Home Search Collections Journals About Contact us My IOPscience

Bouncing universes in scalar-tensor gravity models admitting negative potentials

This content has been downloaded from IOPscience. Please scroll down to see the full text.

JCAP07(2015)002

(http://iopscience.iop.org/1475-7516/2015/07/002)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 130.237.29.138
This content was downloaded on 07/07/2015 at 00:45

Please note that terms and conditions apply.

J ournal of Cosmology and Astroparticle Physics
An IOP and SISSA journal

Bouncing universes in scalar-tensor

gravity models admitting negative

JCAP07(2015)002
potentials

B. Boisseau,a H. Giacomini,a D. Polarskib and A.A. Starobinskyc,d

a Université de Tours, Laboratoire de Mathématiques et Physique Théorique,
CNRS/UMR 7350,
37200 Tours, France
b Université Montpellier & CNRS, Laboratoire Charles Coulomb, UMR 5221,

F-34095 Montpellier, France

c L.D. Landau Institute for Theoretical Physics RAS,

Moscow 19334, Russian Federation

d Kazan Federal University,

Kazan 420008, Republic of Tatarstan, Russian Federation

E-mail: bruno.boisseau@lmpt.univ-tours.fr, hector.giacomini@lmpt.univ-tours.fr,
david.polarski@umontpellier.fr, alstar@landau.ac.ru

Received April 30, 2015

Revised June 3, 2015
Accepted June 10, 2015
Published July 6, 2015

Abstract. We consider the possibility to produce a bouncing universe in the framework of

scalar-tensor gravity models in which the scalar field potential may be negative, and even
unbounded from below. We find a set of viable solutions with nonzero measure in the space
of initial conditions passing a bounce, even in the presence of a radiation component, and
approaching a constant gravitational coupling afterwards. Hence we have a model with a
minimal modification of gravity in order to produce a bounce in the early universe with
gravity tending dynamically to general relativity (GR) after the bounce.

Keywords: modified gravity, cosmic singularity, dark energy theory

ArXiv ePrint: 1504.07927

c 2015 IOP Publishing Ltd and Sissa Medialab srl

doi:10.1088/1475-7516/2015/07/002
Contents

1 Introduction 1

2 A bouncing model 2

3 Study of the bouncing solution 4

4 Conclusions 9

JCAP07(2015)002
1 Introduction

Since the realization that the strong curvature singularity arising in the past of our Universe
in Friedmann-Lemaı̂tre-Robertson-Walker (FLRW) models may not be avoided in generic
solutions of the Einstein gravity with matter in the form of an ideal fluid (only its type
changes from an isotropic one to the generic BKL vacuum singularity), there were many
attempts to construct solutions avoiding this singularity by having at least one bounce at
some high value of curvature either using more complicated field-theoretical models of matter
or by modifying gravity. The simplest example to obtain one bounce in a set of solutions
having a nonzero measure in the space of initial conditions is presented by a massive scalar
field in a closed (i.e. positively spatially curved) FLRW universe [1]. This does not remove the
curvature singularity in a generic solution, even in the class of FLRW models, but shifts it to
the past, beyond any given finite number of bounces of the scale factor. Non-singular FLRW
solutions with an infinite number of bounces exist, too, but they are degenerate: they have a
zero measure in the space of initial conditions [2], see also [3]. This occurs because a minimally
coupled scalar field with a non-negative potential can violate the strong energy condition, but
not the weak and null ones.1 However, having a bounce due to a positive spatial curvature
requires severe fine tuning of initial conditions before the contraction stage in the past, see
e.g. the calculations of the measure of those ones leading to a bounce in the above mentioned
cosmological model in [1] and [6]. That is why it is much more interesting to obtain a non-
degenerate bounce in FLRW models in the absence of spatial curvature. Previously known
examples of such FLRW models are based on such radical modifications of general relativity
(GR) as theories with scalar [7, 8] or tensor ghosts, loop quantum gravity (see e.g. [9]) or
gravity described by an effectively non-local Lagrangian (see e.g. [10, 11], and [12] for a recent
review). By contrast, in our paper we would like to restrict ourselves to the well-known and
very modest modification of GR — scalar-tensor gravity without ghosts. On the other hand,
we want to abandon the common assumption that the scalar field potential V (φ) in this
gravity theory is non-negative, we will even consider the case when it is unbounded from
below. In section 2, the bouncing model is presented and general equations and results are
given. In section 3, the existence of physically viable bouncing solutions is shown and their
behaviour is studied in details. Section 4 contains our conclusions.
1
In the case of open (negatively spatially curved) FLRW models, a bouncing solution was found in [4]
for a non-minimally coupled scalar field with a quartic potential (so this model represents a particular kind
of scalar-tensor gravity models). However, later it was shown in [5] that this regular solution is degenerate,
too: it is unstable against arbitrarily small anisotropic and/or inhomogeneous metric perturbations preventing
transition to the regime with a tensor ghost (repulsive gravity).

–1–
2 A bouncing model

Let us consider a universe where gravity is described by scalar-tensor theory. The Lagrangian
density in the Jordan frame of the gravitational sector is given by
1
L= [F (Φ)R − Z(Φ) g µν ∂µΦ∂νΦ − 2U (Φ)] . (2.1)
2
Everywhere below we will use the freedom to take Z = 1 or Z = −1, corresponding physically

−2
to ωBD > 0 or ωBD < 0 (ωBD = ZF dF dΦ ). For ωBD < 0, the theory is ghost-free
provided − 32 < ωBD < 0. We consider further spatially flat FLRW universes with metric
ds2 = −dt2 + a2 (t)dx2 yielding the following modified Friedmann equations

JCAP07(2015)002
1
−3F H 2 + Z Φ̇2 − 3H Ḟ + U = 0 , (2.2)
2
2F Ḣ + Z Φ̇2 + F̈ − H Ḟ = 0 . (2.3)

with H ≡ aȧ . Here and below a dot, resp. a prime, stands for the derivative with respect to
t, resp. to Φ. Equations (2.2), (2.3) contain the equation of motion of Φ

Z (−Φ̈ − 3H Φ̇) + 3F ′ (Ḣ + 2H 2 ) − U ′ = 0 . (2.4)

We start our analysis with Z = 1 and we specialize to the following model

1
F = − Φ2 + κ−2 , (2.5)
6
Λ
U = 2 − cΦ4 , (2.6)
κ
where κ−2 > 0, Λ > 0 and c are constant parameters, only c being dimensionless. Surpris-
ingly, while we arrived at this model by requiring that a combination of (2.2), (2.3) and (2.4),
without derivation with respect to t, be zero, this model has been considered in the past in
completely different contexts (see e.g. [13]). Thus (2.4) corresponds to a massless scalar field
conformally coupled to Einstein gravity with an interaction potential −cΦ4 . Therefore, the
energy-momentum tensor of the field Φ is the sum of “dark” radiation with a negative energy
density and a cosmological constant.
As a result, the system (2.2), (2.3) reduces to
A
3H 2 = Λ + κ2 4 , (2.7)
a
1 dχ 2
 
− cχ4 = A , (2.8)
2 dη
with an effective positive cosmological constant Λ and the bare gravitational constant 8πG =
F −1 (Φ = 0) ≡ κ2 , while χ = aΦ and η = dt/a(t). The crucial point is that the constant A
R

— the energy density of the field χ in Minkowski space-time — can be negative here.
For A < 0, we have in front of us an explicit illustration of the possibility for scalar-
tensor models to accommodate an effective dark energy component of the phantom type
(weff < −1) [14, 15].2 Our system can be completely integrated. As we are interested
2
A phenomenological dark energy model in GR based on (2.7) was considered in [16], leading to (2.9)
and (2.10).

–2–
in bouncing solutions, we must take A < 0 and from (2.8), we have c > 0 so that U is
necessarily an inverted potential, unbounded from below. In spite of looking unphysical at
first sight, scalar fields with such an interaction have been often considered both in quantum
field theory and cosmology, see e.g. [17].
Integrating (2.7), a bouncing solution is obtained
 r 
1 Λ
a = a0 cosh 2 2 t , (2.9)
3
r  r 
Λ Λ
H = tanh 2 t , (2.10)
3 3

JCAP07(2015)002
 1
−Aκ2 4
where a0 = Λ is the value of a at the bounce located at t = 0 with a trivial redefinition
of t. It satisfies Ḣ > 0 and has a constant Ricci scalar R = 6(Ḣ+2H 2 ) = 4Λ. Integrating (2.8)
the analytical expression for χ is obtained in terms of the Jacobi elliptic function dn(u|m) [18]
as follows
 1
−A 4 1
χ(η) = − , (2.11)
c dn(u|2)
with
1 !
√

Λ 4 1 1
u = dn−1 − | 2 + 2 (−Ac) 4 η , (2.12)
cκ2 Φ0

where dn−1 (u|m) is the inverse function of dn(u|m) and Φ0 = Φ(t = 0). The conformal time
η(t), chosen here such that η(0) = 0, is given by
r !
i Λ
η(t) = − q F i t|2 , (2.13)
a Λ 3
0 3

where F (x|m) is the elliptic integral of the first kind [18]. This function is odd and then η(t)
is real. It is possible to show that

Γ2 14


1
η(∞) = q √ . (2.14)
2Λ 4 π
a 0 3

So we get finally for Φ̇0 < 0

Λ

14
− cκ2 1
Φ(t) = s   . (2.15)
q   √
  q
−1 Λ

14 1 9c

1 Λ
cosh 2 Λ3 t dn dn − |2 − i 2 F i 3 t|2 |2
4
cκ2 Φ0 κ2 Λ

We note that under the transformation Z → −1, F → −F , U → −U , the same so-

lution (2.15) is obviously obtained and that (2.7) and (2.8) remain valid as well be-
cause (2.2), (2.3) and (2.4) are invariant under this transformation.
Another attractive feature of this model is the possibility to have a bounce in the
presence of a true radiation component resulting in a trivial change A → A′ in (2.7). This

–3–
can model a bounce in the primordial universe. The bounce will survive provided the sum
of “dark” and true radiation remains negative or
Λ ′4
A + ρrad,0 a′4 ′
0 ≡A =− a < 0, (2.16)
κ2 0
′
which shows that aA4 = aA′4 is invariant, the scale factor a′0 at the bounce decreases in the
0 0
presence of radiation and also when A increases towards zero. Let us finally note that the
bounce disappears for A′ ≥ 0.

3 Study of the bouncing solution

JCAP07(2015)002
In spite of the fact that we have obtained an explicit closed form expression for Φ(t) we will
study the behaviour of the bouncing solution by applying the qualitative theory of differential
equations. This will provide us with a more intuitive and direct approach and will enable
us to obtain results about the qualitative behaviour of our system otherwise difficult to find
from (2.15). Solving for the wave equation recast into the form
Φ̈ + 3H Φ̇ + 4cΦ(Φ̃2 − Φ2 ) = 0 , (3.1)
Λ

21
where Φ̃ ≡ 6c , and evaluating (2.2) at the bounce
1 2
Φ̇ − c(Φ40 − Φ40,min ) = 0 , (3.2)
2 0

14  1
Λ −A 4
where Φ0,min ≡ κ2 c
= ca40
, our system is automatically solved.
The inequality Φ2 < 6κ−2 is readily obtained from the requirement F > 0 while Φ = 0
is impossible from (2.8). Choosing Φ0 > 0, the following inequality is further obtained
Φ̃ < Φ0,min ≤ Φ0 < Φmax , (3.3)
with 6 κ−2 ≡ Φ2max . Indeed, from (2.2), the condition U (Φ0 ) ≡ U0 ≤ 0 implies immediately
Φ0 ≥ Φ0,min , while Φ < Φmax , valid at all times, follows from the physical condition F > 0
(see figure 1). A non vanishing interval allowed for Φ0 requires Φ0,min < Φmax or
Λ κ2
< 1, (3.4)
36c
implying also Φ̃ < Φ0,min . From (3.2), with Φ̇0 < 0 (the justification of this choice will be
clear below), we obtain
q
Φ̇0 = − 2c(Φ40 − Φ40,min ), (3.5)

with Φ̇0 decreasing when Φ0 is increasing.

We study the dynamics for t ≥ 0. For t < 0 the dynamics follows immediately due to
the invariance of the system under reflection t → −t. We will come back to this point at the
end of our analysis. We will analyse now the possible existence of local maxima and minima
for Φ(t). From (2.8) we readily obtain for Φ̇ = 0
q
1 2 H4 −4
2 2H + 4 − 4cAa
Φ = . (3.6)
2c

–4–
 0.1

F
- 0.6 - 0.4 - 0.2 0.2 0.4 0.6

- 0.1

- 0.2

Figure 1. The functions F (full) and U (dashed) are displayed for the model parameters κ2 = 20,

JCAP07(2015)002
c = 3 and Λ = 3. The two curves coincide at Φ = 0 because we√have actually plotted ΛF . While
1
U vanishes at Φ ≡ Φ0,min = ( κΛ2 c ) 4 , F does so at Φ ≡ Φmax = κ6 . For these parameters we have
Φ0,min = 0.4728, Φ0,cr = 0.4936860, Φmax = 0.5477. For Φ > 0, the range of possible values at the
bounce lie in the interval Φ0,min ≤ Φ < Φmax but only values in the interval Φ0,cr ≤ Φ < Φmax and
with a negative slope will yield physically viable solutions.

The choice of the positive sign before the square root is compulsory since Φ2 must be positive.
It is straightforward to show that (3.4) implies Φ < Φmax for Φ given by (3.6). Hence the
inequality Φ < Φmax follows when Φ̇ = 0. We have further from (2.3) when Φ̇ = 0 that
−ΦΦ̈ + 6F Ḣ = 0, therefore Φ̈ > 0. So we have shown the impossibility for Φ to have a
maximum.
On the other hand if Φ has a finite limit for t → ∞, we must have Φ̇∞ = 0 and Φ̈∞ = 0,
therefore (3.1) gives 4cΦ∞ (Φ̃2 − Φ2∞ ) = 0. Hence we have either Φ∞ = 0 or Φ∞ = Φ̃.
These results imply in particular that had we chosen Φ̇0 > 0, Φ would eventually enter the
unphysical domain Φ ≥ Φmax , so this choice is not allowed.
If the solution has a minimum, it is unique as Φ does not have any maximum. This
minimum must be larger that Φ̃ as a direct consequence of (3.1). Hence after the minimum,
Φ̇ > 0 forever since there is no maximum and then Φ goes to infinity — therefore crossing the
value Φmax — because the only possible finite limits Φ∞ are lower than the minimum and
cannot be reached. We will now show that in this case Φ tends to infinity in a finite time
√ t∞
(see figure 2). When Φ → ∞, keeping the leading terms in (2.2) we easily find Φ̇ ∼ 2cΦ2 ,
and after integration
1
Φ∼ √ . (3.7)
− 2ct + k

with k > 0 and t∞ = √k2c . As a consequence F becomes zero in a finite time too, rendering
this solution unphysical. When Φ0 = Φ0,min (U0 = 0), the bounce is a minimum from
eqs. (3.1) and then Φ goes to infinity in a finite time.
Let us consider now the case where there is no minimum, i.e. Φ̇ < 0 for t ≥ 0. From (3.1)
we get Φ̈ = −3H Φ̇ − 4cΦ(Φ̃2 − Φ2 ) > 0 as long as Φ > Φ̃. However when Φ(t) < Φ̃, the sign
of Φ̈(t) can change. Two subcases will therefore arise:
1) Φ(t) is always greater than Φ̃ and, since Φ̇(t) < 0, tends asymptotically to Φ = Φ̃ with
positive concavity (Φ̈(t) > 0).

2) Φ(t) crosses Φ̃ and, since Φ̇(t) < 0, tends asymptotically to 0. Below Φ̃, Φ̈(t) can
however change sign.

–5–
 1.5

1.0

0.5

0.5 1.0 1.5 2.0 2.5

Figure 2. The function Φ(t) is shown starting at t = 0 at the bounce for the same model parameters

JCAP07(2015)002
as in figure 1. The initial value Φ0 is slightly below the critical value Φ0,cr = 0.4936860 . . . so this
solution will diverge in a finite time.

In order to differentiate these two possibilities, we study the system in the neighborhood
of t = ∞ where (3.1) can be recast as an autonomous planar system, viz.

y˙1 = y2 , (3.8)
√ Λ
y˙2 = − 3Λ y2 − 2 y1 + 4cy13 , (3.9)
3

with y1 ≡ Φ. This system has two hyperbolic fixed (critical)qpoints: (0, 0) and q (Φ̃, 0). The
eigenvalues associated to (0, 0) are both negative, λ1 = − 3 and λ2 = − 4Λ
Λ
3 , so it is
q
Λ
a stable node. The eigenvalues associated with (Φ̃, 0) have opposite sign, λ1 = 3 and
q
λ2 = −2 4Λ 3 , hence it is a saddle point. So the solution Φ → Φ̃ is “unstable” and only with
an exact initial value Φ0 can one reach the fixed point (Φ∞ , Φ̇∞ ) = (Φ̃, 0). In contrast, once
we find some Φ0 value which yields a solution tending to 0, we can vary Φ0 in a neighbourhood
of this value and still go to the fixed point (0, 0).
By continuity, if we take Φ0 slightly larger than Φ0,min , Φ(t) will still go to infinity in a
finite time. Further increasing Φ0 , two possibilities arise:

a) For Φ0,min ≤ Φ0 < Φ0,cr , where Φ0,cr is some critical value, Φ(t) diverges in a finite
time.

b) For Φ0,cr < Φ0 < Φmax solutions will tend to zero.

In the absence of Φ0,cr only the first behaviour holds. For the specific case Φ0 = Φ0,cr , the
solution will tend to Φ̃. So this solution which tends to the saddle point is the separatrix which
separates the two types of behaviours. Hence for given parameters κ, Λ and c satisfying (3.3),
the two types of behaviours are possible provided the existence of the separatrix.
Determination of the initial value Φ0,cr yielding the separatrix is a well-known problem in
the theory of ordinary differential equations, known as the connexion problem [19]. For given
values of the parameters of the system, it is not possible in general to determine analytically
whether Φ0,cr exists, neither to find a closed form expression for it when it does exist. We
were able to find numerically Φ0,cr for certain values of the parameters. For instance, for
κ2 = 20, c = 3 and Λ = 3, we find Φ0,cr = 0.4936860 . . .. An example of viable solutions is
shown on figure 3.

–6–
 F F
0.5
0.5

0.4
0.4

0.3
0.3

0.2
0.2

0.1 0.1

t t
2 4 6 8 10 12 2 4 6 8 10 12

JCAP07(2015)002
Figure 3. The function Φ(t) is shown starting at t = 0 at the bounce for the same model parameters
as in figure 1. On the left panel, the initial value is Φ0 = 0.54 while on the right panel Φ0 = 0.50.
As explained in the text Φ̈ is positive above Φ̃ whose value here is Φ̃ = 0.40824829 . . ., but can
change sign below Φ̃. The curve on the right has a more pronounced feature because it starts closer
to Φ0,cr = 0.4936860 . . .. Both curves tend asymptotically to zero and constitute viable bouncing
solutions. The curve starting precisely at Φ0,cr will tend asymptotically to Φ̃.

Using the analytical results given earlier, one can derive an analytical expression for Φ0,cr
 1
Λ 4 1
Φ0,cr = −  . (3.10)
cκ2

2j+1 9c

1  Γ2 ( 14 )
dn √
2
− Λκ2
4 √
4 π
|2

Let us explain how (3.10) is obtained. As we look for the initial value Φ0,cr for which Φ → Φ̃,
it is clear from (2.15) that u(t = ∞) must then correspond to a zero of the function dn(u | 2),
which are given by

(2j + 1) Γ2 41


√ , j = 0, 1, . . . . (3.11)
4 2π

Using (2.14), we obtain

1 1 !
 c  1 √ Γ2 1

(2j + 1)Γ2


4 −1 Λ 4 1 4
√ = dn − |2 + 3 √4 . (3.12)
4 2π cκ2 Φ0,cr Λκ2 4 π

Inverting this expression we obtain (3.10).

The value of j must be chosen in such a way that Φ0,cr lies inside the allowed interval (3.3)
and that (3.12) is satisfied. As the function dn(u|2) is periodic, only a small number of values
of j must be considered. The numerical value of Φ0,cr given above was obtained from (3.10)
for j = 1. For some values of the model parameters and any value of j, Φ0,cr is not in the
allowed interval and then the resulting model is not viable.
In the presence of radiation one checks easily that the allowed range of initial values
shrinks with a new larger minimal value Φ′0,min satisfying

A
Φ′4 4
0,min = Φ0,min . (3.13)
A′

–7–
We have of course A < A′ < 0. The quantities Φ̃, Φmax on the other hand depend
on the model parameters and remain unchanged. One can just repeat the proof of the
existence of viable bouncing solutions. Hence, as long as we have Φ′0,min < Φ0,cr , the
set of viable
 bouncing solutions  remains. In the new expression for Φ(t) we will have
√ A 9c
1

−1

1
A Λ 4 1
q
Λ
u → dn − A′ cκ2 Φ0 | 2 − i 2 A′ κ2 Λ 4 F i 3 t | 2 while the same corrective

1
factor AA′ 4 appears in front of (2.15) as well. We have the obvious corresponding change
in (3.12) and (3.10).
Let us consider next the possibility to have a viable bounce when Z = −1 in the absence
of radiation. If we want to keep the conformal invariance, the absence of ghosts (ωBD > − 32 )

JCAP07(2015)002
and the existence of a bounce, we are led to the following ansatz:

FZ=−1 (Φ) = −FZ=1 (Φ) UZ=−1 (Φ) = −UZ=1 (Φ) , (3.14)

with the parameters κ−2 , Λ still positive. The quantities FZ=1 , resp. UZ=1 , correspond
to (2.5), resp. (2.6). It is easily checked that (2.7) is retrieved leading to a bounce provided
the quantity A is negative. This implies here too that c > 0, but now UZ=−1 no longer
represents an inverted potential and we have UZ=−1 (Φ = 0) < 0. As we have emphasized
earlier, with (3.14) the same equation of motion for the field Φ is obtained therefore leading
to the same solution (2.15). However, the crucial √difference is that the domain of physical
validity differs in both cases: now all values Φ > κ6 satisfy FZ=−1 > 0. At the bounce, it
is seen from (2.2) that the potential must satisfy UZ=−1 (Φ0 ) ≥ 0 yielding in this case too
Φ0 ≥ Φ0,min .
Up to these differences, we can essentially repeat the analysis performed for Z = 1. It
is easy to show using (2.4) that Φ̇ = 0 implies Φ̈ > 0 for Φ > Φ̃, and Φ̈ < 0 for Φ < Φ̃. On
the other hand from (2.3), Φ̇ = 0 implies Φ̈ < 0. Hence Φ̇ = 0 is not possible for Φ > Φ̃
while it would correspond to a maximum for Φ < Φ̃. On the other hand, from (2.4) only
Φ̃ remains an acceptable finite limit of Φ because Φ = 0 is now excluded. Imposing the
2
condition Λκ36c > 1, we now obtain

√
6
< Φ0,min < Φ̃ . (3.15)
κ

If Φ0 > Φ̃, Φ̇0 > 0 implies Φ → ∞ while Φ̇0 < 0 can in principle produce two possible
behaviours namely Φ → Φ̃ (in that case Φ0 = Φ0,cr ) or Φ → 0, the latter being unviable.
The first possibility cannot occur. Indeed with Φ̇0 < 0, some initial values above as well as
below Φ0,cr would tend to zero. But the particular value Φ0,cr must separate two different
behaviours, hence this is not possible. When Φ0 < Φ̃, Φ → 0 for a negative initial slope
because there is no minimum. With Φ̇0 > 0, three different behaviours can be found: Φ → 0
when Φ goes through a maximum (below Φ̃), Φ → ∞ when there is no maximum and finally
Φ → Φ̃ for Φ0,cr which now separates the two behaviours mentioned earlier, its existence
is therefore allowed in this case. It is interesting that while a viable bounce can exist for
Z = −1 with a potential bounded from below, it has measure zero in the initial conditions,
only one particular set of initial conditions at the bounce is acceptable. The expression of
Φ0,cr for this case is also given by (3.10).

–8–
4 Conclusions

We have shown that for Z = 1 only three types of solutions are possible. All stable solutions
vanishing asymptotically are especially
√ −1interesting, their initial value Φ0 at the bounce can
vary in some range Φ0,cr < Φ0 < 6κ . The solution with Φ0 = Φ0,cr is unique (measure
zero) and moreover unstable. The stable solutions lead to the General Relativity limit if we
assign to the bare gravitational constant κ2 its numerical value in GR. Consistency of our
2
model requires then the inequality Λ8πκ ≪ 1. If we want to embed our model in a realistic
cosmology, this attractive feature is even reinforced by the possibility to add a radiation
component to our bouncing model.

JCAP07(2015)002
Further, due to the invariance of the equations under the change t → −t, the solutions
for t < 0 are obtained from those found for t > 0 changing the sign of Φ̇. Hence, regular
solutions for t > 0 will diverge in the past both for Z = 1 and Z = −1. Still, starting with
regular initial values at −∞ < t < 0, it is possible to have a bouncing universe regular in
the future. Note that a solution symmetric around the bounce requires Φ̇0 = 0 which will
diverge both in the past and in the future.
It is well known that a bounce in the early universe can cure many of the problems
occurring in Big Bang cosmology besides of course avoiding the Big Bang singularity itself.
Without exploring further this scenario, we have shown that surprisingly such a bounce can
occur for a FLRW flat universe within one of the oldest and best understood alternatives to
GR in a model tending dynamically to GR after the bounce. The key point is to admit a
negative scalar field potential that can lead to a transient negative energy density of the scalar
field for some range of parameters.3 Of course, to make the model cosmologically viable, one
has to generalize it somehow to open channels for decay of the effective positive (and large)
cosmological constant Λ into other quantum fields including those of the Standard Model of
elementary particles at late times.

Acknowledgments

A.S. was partially supported by the RFBR Grant No. 14-02-00894 and by the Russian
Government Program of Competitive Growth of the Kazan Federal University.

References

[1] A.A. Starobinsky, On a nonsingular isotropic cosmological model, Sov. Astron. Lett. 4 (1978)
82.
[2] D.N. Page, A Fractal Set of Perpetually Bouncing Universes?,
Class. Quant. Grav. 1 (1984) 417 [INSPIRE].
[3] A.Y. Kamenshchik, I.M. Khalatnikov and A.V. Toporensky, Simplest cosmological model with
the scalar field, Int. J. Mod. Phys. D 6 (1997) 673 [gr-qc/9801064] [INSPIRE].
[4] V.N. Melnikov and S.V. Orlov, Nonsingular cosmology as a quantum vacuum effect,
Phys. Lett. A 70 (1979) 263 [INSPIRE].
3
In this connection, the early paper [20] has to be mentioned, too, where a FLRW bounce in f (R) gravity
was achieved in case of the scalar degree of freedom in this theory (scalaron) being tachyonic permanently
that corresponds to a negative (and even unbounded from below) effective scalar field potential in the Einstein
frame, see also [21].

–9–
 [5] A.A. Starobinsky, Can the effective gravitational constant become negative?, Sov. Astron. Lett.
7 (1981) 36.
[6] G.W. Gibbons and N. Turok, The Measure Problem in Cosmology,
Phys. Rev. D 77 (2008) 063516 [hep-th/0609095] [INSPIRE].
[7] D.A. Tretyakova, A.A. Shatskiy, I.D. Novikov and S. Alexeyev, Non-singular Brans-Dicke
cosmology with cosmological constant, Phys. Rev. D 85 (2012) 124059 [arXiv:1112.3770]
[INSPIRE].
[8] J.D. Barrow and D. Sloan, Bouncing Anisotropic Universes with Varying Constants,
Phys. Rev. D 88 (2013) 023518 [arXiv:1304.6699] [INSPIRE].
[9] A. Ashtekar, T. Pawlowski and P. Singh, Quantum Nature of the Big Bang: Improved

JCAP07(2015)002
dynamics, Phys. Rev. D 74 (2006) 084003 [gr-qc/0607039] [INSPIRE].
[10] T. Qiu, J. Evslin, Y.-F. Cai, M. Li and X. Zhang, Bouncing Galileon Cosmologies,
JCAP 10 (2011) 036 [arXiv:1108.0593] [INSPIRE].
[11] D.A. Easson, I. Sawicki and A. Vikman, G-Bounce, JCAP 11 (2011) 021 [arXiv:1109.1047]
[INSPIRE].
[12] D. Battefeld and P. Peter, A Critical Review of Classical Bouncing Cosmologies,
Phys. Rept. 571 (2015) 1 [arXiv:1406.2790] [INSPIRE].
[13] C. Martinez, R. Troncoso and J. Zanelli, de Sitter black hole with a conformally coupled scalar
field in four-dimensions, Phys. Rev. D 67 (2003) 024008 [hep-th/0205319] [INSPIRE].
[14] B. Boisseau, G. Esposito-Farese, D. Polarski and A.A. Starobinsky, Reconstruction of a scalar
tensor theory of gravity in an accelerating universe, Phys. Rev. Lett. 85 (2000) 2236
[gr-qc/0001066] [INSPIRE].
[15] R. Gannouji, D. Polarski, A. Ranquet and A.A. Starobinsky, Scalar-Tensor Models of Normal
and Phantom Dark Energy, JCAP 09 (2006) 016 [astro-ph/0606287] [INSPIRE].
[16] D. Polarski, Past and future of some universes, arXiv:1303.4470 [INSPIRE].
[17] V.A. Rubakov, Harrison-Zeldovich spectrum from conformal invariance, JCAP 09 (2009) 030
[arXiv:0906.3693] [INSPIRE].
[18] M. Abramowitz and I.A. Stegun, Handbook of mathematical functions with formulas, graphs,
and mathematical tables, M. Abramowitz and I.A. Stegun eds., Dover Publications Inc.,
New York U.S.A. (1992).
[19] B. Boisseau, P. Forgacs and H. Giacomini, An Analytical approximation scheme to two point
boundary value problems of ordinary differential equations, J. Phys. A 40 (2007) F215
[hep-th/0611306] [INSPIRE].
[20] T.V. Ruzmaikina and A.A. Ruzmaikin, Quadratic Corrections to the Lagrangian Density of the
Gravitational Field and the Singularity, Sov. Phys. JETP 30 (1970) 372.
[21] M. Ivanov and A.V. Toporensky, Stable super-inflating cosmological solutions in f(R)-gravity,
Int. J. Mod. Phys. D 21 (2012) 1250051 [arXiv:1112.4194] [INSPIRE].

– 10 –