1: \documentclass[aps,prl,amssymb,showpacs,twocolumn]{revtex4}

2: \usepackage{graphicx}

3: \renewcommand{\baselinestretch}{0.995}

4: 

5: \begin{document}

6: 

7: \preprint{IMSc/2003/10/34}

8: \preprint{AEI--2003--082}

9: \preprint{CGPG--03/10--4}

10: \preprint{gr--qc/0311003}

11: 

12: \title{Quantum suppression of the generic chaotic behavior close to 

13: cosmological singularities}

14: 

15: \author{Martin Bojowald}

16: \email{mabo@aei.mpg.de}

17: \affiliation{Max-Planck-Institut f\"ur Gravitationsphysik, Albert-Einstein-Institut, Am M\"uhlenberg 1, D-14476 Golm, Germany}

18: \affiliation{Center for Gravitational Physics and Geometry, The Pennsylvania

19: State University, 104 Davey Lab, University Park, PA 16802, USA}

20: 

21: \author{Ghanashyam Date}

22: \email{shyam@imsc.res.in}

23: \affiliation{The Institute of Mathematical Sciences,

24: CIT Campus, Chennai-600 113, INDIA}

25: 

26: \pacs{04.60.Pp,98.80.Jk,98.80.Bp}

27: 

28: \newcommand{\lP}{\ell_{\mathrm P}}

29: 

30: \newcommand{\md}{{\mathrm{d}}}

31: \newcommand{\Kern}{\mathop{\mathrm{ker}}}

32: \newcommand{\tr}{\mathop{\mathrm{tr}}}

33: \newcommand{\sgn}{\mathop{\mathrm{sgn}}}

34: 

35: \newcommand*{\R}{{\mathbb R}}

36: \newcommand*{\N}{{\mathbb N}}

37: \newcommand*{\Z}{{\mathbb Z}}

38: \newcommand*{\Q}{{\mathbb Q}}

39: \newcommand*{\C}{{\mathbb C}}

40: 

41: 

42: \begin{abstract}

43: In classical general relativity, the generic approach to the initial

44: singularity is very complicated as exemplified by the chaos of the

45: Bianchi IX model which displays the generic local evolution close

46: to a singularity. Quantum gravity effects can potentially change the

47: behavior and lead to a simpler initial state. This is verified here in

48: the context of loop quantum gravity, using methods of loop quantum

49: cosmology: the chaotic behavior stops once quantum effects become

50: important. This is consistent with the discrete structure of space

51: predicted by loop quantum gravity.

52: \end{abstract}

53: 

54: \maketitle

55: 

56: According to the celebrated singularity theorems of classical general

57: relativity, the backward evolution of an expanding universe leads to a

58: singular state where the classical theory ceases to apply. An

59: extensive analysis of the approach to the singularity, in the general

60: context of inhomogeneous cosmologies, has culminated in the

61: BKL-scenario \cite{BKL}.  According to this scenario, as the

62: singularity is approached, the spatial geometry can be viewed as a

63: collection of small patches each of which evolves essentially

64: independently as a homogeneous model, most generally the Bianchi IX

65: model. This is justified by the observation that interactions between

66: the patches are negligible because time derivatives dominate over

67: space derivatives close to a singularity.

68: 

69: The approach to the singularity of a Bianchi IX model is described by

70: a particle moving in a potential with exponential walls (corresponding

71: to the increasing curvature) bounding a triangle

72: (Fig.~\ref{ClassFig}). During its

73: evolution the particle is reflected at the walls resulting in an

74: infinite number of oscillations (of the scale factors) when the

75: singularity is approached. This classical behavior can be shown to

76: lead to a chaotic evolution by using an analogy with a billiard valid

77: in the asymptotic limit close to the singularity \cite{chaos}.

78: 

79: To appreciate implications of the chaotic approach to the Bianchi IX

80: singularity, observe that at any given time the spatial slice can be

81: decomposed into a collection of almost homogeneous patches, the size

82: of the patches being controlled by the magnitude of space derivatives

83: in the equations of motion. During subsequent evolution when the

84: curvatures grow, these patches have to be subdivided to maintain the

85: homogeneous approximation.  This subdivision is also controlled by

86: evolution of the individual patches.  Since the patches are

87: homogeneous only to a certain approximation, a subdivision of a given

88: patch at a certain time leads to new patches with slightly different

89: initial conditions. The chaotic approach to the Bianchi IX singularity

90: then implies that their geometries will depart rapidly from each

91: other, and the patches have to be fragmented more and more the closer

92: one comes to the singularity. This rapid fragmentation suggests a very

93: complicated and presumably fractal structure of the spatial geometry

94: at the classical singularity.  Note that both of these features, the

95: breaking up of a spatial slice into approximately homogeneous patches

96: and the unending oscillatory approach to the singularity of individual

97: patches, are ultimately consequences of the unbounded growth of the

98: spatial curvature, i.e. the singularity.

99: %

100: \begin{figure}[b]

101: \includegraphics[width=8cm,height=6cm]{class30.eps}

102: \caption{The potential is shown for $V= 30$ in Planck units.

103: Along the $z$-axis is plotted the logarithm of the potential (shifted 

104: so that it is larger than 1 everywhere).  Its form does not change

105: with volume due to the factorized volume dependence.}

106: \label{ClassFig}

107: \end{figure}

108: 

109: However, the classical evolution towards the singularity is expected

110: to break down when the curvatures become too large and to be replaced

111: by quantum dynamics. If we truncate the classical model at a certain

112: lowest volume, technically chaos does not occur. It is a natural

113: question to ask if the qualitative features of the chaotic approach

114: continue to survive quantum modifications. The answer to this

115: question is neither obvious nor independent of concrete theories of

116: quantum gravity and must be explored within the context of a specific

117: candidate theory.

118: 

119: For instance, effective actions with extra terms and additional fields

120: motivated from string or M theories do not lead to a non-chaotic

121: behavior \cite{StringChaos} and thus retain their implications for

122: fragmentation in the context of a general inhomogeneous singularity.

123: On the other hand, loop quantum gravity \cite{Rov:Loops,ThomasRev}

124: predicts a discrete structure of space. For these theories, an

125: unlimited fragmentation of space would be inconsistent with the

126: discrete structure and one would expect a qualitative modification of

127: the chaotic approach.  This translates into a self consistency test

128: for such theories. {\em Thus, a quantum theory of gravity with a

129: discrete structure of space must, for self-consistency, provide a

130: mechanism which prevents the unlimited fragmentation in the expected

131: general approach to a classical singularity.}

132: 

133: In this paper we use methods of loop quantum cosmology

134: \cite{LoopCosRev,HomCosmo,Spin}, a part of loop quantum gravity, to study 

135: this issue. This allows us to obtain explicit, non-perturbative

136: modifications of the classical behavior at small volume which can be

137: analyzed for their implications for chaos. Loop quantum cosmology has

138: already lead to a resolution of conceptual problems such as a

139: non-singular evolution \cite{Sing} and also given a new scenario for

140: inflation \cite{Inflation} which is based on the small-volume

141: modifications. The proof of absence of singularities extends to

142: homogeneous models, in particular the Bianchi IX model

143: \cite{HomCosmo,Spin}.  The investigations of this paper will provide a

144: new consistency check of loop effects and their physical viability.

145: Specifically, quantum modifications for the vacuum Bianchi IX model

146: (central to the issue of a chaotic approach) are presented and shown

147: to prevent the chaotic behavior.

148: 

149: The dynamics of the Bianchi IX model can be formulated on its

150: minisuperspace spanned by the positive scale factors $a_I$,

151: $I=1,\ldots, 3$, related to the diagonal metric components

152: $g_{II}=a_I^2$. It is given by the Hamiltonian constraint 

153: \begin{eqnarray}

154:  H&=& \frac{2}{\kappa}\left[~(\Gamma_J\Gamma_K-\Gamma_I)a_I

155:    -\frac{1}{4} a_I\dot{a}_J\dot{a}_K + \text{cyclic}

156:    \right]\, \label{HH}

157: \end{eqnarray}

158: where $\kappa=8\pi G$ is the gravitational constant and

159: \begin{equation} \label{SpinConn}

160:  \Gamma_I = \frac{1}{2}\left(\frac{a_J}{a_K}+ \frac{a_K}{a_J}-

161:    \frac{a_I^2}{a_Ja_K}\right)

162:  \mbox{ for $\epsilon_{IJK}=1$}

163: \end{equation}

164: are the spin connection components. Thus, the potential term obtained

165: from (\ref{HH}) is given by,

166: \begin{equation}\label{ClassPot}

167: W(a_1, a_2, a_3)  =  \frac{1}{2}\left[ \sum_I a_I^4  -  2

168: (a_1a_2a_3)^2 \sum_I a_I^{-2} \right] 

169: \end{equation}

170: 

171: In order to diagonalize the kinetic term of (\ref{HH}) one can

172: introduce the Misner variables \cite{Mixmaster}: the logarithmic

173: volume $\Omega := -\frac{1}{3} \log V = -\frac{1}{3}\log(a_1a_2a_3)$

174: and the anisotropies $\beta_{\pm}$, introduced via $ a_1 :=

175: e^{-\Omega+\beta_++\sqrt{3}\beta_-} , a_2 :=

176: e^{-\Omega+\beta_+-\sqrt{3}\beta_-} , a_3 := e^{-\Omega-2\beta_+} \,.$

177: In terms of these variables the potential (\ref{ClassPot}) is obtained

178: as,

179: \begin{eqnarray} \label{PotFar}

180: {\cal W}(\Omega, \beta_+, \beta_-) & = & \frac{1}{2}e^{-4\Omega} \left[

181: e^{-8\beta_+} - \right. \nonumber \\

182: & & 4 e^{- 2\beta_+}

183: \text{cosh}( 2 \sqrt{3} \beta_- ) + \nonumber \\

184: & & \left. 2 e^{4\beta_+}\left\{\text{cosh}(4 \sqrt{3} \beta_-) - 1

185: \right\}

186: \right] 

187: \end{eqnarray}

188: Evidently, the $\Omega$-dependence factorizes and one obtains an

189: anisotropy potential ${\cal V}(\beta_+,\beta_-)$ which exhibits

190: exponential walls for large anisotropies as displayed in

191: Fig.~\ref{ClassFig}.  A typical wall can be derived from the potential

192: by setting $\beta_-=0$ and taking $\beta_+$ to be negative, e.g.

193: $W(a_1,a_1,V/a_1^2)=\frac{1}{2}e^{-4\Omega}(e^{-8\beta_+}-

194: 4e^{-2\beta_+})\sim \frac{1}{2}e^{-4\Omega-8\beta_+}$.

195: 

196: The wall picture implies that the universe, described by a particle

197: moving in such a potential, runs through almost free (Kasner) epochs

198: where the potential can be ignored, interrupted by reflections at the

199: walls where the expansion/contraction behavior of different directions

200: changes. The infinite number of these reflections implies that the

201: system behaves chaotically.

202: 

203: One can see that the infinite height of the walls is a consequence of

204: the diverging intrinsic curvature in $\Gamma_I$ for small and large

205: $a_I$. In the classical evolution each $a_I$ will eventually become

206: arbitrarily small as the singularity is approached, e.g.\ $a_1$ and

207: $a_2$ for the typical wall above. However, at a certain stage of the

208: evolution quantum gravity is expected to become important which will

209: lead to a modification of the behavior. Clearly, a necessary condition 

210: for the chaotic behaviour to be prevented by the quantum modification is 

211: that quantum gravity should effectively contain an upper limit on the 

212: curvature such that the walls would have only finite height, changing the 

213: whole scenario.

214: 

215: A quantum theory where such a maximal curvature {\em follows} 

216: is loop quantum cosmology \cite{InvScale,QSDV}. The origin of

217: this upper bound for the curvature lies in the quantization of the

218: relevant quantities and can be compared conceptually to the finite

219: ground state energy of the hydrogen atom obtained after quantization.

220: In the classical equations of motion one can incorporate this

221: feature of loop quantum cosmology by replacing the $\Gamma_I$ with

222: effective coefficients which are derived from the quantization

223: \cite{Spin}.  This leads to the corresponding effective potential.

224: 

225: The new, effective potential is more complicated than the original

226: one.  Nonetheless, for large volume, the effective potential

227: approximates the original one. It is significantly different for small

228: volumes and is responsible for the breakdown of chaos. The volume

229: dependence does not factorize, making the analysis of the classical

230: motion (with volume as internal time) harder. One should also keep in

231: mind that the analysis done here uses only the effective classical

232: description which includes some {\em non-perturbative} quantum effects

233: in the potential. The full quantum evolution is much more complicated

234: and is given by a partial difference equation for the wave function

235: \cite{Spin}.

236: 

237: Using the effective potential in classical equations of motion is

238: sufficient to shed light on the above consistency requirement, even

239: though it is not valid close to the classical singularity.  The

240: underlying dynamical equation of loop quantum cosmology contains a

241: parameter, $j$, that appears as a quantization ambiguity

242: \cite{Ambig}. Its value controls the size of the universe where the

243: maximal curvature is attained. By choosing it to be sufficiently

244: large, one can move the quantum effects in the effective potential

245: into the semiclassical domain. Those large values may not be expected

246: or natural from a physical point of view, but they allow us to study

247: the consistency issue in a simplified setting. A necessary requirement

248: for consistency of the quantum theory then is that the effective

249: classical description stops the Bianchi IX oscillations at some

250: volume.

251: 

252: To be more explicit, one first has to introduce densitized triad

253: variables $|p^I|=V/a_I$, with the volume $V=a_1a_2a_3=

254: \sqrt{|p^1p^2p^3|}$, which become basic operators in loop quantum

255: gravity. For convenience, we have taken them to be dimensionless

256: by setting $\frac{1}{2}\gamma\lP^2=1$ in the notation of

257: \cite{Spin}.  Their inverses $|p^I|^{-1}$ do have well-defined

258: quantizations as finite operators, despite the classical curvature

259: divergence at $p^I=0$ which is eliminated by quantum effects

260: \cite{InvScale}. One can model its effect in the classical behavior by

261: replacing $(p^I)^{-1}$ in the spin connection components by a function

262: $F(p^I/2j)$ to get an effective spin connection which includes

263: non-perturbative quantum effects. The function $F$ is derived

264: directly from a quantization \cite{Ambig}. The ambiguity parameter $j$

265: appears explicitly and controls the peak of $F$. 

266: 

267: The potential $W$ then gets replaced by an effective one, $W_j$, by

268: using the effective instead of the classical spin connection

269: components. One can again obtain the typical wall behavior, this time

270: by evaluating $W_j(p^1,p^1,(V/p^1)^2)$ as a function of $p^1$ at fixed

271: volume.  At these arguments, with $p^1$ large enough but $V/p^1$ not

272: too small, the complicated expression simplifies to just two dominant

273: terms.  As a function of $p^1$ and the volume, the typical wall

274: becomes (see Fig.~\ref{WallProf})

275: %

276: \begin{equation} \label{Wall}

277:  W_j(p^1,p^1,2jq) \approx 

278: \frac{V^4F^2(q)}{32j^4q^2} \{3

279: - 2q F(q)\} ,

280: \end{equation}

281: where we have used the notation $q := \frac{1}{2j}(V/p^1)^2$.

282: 

283: \begin{figure}[ht]

284: \begin{center}

285:  \includegraphics[width=7cm,height=5cm,keepaspectratio]{BianchiIXWallProf.eps}

286: \end{center}

287: \caption{Effective wall (\ref{Wall}) of finite height for $j=1$, $V=10$ as a

288: function of $x=-\beta_+$, compared to the classical exponential wall

289: (upper dashed curve). Also shown is the exact wall

290: $W_j(p^1,p^1,(V/p^1)^2)$ (lower dashed curve), which for $x$

291: smaller than the peak value coincides well with the approximation

292: (\ref{Wall}) up to a small, nearly constant shift.}

293: \label{WallProf}

294: \end{figure}

295: %

296: For values at the peak or larger, $F(q)\sim q^{-1}$, and the classical

297: wall $\frac{1}{2}e^{-4\Omega-8\beta_+}$ is reproduced. The peak of the

298: finite walls is reached for a constant argument $q$ of $F$ which in

299: Misner variables implies that $e^{-2\Omega+2\beta_+}$ is

300: constant. Thus, the wall maxima lie on the line $\beta_+=\Omega+{\rm

301: const}$ in the classical phase space, and the height of the wall drops

302: off as $e^{-12\Omega}\propto V^4$ with decreasing volume. At very

303: small volume, however, the walls collapse even more rapidly, and a

304: numerical analysis shows that the potential becomes negative

305: everywhere at a dimensionless volume of about $(2.172 j)^{3/2}$ in

306: Planck units, i.e.\ just around the elementary discrete volume for the

307: smallest value $j=1/2$.

308: 

309: In Fig.~\ref{ExactFig} are shown three snapshots of the effective

310: potential at decreasing volumes in the vicinity of the isotropy point

311: in the anisotropy plane. The potential at larger volumes clearly

312: exhibits a wall (positive potential) of finite height and finite extent.

313: As the volume is decreased the wall moves {\em inwards} and its height 

314: decreases. Progressively the wall disappears completely making the

315: potential negative everywhere. Eventually, the potential approaches zero 

316: from below. 

317: 

318: This immediately shows that the classical reflections will stop after

319: a finite amount of time, rendering the classical argument about chaos

320: inapplicable. The volume where the transition from the classical

321: behavior to the modified behavior takes place, i.e.\ the first time

322: the universe can ``jump over the wall,'' depends on the initial

323: conditions but it will certainly happen --- the latest at the

324: elementary discrete volume for the smallest value of the ambiguity

325: parameter.

326: 

327: \begin{figure}

328: \begin{center}

329: \includegraphics[height=6cm]{exact150.eps}

330: \includegraphics[width=8cm,height=6cm]{exact100.eps}

331: \includegraphics[width=8cm,height=6cm]{exact30.eps}

332: \end{center}

333: \caption{The modified effective potential is plotted for $j = 5$ at

334: volumes 150, 100, 30 in Planck units for $-1.5 \le \beta_{\pm} \le

335: 1.5$.}

336: \label{ExactFig}

337: \end{figure}

338: %

339: Starting at large volumes, the universe will pass through several phases

340: of Kasner evolution punctuated by reflections. This will continue until

341: the small volume modification of the effective potential comes into

342: play. The stability of a Kasner trajectory in the presence of the modified

343: potential will decide the nature of subsequent deviations after the `last'

344: Kasner epoch. Pleasingly, the Kasner trajectories turn out to be stable

345: in the presence of the effective potential \cite{ChaosLQC}. By contrast, 

346: these are unstable in the presence of the classical potential.

347: 

348: Intuitively, this gives us a natural and consistent picture of the

349: approach to a classical singularity in a quantum theory with a

350: discrete structure of space: At larger volume the system follows the

351: complicated classical evolution with different Kasner phases

352: interrupted by reflections at the walls. As described before, in the

353: context of an inhomogeneous space this leads to a fragmentation into

354: smaller and smaller patches.

355: Once quantum effects are taken into account, the reflections stop just

356: when the volume of a given patch is about the size of a Planck volume,

357: i.e.\ at the scale of the discreteness. Below that scale, a further

358: fragmentation does not take place and the discrete structure is

359: preserved.  We again emphasize that the modifications used here become

360: important at the latest when the Planck scale is reached implying

361: consistency with the expectations from a discrete structure. Below

362: this scale, though, the effective classical description should be

363: superseded by the quantum description. We also emphasize that the

364: existence of a maximal curvature is a {\em consequence} of quantization 

365: and {\em not} put in, in an ad-hoc manner. The results thus constitute a 

366: consistency test for loop quantum gravity.

367: 

368: This also indicates that the results of

369: \cite{Sing,IsoCosmo,HomCosmo,Spin} which prove a non-singular quantum

370: evolution of homogeneous models can be generalized to the full theory,

371: removing also inhomogeneous classical singularities. At the present

372: stage, however, the results for the general case are to be regarded as

373: preliminary and have to be supported by more general techniques

374: directly in inhomogeneous quantum models.

375: 

376: We thank G.~Hossain, R.~Penrose and A.~Rendall for discussions. This

377: work was supported in part by NSF grant PHY00-90091 and the Eberly

378: research funds of Penn State.

379: 

380: \begin{thebibliography}{10}

381: 

382: \bibitem{BKL}

383: V.~A. Belinskii, I.~M. Khalatnikov, and E.~M. Lifschitz, Adv.\ Phys. {\bf 13},

384:   639  (1982).

385: 

386: \bibitem{chaos}

387: T. Damour, H. Henneaux and H. Nicolai, Class. Quantum Grav. {\bf 20}, R145

388: (2003).

389: 

390: \bibitem{Rov:Loops}

391: C. Rovelli, Living Reviews in Relativity {\bf 1}, 1 (1998).

392: 

393: \bibitem{ThomasRev}

394: T. Thiemann, gr-qc/0110034.

395: 

396: \bibitem{StringChaos}

397: T. Damour and M. Henneaux, Phys.\ Rev.\ Lett. {\bf 85},  920  (2000).

398: 

399: \bibitem{LoopCosRev}

400: M. Bojowald and H.~A. Morales-T\'ecotl,  in {\em Proceedings of the Fifth

401:   Mexican School (DGFM): The Early Universe and Observational Cosmology}

402:   (Lect.\ Notes Phys., Springer-Verlag, 2003), gr-qc/0306008.

403: 

404: \bibitem{HomCosmo}

405: M. Bojowald, Class.\ Quantum Grav. {\bf 20},  2595  (2003).

406: 

407: \bibitem{Spin}

408: M. Bojowald, G. Date, and K. Vandersloot, 

409: Class.\ Quantum Grav.\ {\bf 21}, 1253 (2004).

410: 

411: \bibitem{Sing}

412: M. Bojowald, Phys.\ Rev.\ Lett. {\bf 86},  5227  (2001).

413: 

414: \bibitem{Inflation}

415: M. Bojowald, Phys.\ Rev.\ Lett. {\bf 89},  261301  (2002).

416: 

417: \bibitem{Mixmaster}

418: C.~W. Misner, Phys.\ Rev.\ Lett. {\bf 22},  1071  (1969).

419: 

420: \bibitem{InvScale}

421: M. Bojowald, Phys.\ Rev.\ D {\bf 64},  084018  (2001).

422: 

423: \bibitem{QSDV}

424: T. Thiemann, Class.\ Quantum Grav. {\bf 15}, 1281 (1998).

425: 

426: \bibitem{Ambig}

427: M. Bojowald, Class.\ Quantum Grav. {\bf 19},  5113  (2002).

428: 

429: \bibitem{ChaosLQC}

430: M. Bojowald, G. Date, and G.~M. Hossain, The Bianchi IX model in loop

431: quantum cosmology, in preparation.

432: 

433: \bibitem{IsoCosmo}

434: M. Bojowald, Class.\ Quantum Grav. {\bf 19},  2717  (2002).

435: 

436: \end{thebibliography}

437: 

438: \end{document}

439: