Arxiv:1301.4703V2 [Gr-Qc] 27 Jan 2013 Models Are Unstable to the Addition of Radiation

Anisotropy in a Nonsingular Bounce

Yi-Fu Cai,1, ∗ Robert Brandenberger,1, † and Patrick Peter2, ‡ 1Department of Physics, McGill University, Montréal, QC H3A 2T8, Canada 2 GRεCO – Institut d’Astrophysique de Paris, UMR7095 CNRS, Université Pierre & Marie Curie, 98 bis boulevard Arago, 75014 Paris, France (Dated: June 26, 2018) Following recent claims relative to the question of large anisotropy production in regular bouncing scenarios, we study the evolution of such anisotropies in a model where an Ekpyrotic phase of contraction is followed by domination of a Galileon-type Lagrangian which generates a non-singular bounce. We show that the anisotropies decrease during the phase of Ekpyrotic contraction (as expected) and that they can be constrained to remain small during the non-singular bounce phase (a non-trivial result). Speciﬁcally, we derive the e-folding number of the phase of Ekpyrotic contraction which leads to a present-day anisotropy in agreement with current observational bounds.

PACS numbers: 98.80.-k,98.80.Cq

I. INTRODUCTION ferentiated from inﬂation is the shape of the bispectrum [12].

The idea of a non-singular bouncing universe as an al- Bouncing cosmologies, however, face several problems. ternative to the singular Big Bang paradigm has been First of all, new physics is required to provide the bounce attractive on philosophical grounds for a long time. (a pure GR bounce sourced by a regular fluid also leads More recently [1, 2] it has been realized that a non- to dynamical instabilities [13], while with a scalar field singular bouncing cosmology with a matter-dominated a positive spatial curvature [14] or non-canonical kinetic initial phase of contraction - a “matter bounce” cosmology term [15] is required). Such new physics can either come - may provide an alternative to the inflationary paradigm from introducing new forms of matter such as phantom or of cosmological structure formation (see e.g. [3] for re- quintom fields [16], ghost condensates [17, 18], Galileons cent reviews). Since the curvature perturbation ζ grows [19], S-branes [20] or effective string theory actions [21]. on super-Hubble scales in the contracting phase, fluctu- It can also arise from corrections to Einstein gravity, as ations on long wavelengths get boosted by a larger fac- e.g. in the non-singular universe construction of [22], tor than those on small scales. In a matter-dominated specifically applied to the bouncing scenario in [23], in phase of contraction the relative enhancement factor is the ghost-free higher derivative gravity model of [24], in precisely such as to transform an initial vacuum spectrum terms of torsion gravity [25], or in Hořava-Lifshitz gravity into a scale-invariant one. Studies in many non-singular [26]. Finally, a bouncing phase can also originate from models [4] have shown that the scale-invariance of the pure quantum cosmological effects [27]; those also pro- spectrum of ζ1 before the bounce is maintained through vide a natural framework in which a dust-dominated con- the bounce on length scales which are larger than the traction is easily implemented to yield a scale-invariant duration of the bouncing phase (the phase in which the spectrum of perturbation [28]. violation of the usual energy conditions is localized)2. A more serious problem for bouncing cosmologies is Hence, the matter bounce scenario leads to the same the fact that the contracting phase is unstable to various shape of the spectrum of primordial curvature fluctua- effects. As discussed e.g. in [29], many matter bounce tions on super-Hubble scales as inflationary cosmology.

arXiv:1301.4703v2 [gr-qc] 27 Jan 2013 models are unstable to the addition of radiation. This Starting with vacuum initial conditions, the fluctuations arises since the energy density of radiation scales as a−4, are also Gaussian and coherent. A specific prediction of where a(t) is the cosmological scale factor, whereas the the matter bounce with which the scenario can be dif- density of matter grows as a−3. In some models (e.g. [16]) this will lead to a “Big Crunch” singularity instead of a smooth bounce. Some constructions (e.g. those of [18, 22], [20] and [30]) are free from this problem. How- ∗Electronic address: [email protected] †Electronic address: [email protected] ever, a problem for all constructions mentioned so far is ‡Electronic address: [email protected] the instability to the growth of anisotropic stress, whose 1 The variable ζ is the curvature fluctuation in comoving gauge. associated energy density grows as a−6. This is the fa- It is proportional to the canonical fluctuation variable v - the mous BKL instability of a contracting universe [31]. Sasaki-Mukhanov variable [5, 6] - in terms of which the action for cosmological fluctuations has canonical form. See e.g. [7] for A solution of this anisotropy problem can be realized in a comprehensive review and [8, 9] for introductory overviews of the theory of cosmological perturbations. the Ekpyrotic scenario [32], in which it is postulated that 2 Note that there are counterexamples as discussed in [10] and a matter field with equation of state parameter w 1 more recently in [11]. (where w = p/ρ is the ratio of pressure p to energy den- 2 sity ρ) is dominant in the contracting phase3. Such an not apply to all non-singular bouncing cosmologies. equation of state can be realized by treating the dominant The outline of this paper is as follows. In the next sec- form of matter as a scalar field with negative exponen- tion we review the bounce model introduced in [38] and tial potential. Since the energy density of the dominant derive the resulting equations of motion for a homoge- matter then scales with a−q with q 6, anisotropies be- neous but anisotropic universe. In Section 3 we analyt- come negligible and the BKL instability is avoided [37]4. ically study the background dynamics in each phase of In a recent paper [38], a subset of the present authors the cosmological evolution from the initial matter phase introduced a scalar field with an Ekpyrotic potential to of contraction through the Ekpyrotic phase to the bounc- construct a matter bounce scenario which is free from the ing phase and the subsequent fast-roll expanding period. BKL instability problem. Specifically, we determine the decay or growth rates of The Ekpyrotic scenario in its original formulation [32] the anisotropy parameter in each phase. In Section 4 we involves a singular bounce. In addition, the curvature solve the dynamical system numerically and present our final results. We close with a general discussion. spectrum of ζ is an nS = 3 spectrum rather than a scale- invariant n = 1 one [39–42]. Hence, without non-trivial A word on notation: We define the reduced Planck S p matching of ζ across the bounce, one cannot obtain a mass by MPl = 1/ 8πGN where GN is Newton’s gravi- scale-invariant spectrum at late time5. To solve this tational constant. The sign of the metric is taken to be problem, a new and non-singular version of the Ekpyrotic (+, −, −, −). Note that we take the value of the mean scenario [46] was proposed in which a second scalar field scale factor at the bounce point to be aB = 1 throughout is introduced which does not influence the background the paper. dynamics but develops a scale-invariant spectrum which starts out as an isocurvature mode but which is trans- ferred to the adiabatic mode during the evolution. The II. A NONSINGULAR BOUNCE MODEL second field can also be given a “ghost condensate” La- grangian [47] in which case it mediates a non-singular We consider a nonsingular bounce model in which the bounce. However, as has been pointed out in [48], in universe is filled with two matter components, a cosmic this “New Ekpyrotic” scenario the anisotropies which are scalar field φ and a generic matter fluid, as proposed in highly suppressed during the contracting phase again Ref. [38] (which, in turn, is based on the theory devel- raise their head and lead to a BKL instability. oped in [49]). The Lagrangian of φ is given by In our previous work [38], we argued qualitatively that in the model we considered the anisotropies remained L [φ (x)] = K(φ, X) + G(φ, X)φ, (1) negligibly small during the bouncing phase. The reason for the difference compared to what happens in the model where K and G are functions of φ and its canonical ki- of [46] is that in our model the kinetic condensate which netic term grows as the bounce is approached does not need to de- 1 µ crease again by the time of the bounce point. This leads X ≡ ∂µφ∂ φ, (2) to a shorter bounce time scale and to different dynamics. 2 In this paper we carefully study the development of while the other kinetic terms of φ include the operator anisotropies in the bouncing cosmology with an Ekpy- µν rotic phase of contraction introduced in [38]. We work φ ≡ g ∇µ∇ν φ. (3) in the context of a homogeneous but anisotropic Bianchi cosmology in which the scale factors in each spatial di- Variation of the above scalar field Lagrangian mini- mension evolve independently. We are able to show that mally coupled to Einstein gravity leads to the following no BKL type instability develops, in agreement with corresponding energy momentum tensor what the study of [38] indicated. Our work thus shows φ σ that the arguments against non-singular (as opposed to Tµν = (−K + 2XG,φ + G,X ∇σX∇ φ)gµν singular) bouncing cosmologies put forwards in [48] do +(K,X + G,X φ − 2G,φ)∇µφ∇ν φ −G,X (∇µX∇ν φ + ∇ν X∇µφ), (4)

in which we use the notation that F and F denote 3 ,φ ,X There are other approaches to address the anisotropy prob- derivatives of whatever functional F(φ, X) may be with lem. For example, nonlinear matter terms may smooth out the αβ respect to φ and X, respectively. anisotropies [33]. Adding quadratic Rαβ R terms to the gravitational action can also prevent the BKL instability [34]. For the model under consideration we choose: 4 Note, however, that including anisotropic pressures may reintro- duce instabilities towards anisotropy generation [35]. K(φ, X) = M 2 [1 − g(φ)] X + βX2 − V (φ), (5) 5 However, the spectrum of the Bardeen potential Φ is scale- Pl invariant [43], and, as argued in [10] and shown explicitly in some examples [44, 45], it is this spectrum which may pass through where we introduce a positive-definite parameter β so the bounce, thus yielding a scale-invariant spectrum of curva- that the kinetic term is bounded from below at high en- ture fluctuations at late times. ergy scales. Note that the first term of K involves M 2 Pl 3 since in the present paper we adopt the convention that the scalar field φ is dimensionless. The function g(φ) is chosen such that a phase of ghost condensation only occurs during a short time when φ approaches φ = 0. This requires the dimensionless function g to be smaller than unity when |φ| 1 but larger than unity when φ approaches the origin. To obtain a nonsingular bounce, we must make an explicit choice of g as a function of φ. We want g to be negligible when |φ| 1. In order to obtain a violation of the Null Energy Con- dition after the termination of the Ekpyrotic contracting phase, g must become the dominant coefficient in the quadratic kinetic term when φ approaches 0. Thus, we suggest its form to be

2g0 g(φ) = q q , (6) − 2 φ b 2 φ e p + e g p where g0 is a positive constant defined as the value of g Figure 1: Model functions g(φ) and V (φ) as given by Eqs. (6) at the moment when φ = 0, and is required to be larger and (7), with background parameters taken as for the follow- than unity, g0 > 1. ing evolution figures, namely as in Eqs. (65) and (66). We have also introduced a non-trivial potential V for φ. This potential is chosen such that Ekpyrotic contraction is possible. It is well known that the homogeneous tra- function of only X: jectory of a scalar field can be an attractor solution when its potential is an exponential function. One example is G(X) = γX, (8) inflationary expansion of the universe in a positive-valued exponential potential, and the other one is the Ekpyrotic where γ is a positive-definite number. model in which the homogeneous field trajectory for a We now turn to the study of the cosmology of this negative exponential potential is an attractor in a con- model. In order to characterize a homogeneous but tracting universe. For a phase of Ekpyrotic contraction, anisotropic universe, we take the metric to be of the form we take the form of the potential to be X ds2 = dt2 − a2(t) e2θi(t)σiσi, (9) 2V V (φ) = − 0 , (7) i q 2 q 2 − φ bV φ e q + e q where t is cosmic time, σi are linearly independent at

4 all points in space-time and form a three dimensional where V0 is a positive constant with dimension of (mass) . homogeneous space. Thus the potential is always negative and asymptotically In the case of a Ricci flat space, one can consider the approaches zero when |φ| 1. Ignoring the second term projection σi = dxi and thus the metric is of Bianchi of the denominator, this potential reduces to the form type-I form. The factor a(t) can be viewed as the mean used in the Ekpyrotic scenario [32]. Both functions g(φ) scale factor of this universe, and the functions eθi(t) de- and V (φ) are shown on Fig. 1 with the parameters used scribe the correction of anisotropies to the scale factor. in the later parts of this work. Since the values of scale factors can be re-scaled arbitrar- 6 The term G(φ, X) is a Galileon type operator which ily, one can impose an additional constraint is consistent with the fact that the Lagrangian contains higher order derivative terms in φ, but the equation of X θi = 0. (10) motion remains a second order differential equation. Phe- i nomenologically, there are few requirements on the explicit form of G(φ, X). We introduce this operator since Then, one can immediately define a mean Hubble param- we expect that it can be used to stabilize the gradi- eter as follows, ent term of cosmological perturbations, which requires a˙ that the sound speed parameter behaves smoothly and is H ≡ , (11) positive-definite throughout most of the background evo- a lution. For simplicity, we will choose G to be a simple and the individual Hubble parameters along spatial directions are given by, 1 d θi ˙ 6 See [36] for a discussion of Galileon type Lagrangians. Hi ≡ ae = H + θi, (no sum) (12) aeθi dt 4 where the overdot denotes the derivative with respect to values of φ˙. However, these terms will become important cosmic time t. during the bouncing phase where φ˙ reaches a maximal Since we are interested in studying anisotropies rather value. For simplicity, in the following we will consider than inhomogeneities we can treat the matter fields to be matter fluid is cold and thus wm = 0. homogeneous, which implies φ is only a function of cosmic Finally, we can write down Einstein equations in this time. Thus, the kinetic terms of the homogeneous scalar background, given by field background become R M 2 R − g = T φ + T m . (19) 1 Pl µν µν µν µν X = φ˙2, 2 2 φ = φ¨ + 3Hφ,˙ (13) Once expanded in components, this tensor equation yields the effective Friedmann equations, so that, for this background, the energy density of the ρ 1 X scalar field is H2 = T + θ˙2, (20) 3M 2 6 i Pl i 1 2 ˙2 3 ˙4 ˙3 ρφ = M (1 − g)φ + βφ + 3γHφ + V (φ), (14) ρ + p 1 X 2 Pl 4 H˙ = − T T − θ˙2, (21) 2M 2 2 i and the pressure is Pl i where ρ and p represent the total energy density and 1 1 T T p = M 2 (1 − g)φ˙2 + βφ˙4 − γφ˙2φ¨ − V (φ), (15) φ Pl pressure in the Bianchi type-I universe, i.e., the sum of 2 4 the contributions of the scalar field and the fluid. as follows by computing the diagonal components of the Moreover, combining the spatial component of Ein- stress-energy tensor (4). stein equation with the constraint equation (10) yields Additionally, the matter fluid contributes its own en- ¨ ˙ ergy density ρm and pressure pm, and usually they are θi + 3Hθi = 0, (22) associated with a constant equation-of-state parameter 1 from which it follows that wm = pm/ρm. Namely, for normal radiation, wm = 3 , while for normal matter, wm = 0. a3 θ˙ (t) = M B , (23) To derive the equation of motion for φ, one can either i θ,i a3(t) vary the Lagrangian with respect to φ or, equivalently, require that the covariant derivative of its stress-energy where aB is the mean scale factor of the universe at the tensor vanishes. This yields bouncing point. The coefficients Mθ,i are integral constants with a dimension of mass. According to the con- ¨ ˙ Pφ + Dφ + V,φ = 0, (16) straint equation (10), one can read off that where we have introduced X Mθ,i = 0. (24) 2 3γ i P = (1 − g)M 2 + 6γHφ˙ + 3βφ˙2 + φ˙4, (17) Pl 2M 2 Pl Plugging Eq. (23) into Eq. (20) shows that one can 2 2 1 2 ˙ ˙2 introduce an effective energy density of anisotropy D = 3(1 − g)M H + 9γH − M g,φ φ + 3βHφ Pl 2 Pl M 2 2 4 5 Pl X ˙2 −6 3 9γ Hφ˙ 3βγφ˙ ρθ ≡ θi ∝ a , (25) − (1 − g)γφ˙3 − − 2 2 2M 2 2M 2 i Pl Pl 6 3 X ˙2 ˙ 3G,X ˙ whose evolution as 1/a implies an effective equation-of- − G,X θi φ − (ρm + pm)φ. (18) 2 2M 2 state parameter equal to wθ = 1. We see that this effec- Pl i tive energy density increases faster than that of pressure- From Eq. (16), it is clear that the function P determines less matter or radiation in a contraction universe. This the positivity of the kinetic term of the scalar field and is the source of the BKL instability of the contracting thus can be used to determine whether the model con- phase of many bouncing cosmologies. tains a ghost or not at the perturbative level; the function D on the other hand, represents an effective damping term. By keeping the first terms of the expressions of III. BACKGROUND EVOLUTION P and D and setting g = 0, one can recover the standard Klein-Gordon equation in the FRW background. The initial conditions of our model are chosen (as in Neglecting the other terms is a good approximation when [38]) such that we start in a contracting phase dominated the velocity of φ is sub-Planckian. Note that the friction by regular matter. Since the energy density of the Ekpy- term D contains the contributions from anisotropic fac- rotic scalar field φ grows faster than that of regular mat- tors and matter fluid, which can be suppressed for small ter, φ will at some time begin to dominate the energy 5 density. At this point, the Ekpyrotic phase of contrac- Integrating the above yields the following expressions tion begins, and lasts until the nonsingular bounce in- for the anisotropy factors terval begins (this is the phase where the new physics t 3 3 Z 2a Mθ,i t − t˜ 2a Mθ,i effects dominate), followed by a period of fast-roll expan- θ = θ˙ (t0)dt0 = − B E E = − B H(t). i i 3a3 H ˜ 3a3 H2 sion, which in turn ends at a transition to the expansion −∞ E E t − tE E E of Standard Big Bang cosmology. We choose the initial (30) conditions for the density of regular matter and for the From Eq. (30), we can observe that the anisotropy fac- value of φ such that the temperature at which the Ekpy- tors θi approach zero in the limit of t → −∞. As the rotic phase begins is higher than that at the time of equal universe contracts with a matter-dominated equation of matter and radiation in the Standard Big Bang expand- state, the absolute value of H increases and thus θi be- ing phase. In this way, we ensure that initial vacuum comes larger as well. perturbations develop into a scale-invariant spectrum of Before the Ekpyrotic phase, the universe is dominated cosmological fluctuations on all scales which are currently by the pressureless matter fluid and the energy density of normal matter fluid evolves as ρ ' 3M 2 H2. How- probed. m Pl In the following we study the evolution of the ever, Eq. (29) shows that the effective energy density of 4 anisotropy in each of the periods of cosmological evo- anisotropies evolves as ρθ ∼ H and will therefore come lution mentioned above. to dominate. This is nothing but another way of stating the BKL instability in a contracting matter-dominated universe. In the following we will see that the Ekpyrotic phase of contraction cures this problem. However, in or- A. Matter contraction der to be able to enter this phase, the initial anisotropy cannot be too large for the model not to break down al- We start by considering the period when the universe together. The requirement is that ρθ is less than ρm at is dominated by a pressureless matter fluid, i.e., when the transition time t , which implies the background equation of state parameter is roughly E w = 0. In this phase, the mean scale factor evolves as X a6 M 2 6H2 E , (31) θ,i . E a6 2/3 i B t − t˜ a(t) ' a E , (26) E t − t˜ a requirement which we implement in our numerical dis- E E cussion below. Let us now move on to the discussion of the evolution where tE denotes the final moment of matter contraction and the beginning of the Ekpyrotic phase, and a is the of the anisotropy factors during the Ekpyrotic phase of E contraction. value of the mean scale factor at the time tE . In the ˜ above, tE is an integration constant which is introduced to match the mean Hubble parameter continuously at the B. Ekpyrotic contraction time tE , i.e.,

2 We assume a homogeneous scalar ﬁeld φ which is ini- t˜ ' t − . (27) E E 3H tially placed in the regime of φ −1 in the phase of E matter contraction. In this case, the Lagrangian for φ ap- Correspondingly, the mean Hubble parameter during the proaches the conventional canonical form and thus yields period of matter contraction is given by an attractor solution which is given by rq 2V (t − t˜ )2 2 φ(t) ' − ln 0 B− , (32) H(t) = . (28) 2 q(1 − 3q)M 2 ˜ Pl 3(t − tE ) ˜ where tB− is an integration constant which is chosen in Following Eq. (23), one can immediately write down order that the mean Hubble parameter at the end of the the time derivatives of the anisotropy factors as follows phase of Ekpyrotic contraction matches with the one at the beginning of the bouncing phase. This attractor so- a3 H2(t) θ˙ (t) = M B , (29) lution corresponds to an eﬀective equation of state i θ,i a3 H2 E E 2 w ' −1 + . (33) 3q where HE is the value of mean Hubble parameter at the moment t . Note that the dimensional parameters of E During the phase of Ekpyrotic contraction, the mean ˙ anisotropy Mθ,i are therefore related to the values of θi scale factor evolves as at the moment tE ; a smooth bounce will thus take place q t − t˜ only if these values satisfy some constraints which we a(t) ' a B− , (34) B− ˜ derive below. tB− − tB− 6

p p where aB− is the value of mean scale factor at the time tB− φ− ∼ − p/2 ln(2g0) and φ+ ∼ p/2 ln(2g0)/bg (assum- which corresponds to the end of Ekpyrotic contraction ing one term in the denominator of g(φ) dominates over and the beginning of the bouncing phase. Therefore, the the other at each transition time), the value of the func- mean Hubble parameter is given by tion g(φ) becomes larger than unity and thus the universe q enters a ghost condensate state. The occurrence of the H(t) ' , (35) ghost condensate naturally yields a short period of null t − t˜ B− energy condition violation and this in turn gives rise to a nonsingular bounce. where, in order to make H(t) continuous at the time tB− , one must set As shown in Ref. [38], we have two useful parameteri- q zations to describe the evolution of the scale factor in the t˜ = t − . (36) B− B− H bounce phase. One is the linear parametrization of the B− mean Hubble parameter Additionally, we require the mean scale factor to evolve smoothly and continuously at the time tE . This leads to H(t) ' Υt, (40) the relation q and the other is the evolution of the background scalar HB− aE ' aB− . (37) 2 2 HE ˙ ˙ −t /T φ(t) ' φB e , (41) Given these preliminaries, we find that the time derivatives of the anisotropy factors evolve as where the coefficient Υ is set by the detailed microphysics of the bounce. The coefficient T can be determined by a3 H 3q θ˙ (t) ' M B , (38) matching the detailed evolution of the scalar field at the i θ,i a3 H B− E beginning or the end of the bouncing phase, which will be addressed in next subsection. Thus, during the bounce and thus the effective energy density of anisotropy evolves the mean scale factor evolves as 6q as ρθ ∼ H . Hence, whereas ρθ still increases during the 1 Υt2 phase of Ekpyrotic contraction, the growth rate is much a(t) ' a e 2 . (42) slower than in the matter-dominated phase (given the B small value of q). In particular, the energy density in φ Note that a nonsingular bounce requires that the total increases much faster than that due to the anisotropies: energy density vanishes at the bounce point. The total the BKL instability is tamed in this manner by the Ekpy- energy density includes the contributions from the mat- rotic contraction phase. ˙ ter fields and the anisotropy factors. This leads to the Integrating the above expressions for θi we obtain ˙ following result for the value of φB Z t ˙ 0 0 θi(t) = θi,E + θi(t )dt 2 " s # (g0 − 1)M 12β(V + ρ + ρ ) t 2 Pl 0 m θ E φ˙ ' 1 + 1 + B 2 4 3β (g0 − 1) M 3 " 1−3q# Pl (2 − 3q)a Mθ,i 3q H ' B −1 + E . 3 2(g0 − 1) 2 3(1 − 3q)a HE 2 − 3q H ' M , (43) E 3β Pl (39)

From the expression (39) we see that the absolute values where we have made use of approximations that ρm and ρ are much less than V and V M 4 in the second of βi are monotonically increasing when cosmic time t θ 0 0 Pl line. These approximations must be valid for the model evolves from tE to the onset of the bouncing phase at time to hold since both ρm and ρθ are greatly diluted in Ekpy- tB− . However, they very rapidly converge to their limit- ing values, and since any constant part of the anisotropy rotic phase and V0 is the maximal absolute value of the factors can be absorbed in a rescaling of the spatial co- potential of φ which, according to the observational con- ordinates, it turns out that the actual anisotropy, effec- straint from the amplitude of cosmological perturbations, tively measured as the distance to an effective FLRW must be far below the Planck scale. space, is effectively decreasing as the amplitude of the Now we calculate the anisotropy factors in the bounc- mean Hubble parameter increases: the Ekpyrotic phase ing phase. We first study their time derivatives, which thus provides a simple solution to the BKL instability are given by growth in a contracting phase. 3 2 ˙ − 2 Υt θi(t) ' Mθ,ie . (44)

C. Bounce phase Thus we see that the integration constants Mθ,i can ˙ be interpreted as the values of θi at the bounce point ˙ In our model the scalar ﬁeld evolves monotonically tB = 0, and they are the maximal values θi will ever from φ −1 to φ 1. For values of φ between take throughout the whole cosmic evolution. It is again 7 easy to perform the integrals and obtain the anisotropy phase we ﬁnd the following approximate solution for the factors: evolution of φ:

! t 3 r r a 2 2 π 3Υ B+ −t /T H(t) θ (t) ' θ + M Erf t , (45) φ˙(t) ' φ˙ ' φ˙ e B+ , (52) i i,B− θ,i B+ a3(t) B H 6Υ 2 B+ tB− where we have applied (41) in the second equality. This where Erf is the Gauss error function. Around the implies that bounce point, we can perform a Taylor expansion of (45) up to leading order and thus obtain the following approx- M 2 M 2 φ˙2 H2 ρ ' Pl φ˙2 ' Pl B . (53) imate expression, φ 2t2 /T 2 2 2 B+ H 2e B+ θi(t) ' θi,B− + Mθ,i t − tB− , (46) Moreover, the Friedmann equation requires that ρφ ' 3M 2 H2 in the fast-roll phase, so that T 2 is given by which is a linear function of cosmic time. We ﬁnd that Pl the growth of the anisotropy at linear order only depends 2H2 on the coeﬃcients M . However, at nonlinear order in B+ θ,i T 2 ' . (54) cosmic time, the growth would also depend on the slope " 2 # M (g0 − 1) of the bounce phase characterized by the parameter Υ. Υ2 ln Pl 9βH2 Finally, we have the relations B+

Υ t2 During the fast-roll expansion epoch, one can solve for a ' a e 2 B− , (47) B− B the time derivatives of the anisotropy factors

HB− ' ΥtB− , (48) a3 H(t) θ˙ (t) ' M B , (55) by matching the mean scale factor and the mean Hubble i θ,i a3 H parameter at the beginning of the bouncing phase. B+ B+ from which it follows that the effective energy density of anisotropy evolves as ρ = ρ H2/H2 , and thus θ θ,B+ B+ D. Fast-roll expanding phase its ratio relative to the total energy density does not change (recall that the absolute value of this ratio is also After the bounce, the universe enters the expanding very small because of the decrease of the shear contri- phase, which typically begins with a period of fast-roll bution during the Ekpyrotic phase of contraction dis- expansion for the background during which the universe cussed above). On the other hand, the matter fluid, is still dominated by the scalar field φ. During this stage, having wm = 0, sees its energy density going as ρm = the motion of φ is dominated by its kinetic term while the ρm,B+ H/HB+ , and therefore its contribution to the back- potential is negligible. Thus, the background equation of ground will catch up with that of the scalar field and the state parameter is w ' 1. Correspondingly, the mean anisotropy, bringing the fast roll phase to an end and ini- scale factor evolves as tiating the transition to the usual expansion history of 1/3 Standard Big Bang cosmology. t − t˜ a(t) ' a B+ , (49) The anisotropy factors in the fast roll phase are given B+ ˜ tB+ − tB+ by 3 where t represents the end of the bounce phase and a Mθ,i t − t˜ B+ θ (t) ' θ + B ln B+ , (56) the beginning of the fast-roll period, and a is the value i i,B+ 3 B+ 3a H t − t˜ of the mean scale factor at that moment. Then one can B+ B+ B+ B+ write down the mean Hubble parameter in the fast-roll so they are safely growing only logarithmically during phase this phase. To conclude this section, it is clear that the anisotropy 1 H(t) ' , (50) contribution can easily be made to remain small through- ˜ 3(t − tB+ ) out the entire evolution, i.e. including the matter contraction and the bounce phases. We now turn to the and the continuity of the mean Hubble parameter at t B+ actual constraints that should be imposed on the param- yields eters of our model. 1 t˜ = t − . (51) B+ B+ 3H B+ E. Theoretical constraints Recall that, in Eq. (41), we made use of a Gaussian parametrization of the scalar field evolution in the bounce In this subsection, we study the theoretical constraints phase, with characteristic timescale T . In the fast roll on the scenario of anisotropic nonsingular bounce. It is 8 convenient to introduce an effective e-folding number of we expect this may impose a much more stringent con- the Ekpyrotic phase as follows straint on the anisotropy parameters. For that, one could require that the anisotropy contribution be smaller a H H B− B− B− that the observed level of perturbations, namely Ωpert ∼ NE ≡ ln = (1 − q) ln , (57) −10 aE HE HE 10 . It is interesting to note that increasing the Ekpyrotic e- which characterizes the variation of the mean value of fold number to, say, NE ∼ 15 while keeping q ∼ 0.1, leads the conformal Hubble scale aH during this period. −10 2 to relation (31) to yield Ωθ ∼ 10 ∼ Ωpert Ωpert. Recall that the anisotropy is not allowed to dominate This means that after a sufficiently long Ekpyrotic con- before the onset of the Ekpyrotic phase. This is the contraction, the anisotropy contribution is totally negligible, straint (31). Applying (37) and (57), we can further even compared with the amplitude of primordial pertur- derive the following relation for the anisotropy at the bations. At this particular level however, the anisotropy bounce time: contribution could be comparable to second order in 2 terms of the primordial curvature perturbation expan- X Mθ,i −2(1−3q)N /(1−q) ' e E , (58) sion, and thus could contribute to the primordial non- 6H2 i B− Gaussianities of local shape with fNL ∼ O(1). If one wants to make successful contact with late time which is, as expected, exponentially suppressed by the cosmology, there is a second constraint that should be im- number of e-foldings of Ekpyrotic contraction. Note that plemented on the model, namely that the fast roll phase one can introduce a relative density parameter to de- ends before the time of Big Bang Nucleosynthesis (BBN). scribe the contribution of the anisotropies as follows Roughly speaking, the energy density of regular matter ρ does not change much during the phase of Ekpyrotic con- Ω ≡ θ . (59) θ 3M 2 H2 traction (for small values of q). On the other hand, the Pl density of φ grows rapidly. In the fast roll phase of ex- This parameter must be less than the amplitude of the pansion, the decrease in the density of regular matter is observed anisotropies in the cosmic microwave back- no longer negligible. Hence, the energy density of matter −5 will be much lower at the time t when ρm(t ) = ρφ(t ) ground (CMB), which is of order Ωpert ∼ O(10 ), F F F than at the time t when Ekpyrotic contraction begins. hereby defining Ωpert as the amount of energy density in E perturbations, as given by CMB observations. This leads In fact, it is straightforward to derive that to the following constraint on the number of e-foldings −(1−3q)NE /(1−q) HF . |HE |e , (62) 2 X Mθ,i < Ωpert, (60) showing that the value of H should be much less than 6H2 F i B− |HE |. The Hubble rate HF is associated with the initial temperature T when the expansion begins to follow where we have made use of (55). Provided that this F the Standard Big Bang evolution (it is the equivalent of constraint is satisfied, the current anisotropy will be be- the temperature of reheating in inflationary cosmology). low the current upper bound provided that it was ini- Specifically, the relation is tially small enough not to prevent the onset of Ekpyrotic contraction. Combining this inequality and the approxi- g1/2πT 2 mate relation (58), we can obtain a lower bound on the s F HF ' , (63) e-folding number of the Ekpyrotic phase 9.5MPl 1 − q 1 where gs is the effective partible number for radiation. NE > ln . (61) 2(1 − 3q) Ωpert As a consequence, in analogy with inflationary cosmology, the constraint (62) leads to an upper bound on the Note that the nonsingular bounce model studied in the effective “reheating” temperature: present paper belongs to “fast bounce” category and the 1 absolute values of HB− and HB− are of the same order. ! 2 3MPl |HB− | −(2−3q)N /[2(1−q)] Thus, we can simply consider HB− to be the energy scale T e E (64) F . 1/2 of the bounce. As an example, if we choose q = 0.1 and gs −5 Ωpert = 10 , then the relation (61) yields NE > 7.4. This result implies that the Ekpyrotic phase is rather in our nonsingular bounce model. From the BBN con- efficient at lowering the contribution of anisotropies since straint, we find that the lower limit of the “reheating” a mere few e-folds of Ekpyrotic contraction are enough temperature is of the order O(MeV). If we consider this to damp any reasonable initial anisotropy to negligible lower bound and take gs ∼ 100, NE ∼ 30 and q ∼ 0.1, −17 level. then we find that HB+ > 10 MPl which can easily be

Note that the e-folding number NE can also be con- implemented in the model, as we shall see in the following strained on observable scales of CMB experiments, and numerical calculations. 9

F. Numerical estimates

- 4 To illustrate that a nonsingular bounce can be achieved 1 . 5 x 1 0 5 . 0 0 x 1 0 - 5 2 . 5 0 x 1 0 - 5 - 4 in our model, we numerically solved the background 1 . 0 x 1 0 0 . 0 0

equations of motion. Expressing all relevant functions - 2 . 5 0 x 1 0 - 5 - 5 and parameters in the corresponding units of the reduced 5 . 0 x 1 0 - 5 . 0 0 x 1 0 - 5 - 0 . 0 8 - 0 . 0 4 0 . 0 0 0 . 0 4 0 . 0 8

Planck mass MPl , we set 0 . 0 H

−7 −3 - 5 H V0 = 10 , g0 = 1.1, β = 5, γ = 10 , - 5 . 0 x 1 0 H 1 H bV = 5, bg = 0.5, p = 0.01, q = 0.1 (65) - 1 . 0 x 1 0 - 4 2 H 3 to illustrate the calculations. - 1 . 5 x 1 0 - 4

Moreover, we consider the following parameters of the 4 4 4 4 - 4 . 0 x 1 0 - 2 . 0 x 1 0 0 . 0 2 . 0 x 1 0 4 . 0 x 1 0 matter fluid and the anisotropy t −10 −6 ρm,B = 2.8 × 10 ,Mθ,1 = 2.2 × 10 , −6 −6 Mθ,2 = 3.4 × 10 ,Mθ,3 = −5.6 × 10 , (66) Figure 2: Time evolution of the Hubble parameters H (black line) and Hi (red dashed, blue dotted and magenta dot-dashed and choose as the initial conditions for the scalar field lines for the Hubble expansion rates along the x1, x2, and x3 the following: axes, respectively), in units of the reduced Planck mass MPl , with background parameters given by Eqs. (65) and (66), ˙ −6 φini = −2, φini = 7.8 × 10 . (67) and initial conditions as in (67). The main plot shows that a nonsingular bounce occurs, and that the time scale of the The actual computation also requires the initial value bounce is short (it is a “fast bounce” model). The inner insert of the mean Hubble parameter, which is determined by shows a blowup of the smooth Hubble parameters during the imposing the Hamiltonian constraint equation. Figs. 2 bounce phase: this zoomed-in view of the Hubble parameters and 3 show the evolution of the Hubble parameters and around the bounce point shows that the Hubble rates vanish at different times, so that the scale factors bounce at different “effective” energy densities for matter components and times as well. for the anisotropy, respectively. From Fig. 2, one can see that Hubble parameters along all spatial coordinates evolve smoothly through the t t t bouncing point with an approximate dependence on cos- E B F ρ mic time which is linear. The maximal value of the mean φ ρ Hubble parameter, which we denote as the bounce scale 1 0 - 8 m ρ θ1 HB , is mainly determined by the value of the potential ρ −4 θ2 parameter V0. Specifically, HB is of order O(10 M ) in - 1 0 ρ Pl 1 0 θ3 our numerical result. We also note that the bounces oc- curring in the three spatial directions do not occur at ex- ρ actly the same moment – a consequence of the existence 1 0 - 1 2 of anisotropy. This could leave a smoking gun signature for detecting nonsingular bounce cosmology in high accu- 1 0 - 1 4 racy CMB experiments since the difference in the times of the bounces along various spatial coordinates would 4 4 4 4 - 3 . 0 x 1 0 0 . 0 3 . 0 x 1 0 6 . 0 x 1 0 9 . 0 x 1 0 affect the ultraviolet (UV) modes of primordial pertur- t bations passing through the bouncing phase. We leave this issue for a forthcoming investigation. From Fig. 3, one can easily see that the universe in Figure 3: Time evolution of the “Effective” energy densities of our model experiences four phases, which are matter con- the scalar field ρφ (full black line), the matter fluid ρm (dot- traction, Ekpyrotic contraction, the bounce, and fast-roll dashed green line) and the anisotropy factors ρθi (red dashed, expansion, in turn. At the beginning, the universe is blue dotted and magenta dot-dashed lines), with same initial conditions and background parameters are in Fig. 2. dominated by the matter fluid. At some point (time tE ) during the phase of contraction, the contribution of the scalar field becomes dominant, and the universe enters the Ekpyrotic phase. Note that the effective energy den- bounce, the scalar field φ enters a fast roll phase with sity of anisotropies grows faster than ρm but slower than an effective equation of state equal to unity. As a con- ρφ during the matter contraction phase. Thus, if ρθi sequence, the energy densities ρφ and ρθi dilute at the same rate, and finally the matter fluid catches up with does not dominate over the background before tE , it will never become dominant throughout the whole evolution, the density of φ at the time tF . as already discussed in the previous sections. After the Fig. 4 shows the evolution of the anisotropy factors 10

t t t t t t E B F 1 0 0 E B F

2 . 5 0 x 1 0 - 6 1 0 - 1 Ω 0 . 0 0 φ - 2 . 1 0 Ω θ Ω m Ω - 2 . 5 0 x 1 0 - 6 θ. θ 1 1 0 - 3 θ. 2 θ. - 6 3 - 5 . 0 0 x 1 0 1 0 - 4

4 . 0 x 1 0 - 6 σ 0 . 1 1 σ - 6 2 2 . 0 x 1 0 σ 0 . 0 3 θ 1 0 . 0 θ θ 2 σ - 0 . 1 θ 3 - 2 . 0 x 1 0 - 6

- 0 . 2 - 4 . 0 x 1 0 - 6

- 3 . 0 x 1 0 4 0 . 0 3 . 0 x 1 0 4 6 . 0 x 1 0 4 9 . 0 x 1 0 4 - 3 . 0 x 1 0 4 0 . 0 3 . 0 x 1 0 4 6 . 0 x 1 0 4 9 . 0 x 1 0 4 t t

Figure 4: Time evolution of the anisotropy factors θi (lower Figure 5: Time evolution of the density parameters Ωφ, Ωm, panel) and their time derivatives θ˙i (upper panel), with same and Ωθ (upper panel), and of the shear function σi (lower initial conditions and background parameters are in Fig. 2. panel), with same initial conditions and background parame- The anisotropies increase during the contracting phase but ters are in Fig. 2. rapidly approach constant values in the following expanding phase. Indeed, the parameters chosen do not satisfy the bounds imposed by CMB observations, so the effects of setting θi and their time derivatives. Although the anisotropy non vanishing initial anisotropies can be enlarged so as functions grow during the contraction, they evolve to- to be visible at all. Assuming parameter values taking wards constant values in the expanding epoch. There- into account the experimental constraints would lead to fore, after the time tF , one can rescale all scale factors by figures exactly similar to those obtained for a regular absorbing the asymptotic factors in a redefinition of the isotropic bouncing model. coordinates, and we finally get an isotropic universe. It implies that at the level of homogeneous cosmology the anisotropies do not destabilize our nonsingular bounce IV. CONCLUSION AND DISCUSSION model. This can also be read from the upper panel of ˙ Fig. 4 which shows that θi approach zero after a suffi- We have studied homogeneous but anisotropic cos- ciently long period of expansion. mological solutions of the Galileon type action with an In order to better characterize the anisotropy quanti- Ekpyrotic scalar field potential, as originally introduced tatively, we can define the so-called shear parameters in [38]. The model was constructed to yield non-singular ˙ 2θi isotropic bouncing cosmological solutions. In general, the σi ≡ θie , (68) contracting phase of a bouncing cosmology suffers from and the density parameters a BKL instability against the growth of anisotropies. By ρ studying the solutions of Bianchi-type homogeneous so- Ω ≡ I , (69) I P ρ lutions in which the scale factors in different spatial di- I I mensions are allowed to be independent we have stud- where the subscript “I” represents φ, m and θ, respec- ied the evolution of the anisotropies in our model. We tively. Fig. 5 shows the numerical solution we obtained have shown that the anisotropies are damped in the phase for their time development. The shear functions increase of Ekpyrotic contraction, as was already conjectured on up to their maximal values at the bounce point, after qualitative grounds in [38]. They remain small during the which they rapidly decrease to end up vanishingly small non-singular bounce phase, and are further suppressed when the universe connects with the Standard Big Bang once space begins to expand again. This (counter) ex- evolution. From the evolution of the density parameters, ample shows that a no-go theorem concerning smooth we see that the contribution of the anisotropy only grows bounces cannot be formulated, at least in terms of BKL relative to the dominant density in the phase of matter instabilities. contraction, but it then rapidly decreases in the Ekpy- It would be of interest to study the evolution of lin- rotic phase and in the fast roll phase. earized inhomogeneities in our anisotropic background Note that all numerical calculations shown here are and to see if there are any late time signatures of meant to illustrate the discussion of the previous sections. the initial (pre-Ekpyrotic) contracting phase when the 11 anisotropies might be quite large. This is left for future Chair program. The research of P.P. is supported by work. CNRS and a PICS collaboration grant (#05893). One of us (R.B.) thanks Paul Steinhardt and Bing-Kan Xue for discussions which stimulated us to perform this im- Acknowledgments proved analysis of anisotropies in our bouncing cosmology model. The work of R.B. and Y.C. is supported in part by an NSERC Discovery Grant and by the Canada Research

[1] D. Wands, “Duality invariance of cosmological perturba- variance,” Phys. Rev. D 66, 083503 (2002) [arXiv:hep- tion spectra,” Phys. Rev. D 60, 023507 (1999) [arXiv:gr- ph/0203275]. qc/9809062]. [11] R. Brandenberger, C. Kounnas, H. Partouche, S. Patil [2] F. Finelli and R. Brandenberger, “On the generation of a and N. Toumbas, “Fluctuations in Non-Singular Bounc- scale-invariant spectrum of adiabatic fluctuations in cos- ing Cosmologies from Type II Superstrings”, to be pub- mological models with a contracting phase,” Phys. Rev. lished. D 65, 103522 (2002) [arXiv:hep-th/0112249]. [12] Y. F. Cai, W. Xue, R. Brandenberger and X. Zhang, [3] R. H. Brandenberger, “Alternatives to Cosmological “Non-Gaussianity in a Matter Bounce,” JCAP 0905, 011 Inflation,” arXiv:0902.4731 [hep-th]; (2009) [arXiv:0903.0631 [astro-ph.CO]]. R. H. Brandenberger, “Cosmology of the Very Early [13] P. Peter and N. Pinto-Neto, “Has the Universe always Universe,” AIP Conf. Proc. 1268, 3-70 (2010). expanded?” Phys. Rev. D 65, 103522 (2002). [arXiv:1003.1745 [hep-th]]; [14] P. Peter, N. Pinto-Neto and D. Gonzalez, “Propagation R. H. Brandenberger, “Introduction to Early Uni- of a diabatic and entropy perturbations in a bouncing verse Cosmology,” PoS ICFI2010, 001 (2010) Universe,”, JCAP 12, 003 (2003); [arXiv:1103.2271 [astro-ph.CO]]; J. Martin and P. Peter, “Parametric amplification of met- R. H. Brandenberger, “The Matter Bounce Alternative to ric fluctuations through a bouncing phase,” Phys. Rev. Inflationary Cosmology,” arXiv:1206.4196 [astro-ph.CO]. D 68, 103517 (2003); [4] Y. F. Cai, T. Qiu, R. Brandenberger, Y. S. Piao and J. Martin and P. Peter, “On the causality argument X. Zhang, “On Perturbations of Quintom Bounce,” JCAP in bouncing cosmologies,” Phys. Rev. Lett. 92, 061301 0803, 013 (2008) [arXiv:0711.2187 [hep-th]]; (2004); Y. F. Cai, T. T. Qiu, R. Brandenberger and X. m. Zhang, J. Martin and P. Peter, “Properties of the transition ma- “A Nonsingular Cosmology with a Scale-Invariant Spec- trix in bouncing cosmologies,” Phys. Rev. D 69, 107301 trum of Cosmological Perturbations from Lee-Wick The- (2004); ory,” Phys. Rev. D 80, 023511 (2009) [arXiv:0810.4677 F. T. Falciano, M. Lilley and P. Peter, “A classical [hep-th]]; bounce: constraints and consequences,” Phys. Rev. D X. Gao, Y. Wang, W. Xue and R. Brandenberger, 77, 083513 (2008); “Fluctuations in a Hořava-Lifshitz Bouncing Cosmology,” F. Finelli, P. Peter and N. Pinto-Neto, “Spectra arXiv:0911.3196 [hep-th]; of primordial fluctuations in two-perfect-fluid regular S. Alexander, T. Biswas and R. H. Brandenberger, “On bounces,” Phys. Rev. D 77, 103508 (2008). the Transfer of Adiabatic Fluctuations through a Nonsin- [15] L. R. Abramo and P. Peter, “K-Bounce,” JCAP 09, 001 gular Cosmological Bounce,” arXiv:0707.4679 [hep-th]; (2007) R. Brandenberger, H. Firouzjahi and O. Saremi, “Cos- [16] Y. F. Cai, T. Qiu, Y. S. Piao, M. Li and X. Zhang, mological Perturbations on a Bouncing Brane,” JCAP “Bouncing Universe with Quintom Matter,” JHEP 0710, 0711, 028 (2007) [arXiv:0707.4181 [hep-th]]. 071 (2007) [arXiv:0704.1090 [gr-qc]]; [5] M. Sasaki, “Large Scale Quantum Fluctuations in the In- Y. -F. Cai and X. Zhang, “Evolution of Metric Pertur- flationary Universe,” Prog. Theor. Phys. 76, 1036 (1986). bations in Quintom Bounce model,” JCAP 0906, 003 [6] V. F. Mukhanov, “Quantum Theory of Gauge Invariant (2009) [arXiv:0808.2551 [astro-ph]]. Cosmological Perturbations,” Sov. Phys. JETP 67, 1297 [17] P. Peter and N. Pinto-Neto, “Primordial perturbations in (1988) [Zh. Eksp. Teor. Fiz. 94N7, 1 (1988)]. a non singular bouncing universe model,” Phys. Rev. D [7] V. F. Mukhanov, H. A. Feldman and R. H. Branden- 66, 063509 (2002). berger, “Theory of cosmological perturbations. Part 1. [18] C. Lin, R. H. Brandenberger and L. Levasseur Perreault, Classical perturbations. Part 2. Quantum theory of per- “A Matter Bounce By Means of Ghost Condensation,” turbations. Part 3. Extensions,” Phys. Rept. 215, 203 JCAP 1104, 019 (2011) [arXiv:1007.2654 [hep-th]]. (1992). [19] T. Qiu, J. Evslin, Y. F. Cai, M. Li and X. Zhang, “Bounc- [8] R. H. Brandenberger, “Lectures on the theory of cosmo- ing Galileon Cosmologies,” JCAP 1110, 036 (2011) logical perturbations,” Lect. Notes Phys. 646, 127 (2004) [arXiv:1108.0593 [hep-th]]; [arXiv:hep-th/0306071]. D. A. Easson, I. Sawicki and A. Vikman, “G-Bounce,” [9] P. Peter and J.-P. Uzan, “Primordial Cosmology,” Oxford JCAP 1111, 021 (2011) [arXiv:1109.1047 [hep-th]]. University Press (2009). [20] C. Kounnas, H. Partouche and N. Toumbas, “Thermal [10] R. Durrer and F. Vernizzi, “Adiabatic perturbations in duality and non-singular cosmology in d-dimensional pre big bang models: Matching conditions and scale in- superstrings,” Nucl. Phys. B 855, 280 (2012) 12

[arXiv:1106.0946 [hep-th]]; [34] J. Middleton and J. D. Barrow, “The Stability of an I. Florakis, C. Kounnas, H. Partouche and N. Toumbas, Isotropic Cosmological Singularity in Higher-Order Grav- “Non-singular string cosmology in a 2d Hybrid model,” ity,” Phys. Rev. D 77, 103523 (2008) [arXiv:0801.4090 Nucl. Phys. B 844, 89 (2011) [arXiv:1008.5129 [hep-th]]. [gr-qc]]; [21] J. C. Fabris, R. G. Furtado, P. Peter and N. Pinto-Neto, J. D. Barrow and J. D. Middleton, “Stable isotropic cos- “Regular cosmological solutions in low energy effective mological singularities in quadratic gravity,” Phys. Rev. action from string theories,” Phys. Rev. D 67, 124003 D 75, 123515 (2007) [gr-qc/0702098 [GR-QC]]. (2003). [35] J. Barrow and K. Yamamoto, “Anisotropic Pressures at [22] V. F. Mukhanov and R. H. Brandenberger, “A Nonsin- Ultra-stiff Singularities and the Stability of Cyclic Uni- gular universe,” Phys. Rev. Lett. 68, 1969 (1992). verses,” Phys. Rev. D 82, 063516 (2010) [arXiv:1004.4767 [23] L. R. Abramo, P. Peter and I. Yasuda, “Non singular [gr-qc]]. bounce in modified gravity,” Phys. Rev. D 81, 023511 [36] C. Deffayet, G. Esposito-Farèse, A. Vikman “Co- (2010). variant Galileon,” Phys. Rev. D 79, 084003 (2009) [24] T. Biswas, A. Mazumdar and W. Siegel, “Bouncing uni- [arXiv:0901.1314]; verses in string-inspired gravity,” JCAP 0603, 009 (2006) [37] J. K. Erickson, D. H. Wesley, P. J. Steinhardt and [arXiv:hep-th/0508194]. N. Turok, “Kasner and mixmaster behavior in universes [25] Y. -F. Cai, S. -H. Chen, J. B. Dent, S. Dutta and with equation of state w ≥ 1,” Phys. Rev. D 69, 063514 E. N. Saridakis, “Matter Bounce Cosmology with the (2004) [hep-th/0312009]. f(T) Gravity,” Class. Quant. Grav. 28, 215011 (2011) [38] Y. -F. Cai, D. A. Easson and R. Brandenberger, “Towards [arXiv:1104.4349 [astro-ph.CO]]. a Nonsingular Bouncing Cosmology,” JCAP 1208, 020 [26] R. Brandenberger, “Matter Bounce in Horava- (2012) [arXiv:1206.2382 [hep-th]]. Lifshitz Cosmology,” Phys. Rev. D 80, 043516 (2009) [39] D. H. Lyth, “The primordial curvature perturbation in [arXiv:0904.2835 [hep-th]]; the ekpyrotic universe,” Phys. Lett. B 524, 1 (2002) Y. -F. Cai and E. N. Saridakis, “Non-singular cosmology [arXiv:hep-ph/0106153]; in a model of non-relativistic gravity,” JCAP 0910, 020 D. H. Lyth, “The failure of cosmological perturbation the- (2009) [arXiv:0906.1789 [hep-th]]. ory in the new ekpyrotic scenario,” Phys. Lett. B 526, [27] J. Acacio de Barros, N. Pinto-Neto and M. A. Sagorio- 173 (2002) [arXiv:hep-ph/0110007]. Leal, “The causal interpretation of dust and radiation [40] R. Brandenberger and F. Finelli, “On the spectrum of fluid in non singular quantum cosmologies,” Phys. Lett. fluctuations in an effective field theory of the ekpy- A 241, 229 (1998). rotic universe,” JHEP 0111, 056 (2001) [arXiv:hep- [28] P. Peter, E. Pinho and N. Pinto-Neto, “Tensor perturba- th/0109004]. tions in quantum cosmological background,” JCAP 07, [41] J. c. Hwang, “Cosmological structure problem in the 014 (2005); ekpyrotic scenario,” Phys. Rev. D 65, 063514 (2002) P. Peter, E. Pinho and N. Pinto-Neto, “Gravitational [arXiv:astro-ph/0109045]. wave background in perfect fluid quantum cosmologies,” [42] J. Martin, P. Peter, N. Pinto Neto and D. J. Schwarz, Phys. Rev. D 73, 104017 (2006); “Passing through the bounce in the ekpyrotic models,” P. Peter, E. Pinho and N. Pinto-Neto, “Non inflationary Phys. Rev. D 65, 123513 (2002) [hep-th/0112128]; “Com- model with scale invariant cosmological perturbations,” ment on Density perturbations in the ekpyrotic scenario,” Phys. Rev. D 75, 023516 (2007); Phys. Rev. D 67, 028301 (2003). P. Peter and N. Pinto-Neto, “Cosmology without infla- [43] J. Khoury, B. A. Ovrut, P. J. Steinhardt and N. Turok, tion,” Phys. Rev. D 78, 063506 (2008). “Density perturbations in the ekpyrotic scenario,” Phys. [29] J. Karouby and R. Brandenberger, “A Radiation Bounce Rev. D 66, 046005 (2002) [arXiv:hep-th/0109050]. from the Lee-Wick Construction?,” Phys. Rev. D 82, [44] A. J. Tolley, N. Turok and P. J. Steinhardt, “Cosmological 063532 (2010) [arXiv:1004.4947 [hep-th]]; perturbations in a big crunch / big bang space-time,” J. Karouby, T. Qiu and R. Brandenberger, “On the Insta- Phys. Rev. D 69, 106005 (2004) [hep-th/0306109]. bility of the Lee-Wick Bounce,” Phys. Rev. D 84, 043505 [45] T. J. Battefeld, S. P. Patil and R. Brandenberger, “Non- (2011) [arXiv:1104.3193 [hep-th]]. singular perturbations in a bouncing brane model,” Phys. [30] T. Cailleteau, P. Singh and K. Vandersloot, “Non- Rev. D 70, 066006 (2004) [hep-th/0401010]. singular Ekpyrotic/Cyclic model in Loop Quan- [46] A. Notari and A. Riotto, “Isocurvature perturbations in tum Cosmology,” Phys. Rev. D 80, 124013 (2009) the ekpyrotic universe,” Nucl. Phys. B 644, 371 (2002) [arXiv:0907.5591 [gr-qc]]. [hep-th/0205019]; [31] V. A. Belinsky, I. M. Khalatnikov and E. M. Lifshitz, F. Finelli, “Assisted contraction,” Phys. Lett. B 545, 1 “Oscillatory approach to a singular point in the relativis- (2002) [hep-th/0206112]; tic cosmology,” Adv. Phys. 19, 525 (1970). E. I. Buchbinder, J. Khoury and B. A. Ovrut, “New [32] J. Khoury, B. A. Ovrut, P. J. Steinhardt and N. Turok, Ekpyrotic cosmology,” Phys. Rev. D 76, 123503 (2007) “The ekpyrotic universe: Colliding branes and the origin [hep-th/0702154]; of the hot big bang,” Phys. Rev. D 64, 123522 (2001) P. Creminelli and L. Senatore, “A Smooth bouncing cos- [arXiv:hep-th/0103239]; mology with scale invariant spectrum,” JCAP 0711, 010 J. Khoury, B. A. Ovrut, N. Seiberg, P. J. Steinhardt and (2007) [hep-th/0702165]; N. Turok, “From big crunch to big bang,” Phys. Rev. D J. -L. Lehners, P. McFadden, N. Turok and P. J. Stein- 65, 086007 (2002) [arXiv:hep-th/0108187]. hardt, “Generating ekpyrotic curvature perturbations be- [33] V. Bozza and M. Bruni, “A Solution to the anisotropy fore the big bang,” Phys. Rev. D 76, 103501 (2007) [hep- problem in bouncing cosmologies,” JCAP 0910, 014 th/0702153 [HEP-TH]]. (2009) [arXiv:0909.5611 [hep-th]]. [47] N. Arkani-Hamed, H. -C. Cheng, M. A. Luty and 13

S. Mukohyama, “Ghost condensation and a consistent in- perfect Dark Energy from Kinetic Gravity Braiding,” frared modification of gravity,” JHEP 0405, 074 (2004) JCAP 1010, 026 (2010) [arXiv:1008.0048 [hep-th]]; [hep-th/0312099]. T. Kobayashi, M. Yamaguchi and J. ’i. Yokoyama, “Gen- [48] B. Xue and P. J. Steinhardt, “Evolution of curvature and eralized G-inflation: Inflation with the most general anisotropy near a nonsingular bounce,” arXiv:1106.1416 second-order field equations,” Prog. Theor. Phys. 126, [hep-th]; 511 (2011) [arXiv:1105.5723 [hep-th]]. B. Xue and P. J. Steinhardt, “Unstable growth of curva- [50] A. Nicolis, R. Rattazzi and E. Trincherini, “The Galileon ture perturbation in non-singular bouncing cosmologies,” as a local modification of gravity,” Phys. Rev. D 79, Phys. Rev. Lett. 105, 261301 (2010) [arXiv:1007.2875 064036 (2009) [arXiv:0811.2197 [hep-th]]. [hep-th]]. [49] C. Deffayet, O. Pujolas, I. Sawicki and A. Vikman, “Im-