1: \documentclass[12pt]{article}

2: %\usepackage{amsmath}

3: %\usepackage{amsfonts}

4: \usepackage{amssymb}

5: \usepackage{cite}

6: \setlength{\topmargin}{-.3in} \setlength{\oddsidemargin}{.0in}

7: %\setlength{\evensidemargin}{-.21in}

8: \setlength{\textheight}{8.5in} \setlength{\textwidth}{6.35in}

9: \setlength{\footnotesep}{\baselinestretch\baselineskip}

10: \newlength{\abstractwidth}

11: \setlength{\abstractwidth}{\textwidth}

12: \addtolength{\abstractwidth}{-6pc}

13: \usepackage[dvips]{graphicx}

14: \usepackage{subfigure}

15: 

16: \flushbottom \thispagestyle{empty} \pagestyle{plain}

17: 

18: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

19: %%%%%%%%%%%%%%%%%%%%%%

20: \renewcommand{\thefootnote}{\fnsymbol{footnote}}

21: \renewcommand{\thanks}[1]{\footnote{#1}} % Use this for footnotes

22:  \newcommand{\starttext}{

23: \setcounter{footnote}{0}

24: \renewcommand{\thefootnote}{\arabic{footnote}}}

25: \renewcommand{\theequation}{\thesection.\arabic{equation}}

26: \newcommand{\be}{\begin{equation}}

27: \newcommand{\bea}{\begin{eqnarray}}

28: \newcommand{\eea}{\end{eqnarray}}

29: \newcommand{\beq}{\begin{equation}}

30: \newcommand{\ee}{\end{equation}}

31: \newcommand{\eeq}{\end{equation}}

32: \newcommand{\N}{{\cal N}}

33: \newcommand{\<}{\langle}

34: \newcommand{\oc}{{\cal O}_C}

35: \renewcommand{\a}{\alpha}

36: \renewcommand{\b}{\beta}

37: \newcommand{\half}{{1\over 2}}

38: \newcommand{\comment}[1]{}

39: \renewcommand{\>}{\rangle}

40: \def\ba{\begin{eqnarray}}

41: \def\ea{\end{eqnarray}}

42: \newcommand{\PSbox}[3]{\mbox{\rule{0in}{#3}\special{psfile=#1}\hspace{#2}}}

43: 

44: \def\corrxy{\langle \phi(x) \phi(y) \rangle}

45: \def\corrf{\langle \phi(\tau_1) \phi(\tau_2) \rangle}

46: \def\t0{$t=0$}

47: \def\14{{1\over4}}

48: \def\12{{1 \over 2}}

49: \def\eq{&=&}

50: \def\tro{\tilde{\rho}}

51: \def\d{\partial}

52: \def\dt{\partial_{\tau}}

53: \def\ds{\partial_{\sigma}}

54: \def\h3{h^{3\over 2}}

55: \def\R{\bar{R}}

56: \def\qft{quantum field theory}

57: \def\>{\rangle}

58: \def\<{\langle}

59: \def\sc {Schwarzschild}

60: \def\ls{\sqrt{\alpha'}}

61: \def\des{de Sitter Space}

62: \def\f{\Phi}

63: \def\cc{cosmological constant}

64: \def\st{string theory}

65: \def\sms{supermoduli--space}

66: \def\lmb{\lambda}

67: \def\lo{\lambda_0}

68: \def\sb{supersymmetry breaking}

69: \def\sbs{supersymmetry breaking scale}

70: \def\bdg{Banks Dine Gorbatov}

71: \def\0cc{$\Lambda = 0$}

72: \def\pcc{$\Lambda > 0$}

73: \def\sds{Schwarzschild-de Sitter}

74: \def\rd{\dot r}

75: \def\Om{   \Omega_{D-1}^2   }

76: \def\om{   \Omega_{D-2}^2   }

77: \def\pt{\tilde \phi}

78: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

79: %%%%%%%%%%

80: 

81: \begin{document}

82: \renewcommand{\theequation}{\thesection.\arabic{equation}}

83: \begin{titlepage}

84: \bigskip

85: \rightline{UCB-PTH-05/33} \rightline{LBNL-58948} 

86: \rightline{hep-th/0511084}

87: 

88: \bigskip\bigskip\bigskip\bigskip

89: 

90: \centerline{\Large \bf {Asymptotic states of the bounce geometry}}

91: 

92: 

93: \bigskip\bigskip

94: \bigskip\bigskip

95: 

96: \centerline{\bf Raphael Bousso and Ben Freivogel}

97: \medskip

98: \centerline{\small Department of Physics and Center for Theoretical

99: Physics} 

100: \centerline{\small 

101: University of California, Berkeley, California 94720, U.S.A.}

102: \medskip

103: \centerline{\small Lawrence Berkeley National Laboratory, 

104: Berkeley, California 94720, U.S.A.}

105: \medskip

106: \medskip

107: 

108: \bigskip\bigskip \begin{abstract}

109: %

110: We consider the question of asymptotic observables in cosmology.  We

111: assume that string theory contains a landscape of vacua, and that

112: metastable de~Sitter regions can decay to zero cosmological constant

113: by bubble nucleation.  The asymptotic properties of the corresponding

114: bounce solution should be incorporated in a nonperturbative quantum

115: theory of cosmology.  A recent proposal for such a framework defines

116: an S-matrix between the past and future boundaries of the bounce.  We

117: analyze in detail the properties of asymptotic states in this

118: proposal, finding that generic small perturbations of the initial

119: state cause a global crunch. We conclude that late-time amplitudes

120: should be computed directly. This would require a string theory

121: analogue of the no-boundary proposal.

122: %

123: \end{abstract}

124: 

125: \end{titlepage}

126: \starttext \baselineskip=18pt \setcounter{footnote}{0}

127: 

128: 

129: 

130: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

131: %%%%%%

132: %%%%%%%%%%%%%%%%

133: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

134: %%%%%%

135: %%%%%%%%%%%%%%%%

136: 

137: 

138: \setcounter{equation}{0}

139: 

140: \section{Introduction}

141: 

142: The observed acceleration of the expansion of the universe has

143: modified the cosmological constant problem.  In Ref.~\cite{BP} it was

144: argued that string theory contains an exponentially large number of

145: long-lived false vacua, all of which are realized dynamically as huge

146: regions in an infinite universe, and some of which have an effective

147: cosmological constant of order the observed value.  Recent progress on

148: stabilizing string compactifications%

149: ~\cite{KKLT,DasRaj99,GidKac01,DenDou05,DenDou04,ConQue05,%

150:   BerMay05,BalBer05,SalSil04,SalSil204,MalSil02,Ach02,CarLuk04,%

151:   CurKra05,GurLuk04,BecBec03,BecCur04,BruAlw04,GukKac03,%

152:   BucOvr03,DewGir05} %

153: lends support to this scenario and provides increasingly concrete

154: explorations of the stringy ``landscape''~\cite{Sus03}.

155: 

156: Another, more conceptual problem is the question of observables in

157: cosmology.  A number of authors have pointed out that the definition

158: of exact observables is problematic in cosmology (see, e.g.,

159: Refs.~\cite{BanFis01a,Wit01,HelKal01,KraOog03,Ban04,Bou05,FreHub05})

160: because of obstructions such as event horizons, thermal radiation, and

161: entropy bounds.  In particular, in Ref.~\cite{Bou05} it was argued

162: that an S-matrix cannot be measured by any real observer inside a

163: large class of universes.  However, it was also noted that event

164: horizons do not automatically preclude the existence of other exact

165: observables.  In any case, if exact observables can be defined at all,

166: they will be defined at asymptotically late time.

167: 

168: There is no reason why a rigorous quantum gravity theory with

169: well-defined observables should exist for all imaginable cosmologies;

170: indeed, it would be satisfying if its existence were to provide a

171: criterion constraining the type of universe we find ourselves

172: in. Assuming that the landscape picture is

173: correct,\footnote{Disclaimers should include the fact that corrections

174:   to metastable de~Sitter solutions are not yet controlled with the

175:   same level of rigor as in supersymmetric backgrounds.  Because of

176:   our potential ignorance of vast sectors of string theory, current

177:   surveys of vacua may not be representative.  For further criticism,

178:   see, e.g., Ref.~\cite{BanDin03}.} it makes sense to search for a

179: nonperturbative theoretical framework that rigorously defines its

180: dynamics and observables.

181: 

182: A first step in this direction was taken in Ref.~\cite{FreSus04},

183: which noted that the asymptotic structure of supersymmetric regions

184: with vanishing cosmological constant may allow for constructions of

185: observables that would be impossible in an eternal de~Sitter universe.

186: (Regions with negative vacuum energy do not share this property

187: because they collapse when accessed cosmologically.)  The

188: corresponding semiclassical solution is the Coleman-De~Luccia

189: ``bounce''.  It consists of a bubble of a $\Lambda = 0$ vacuum inside

190: de~Sitter space. The bubble of true vacuum expands but the de~Sitter

191: region inflates fast enough that it is not consumed. Asymptotically at

192: early and late times, the geometry has infinite regions of true and

193: false vacuum.  It was suggested that this system can be described

194: quantum mechanically in terms of an S-matrix relating asymptotic

195: states in the early and late time $\Lambda =0$ regions.

196: 

197: In the present paper we present results that suggest a partial

198: modification of this strategy.  In Sec.~\ref{sec-bound} we identify an

199: entropy bound limiting the number of states in the semiclassical

200: geometry expected to dominate the path integral.  This is concrete

201: evidence that the proposed S-matrix would have finite rank, as first

202: suggested by Banks~\cite{Ban04}.  We point out that the past half of

203: the Coleman-De~Luccia evolution violates the second law of

204: thermodynamics.  We find that this makes it difficult to characterize

205: the asymptotic states relevant to the proposed S-matrix.

206: 

207: In Sec.~\ref{sec-instability} we analyze how the entropy bound is

208: reconciled at the semiclassical level with the infinite number of

209: particles present in the universe at early and late times.  Here we

210: find that the picture is more complex than previously

211: noted~\cite{Ban04}.  By construction, the time-symmetric

212: Coleman-De~Luccia solution involves an era in which the coarse grained

213: entropy decreases.  This leads to violent instabilities.  Even a small

214: perturbation in the initial state will generically produce a big

215: crunch rather than a future asymptotic region.  In this respect the

216: Coleman-De~Luccia geometry considered here is not better, but worse

217: than eternal de~Sitter space.

218: 

219: In Sec.~\ref{sec-halo} we characterize ``allowed'' states that

220: correspond to a semiclassical evolution from a past to a future

221: asymptotic region.  Many such states do not appear to admit a simple

222: description near the boundary at all.  Those that do are best thought

223: of as excitations of zero mass in a cosmological fluid.  The task is

224: to leave all fields outside a compact region undisturbed, in order to

225: limit the crunch to a finite black hole.  We achieve this by treating

226: all matter as dust and creating an overdense region surrounded by an

227: underdense region (a ``halo'').

228: 

229: Working with such states is awkward, and the relation between their

230: S-matrix elements and bulk observables accessible to physical

231: observers is obscure.  In light of the inhomogeneity of the allowed

232: perturbations we found, the corresponding S-matrix is unlikely to

233: describe the perturbations that are actually observed in the universe

234: at late times.  Unless the dust approximation is exact, even halo

235: states involve infinitely fine-tuned small adjustments in the entire,

236: infinite initial $\Lambda=0$ region.  Moreover, the choice of initial

237: state has no natural interpretation and may be redundant.  Note that

238: all of these difficulties relate to the {\em past\/} asymptotic

239: region.

240: 

241: Yet, in light of the difficulty in defining observables in cosmology,

242: we would like to exploit the more advantageous asymptotic properties

243: of the Coleman-De~Luccia solution.  They are associated with the {\em

244:   future\/} asymptotic region.  It is particularly promising that the

245: solution contains an open (rather than closed or flat) FRW universe,

246: since this cuts off the growth of perturbations after curvature

247: domination. Arbitrarily large regions of arbitrarily flat space become

248: available at late times.  In particular, there will be well-separated,

249: noninteracting particles. Just as string theory computes the S-matrix

250: in flat space, it may well compute amplitudes for out-states in the

251: Coleman-De~Luccia geometry.

252: 

253: Thus, our results suggest that the second-law-violating, contracting

254: half of the Coleman-De~Luccia universe does more to obstruct than to

255: facilitate a nonperturbative description of the landscape and its

256: dynamics.  They point instead to a framework in which amplitudes are

257: computed for boundary conditions specified {\em at late times only}.

258: An example of this in semiclassical gravity is the Hartle-Hawking

259: wavefunction~\cite{HarHaw83}, defined by the saddlepoint approximation

260: to a path integral over compact geometries.  In the context of string

261: theory, however, one would seek an exact nonperturbative definition of

262: amplitudes in the asymptotic future.

263: 

264: 

265: \section{Metric and causal structure}

266: \label{sec-metric}

267: 

268: In this section we review the fully extended Coleman-De~Luccia

269: solution~\cite{ColDel80}, or ``bounce''.  We present the metric and

270: conformal diagram.  More details can be found, e.g., in

271: Refs.~\cite{Ban02,FreSus04,Bou05}. We work in four dimensions for

272: simplicity, but our analysis is independent of dimension aside from

273: trivial factors.

274: 

275: The bounce is a solution to the Einstein equations in the presence of

276: a scalar field potential with at least one false vacuum, as shown in

277: Fig.~\ref{fig-pot}.  We assume that the cosmological constant vanishes

278: exactly in the true vacuum.  The simplest classical solutions are

279: Minkowski space (where the scalar is everywhere in the true vacuum)

280: and de~Sitter space (with the scalar in the false vacuum).  Both of

281: these have maximal symmetry, generated respectively by the Poincar\'e

282: group and the de~Sitter group, $SO(1,4)$.  The de~Sitter curvature

283: radius $R$ is inversely related to the cosmological constant $\Lambda$

284: in the false vacuum, $R= \sqrt{3/\Lambda}$.

285: %

286: \begin{figure}[!htb]

287: \center

288: \includegraphics [scale=.5, clip]

289: {potential} \caption{A potential which has a metastable

290:   minimum with positive energy and a stable minimum with

291:   zero energy.}

292: \label{fig-pot}

293: \end{figure}

294: 

295: But there is also a field configuration in which the scalar crosses

296: the barrier from true to false vacuum, forming a domain wall.  The

297: bounce is the most symmetric solution of the Einstein equation with

298: this field configuration.  Its isometry group is $SO(1,3)$. For

299: simplicity, we will describe the solution in the thin wall limit where

300: the width of the bubble is negligible compared to its minimum radius.

301: (See Ref.~\cite{ColDel80} for the conditions on the potential under

302: which this limit is valid.)

303: 

304: Inside the domain wall, the spacetime is approximately empty, so the

305: geometry is simply a region of Minkowski space glued across the thin

306: domain wall to a portion of de Sitter space.  This approximation

307: actually requires infinite fine tuning, so it will important to

308: include the effects of matter present in the $\Lambda=0$ region.

309: However, it is instructive to describe first the geometry of the

310: idealized, vacuous bounce.  It will then not be difficult to go beyond

311: this excessive simplification.

312: 

313: The domain wall is a hyperboloid, whose worldvolume describes a

314: three-dimensional de Sitter spacetime.  Its minimum area is $4\pi

315: r_0^2$, where the parameter $r_0$ is determined by the tension

316: $\sigma$ of the domain wall and the ($4$-dimensional) de Sitter radius

317: $R$, 

318: \begin{equation} 

319: r_0 = {8 \pi G \sigma R^2 \over (4 \pi G

320:   \sigma R)^2 + 1 }~.  

321: \end{equation}

322: 

323: In the flat spacetime enclosed by the hyperboloid the position of the

324: domain wall satisfies

325: %

326: \be r^2 - t^2 = r_0^2~, \ee

327: %

328: where $r$ and $t$ are the standard Minkowski coordinates.  The wall

329: expands from a minimum radius $r_0$, at $t=0$, to infinite size as

330: $t\to\pm\infty$.  To describe the motion of the domain wall from the

331: de Sitter side, it is simplest to embed de Sitter space in

332: $5$-dimensional Minkowski space as the surface

333: %

334: \be X_\mu X^\mu = -R^2, ~~~~~\mu = 0,1,...,4~. \ee

335: %

336: The domain wall is the intersection of the de Sitter space with

337: the plane

338: %

339: \be X_4 = (R^2 - r_0^2)^{1/2}~. \ee

340: %

341: The part of the space farther from the origin than the plane is

342: discarded and replaced by a piece of flat space. An embedding

343: picture of the solution is shown in Fig.~\ref{embd}.

344: %

345: \begin{figure}[!htb]

346: \center

347: \includegraphics [scale=.5]

348: {bounceEmbedding.eps} \caption{The Coleman-De~Luccia bounce geometry,

349:   as an embedding in Minkowski space.  The flat piece has zero cosmological

350:   constant, while the curved piece has positive cosmological constant.

351:   A domain wall separates the two regions.} \label{embd}

352: \end{figure}

353: 

354: The solution is invariant under the symmetry group of the hyperboloid,

355: $SO(1, 3)$. On the de Sitter side, the full $SO(1,4)$ de Sitter

356: symmetry is broken down to those boosts and rotations which leave

357: $X_4$ unchanged. On the Minkowski side, the symmetry arises because

358: the domain wall picks out an origin, breaking translation invariance

359: while preserving the Lorentz group. The orbits of the symmetry group

360: are surfaces of constant invariant distance from the origin. These

361: surfaces can be spacelike separated from the origin, like the domain

362: wall (region I of Fig.~\ref{fig-benpen}). They can also be timelike

363: separated from the origin (region II of Fig.~\ref{fig-benpen}).

364: %

365: \begin{figure}[!!htb] \center \includegraphics [scale=.8, clip]

366:       {bouncepenrose} \caption{Conformal diagram of the

367:         Coleman-De~Luccia solution, including orbits of the symmetry

368:         group $SO(1,3)$. The thick orbit is the domain wall.  The

369:         radius of spheres goes to zero at the left and right

370:         boundaries and to infinity on the remaining boundaries.}

371:       \label{fig-benpen} \end{figure}

372: %

373: The surviving $SO(1, 3)$ symmetry is not an artifact of the vacuous

374: bounce.  It is present in the full solution (even beyond the thin wall

375: approximation). 

376: 

377: If the thin wall approximation is well satisfied, the energy density

378: in the center of the true vacuum bubble (point $b$ in

379: Fig.~\ref{fig-benpen}) will almost vanish.  But it will not be exactly

380: zero, because the distance of $b$ to the domain wall is finite, so

381: that $b$ lies on the exponential tail along which $\phi$ attempts to

382: approach the true vacuum.  By continuity, the same nonzero energy

383: density will be present on the infinitesimally later point $c$, and

384: hence {\em on the entire symmetry orbit of c}.  The presence of a

385: constant energy density on a hyperbolic spacelike slice means that

386: region II (III) is not flat space, but an expanding (contracting) open

387: FRW cosmology~\cite{ColDel80}, with metric 

388: %

389: \begin{equation} 

390: ds^2 = -d\tau^2 + a^2(\tau) [d\rho^2 + \sinh^2 \rho d\om].  

391: \label{eq-iop}

392: \end{equation} 

393: %

394: Thus, the idealization as flat spacetime is not even approximately

395: correct in regions II and III, in the sense that the total energy in

396: those regions is infinite.

397: 

398: As the field $\phi$ completes its descent into the true vacuum, it

399: will need to dissipate this extra energy.  For simplicity, we think of

400: the conversion of potential energy to particles (``reheating'') as

401: occurring on a definite time slice of the FRW geometry, the reheating

402: surface.  By coupling $\phi$ to other fields, particles are produced

403: at the reheating surface. The details can vary, but it is inevitable

404: that somehow the initially uniform potential energy will evolve into

405: an incoherent, highly entropic form as the universe expands. Because

406: the symmetry orbits are noncompact hyperboloids, an infinite number of

407: particles will be produced.

408: 

409: 

410: 

411: \section{Entropy}

412: \label{sec-bound}

413: 

414: The bounce solution has an infinite amount of matter, and hence, an

415: infinite amount of coarse-grained entropy in the FRW regions.  In this

416: section, we apply the covariant entropy bound~\cite{CEB1,CEB2} to show

417: that the total number of allowed microstates is nevertheless finite.

418: 

419: A light-sheet is a convergent null hypersurface orthogonal to an

420: arbitrary two-dimensional spatial surface with area $A$.  The bound

421: says that there can be at most $e^{A/4G}$ different quantum states on

422: a light-sheet, i.e., the entropy will not exceed $A/4G$ (see

423: Ref.~\cite{RMP} for a detailed review).

424: %

425: \begin{figure}[!htb] \center \includegraphics [scale=.6] {cdlbp3.eps}

426: \caption{A wedged conformal diagram~\cite{CEB2} for the bounce

427:   geometry.  Normal ($>$, $<$), trapped ($\vee$), and anti-trapped

428:   ($\wedge$) regions are separated by apparent horizons (thick lines).

429:   Their shape is determined by matter sources (the domain wall, and

430:   the matter created by it in regions II and III).  We treat the

431:   domain wall (dotted line) as a delta function source.  $L_1$ and

432:   $L_2$ are light-sheets of the sphere $P$.  The tightest entropy

433:   bound on the spacetime is obtained from $L$, a single light-sheet

434:   whose maximal area is the de~Sitter horizon in the top right

435:   corner.}  \label{fig-cdlbp} \end{figure}

436: 

437: In a spherically symmetric spacetime, each sphere can be classified

438: depending on the contracting null directions, i.e., the directions in

439: which light-sheets exist.  The corresponding domains are indicated in

440: Fig.~\ref{fig-cdlbp}: spheres are either trapped ($\vee$), antitrapped

441: ($\wedge$), or normal ($>$ and $<$, with the open side pointing

442: towards $r=0$).

443: 

444: As is common for a cosmological solution, the bounce geometry contains

445: light-cones of finite maximal area.  By following the wedges back to

446: the tips, one is led to an apparent horizon, an area maximal with

447: respect to orthogonal light-rays.  Such a surface, for example the

448: sphere $P$ in Fig.~\ref{fig-cdlbp}, admits light-sheets in two

449: opposing directions.  These two light-sheets, $L_1$ and $L_2$, form a

450: complete Cauchy surface, in the sense that every timelike curve must

451: pass through them.  Hence the total number of states in the universe

452: is bounded by $e^{A_P/2G}$.

453: 

454: Note that $P$ is a sphere on the worldvolume of the domain wall.  Its

455: area increases monotonically and without bound as $P$ is moved down.

456: This explains why we can have an infinite coarse-grained entropy at

457: early times in the FRW region.  We get the best possible bound by

458: moving $P$ all the way up until the light-sheet becomes $L$.  This is

459: a {\em single\/} light-sheet starting at the late-time cosmological

460: horizon of the de~Sitter region (the top right corner).  Hence the

461: optimal bound on the total number of states is given by

462: \begin{equation}

463: N \leq \exp\left(\frac{\pi R^2}{G} \right)~,

464: \label{eq-bound}

465: \end{equation}

466: where $R$ is the de~Sitter radius.  (One might worry about what

467: happened to the second light-sheet: is this estimate not off by a

468: factor of 2?  There is indeed a de~Sitter horizon volume in the top

469: right corner which is not captured by $L$.  However, the

470: D-bound~\cite{Bou00b} implies that no matter entropy is present there.)

471: 

472: Thus there is a huge discrepancy: Unbounded coarse-grained entropy is

473: allowed, and indeed present, at early and late times in the FRW

474: regions II and III; yet, the total number of microstates allowed to

475: pass though the light-sheet $L$ in the center of the geometry is

476: finite.  It follows that of the infinite number of microstates

477: corresponding to the initial macroscopic configuration, all but a

478: final number will not reach the light-sheet $L$.  What enforces this

479: pruning dynamically?

480: 

481: 

482: \section{Instabilities}

483: \label{sec-instability}

484: 

485: In this section we discuss why many initial states that are

486: macroscopically identical to that of the background lead to a

487: completely different future evolution.  It is instructive to consider

488: first the idealized case where no energy density is present in the

489: true vacuum region of the background.  This case turns out to be

490: somewhat similar to de~Sitter space~\cite{Ban00,Bou00a}: the

491: background is stable against sufficiently small perturbations, but

492: introducing too much matter will make it collapse to a big

493: crunch.  The extra mass produced near the time-symmetric slice is a

494: good diagnostic for the permissibility of a perturbation.

495: 

496: Interestingly, this is no longer the case for the true bounce

497: solution, which is filled with homogeneous matter energy density.  We

498: find that this background is far more unstable.  The process whereby

499: matter energy is absorbed into vacuum energy, permitting a bounce, is

500: extremely delicate since it decreases coarse-grained entropy.

501: Arbitrarily small perturbations at early times typically suffice to

502: derail this absorption completely, causing a big crunch.  (On the

503: other hand, one might imagine introducing enormous extra mass at early

504: times; this could be consistent with a bounce as long as it all ends

505: up in a coherent excitation of the field $\phi$.)

506: 

507: 

508: \subsection{Instabilities of the vacuous bounce}

509: \label{sec-babyinst}

510: 

511: In this subsection we assume that by fine tuning the

512: potential, we have arranged that there are no particles in the

513: background.  In this case the matter entropy vanishes at all times.

514: 

515: In simple examples, it is possible to see how gravitational

516: backreaction enforces the bound (\ref{eq-bound}). For example, we

517: could add a black hole at rest at the center of the flat region. We

518: will see that there is a bound on how big the black hole can be.

519: 

520: It is simplest to analyze the situation on the time-symmetric slice

521: \t0; because we are adding a black hole at rest at the center it does

522: not break the time reversal invariance of the background.  The

523: solution will be Schwarzschild out to the domain wall. Outside the

524: domain wall, the geometry must be de Sitter or Schwarzschild-de Sitter

525: because of the spherical symmetry. The radius of the domain wall,

526: $r_0$, will be bigger~\cite{FreSus04} than the Schwarzschild radius of

527: the mass, $2GM$. In order to match to de Sitter space along the

528: time-symmetric slice, the domain wall radius must be less than the de

529: Sitter radius $R$. In order to satisfy both of these constraints, we

530: need $2GM < R$. A bound on the mass of the black hole is equivalent to

531: a limit on the entropy, \be S < {\pi R^2 \over G}, \ee which agrees

532: precisely with the prediction we derived from the covariant entropy

533: bound.

534: 

535: Another example is to add $N$ photons to the initial state.  Because

536: the spatial slices are infinite, their wavelength can be made

537: arbitrarily long.  Hence $N$ can be large without causing large energy

538: density.  The entropy of the state is $N$, and if the entropy exceeds

539: the entropy bound (\ref{eq-bound}) then something must go wrong. What

540: happens is that as the space contracts towards $t=0$, the photons are

541: blueshifted. At $t=0$, the wavelength of the photons is limited by the

542: size of the space, roughly the de Sitter radius $R$. $N$ photons with

543: wavelength $R$ have energy $N/R$.  If $N$ exceeds the entropy bound

544: $S_{max}$ then the energy exceeds roughly $R/G$.

545: 

546: There must be a huge backreaction, because the biggest black hole that

547: can fit in de Sitter space has a mass of order

548: \begin{equation}

549: M_{\rm BH}\approx R/G~.

550: \end{equation}

551: If we exceed the entropy bound, we are inserting more energy than can

552: fit through the geometry, even if we allow black hole formation. The

553: result must be a big crunch.

554: Thus we find again that the conditions for avoiding a crunch and for

555: avoiding a violation of the bound coincide.

556: 

557: Of course, not all states that lead to a crunch have large entropy.

558: Consider adding two small particles of mass $m$ which are comoving

559: with respect to the open-FRW coordinate system picked out by the

560: symmetries of the geometry.  As we shall see, they can be more simply

561: characterized from the point of view of Minkowski space, as two

562: particles colliding at the event $r=0$, $t=0$, as shown in

563: Fig.~\ref{collision}.

564: %

565: \begin{figure}[!!htb] \center \includegraphics [scale=.7, clip]

566:       {collisionconf} \caption{Two ordinary particles collide with a

567:         large center of mass energy if they begin far apart. The

568:         backreaction will cause a crunch (heavy line).  The regions

569:         above this line are included only to guide the eye.  The

570:         diagram has been doubled: every point represents a

571:         hemisphere, not a sphere.}  \label{collision} \end{figure}

572: %

573: The FRW coordinates cover a region of flat space, with metric

574: \be

575: ds^2 = -d\tau^2 + \tau^2(d\rho^2 + \sinh^2\rho\, d\om)~.

576: \ee

577: We use Greek letters $(\tau,\rho)$ for the FRW coordinates and Roman

578: letters $(r,t)$ for the usual flat space coordinates. The two

579: coordinate systems are related by

580: \bea

581: \tau^2 & = & t^2 - r^2 \\

582: \sinh^2\rho & = & {r^2 \over t^2 - r^2}~.

583: \eea

584: A particle at rest in FRW coordinates

585: follows the trajectory

586: \begin{equation}

587: r(t) = t \tanh \rho_0~,

588: \end{equation}

589: where $\rho_0$ is the value of its FRW radial coordinate. The

590: trajectory passes through the origin at $t=0$ no matter where in the

591: FRW the particle began. So two particles at rest relative to the FRW

592: coordinates will collide at the origin. If they start at opposite

593: points on the 2-sphere, their relative velocity will be

594: \begin{equation}

595: v_{\rm rel} = 2 \tanh \rho_0 ~.

596: \end{equation}

597: The center of mass energy is

598: \begin{equation}

599: E_{\rm cm} = 2 m \cosh \rho_0~,

600: \end{equation}

601: where $m$ is the mass of the particles, so the energy is large if they

602: start far apart.  To avoid a global crunch, this energy must not

603: exceed the mass of the maximal black hole, leading to the

604: inequality

605: \begin{equation}

606: m \cosh \rho_0 \lesssim R/G~.

607: \end{equation}

608: This result is interesting because it indicates that two innocent

609: perturbations far from each other tend to collide catastrophically.

610: 

611: \subsection{Instabilities of the bounce}

612: \label{sec-fullinst}

613: 

614: As discussed in Sec.~\ref{sec-metric}, the FRW regions (II and III) of

615: the bounce contain an infinite number of particles. The light-sheet

616: $L$ can contain only a finite number of particles, so as time evolves,

617: all but a finite number of particles must be absorbed. For

618: definiteness, we assume that in the background all particles are

619: produced on a reheating surface in region II, so by time-reversal

620: symmetry, they must be absorbed on a ``recooling'' surface in region

621: III, prior to $t=0$.

622: 

623: Recooling is an extremely delicate process that decreases the entropy

624: drastically.  It is comparable to a broken glass reassemling

625: spontaneously, or all air molecules collecting in one small corner of

626: a room.  Only a miniscule subset of phase space trajectories

627: correspond to these processes.  Tiny perturbations do not take these

628: special states into each other, but into states with a generic

629: evolution.  For example, a tiny mass added in the far past will

630: gravitationally influence a vast number of other particles and disturb

631: the recooling process in the entire future light-cone of the

632: perturbation (see Fig.~\ref{unstable}).  If the perturbation is

633: %

634: \begin{figure}[!htb] \center \includegraphics [scale=.4]

635:       {unstable.eps} \caption{The past half of the bounce solution.  A

636:         perturbation (dot) will affect everything in its future

637:         lightcone (shaded), in general destroying the delicate

638:         ``recooling" process (at the heavy line).  Comparison to the

639:         apparent horizon (wiggly line) shows that this can cause a

640:         large number of horizon volumes to remain matter-dominated.}

641:       \label{unstable} \end{figure}

642: %

643: introduced sufficiently early, this will result in an enormous number

644: of particles remaining unabsorbed.  {\em Their\/} energy, not the

645: energy of the initial perturbation, decides whether the universe will

646: bounce or crunch.

647: 

648: In fact, recooling can be thrown off just by moving some particles

649: around, leaving the energy invariant (at least at the perturbative

650: level at which we might hope to define energy in this background); or

651: by removing some particles from the background, effectively decreasing

652: the initial energy and entropy.

653: 

654: But the entropy bound does not forbid all states; there should be a

655: finite number that do not destroy the geometry.  In the next section,

656: we construct examples of such states and verify that their entropy

657: approximately saturates the bound.

658: 

659: 

660: \section{Allowed perturbations}

661: \label{sec-halo}

662: 

663: In the saddlepoint approximation to the path integral, amplitudes

664: between points near past and future infinity should correspond to

665: non-crunching perturbations.  In this section we provide approximate,

666: semiclassical constructions of two classes of such states.

667: 

668: \subsection{States defined at \t0}

669: 

670: A natural place to define amplitudes is the past and future FRW

671: infinities.  But we have seen in the previous section that it is

672: tricky to identify in-states that do not collapse.  At least at the

673: semi-classical level, however, such states can be found by perturbing

674: the geometry on the time-symmetric slice, $t=0$.  (In a full quantum

675: theory, amplitudes would receive contributions from various

676: configurations at $t=0$.)

677: 

678: Unlike the FRW regions, the true vacuum portion of the time-symmetric

679: slice is well approximated by empty flat space.  Hence we can use the

680: results of Sec.~\ref{sec-babyinst}.  There are small perturbations

681: with a few extra particles, but most states will correspond to black

682: holes.  As discussed in Sec.~\ref{sec-babyinst}, the largest black

683: hole that fits has radius of order $R$, so the entropy of the possible

684: perturbations will be of order $R^2/G$, in agreement with the bound

685: (\ref{eq-bound}).

686: 

687: When those states propagate forward, they undergo complex interactions

688: with the particles produced at reheating.  Some of the excitations

689: will cross the event horizon of the FRW observer and enter region

690: IV. In Ref.~\cite{FreSus04} it was conjectured that the evolution is

691: nevertheless unitary.  But if scattering off of black holes offers any

692: guidance, the information will arrive at ${\cal I}^+$ in extremely

693: scrambled form.

694: 

695: Thus, it will be hard to look at an out-state and tell whether it

696: originated from an allowed perturbation at $t=0$, i.e., whether it

697: corresponds semiclassically to a geometry with that contains also a

698: past asymptotic region.  But by time-reversal symmetry of the bounce

699: background, this means it is also very difficult to set up an initial

700: perturbation that will not crunch, but will rather bounce and produce

701: a late-time geometry asymptotic to the future regions of the bounce.

702: 

703: In the next subsection, we shall see that it is nevertheless possible

704: to characterize some of the suitable in-states approximately, and to

705: verify once more that their entropy is in agreement with the bound

706: (\ref{eq-bound}).

707: 

708: 

709: \subsection{Halo states}

710: 

711: We are interested in an approximate construction, near ${\cal I}^-$,

712: of asymptotic states that can lead to a nonsingular bounce geometry.

713: We will assume that the matter present in the FRW regions is dust.  We

714: will also assume that its density is low enough so that the FRW

715: evolution is always curvature dominated.

716: 

717: The basic idea is to construct solutions which differ from the bounce

718: in a finite region, so that the delicate ``recooling'' process is only

719: disturbed for a small number of particles.  If we could neglect

720: gravitational signals propagating out from the perturbation, this

721: would be readily accomplished by adding, subtracting, or rearranging

722: particles in a bounded region of space.  But we must be careful to

723: avoid any changes, no matter how small, in the gravitational field

724: outside the region considered.

725: 

726: We achieve this by imposing two conditions.  The first is that the

727: perturbation be spherically symmetric, so that Birkhoff's theorem

728: applies.  The second is that the total mass in the perturbed region

729: remain unchanged compared to the background.  A simple example is an

730: overdense region mass surrounded by an vacuous shell, or

731: ``halo''.\footnote{In reality, the dust approximation is never exact,

732:   and with realistic quantum fields it is impossible to suppress the

733:   propagation of signals out from the perturbed region entirely.  Here

734:   we assume that the dust approximation is good and that the

735:   corrections (the weak signals leaking out) can be precisely

736:   cancelled by subtle perturbations of the global background, without

737:   affecting the basic geometry and coarse-grained entropy of our

738:   solutions.}

739: 

740: Our solution consists of three pieces joined along timelike

741: hypersurfaces, as shown in Fig.~\ref{halo}.  We take the outer piece

742: to be a portion of the background, i.e., an open contracting dusty FRW

743: universe, except for the region $\rho<\rho_0$, which we excise and

744: throw away.  [Here $\rho$ is the comoving radial coordinate; see

745:   Eq.~(\ref{eq-iop}).]  By starting with this exterior piece and

746: working our way in, the absence of any perturbation outside $\rho_0$

747: is guaranteed by construction. This will be important.

748: \begin{figure}[!htb] \subfigure[]{ \includegraphics [scale=.7]

749:   {int.eps} } \hspace{0.3 in} \subfigure []{\includegraphics [scale =

750:     .7]{ext.eps}} \hspace{.3 in} \subfigure []{\includegraphics

751:   [scale=.7] {intandext.eps}} \caption{Construction of an in-state

752:   which will {\em not} cause a global collapse. The full solution (c)

753:   is obtained by gluing a piece of the background solution (b) to a

754:   collapsing ball of dust surrounded by vacuum (a). The full solution

755:   keeps the region of (a) to the left of the line $\beta$ and the

756:   region of (b) to the right of $\beta$. $\alpha$ marks the boundary

757:   between the collapsing dust (shaded) and vacuum. In the center of

758:   the diagram the particles of the outside solution have been

759:   absorbed, rendering the choice of $\beta$ arbitrary (dashed

760:   line). In this region $\beta$ is not a geodesic, whereas $\alpha$ is

761:   a geodesic throughout the diagram.  Heavy lines represent the domain

762:   wall (timelike) and the reheating/recooling surfaces (spacelike).

763:   The black hole horizon and singularity are shown in the top left

764:   corner.}  \label{halo} \end{figure}

765: 

766: We take the middle piece to be vacuous.  Then Birkhoff's theorem

767: implies that the middle piece is given by a portion of a Schwarzschild

768: solution.  The Schwarzschild radius characterizing this solution is

769: that of the black hole which the excised piece (which we discarded)

770: would have formed if complemented by a vacuum exterior.  Equivalently,

771: the correct solution is fixed by the requirement that the

772: Schwarzschild geometry contain a geodesic identical to the path of the

773: innermost dust shell of the outer piece.  This geodesic, $\beta$,

774: forms the outer limit of the middle piece.

775: 

776: We have thus obtained an open universe containing a black hole.  The

777: fully extended solution contains an Einstein-Rosen bridge to an

778: asymptotically flat region.  This would not describe a reasonable

779: perturbation of the in-state.  But this is easily fixed by bringing

780: matter back at a smaller radius.  

781: 

782: Consider a timelike radial geodesic, $\alpha$, in the middle piece,

783: such that $\alpha$ begins at past timelike infinity and lies

784: everywhere inside $\beta$, as shown in Fig.~\ref{halo}.  The inner

785: piece (inside of $\alpha$) can then be taken to be the central portion

786: of an open dust-dominated FRW universe uniquely determined by the

787: requirement that it contain the geodesic hypersurface $\alpha$ that

788: forms its boundary.  Note that the inner piece is {\em not\/} a

789: portion of the original open FRW universe---it will have a different

790: density at the time when curvature starts to dominate.\footnote{One

791:   could also construct halo solutions with white-hole initial

792:   conditions by taking the inner piece to be a portion of a closed

793:   recollapsing dust dominated FRW universe.}

794: 

795: Because they involve underdense shells, we will call these states halo

796: states.  They differ from the background only within a finite comoving

797: radius, and because they contain only dust, no signal will propagate

798: out to the unperturbed region.  Hence, it is easy to evolve the halo

799: states from past infinity to the recooling surface.  In the

800: unperturbed region, the conversion of particles into vacuum energy

801: will proceed exactly as in the vacuum.\footnote{At the recooling

802:   surface, spatial gradients will arise in the field which makes up

803:   the domain wall, allowing signals to propagate out.  This is

804:   acceptable, since our goal was only to avoid disturbing the

805:   recooling.  However, there will be corrections to our solution after

806:   recooling.}

807: 

808: The halo region will not be absorbed, so that a collapsing ball of

809: dust will enter the central region of the spacetime.  Thus halo states

810: form black holes, and black holes cannot fit if they are too large.

811: It follows that we can estimate the number of (bouncing) halo states

812: by the entropy of the black holes they form, which is of order

813: $R^2/G$.  Thus we find that the entropy bound (\ref{eq-bound}) is

814: approximately saturated by the non-crunching states.

815: 

816: Of course, the idealization as dust is not realistic in quantum field

817: theory.  Even if the background outside $\rho_0$ is truly unperturbed

818: near ${\cal I}^-$, some weak signals will propagate out from the halo

819: and disturb the absorption process on large scales.  However, we do

820: expect that there are allowed perturbations whose main feature is a

821: halo of size $\rho_0$, plus extremely subtle additional perturbations

822: outside the halo which cancel against such signals.

823: 

824: Not all allowed states will be of halo form.  We have merely pointed

825: out that the halo states appear to be the only allowed states that can

826: be described simply near the boundary of spacetime, and that they can

827: account for a significant portion of the allowed entropy.  The halo

828: states are highly inhomogeneous and thus of no relevance to a

829: realistic universe. The other allowed states are extremely scrambled

830: near infinity and hence hard to characterize.  

831: 

832: As we discussed in the introduction, this limits the appeal of an

833: S-matrix description (involving the past hat) as a framework for the

834: landscape.  But the attractive properties of the Coleman-De~Luccia

835: solution for defining observables~\cite{FreSus04}---the presence of

836: noninteracting particles and low curvature at late times---remain to

837: be exploited.  They may yet allow for the construction of an exact

838: quantum cosmology whose amplitudes correspond to asymptotic

839: observables in the future hat.

840: 

841: 

842: \subsubsection*{Acknowledgements}

843: 

844: We would like to thank T.~Banks and L.~Susskind for valuable

845: discussions.  This work was supported by the Berkeley Center for

846: Theoretical Physics, by a CAREER grant of the National Science

847: Foundation, and by DOE grant DE-AC03-76SF00098.

848: 

849: 

850: \bibliographystyle{board}

851: \bibliography{all}

852: 

853: \end{document}

854: 

855: 

856: