Anisotropically Inflating Universes John D. Barrow∗ and Sigbjørn Hervik† ∗DAMTP, Centre for Mathematical Sciences, Cambridge University, Wilbeforce Road, Cambridge CB3 0WA, UK. †Department of Mathematics and Statistics, Dalhousie University, Halifax, Nova Scotia, B3H 3J5 Canada. E-mail: [email protected], [email protected] (Dated: February 7, 2008) We show that in theories of gravity that add quadratic curvature invariants to the Einstein- Hilbertactionthereexistexpandingvacuumcosmologies withpositivecosmological constantwhich donotapproachthedeSitteruniverse. Exactsolutionsarefoundwhichinflateanisotropically. This behaviourisdrivenbytheRiccicurvatureinvariantandhasnocounterpartinthegeneralrelativistic limit. These examples show that the cosmic no-hair theorem does not hold in these higher-order extensions of general relativity and raises new questions about theubiquityof inflation in thevery early universeand the thermodynamicsof gravitational fields. 6 PACSnumbers: 95.30.Sf,98.80.Jk,04.80.Cc,98.80.Bp,98.80.Ft,95.10.Eg 0 0 2 I. INTRODUCTION fromtheRiccicurvaturescalar,R Rµν areaddedtothe µν n Lagrangianofgeneralrelativitythennewtypesofcosmo- a logicalsolutionarisewhenΛ>0 whichhavenocounter- J The inflationary universe is the central cosmological parts in general relativity. They inflate anisotropically 0 paradigm which astronomical observations aim to test, and do not approach the de Sitter spacetime at large 3 and by which we seek to understand how the universe times. We give two new exact solutions for spatially ho- might have evolved from a general initial condition into mogeneous anisotropic universes with Λ > 0 which pos- 3 its present state of large-scale isotropy and homogeneity sess this new behaviour. They provide counter-examples v together with an almost flat spectrum of near-Gaussian 7 to the expectation that a cosmic no-hair theorem will fluctuations. The essential feature of this inflationary 2 continuetoholdinsimplehigher-orderextensionsofgen- picture is a period of accelerated expansion during the 1 eral relativity. Other consequences of such higher-order 1 early stages of the universe [1]. The simplest physically- theories have been studied in [13, 14, 15]. The presence 1 motivated inflationary scenario drives the acceleration of such quadratic terms as classical or quantum correc- 5 by a scalar field with a constant potential, and the lat- tions to the description of the gravitational field of the 0 ter can also be described by adding a positive cosmo- / very early universe will therefore produce very different c logical constant to the Einstein equations. In order to outcomes following expansion from general initial condi- q understand the generality of this scenario it is impor- tions to those usually assumed to arise from inflation. - tant to determine whether universal acceleration and r This adds new considerations to the application of the g asymptotic approach to the de Sitter metric always oc- chaotic and eternal inflationary theories [16] in conjunc- : curs. A series of cosmic no-hair theorems of varying v tion with anthropic selection [17]. i strengths and degrees of applicability have been proved X to demonstrate some necessary and sufficient conditions We will consider a theory of gravity derived from an r for its occurrence [2, 3, 4, 5, 6]. Similar deductions a action quadratic in the scalar curvature and the Ricci are possible for power-law [7, 8] and intermediate in- tensor. More specifically, ignoring the boundary term, flationary behaviour, [9], where accelerated expansion is we will consider the D-dimensional gravitationalaction driven by scalar-fieldpotentials that have slow exponen- tial or power-law fall-offs, but we will confine our dis- cussion to the situation that occurs when there is a pos- itive cosmological constant, Λ > 0. So far, investiga- 1 tions have not revealedany strong reasonto doubt that, SG = dDx |g| R+αR2+βRµνRµν −2Λ . 2κ when Λ > 0 and other matter is gravitationally attrac- ZM p (cid:0) (cid:1)(1) tive, any stable, ever-expanding general-relativistic cos- Variationofthisactionleadstothe followinggeneralised mological model will approach isotropic de Sitter infla- Einstein equations (see, e.g., [18]): tionexponentiallyrapidlywithintheeventhorizonofany geodesically-moving observer. Similar conclusions result whenweconsiderinflationinthosegeneralisationsofgen- eralrelativity inwhichthe Lagrangianis a function only G +Φ +Λg =κT , (2) µν µν µν µν of the scalar curvature, R, of spacetime. This similarity is a consequence of the conformal equivalence between these higher-order theories in vacuum and general rela- tivity inthe presenceofa scalarfield[10, 11, 12]. Inthis where T is the energy-momentum tensor of the ordi- µν paper we will show that when quadratic terms formed nary matter sources, which we in this paper will assume 2 to be zero, and Since integration over the Euler density is a topological invariant, the variation of E will not contribute to the 1 G ≡R − Rg , (3) equations of motion. The Friedmann-Robertson-Walker µν µν µν 2 (FRW)universesareconformallyflatso,forasmallvari- Φµν ≡ ation, the invariant C Cµνρσ will not contribute ei- µνρσ 1 ther. Hence, sufficiently close to a FRW metric only the 2αR R − Rg +(2α+β)(g (cid:3)−∇ ∇ )R µν 4 µν µν µ ν R2 term will contribute. The stability of the FRW uni- (cid:18) (cid:19) verseisthereforedeterminedbythesignof(3α+β)[19]. 1 1 +β(cid:3) Rµν − Rgµν +2β Rµσνρ− gµνRσρ Rσρ, Onecancheckthisexplicitlyusingeq. (2). Westartwith 2 4 (cid:18) (cid:19) (cid:18) (cid:19) the metric ansatz: (4) Λ with (cid:3) ≡ ∇µ∇µ. The tensor Φµν incorporates the de- ds2 =−dt2+e2b(t) dx2+dy2+dz2 , H =r3, viation from regular Einstein gravity, and we see that and note that in 4D th(cid:0)e trace of eq.(2)(cid:1)reduces to α=β =0 implies Φ =0. µν First,consideranEinsteinmetric,sothatR =λg . −R+2(3α+β)(cid:3)R+4Λ=0, (8) µν µν This is a solution of eq.(2) with T =0 provided that µν whichcanbe usedto determine the stability of the Ricci λ scalar. We can perturb the Ricci scalar by assuming a Λ= [(D−4)(Dα+β)λ+(D−2)]. (5) smalldeviationfromtheflatdeSittermetricoftheform: 2 b(t)=Ht+b eλ1t+b eλ2t+O(e2λit), Hence, when D = 4 any Einstein space is a solution to 1 2 eq.(2) provided that Λ = (D −2)λ/2. In particular, if where b and b are arbitrary constants. Eq.(8) implies 1 2 Λ > 0, de Sitter spacetime is a solution to eq.(2). If Λ=0, we need λ=0 and de Sitter spacetime cannot be 3H 2 λ =− 1± 1− , (9) a solution. 1,2 2 s 9H2(3α+β)! Now consider solutions to eq.(2) which are non- perturbative and α and β are not small. We know that if (3α + β) 6= 0. For (3α + β) = 0, we must have solutions with β = 0,α 6= 0 are conformally related to b1 = b2 = 0. From this expression we see that if Einstein gravity with a scalar field φ = ln(1 + 2αR) (3α+β) > 0 then the solution will asymptotically ap- that possesses a self-interaction potential of the form proach the flat de Sitter spacetime as t → ∞; however, V(φ) = (eφ −1)2/4α, [10, 11, 12], and their inflation- for (3α+β) < 0 the solution is unstable. For the spe- ary behaviours for small and large |φ|, along with that cial case of β = 0, this result agrees with the stability of theories derived from actions that are arbitraryfunc- analysis of [19]. A construction of an asymptotic series tions of R, are well understood. However, there is no approximation around the de Sitter metric for the case such conformal equivalence with general relativity when β =0hasalsobeenperformed[20,21,22,23,24]. Inthe β 6=0andcosmologieswithΛ>0canthenexhibitquite caseofgeneralrelativity(α=β =0)anumberofresults different behaviour. for the inhomogeneous case of small perturbations from isotropy and homogeneity when Λ > 0 have also been obtained [2, 3, 4, 5, 25, 26, 27, 28]. II. THE FLAT DE SITTER SOLUTION We see that, as long as (3α+β)>0, any FRW model sufficiently close to the flat de Sitter model will asymp- Firstconsiderthe spatially-flatde Sitter universewith totically approach de Sitter spacetime and consequently metric obeys the cosmological no-hair theorem. We should em- phasize that only FRW perturbation modes have been Λ consideredhere. The questionofwhether the flatde Sit- ds2 =−dt2+e2Ht dx2+dy2+dz2 , H = . dS 3 ter universe is stable againstgeneralanisotropic or large r (cid:0) (cid:1) (6) inhomogeneous perturbations when α 6= 0 and β 6= 0 is Thestabilityofthissolutionintermsofperturbationsof stillunsettled. Inthecaseofuniversesthatarenot’close’ the scale-factor depends on the sign of (3α+β). In 4D, toisotropicandhomogeneousFRWmodelsweshallnow we can use the Weyl invariant and the Euler density, E, show that the cosmic no-hair theorem for Λ>0 vacuum defined by [40], cosmologies is not true: there exist ever-expanding vac- uum universes with Λ > 0 that do not approach the de 1 C Cµνρσ = R Rµνρσ −2R Rµν + R2, Sitter spacetime. µνρσ µνρσ µν 3 E = R Rµνρσ −4R Rµν +R2, (7) µνρσ µν III. EXACT ANISOTROPIC SOLUTIONS to eliminate the quadratic Ricci invariant in the action, since We now present two new classes of exact vac- αR2+βR Rµν = 1(3α+β)R2+ β (C Cµνρσ −E). uum anisotropic and spatially homogeneous universes of µν 3 2 µνρσ 3 Bianchi types II and VI with Λ > 0. These are new cosmological constant drives the late-time evolution to- h exact solutions of the eqns. (2) with (α,β)6=(0,0). wards the de Sitter spacetime. An exact statement of Bianchi type II solutions: the theorem can be found in the original paper by Wald [6]. It requires the matter sources (other than Λ) to obey the strong-energy condition. It has been shown a 2 ds2 =−dt2+e2bt dx+ (zdy−ydz) +ebt(dy2+dz2), that if this condition is relaxed then the cosmic no-hair II 2 h i (10) theorem cannot be proved and counter-examples exist where [7, 32, 33, 34]. In [35], the cosmic no-hair conjecture wasdiscussedforBianchicosmologieswithanaxionfield 11+8Λ(11α+3β) 8Λ(α+3β)+1 a2 = , b2 = . (11) with a Lorentz Chern-Simons term. Interestingly, exact 30β 30β BianchitypeII solutions,similartotheonesfoundhere, These solutions are spacetime homogeneous with a 5- were found which avoided the cosmic no-hair theorem. dimensional isotropy group. They have a one-parameter However, unlike for our solutions, these violations were family of 4-dimensional Lie groups, as well as an iso- drivenby anaxionfield whose energy-momentumtensor latedone(with Lie algebrasAq andA1 ,respectively, violatedthe strong and dominant energy condition. The 4,11 4,9 no-hair theorem for spatially homogeneous solutions of in Patera et al’s scheme [29]) acting transitively on the Einstein gravity also requires the spatial 3-curvature to spacetime. An interesting feature of this family of solu- be non-positive. This condition ensures that universes tions is that there is a lower bound on the cosmological constant, given by Λ = −1/[8(α+3β)] = −a2/8 for do not recollapse before the Λ term dominates the dy- min namics but it also excludes examples like that of the which the spacetime is static. For Λ > Λ the space- min Kantowski-Sachs S2 × S1 universe which has an exact time is inflating and shearing. The inflation does not solution with Λ > 0 which inflates in some directions resultinapproachtoisotropyortoasymptoticevolution but is static in others. These solutions, found by Weber close to the de Sitter metric. Interestingly, even in the [36], were used by Linde and Zelnikov [37] to model a case of a vanishing Λ the universe inflates exponentially higher-dimensional universe in which different numbers butanisotropically. Wealsonotefromthe solutionsthat of dimensions inflate in different patches of the universe. the essential term in the action causing this solution to exist is the βR Rµν-term andthe distinctive behaviour However,itwassubsequentlyshownthatthisbehaviour, µν liketheWebersolution,isunstable[38,39]. Wenotethat occurs when α = 0. The solutions have no well defined our new solutions to gravity theories with β 6=0 possess β →0 limit, and do not have a generalrelativistic coun- anisotropicinflationarybehaviourwithoutrequiringthat terpart. They are non-perturbative. Similar solutions thespatialcurvatureispositiveandaredistinctfromthe exist also in higher dimensions. Their existence seem to Kantowski-Sachsphenomenon. be related to so-called Ricci nilsolitons [30, 31]. The Bianchi type solutions given above inflate in the Bianchi type VI solutions: h presenceofapositivecosmologicalconstantΛ. However, they are neither de Sitter, nor asymptotically de Sitter; ds2 =−dt2+dx2 nor do they have initial singularities. Let us examine VIh +e2(rt+ax) e−2(st+ah˜x)dy2+e+2(st+ah˜x)dz2 ,(12) howthesemodelsevadetheconclusionsofthecosmicno- hair theorem. Specifically, consider the type II solution, h i eq.(10).We define the time-like vector n = ∂/∂t orthog- where onal to the Bianchi type II group orbits, and introduce 8βs2+(3+h˜2)(1+8Λα)+8Λβ(1+h˜2) an orthonormal frame. We define the expansion tensor r2 = , 8βh˜2 θµν = nµ;ν and decompose it into the expansion scalar, θ ≡ θµ and the shear, σ ≡ θ −(1/3)(g +n n ), 8βs2+8Λ(3α+β)+3 µ µν µν µν µ ν a2 = . (13) in the standard way. The Hubble scalar is given by 8βh˜2 H = θ/3. For the type II metric, we find (in the or- and r, s, a, and h˜ are all constants. These are also ho- thonormal frame) mogeneous universes with a 4-dimensional group acting 1 θ =2b, σ = diag(0,2b,−b,−b). transitively on the spacetime. Both the mean Hubble µν 6 expansion rate and the shear are constant. Again, we As a measure ofthe anisotropy,we introduce dimension- see that the solution inflates anisotropically and is sup- less variables by normalizing with the expansion scalar: ported by the existence of β 6= 0. It exists when α = 0 and Λ=0 but not in the limit β →0. 3σ 1 1 1 µν Σ = =diag 0, ,− ,− . µν θ 2 4 4 (cid:18) (cid:19) IV. AVOIDANCE OF THE NO-HAIR THEOREM Interestingly, the expansion-normalised shear compo- nents are constants (and independent of the parameters The no-hair theorem for Einstein gravity states that α, β, and Λ) and this shows that these solutions vio- for Bianchi types I − VIII the presence of a positive late the cosmological no-hair theorem (which requires 4 σ /θ → 0 as t → ∞). To understand how this solu- equations,eq.(8),weconsideraperturbationoftheRicci µν tion avoids the no-hair theorem of, say, ref. [6], rewrite scalar: eq.(2) as follows: R≈4Λ+r eλ1t+r eλ2t. 1 2 G =T , T ≡−Λg −Φ +κT . µν µν µν µν µν µν Using(cid:3)R=−(R¨+θR˙),whichisvalidforspatiallyhomo- In this form the higher-order curvature terms can be geneous universes, eq.(8) again implies, to lowest order: e e interpreted as matter terms contributing a fictitious energy-momentum tensor Tµν. For the Bianchi II so- 3H 2 λ =− 1± 1− , lution we find 1,2 2 s 9H2(3α+β)! e for(3α+β)6=0. Thisshowsthattheperturbationofthe T = 1diag 5b2−a2,−3b2+3a2,−7b2−a2,−7b2−a2 µν 4 Ricciscalargivesthesameeigenmodesfortheanisotropic = diag(ρ,p ,p ,p ). (14) (cid:0) 1 2 3 (cid:1) solutions of type II and VIh as it did for perturbations e of de Sitter spacetime in eq.(9). In order to determine where ρ and p are the energy density and the principal eie e e the stability of other modes, like shear and anisotropic pressures,respectively. Thenohairtheoremsrequirethe curvature modes, further analysis is required. dominantenergycondition(DEC)andthestrongenergy e e condition (SEC) to hold. However, since ρ˜+p˜ +p˜ + 1 2 p˜ = −3b2 < 0 the SEC is always violated when b 6= 0. 3 V. DISCUSSION The DEC is violated when ρ˜ < 0 and the weak energy condition (WEC) is also violated because ρ˜+p˜ = ρ˜+ 2 p˜ = −(a2+b2)/2 <0.These violations also ensure that The solutions that we have found raise new questions 3 about the thermodynamic interpretation of spacetimes. the singularitytheoremswillnotholdforthese universes We are accustomed to attaching an entropy to the geo- and they have no initial or final singularities. metric structure created by the presence of a cosmolog- Are these solutions stable? Due to the complexity of ical constant, for example the event horizon of de Sitter the equations of motion it is difficult to extractinforma- spacetime. Do these anisotropically inflating solutions tion about the stability of these non-perturbative solu- have a thermodynamic interpretation? If they are sta- tions in general. In the class of spatially homogeneous ble they may be related to dissipative structures that cosmologies the dynamical systems approach has been appear in non-equilibrium thermodynamics and which extremely powerful for determining asymptotic states of haveappearedbeenidentified insituationswhere de Sit- Bianchi models. A similar approach can be adopted to ter metrics appear in the presence of stresses which vi- the class of models considered here; however, the com- olate the strong energy condition [7, 32, 33, 34]. They plexity of the phase space increases dramatically due to also provide a new perspective on the physical interpre- the higher-derivative terms. Nonetheless, some stabil- tation of higher-order gravity terms in the gravitational ity results can be easily extracted. Consider, for exam- Lagrangian. ple, a perfect fluid with a barotropic equation of state, In summary: we have found exact cosmological solu- p = wρ, where w is constant. Due to the exponen- tionsofa gravitationaltheorythat generalisesEinstein’s tial expansion, the value of the deceleration parameter is q ≡ −(1 + H˙/H2) = −1 for the type II and VI by the addition of quadratic curvature terms to the ac- h tion. These solutions display the new phenomenon of solutions given. Hence, these vacuum solutions will be anisotropic inflation when Λ>0. They do not approach stable against the introduction of a perfect fluid with the de Sitter spacetime asymptotically and provide ex- w > −1. This includes the important cases of dust amples of new outcomes for inflation that is driven by a (w = 0), radiation (w = 1/3) and inflationary stresses p=−ρstressandbeginsfrom’general’initialconditions. (−1<w <−1/3). For perturbations of the shear and the curvature, the situation is far more complicated. Even within the class of Bianchi models in general relativity a full stability Acknowledgment analysis is lacking. However, in some cases, some of the modescanbeextracted. ConsideragaintheBianchitype SH was supported by a Killam Postdoctoral Fellow- II solution, eq. (10). Using the trace of the evolution ship. [1] A.Guth,Phys.Rev.D23,347(1981);A.D.Linde,Phys. 1983). Lett.B129, 177 (1983). [3] W.BoucherandG.W.Gibbons,InThe Very Early Uni- [2] J.D. Barrow, Perturbations of a De Sitter Universe, In verse, ed. G. Gibbons, S.W. Hawking and S.T.C Siklos, The Very Early Universe, eds. G. Gibbons, S.W. Hawk- (Cambridge UP,Cambridge, p. 273, 1983). ingandS.T.CSiklos,(CambridgeUP,Cambridge,p.267, [4] A.A. Starobinskii, Sov. Phys.JETP Lett. 37, 66 (1983). 5 [5] L.G.JensenandJ.Stein-Schabes,Phys.Rev.D35,1146 [22] S. Gottl¨ober, V. Mu¨ller and A.A. Starobinskii, Phys. (1987). Rev. D 43, 2510 (1991). [6] R.Wald, Phys. Rev.D 28, 2118 (1983). [23] A.A. Starobinskii and H.-J. Schmidt, Class. Quantum [7] J.D. Barrow, Phys.Lett. B, 187, 12 (1987). Grav. 4, 695 (1987). [8] F. Lucchin and S. Mataresse, Phys. Rev. D 32, 1316 [24] H.-J. Schmidt,Class. Quantum Grav. 5, 233 (1988). (1985); L.F. Abbott and M.B. Wise, Nucl. Phys. B 244, [25] J.D.BarrowandG.G¨otz,Phys.Lett.B231,228(1989). 541 (1984). [26] A.D. Rendall, Annales HenriPoincar´e A5,1041 (2004) [9] J.D.Barrow, Phys.Lett.B, 235, 40 (1990); J.D. Barrow [27] V. Faraoni, Phys.Rev. D72 061501 (2005). andP.Saich,Phys.LettB249, 406 (1990); J.D. Barrow [28] V.FaraoniandS.Nadeau,Phys.Rev.D72124005(2005). and A.R. Liddle, Phys. Rev. D 47, R5219 (1993); A.D. [29] J.Patera,R.T.Sharp,P.WinternitzandH.Zassenhaus, Rendall, Class. Quant.Grav. 22,1655 (2005). J. Math. Phys. 17, 986 (1976). [10] J.D. Barrow, Nucl.Phys. B 296, 697 (1988). [30] J. Lauret, Math. Ann. 319, 715 (2001), Quart. J. Math. [11] J.D. Barrow and S. Cotsakis, Phys. Lett. B. 214, 515 52, 463 (2001), Math. Z. 241, 83 (2002), Diff. Geom. (1988). Appl. 18, 177 (2003). [12] K.I.Maeda, Phys. Rev.D 39, 3159 (1989). [31] S. Hervik, talk held at the CMS Summer 2004 meeting, [13] A.Berkin, Phys. Rev.D 44, 1020 (1991). Halifax, NS, Canada. [14] E. Bruning, D. Coule, C. Xu, Gen. Rel. Grav. 26, 1197 [32] J.D. Barrow, Phys. Lett. B 180, 335 (1986). (1994). [33] J.D. Barrow, Nucl. Phys. B 310, 743 (1988). [15] H.-J. Schmidt,gr-qc/0407095. [34] J.D. Barrow, Phys. Lett. B 183, 285 (1987). [16] A.D. Linde, Phys. Lett. B 175, 395 (1986); A. Vilenkin, [35] N. Kaloper, Phys.Rev.D 44, 2380 (1991). Phys.Rev.D 27, 2848 (1983). [36] E. Weber,J. Math. Phys.25, 3279 (1984). [17] J.D.BarrowandF.J.Tipler,TheAnthropicCosmological [37] A.D. Linde and M.I. Zelnikov, Phys. Lett. B 215, 59 Principle, (Oxford UP,Oxford, 1986). (1988). [18] S.Deser and B. Tekin,Phys. Rev. D67, 084009 (2003). [38] J.D. Barrow and J. Yearsley, Class. Quantum Grav. 13, [19] J.D. Barrow and A.C. Ottewill, J. Phys. A 16, 2757 2965 (1996). (1983). [39] H.H. Soleng, Class. Quantum Grav. 6, 1387 (1989). [20] V. Mu¨ller, H.-J. Schmidt and A.A. Starobinskii, Phys. [40] These expressions are only valid in 4D. Lett.B 202, 198 (1988). [21] A.BerkinandK.I.Maeda,Phys.Rev.D44,1691(1991).