.104 6.104 The distribution of the produced states is not thermal
but, assuming that the whole energy is transferred to the
light states which interact with each other and with other
n (t) n (t) particles not present in the simplified Lagrangian, we can
naively estimate the maximal reheating temperature as
FIG. 1: Time evolution of number density of produced states  30 1/4
for g = 0.1, m = 0.001M , (t = 0) = M , (t = 0) = 0 in TRmax
, (39)
two scalar system. Scale M 0.04MP L , where MP L denotes g 2
the Planck mass MP L 1.22 1019 GeV, is chosen to be close
to the unification scale and allows us to stay in agreement where R is energy density of the relativistic particles (in
with the observational data. our case or and ) and g describes the number of rel-
ativistic degrees of freedom (g O(102 )). In our system
the coupling is big enough to describe energy density as
In our considerations we neglect the expansion of the = mn and without contradicting our assumptions we
universe which is valid assuming that the mean time the can choose the masses as in Table I. Final estimation of
trajectory spends in the non-adiabatic region is smaller TRmax is also presented in Table I.
than the Hubble time, see Figure 2. This means that:
1 2
< , (37) B. System with the additional light sector
gv 3H(w + 1)
where H is a Hubble parameter and w = p is a barotropic Usually when describing preheating light fields not
parameter describing the content of the universe. For the coupled directly to the inflaton are neglected. But it

TABLE I: Energy densities and upper limits on reheat-

ing temperature for two choices of mass. Mass of 10-8
is set to m = 5 1014 GeV. Number densities for each
state correspond to the results from Figure 1, meaning that 10-13

n 3.96 102 GeV3 and n 8.2 109 GeV3 .
m [GeV] [GeV4 ] TRmax [GeV] 10-18

125 106 1.3 102 10-23

700 5.7 106 2 102 0. 2.104 4.104 6.104
n (t) n (t) n (t)

is important to note that corresponding particles may be

produced through an interaction with some other state FIG. 3: Time evolution of number density of produced states
in the system with additional light sector for g = 0.1, y = 1,
coupled directly to the background that is produced res-
n = 1, m = 0.001M , (t = 0) = M , (t = 0) = 0.
onantly. Furthermore, if there are many additional light
degrees of freedom, one can expect that energy transfer
from the background to the light sector during preheat-
ing might be sizeable. In this section we focus on such 1.0

light fields and discuss the possibility of their produc- 0.8

tion through the indirect interaction with the background

We can describe such a situation by extending (32) 0.4
with n light or massless fields n (m  m , m ) that
are not coupled to the background at the tree-level
0. 1.104 2.104 3.104 4.104 5.104
L = 21 ()2 + 12 ()2 12 m2 2 12 m2 2 14 g 2 2 2
+ n 12 (n )2 n 12 m2 n2 n 14 y 2 2 n2 . (40)
n=1 n=2 n=5 n=7 n=10

We assume again that hi is time-varying and the other

fields do not have a vev: hi = hn i = 0. Then particles FIG. 4: Envelope of the time evolution of the background hi
are produced resonantly and as we mentioned before we for g = 0.1, y = 1, m = 0.001M , (t = 0) = M , (t = 0) = 0
for different numbers of additional light fields: 1, 2, 5, 7.
can expect production of n through the interactions with
The physical mass of n is given by
R d3 p  1  The reason why the energy transfer can be stopped
M2 = m2 + 12 y 2 (2) 3 V h p p i 1
2p in this case can be understood as follows. The physical
+O(y 4 , y 2 g 2 , g 4 ), (41) mass of in the system is given by
R d3 p  1 
M2 = m2 + 12 g 2 hi2 + 12 g 2 (2) 1
where p p2 + M2 and V denotes the volume of 3 V hp p i 2p

the system. We can see that n influence backgrounds P  1 

+ 12 y 2 h np np i 1 4 2 2 4
2p + O(y , y g , g ). (42)
evolution via hp p i operator in their mass term. V
We show the results for only one additional field in
Figure 3. One can see that all the states are produced and
Considering an approximation Xp iXp Xp for
their number density is abundant. If the final number X = , n , one can find that
density of is comparable to the one for (n n ) D E
its presence may even quench the preheating process by 1 1 1 1 (+)
V hXp Xp i 2X 2X V N Xp . (43)
terminating the energy transfer. The reason that can p p

be produced so efficiently is the strong coupling between Thus, once or n are produced at the same time they
and that enhances the back-reaction effects. also generate s effective mass 5 which results in par-
We would expect that most of the energy would be ticle production area becoming narrower. This leads to
transferred to n fields as they are very light and the
process is energetically favourable. But we can prove
that the more light species we include, the larger the
final value of |hi| becomes and, in other words, the less 5 These mass correction terms describe a square of plasma fre-
energy from the background goes to the light fields, see quency discovered by I.Langmuir and L.Tonks in the 1920s which
Figure 4. is a critical value for which the wave of can enter Xs plasma

TABLE II: Energy densities and upper limits on reheating
temperature (both in GeV) for two choices of and mass.
Mass of is set to m = 5 1014 GeV. Number densities for
each state correspond to the results from Figure 3, meaning

n (t)
that n 1.82 109 GeV3 and n 9.91 106 GeV3 . 10-5

m [GeV] m [GeV] n [GeV3 ] [GeV4 ] [GeV4 ] TRmax [GeV]

5 3 3 1
125 100 1.21 10 1.24 10 1.21 10 0.93 10
700 125 1.21 105 6.94 103 1.51 103 1.26 101 10-7
0. 2. 104 4. 104 6. 104

the suppression of particle production and also spoils 10-8

the production of other species. Too many n particles

n (t)
produced through indirect coupling to the background 10-9
prevent the production of particles directly coupled to
the background, . 10-10
It is interesting to investigate the impact of both cou-
plings - g that couples to and the background hi 10-11
and y that couples additional fields n to , on the fea- 0. 2. 104 4. 104 6. 104
tures of preheating. Varying the coupling y for fixed g t
leads to the conclusion that the initial stage of preheat-
ing does not depend on y coupling for and states.
It only influences the final abundance of produced and 10-7
states - the bigger y is, the smaller number density of n (t)
these states we observe, see Figure 5. For the impact of 10-9
y is quite opposite - both initial and final stages of pro-
duction are strongly influenced by the value of y. This 10-11
time the bigger y is, the larger number density of we
observe which also results in more effective energy trans- 10-13
0. 2.104 4.104 6.104
fer to the background as y coupling drops, see Figure 6.
Also, for choices of parameters resulting in n n we
y=0.1 y=0.2 y=0.5
can observe quenching of the energy transfer from the
background. y=0.7 y=1
Our study may seem similar to the process of instant
preheating [16, 17], where the system of three fields - FIG. 5: Time evolution of number density of produced states
background , interacting with the background and , and for g = 0.1, n = 1, m = 0.001M , (t = 0) = M ,
some other field not coupled to , is considered. In- (t = 0) = 0 and different values of y coupling. Values y = 0.7
stant preheating relies on the fact that particles pro- and y = 1 correspond to quenching of parametric resonance.
duced within one-time oscillation of decay immediately
to before the next oscillation of . So states can be

or not, because
(+) also produced even though there is no direct interaction
d3 p hNk i
between and . In our work the mechanism of pro-
(2)3 2Xp V duction is different - due to the quantum corrections, not
is proportional to M if X particles are massive enough (nX the decay, and quenching of the preheating comes from
is Xs number density). Moreover, if one considers the massless a plasma gas effect here rather than the rapid decay.
thermal equilibrium distribution with the temperature T :
1 1
hNXp i = 2 p/T Table II presents TRmax and energy densities for each
V e 1
state for the considered model under assumption that
(factor 2 corresponds to the degrees of freedom for momentum = H or = H, H being the Higgs field playing the
k and k particles), it corresponds to the thermal mass of the role of the mediator or the light field. We can see that
d3 p 1 1
T2 additional light sector that quenches preheating rises TR

hXp pX i . lowering the number density of particles at the same
(2) V 2Xp 6

1.0 10-3




0.4 10-11

0. 1.104 2.104 3.104
0.0 t
0. 1.104 2.104 3.104 4.104 5.104
n (new) n (old)
n (new) n (old)
y=0.1 y=0.2 y=0.5
y=0.7 y=1
FIG. 6: Envelope of the time evolution the background hi for 0.6

g = 0.1, n = 1, m = 0.001M , (t = 0) = M , (t = 0) = 0
and different values of y coupling. For y = 0.7 and y = 1 we 0.4
can observe the quenching of the preheating.

0. 1.104 2.104 3.104
new old
In our previous work [14] we presented a formalism
for describing particle production in a time-dependent
FIG. 7: Comparison between time evolution of number den-
background. It turned out it possesses one drawback -
sity of produced states (upper ) and the background hi
there exists a secularity in the number density of mass- (lower ) obtained with a new and old methods for g = 1,
less states that can be a product of approximating the m = 0.001M , (t = 0) = M , (t = 0) = 0. New denotes
fields by their asymptotic values. In this paper we have the interacting theory described here and old - asymptotic
developed more accurate description by expressing the approximation presented in [14].
number operator in terms of interacting fields. Figure 7
compares the two methods for the Lagrangian (32). The
new method avoids artificial secularity caused by time in-
tegral of the interaction effects with the Green functions
not work well and this indicates that preheating might
seen before. The old method seems to overestimate the
be quenched if there are many degrees of freedom of light
production at the late stage because it includes inverse
fields which are connected to the background indirectly.
decay processes, whereas the new one takes into account
mass correction terms. However, the results with secu-
larity are still applicable at the early stages of particle This work has been supported by the Polish NCN grant
production process. DEC-2012/04/A/ST2/00099, OC was also supported by
As the application of the new method in this paper the doctoral scholarship number 2016/20/T/ST2/00175.
we investigated the role of additional light fields coupled SE is partially supported by the Heising-Simons Funda-
indirectly to the background during resonant particle tion grant No 2015-109. OC thanks Bonn Bethe Centre
production processes such as preheating. In particular, Theory Group for hospitality during the completion of
we considered models with a scalar field interacting this paper.
with the background hi through its mass term and
with n light fields n . In order to describe particle
production in the system, at first we defined number
operator in terms of interacting fields and then we Appendix A: Particle production in free fields
theory with time-varying mass terms
solved numerically their equations of motion. In case of
a few additional light fields, their production can be also
resonant through the quantum correction to their mass Let us consider a real free scalar field with the time-
term and their final amount can be sizeable. However, dependent mass term:
many degrees of freedom of these extra light fields can
prevent s and also n s resonant particle production. 1 1
As a result, energy transfer from the background does L= ()2 m2 (t)2 . (44)
2 2

The solution of the equation of motion can be decom- with the coefficients satisfying
posed into
in 1 in 1
d3 k ikx   |k |2 = k
+ , |k |2 = k , Arg(k k ) = Arg in
k .
(x) = e k ak +
a (45) 2k 2 2k 2
k k
(2)3 (59)
Then the occupation number can be expressed as
where k = k (x0 ) is a time-dependent wave function
which satisfies k 1
Nk (t) = h0|Nk |0i = |k |2 = . (60)
p 2k 2
0 = k + k2 k (k k 2 + m2 ), (46) Appendix B: Two scalar system
- details of the calculation
and ak , ak are annihilation and creation operators. The
vacuum state |0i is defined by the relation ak |0i = 0 and In the system described by the Lagrangian 1, we have a
the commutation relations background field hi and two quantum fields: hi,
. The set of differential equations for distributions reads
[(t, x), (t, x0 )] = i(x x0 ), (47)
[(t, x), (t, x )] = [(t, x), (t, x )] = 0, 0
(48) 0 = hi + M2 hi (61)
[ak , ak0 ] 3
= (2) (k k ), 0
hk k i = hk k i + hk k i (62)
[ak , ak0 ] = [ak , ak0 ] =0 (50)
hk k i = hk k i k 2
hk k i (63)
give an inner product relation of the form: (k , k ) = 1.
hk k i = k 2
(hk k i + hk k i) (64)
Using (45) we can represent the Hamiltonian as
Z   hk k i = h k k i + hk k i (65)
1 2 1 1
H = d3 x + ()2 + m2 2 (51) h k i = h k i 2 h k i
k k k k (66)
2 2 2
h k k i = 2
k (h k k i + hk k i) (67)
d3 k 1 h
= 3
k (t) ak ak + ak ak
(2) 2 where the source terms are absent because of our choice
+k (t)ak ak + k (t)ak ak , (52) of physical masses

d3 p
1 1 1
where M2 = m2 + g 2 h p i ,(68)
2 (2)3 V p 2p
k (t) |k (t)|2 + k2 (t)|k (t)|2 , (53) 1
M2 = m2 + g 2 hi2
k (t) 2k (t) + k2 (t)2k (t). (54) 2
d3 p
1 2 1 1
In order to diagonalize the Hamiltonian + g h p i . (69)
2 (2)3 V p 2p
d3 k 1
H = k (t) a
a + a a
k k (55) In order to obtain the above formulae, we applied an
k k
(2)3 2 approximation
d3 k
1 3 3
= k (t) ak ka + (2) (k = 0) (56) hp1 p2 p3 p4 i = hp1 p2 ihp3 p4 i + O(g 2 ) (70)
(2)3 2

we need a set of operators ak , ak satisfying and assumed the momentum conservation

[ak , ak0 ] = (2)3 (k k0 ), [ak , ak0 ] = [ak , ak0 ] = 0, (57) hXp Xp0 i = (2)3 3 (p p0 ) hXp Xp i (71)
Then the number operator Nk ak ak is well-defined all
for the quantum fields X = , . Momentum conserva-
the time. Following [18], we can obtain the new operators
tion indicates that hXp Xp0 i = Cp (2)3 3 (p p0 ), where
by the Bogoliubov transformation
Cp is a proportionality factor. For p0 = p: hXp Xp i =
ak = k ak + k ak (58) V Cp , hence Cp = V1 hXp Xp i.

[1] L. Kofman, A. Linde, A. Starobinsky, Phys. Rev. Lett. [2] L. Kofman, A. Linde, A. Starobinsky, Phys. Rev. D 56
73 (1994) [hep-th/9405187]. (1997) [hep-ph/9704452].

[3] J. H. Traschen and R. H. Brandenberger, Phys. Rev. D A. Mazumdar, Ann. Rev. Nucl. Part. Sci. 60 (2010) 27
42 (1990) 2491. [hep-th/1001.2600].
[4] A. D. Dolgov and D. P. Kirilova, Sov. J. Nucl. Phys. 51 [13] M. A. Amin, M. P. Hertzberg, D. I. Kaiser and
(1990) 172 [Yad. Fiz. 51 (1990) 273]. J. Karouby, Int. J. Mod. Phys. D 24 (2014) 1530003 [hep-
[5] L. Kofman et al., JHEP 05 (2004) hep-th/0403001. ph/1410.3808].
[6] T. Kobayashi, S. Mukohyama, Phys. Rev. D 81 (2010) [14] S. Enomoto, O. Fuksinska, Z. Lalak, JHEP 03 (2015)
[astro-ph.CO/1003.0076]. [hep-ph/1412.7442].
[7] T. Matsuda, JCAP 1204 (2012) [hep-ph/1204.0303]. [15] K. Enqvist, D. G. Figueroa and R. N. Lerner, JCAP
[8] K. Kohri, T. Matsuda, JCAP 1502 (2015) 1301, 040 (2013) [astro-ph.CO/1211.5028 ].
[astro-ph.CO/1405.6769]. [16] G. Felder, L. Kofman, and A. Linde, Phys Rev D 59
[9] K. Enqvist, M. Sloth, Nucl. Phys. B 626 (2002) (1999) 123523 [hep-ph/9812289].
[hep-ph/0109214]. [17] S. Tsujikawa, B. Bassett, and F. Viniegra JHEP 19
[10] D. Lyth, D. Wands, Phys. Lett. B 524 (2002) (2000) [hep-ph/0006354].
[hep-ph/0110002]. [18] B. Garbrecht, T. Prokopec and M. G. Schmidt, Eur.
[11] T. Moroi, T. Takahashi, Phys. Lett. B 522 (2001) Phys. J. C 38, 135 (2004) [hep-th/0211219].
[12] R. Allahverdi, R. Brandenberger, F. Y. Cyr-Racine and