1: \documentclass[11pt,a4paper]{amsart}

2: 

3: \usepackage[]{graphicx}

4: \usepackage{psfrag}

5: 

6: 

7: 

8: %--------------- COMMANDES ABREGEES ---------------

9: 

10: \newcommand{\ncmd}{\newcommand}

11: \ncmd{\rencmd}{\renewcommand}

12: \ncmd{\dspst}{\displaystyle }

13: 

14: 

15: 

16: %------------------ TEXTE ABREGE ------------------

17: 

18: 

19: 

20: \ncmd{\preu}{\noindent \mbox{\bf Proof  }}

21: \ncmd{\sketch}{\noindent \mbox{\bf Sketch of proof  }}

22: \ncmd{\preud}[1]

23: {\noindent \mbox{\bf Proof of \ref{#1}   }}

24: 

25: \ncmd{\ex}{\noindent \mbox{\bf exemple  }}

26: \ncmd{\exs}{\noindent \mbox{\bf exemples  }}

27: \ncmd{\fin}{\hspace*{\fill} 

28: \quad\hbox{\hskip 1pt\vrule width 4pt height 6pt

29:           depth 1.5pt\hskip 1pt} \medskip }

30: %{\hspace*{\fill} $\dagger$ \medskip}

31: \ncmd{\lemv}{\noindent \mbox{\bf Lemma  }}

32: \ncmd{\thmv}{\noindent \mbox{\bf Theorem  }}

33: 

34: %--------------- SYMBOLES ------------------------

35: 

36: \ncmd{\ab}{\mbox{$\mbox{AdS}_{n}\/$ }}

37: \ncmd{\ei}{\mbox{${\mbox{Ein}}_{n}\/$ }}

38: \ncmd{\hei}{\mbox{$\widehat{\mbox{Ein}}_{n}\/$ }}

39: \ncmd{\tei}{\mbox{$\widetilde{\mbox{Ein}}_{n}\/$ }}

40: 

41: 

42: \ncmd{\mr}{\mbox{$\widetilde{M}\/$ }}

43: 

44: 

45: 

46: 

47: \ncmd{\slrr}{\mbox{$\widetilde{SL}(2,{\mathbb R}) \times \widetilde{SL}(2,{\mathbb R})\/$ }}

48: 

49: 

50: %---------------------ENVIRONNEMENTS--------------------

51: 

52: \newtheorem{prop}{Proposition}[section]

53: \newtheorem{thm}[prop]{Theorem}

54: \newtheorem{lem}[prop]{Lemma}

55: \newtheorem{cor}[prop]{Corollary}

56: \newtheorem{defin}[prop]{Definition}

57: \newtheorem{rac}[prop]{Remark}

58: \newtheorem{ric}[prop]{Example} 

59: 

60: 

61: 

62: \ncmd{\exe}{\begin{ric} \em }

63: \ncmd{\eexe}{\em \end{ric}}

64: \ncmd{\rque}{\begin{rac} \em}

65: \ncmd{\erque}{\em \end{rac}}

66: 

67: 

68: %---------------------COMPTEURS----------------------

69: 

70: % \rencmd{\theprop}{\arabic{prop}}        

71: % prop 1 2 3 etc...

72: % \rencmd{\thesection}{\arabic{section}}

73: \newcounter{subinc}

74: \rencmd{\thesubsubsection}{\thesubsection-\alph{subsubsection}}

75: \newcounter{inc}

76: \rencmd{\theequation}{\arabic{inc}}

77: \newcounter{cintro}

78: 

79: 

80: 

81: 

82: %-----------------------------------------------------

83: 

84: \hyphenation{}

85: \makeindex

86: 

87: %------------------------------------------------------

88: 

89: \begin{document}

90: 

91: \vspace{5cm}

92: 

93: \title[BTZ multi black-holes]{Causal properties of AdS-isometry groups II: \\

94: BTZ multi black-holes}

95: \author{Thierry Barbot}

96: \email{Thierry.Barbot@umpa.ens-lyon.fr}

97: \thanks

98: {Work supported 

99: by CNRS, ACI ``Structures g\'eom\'etriques et Trous Noirs''.}

100: \address{CNRS, UMR 5669\\

101: Ecole Normale Sup\'erieure de Lyon\\ 

102: 46 all\'ee d'Italie, 69364 Lyon}

103: \keywords{Anti-de Sitter space, Causality, BTZ black hole}

104: \subjclass{83C80 (83C57), 57S25}

105: \date{\today}

106: 

107: \begin{abstract}

108: This paper is the continuation of \cite{ba1}. We essentially prove

109: that the familly of strongly causal spacetimes defined in \cite{ba1}

110: associated to generic achronal subsets in $\mbox{Ein}_2$ contains 

111: all the examples of BTZ multi black-holes. It provides new elements

112: for the global description of these multi black-holes.

113: We also prove that any strongly causal spacetime locally modelled on 

114: the anti-de Sitter

115: space admits a well-defined maximal strongly causal

116: conformal boundary locally modelled

117: on $\mbox{Ein}_2$.

118: \end{abstract}

119: 

120: \maketitle

121: 

122: \input{introele}

123: 

124: \section{Strongly causal spacetimes}

125: \label{sec.strong}

126: 

127: A lorentzian manifold $M$ is \emph{strongly causal\/} if 

128: for every point $x$ in $M$ every neighborhood of $x$ 

129: contains an open neighborhood $U$ which 

130: is \emph{causally convex,\/} i.e. such that any causal curve in $M$ 

131: joining two points in $U$ is actually contained in $U$

132: (see \S $2.3$ in \cite{ba1}). An

133: isometric action of a group $\Gamma$ on a strongly causal lorentzian

134: manifold $M$ is strongly causal if any point $x$ of $M$ admits 

135: arbitrarly small neighborhoods $U$ such that for every non trivial element $\gamma$ of

136: $\Gamma$ no point of $U$ is causally related to a point of $\gamma{U}$.

137: Clearly, if $\Gamma$ is a group of isometries of a strongly causal lorentzian

138: manifold $M$ acting freely and properly discontinuously, the quotient

139: manifold $\Gamma\backslash{M}$ is strongly causal if and only if

140: the action of $\Gamma$ is strongly causal.

141: 

142: In this paper we need to prove that certain isometric actions are strongly

143: causal. In this {\S} we provide two lemmas useful for this purpose.

144: 

145: In the first lemma we consider a strongly causal lorentzian manifold $M$

146: admitting a Killing vector field

147: $X$. We assume that $X$ is everywhere spacelike. Let $\phi^t$ be the flow generated by

148: $X$ and assume that the ${\mathbb R}$-action defined by $\phi^t$ is free and

149: properly discontinuous. Let $Q^\phi$ be the orbit space of $\phi^t$ and

150: $\pi: M \rightarrow Q^\phi$ the projection map. 

151: It is easy to show that $\pi$ is a locally trivial fibration with fibers

152: homeomorphic to $\mathbb R$.

153: For every tangent vector $v$

154: of $Q^\phi$ at a point $y$ let $x$ be any element of $\pi^{-1}(y)$ and let $w$ be the

155: unique tangent vector to $M$ at $x$ orthogonal to $X(x)$ and such that $d\pi_x(w) = v$. 

156: Since $\phi^t$ is isometric the norm of $w$ does not depend on the choice of $x$:

157: we define the norm $\mid v \mid$ as this norm $\mid w \mid$. This procedure defines

158: a lorentzian metric on $Q^\phi$. 

159: 

160: 

161: 

162: 

163: 

164: \begin{lem}

165: \label{lem.quocausal}

166: If the lorentzian manifold $Q^\phi$ is strongly causal then the action of $\phi^1$ on $M$ is

167: strongly causal.

168: \end{lem}

169: 

170: \preu

171: Let $x$ be any element of $M$ and $U$ any neighborhood of $x$.

172: Let $y = \pi(x)$. Since $Q^\phi$ is assumed strongly

173: causal, $U' = \pi(U)$ contains a causally convex open

174: neighborhood $V'$ of $y$. Define $V = \pi^{-1}(V')$: this open subset is

175: diffeomorphic to $V' \times {\mathbb R}$, such that the restriction of 

176: $\phi^t$ to $V$ corresponds to the translation on the $\mathbb R$-factor. 

177: We also take the convention that the $\mathbb R$-coordinate of $x$ is

178: zero: $x \approx (y,0)$. Shrinking $V'$ if necessary, we can assume that

179: the spacelike norm of $X$ on $V$ is uniformly bounded by some constant $C$,

180: and moreover that any causal curve in $V'$ has lorentzian length less than

181: $1/2C$. It then follows that the $\mathbb R$-coordinate of any causal

182: curve in $V$ cannot vary more than $1/2$. Let $W$ be a causally convex

183: open neighborhood of $x$ contained in

184: $V' \times ]-1/4, +1/4[ \subset V$.

185: We claim that for every non zero integer $n$ non point of $\phi^nW$ is

186: causally related to a point of $W$. Indeed: assume by contradiction

187: that there is a causal curve $c: I \rightarrow M$ joining an

188: element $(z,t)$ to an element $(z', s+n)$ with $z$, $z'$ in $V'$ and

189: $s$, $t$ in $]-1/4, +1/4[$. The projection $\pi \circ c$ is a causal

190: curve in $Q$ joining two $\pi(z)$, $\pi(z')$. Since $V'$ is causally convex

191: $\pi \circ c$ is contained in $V'$, i.e. $c$ is contained in $V$. Hence

192: the $\mathbb R$-coordinate cannot vary along $c$ more than $1/2$.  

193: We obtain a contradiction since $s+n-t$ has absolute value bigger than $1/2$.\fin

194: 

195: \rque

196: \label{rk.quocausalfini}

197: Lemma~\ref{lem.quocausal} remains true when the action of $\phi^t$ is not free

198: but periodic. In this case the group generated by $\phi^1$ is finite.

199: Details are left to the reader.

200: \erque

201: 

202: We will apply lemma~\ref{lem.quocausal} in the case where $M$ has dimension $3$ and is simply

203: connected. In this situation the map $\pi$ is a fibration and the homotopy sequence of

204: this fibration implies that $Q^\phi$ is a $2$-dimensional manifold which is simply connected.

205: Hence a very nice complement is:

206: 

207: 

208: \begin{lem}

209: \label{lem.2causal}

210: Any simply connected $2$-dimensional lorentzian manifold is strongly causal.

211: \end{lem}

212: 

213: \preu

214: Any simply connected surface is diffeomorphic to the sphere or the plane ${\mathbb R}^2$.

215: Since the sphere does not admit any lorentzian metric we just have to consider the

216: case of the plane ${\mathbb R}^2$. Any reader acquainted with dynamical systems

217: will recognize that the lemma follows quite easily from

218: the Poincar\'e-Bendixon Theorem applied to the lightlike foliations: 

219: every leaf of one of these foliations is a closed embedding of the real line

220: in ${\mathbb R}^2$ and for any point $x$ in ${\mathbb R}^2$ the future (resp. 

221: the past) of $x$ is the domain bounded by $r_1 \cup r_2$ where $r_1$, $r_2$ are

222: the future oriented (resp. past oriented) lighlike geodesic rays starting from $x$.

223: Details are left to these readers.

224: 

225: For other readers, we refer to Theorem $3.43$ of \cite{beem} where 

226: a slightly better statement is proved:

227: lorentzian metrics on the $2$-plane are \emph{stably\/} causal, which

228: is stronger than being strongly causal.

229: \fin

230: 

231: 

232: 

233: 

234: \section{Description of the elementary generic achronal subsets of 

235: $\widehat{\mbox{Ein}}_2$}

236: \label{sub.elementary}

237: 

238: In this section, we complete the descriptions of invisible domains 

239: $E(\widetilde{\Lambda})$ and $\Omega(\widetilde{\Lambda})$ by considering

240: the elementary cases. For the reader convenience, we start with the splitting 

241: case, which is already described in \cite{ba1}, {\S} $8.8$.

242: 

243: 

244: \subsection{The splitting case}

245: \label{subsub.split}

246: It is the case $\widetilde{\Lambda} = \{ x,y \}$, where $x$, $y$ are 

247: two non-causally related points in $\widehat{\mbox{Ein}}_{2}$. 

248: Then, $\{ x,y \}$ is a gap pair and there are two associated ordered 

249: gap pairs that we denote respectively by $(x,y)$ and $(y,x)$ (see definition

250: $8.24$ in \cite{ba1}). 

251: $\widetilde{\Lambda}^+$ is an union 

252: ${\mathcal T}^+_{xy} \cup {\mathcal T}^+_{yx}$ of two future oriented lightlike 

253: segments with extremities $x$, $y$ that we call \emph{upper tents\/.} 

254: Such an upper tent is the union of two lightlike segments, one starting 

255: from $x$, the other from $y$, and stopping at their first intersection point, 

256: that we call the \emph{upper corner.\/}

257: 

258: Similarly, $\widetilde{\Lambda}^-$ is an union 

259: ${\mathcal T}^-_{xy} \cup {\mathcal T}^-_{yx}$ of two 

260: \emph{lower tents\/} admitting a similar 

261: description, but where the lightlike segments starting from $x$, $y$ 

262: are now past oriented

263: (see Figure \ref{ttents}, Figure~\ref{splitfigure}), and sharing a common extremity: the \emph{lower 

264: corner.\/}

265: 

266: 

267: \begin{figure}[ht]

268: \centerline{\includegraphics[width=8cm]{tent.ps}}

269: \caption{\label{ttents}

270:    \textit{Upper and lower tents}} 

271: \end{figure}

272: 

273: 

274: The invisible domain $\Omega(\widetilde{\Lambda})$ from $\widetilde{\Lambda}$ in $\widehat{\mbox{Ein}}_{2}$ is the union of two diamond-shape regions $\widetilde{\Delta}_1$, $\widetilde{\Delta}_2$. The boundary of $\widetilde{\Delta}_1$ is the union ${\mathcal T}^+_{xy} \cup {\mathcal T}^-_{xy}$, and the boundary of $\widetilde{\Delta}_2$ is ${\mathcal T}^+_{yx} \cup {\mathcal T}^-_{yx}$. We project all the picture in some affine region $V \approx {\mathbb  R}^3$ of $S(E)$ such that:

275: 

276: -- $V \cap {\mathbb A}{\mathbb D}{\mathbb S}$ is the interior of the hyperboloid: $\{ x^2 + y^2 < 1 + z^2 \}$,

277: 

278: -- $\Lambda = \{ (1,0,0), (-1, 0, 0) \}$.

279: 

280: Then, $E(\Lambda)$ is region $\{ -1 < x < 1 \} \cap {\mathbb A}{\mathbb D}{\mathbb S}$.

281: One of the diamond-shape region $\widetilde{\Delta}_i$ projects to $\Delta_1 = \{ -1 < x < 1, y >0 , x^2+y^2 = 1+z^2 \}$, the other projects to $\Delta_2 = \{ -1 < x < 1, y <0 , x^2+y^2 = 1+z^2 \}$. The past of $\Delta_1$ in $E(\Lambda)$ is $P_1 = \{ (x,y,z) \in E(\Lambda) / z < y \}$. and the future of $\Delta_1$ in $E(\Lambda)$ is $F_1 = \{ (x,y,z) \in E(\Lambda) / z > -y \}$. We have of course a similar description for the future $F_2$ and the past $P_2$ of $\Delta_2$ in $E(\Lambda)$. Observe:

282: 

283: -- the intersections $F_1 \cap F_2$ and $P_1 \cap P_2$ are disjoint. They are tetraedra in $S(E)$:

284: $F_1 \cap F_2$ is the interior of the convex hull of $\Lambda^+$, 

285: and $P_1 \cap P_2$ is the interior of the 

286: convex hull of $\Lambda^-$.

287: 

288: -- the intersection $F_1 \cap P_1$ (resp. $F_2 \cap P_2$) is the intersection between ${\mathbb A}{\mathbb D}{\mathbb S}$ and the interior of a tetraedron in $S(E)$: the convex hull of $\Delta_1$ (resp. $\Delta_2$).

289: 

290: \begin{defin}

291: \label{defghsimple}

292: $E^+(\Lambda) = F_1 \cap F_2$ is the future globally hyperbolic convex core; $E^-(\Lambda) = P_1 \cap P_2$ is the past globally convex core.

293: \end{defin}

294: 

295: This terminology is justified by the following (easy) fact: $F_1 \cap F_2$ (resp. $P_1 \cap P_2$) is the invisible domain $E(\Lambda^+)$ (resp. $E(\Lambda^-)$). Hence, they are indeed globally hyperbolic.

296: 

297: The intersection between the closure of $E(\Lambda)$ in $S(E)$ and the boundary ${\mathcal Q}$ of ${\mathbb A}{\mathbb D}{\mathbb S}$ is the union of the closures of the diamond-shape regions. Hence, $\Delta_{1,2}$ can be thought as the conformal boundaries at infinity of $E(\Lambda)$. Starting from any point in $E(\Lambda)$, to $\Delta_i$ we have to enter in $F_i \cap P_i$, hence we can adopt the following definition:

298: 

299: \begin{defin}

300: \label{def.end}

301: $F_1 \cap P_1$ is an end of $E(\Lambda)$.

302: \end{defin}

303: 

304: Finally:

305: 

306: \begin{defin}

307: The future horizon is the past boundary of $F_1 \cap F_2$; the past horizon is the

308: future boundary of $P_1 \cap P_2$.

309: \end{defin}

310: 

311: 

312: \begin{prop}

313: \label{simpledecomp}

314: $E(\Lambda)$ is the disjoint union of the future and past 

315: glo\-bally hyperbolic cores $E^\pm(\Lambda)$, of the two ends,  

316: and of the past and future horizons.\fin

317: \end{prop}

318: 

319: 

320: 

321: 

322: 

323: 

324: 

325: \begin{figure}

326: \centerline{\includegraphics[width=8cm]{lambdasplit.ps}}

327: \caption{\label{splitfigure}

328:    \textit{The splitting case. The domain $E(\Lambda)$ is between 

329: the hyperplanes $x^\perp$ and $y^\perp$. These hyperplanes,

330: tangent to the hyperboloid, are not drawn, except their intersections with the hyperboloid,

331: which are the upper and lower tents ${\mathcal T}^\pm_{xy}$, ${\mathcal T}_{yx}^\pm$.}} 

332: \end{figure}

333: 

334: \rque

335: \label{rk.BTZnotations}

336: In the conventions of \cite{BTZ, BTZ2, brill}, the globally hyperbolic convex cores $F_1 \cap F_2$ and $P_1 \cap P_2$ are \emph{regions of type II,\/} also called \emph{intermediate regions.\/} The ends $F_1 \cap P_1$ and $F_2 \cap P_2$ are \emph{outer regions,\/} or \emph{regions of type I.\/}

337: \erque

338: 

339: 

340: 

341: 

342: \subsection{The extreme case}

343: \label{subsub.extreme}

344: 

345: 

346: The extreme case is harder to picture out 

347: since $\Omega(\widetilde{\Lambda})$ and $E(\widetilde{\Lambda})$ 

348: are \emph{not} contained in an affine domain (see figure~\ref{extremefigure}). 

349: Assume that $y$ is in the future of $x$. Observe that $\widetilde{\Lambda}^\pm$ 

350: are then pure lightlike. Hence, $E(\widetilde{\Lambda}^\pm)$ are empty. 

351: The region $\Omega(x,y)$ is a ``diamond'' in $\mbox{Ein}_2$ 

352: (we call it an \emph{extreme diamond\/} ) admitting as 

353: boundary four lightlike segments: the segments $[y, \delta(x)]$, 

354: $[x, \delta(x)]$, $[\delta^{-1}(y), x]$, and $[y, \delta^{-1}(y)]$. 

355: 

356: A carefull analysis shows that $E(x,y)$ is precisely the intersection 

357: between the past and the future of $\Omega(x,y)$.

358: 

359: 

360: 

361: 

362: 

363: \begin{figure}

364: \centerline{\includegraphics[width=8cm]{extremefigure.ps}}

365: \caption{\label{extremefigure}

366:    \textit{The extreme case. The point $\delta(x)$ is in the front of

367:    the upper hyperboloid. The point $x$ is in the rear of the lower hyperboloid.

368:    The domain $E(\Lambda)$ is the domain between the hyperplanes $\delta(x)^\perp$ and 

369:    $\delta^{-1}(y)^\perp$.}} 

370: \end{figure}

371: 

372: 

373: Keeping $x$ fixed, and considering a sequence $y_n$ converging to $y$, 

374: with $y_n$ non-causally related to $x$, then one of the diamond-shape 

375: region $\Delta^n_i$ of the associated $\Omega(x, y_n)$ - let's say, 

376: $\Delta^n_2$ - vanishes. The other converges to the entire region 

377: $\Omega(x,y)$. The various parts of the domains $E(x, y_n)$, namely the 

378: globally hyperbolic convex cores $E^\pm(x,y_n)$ and the ends, vanish, 

379: except one of the end, which becomes closer and closer to the entire $E(x,y)$. 

380: 

381: \subsection{The conical case}

382: \label{subsub.conical}

383: In the conical case, $\Lambda$ is an upper or lower tent. By symetry, 

384: we can consider only the upper case: 

385: $\Lambda  ={\mathcal T}^+_{xy} = [x,z] \cup [z,y]$. Then, 

386: $\Lambda^+ = \Lambda \cup {\mathcal T}^+_{yx}$, and 

387: $\Lambda^-$ is the pure lightlike subset 

388: $\Lambda \cup {\mathcal T}^-_{yx}$. In the notations of \S~\ref{subsub.split}, 

389: $\Lambda^+$ is the future boundary of the diamond $\Delta_1$, and 

390: $\Omega(\Lambda)$ is the diamond $\Delta_2$.

391: Then, in some affine domain, $E(\Lambda) = \{ (x,y,z) / -1 < x < 1, z>y \}$. 

392: It can be described as the future in $V$ of $\Delta_2$. 

393: It is also the intersection between the past of $x$, 

394: the past of $y$, and the complement of the past of $z$. Finally, 

395: $E(\Lambda)$ is the union of $F_1 \cap F_2$, $F_2 \cap P_2$ and 

396: the component of the past horizon of $E(x,y)$ separating these two regions 

397: (see figure \ref{splitfigure}).

398: 

399: 

400: 

401: 

402: 

403: \input{fin2}

404: 

405: \input{group2}

406: 

407: \input{coincide}

408: 

409: \input{btzmeme2}

410: 

411: \input{biblio2}

412: 

413: \end{document}

414: 

415: 

416: 

417: 

418: 

419: 

420: 

421: 

422: 

423: 

424: 

425: 

426: 

427: 

428: 

429: 

430: 

431: 

432: 

433: 

434: 

435: 

436: 

437: 

438: 

439: 

440: 

441: 

442: 

443: 

444: 

445: 

446: 

447: 

448: 

449: 

450: 

451: 

452: 

453: 

454: 

455: 

456: 

457: 

458: 

459: 

460: 

461: 

462: 

463: 

464: 

465: 

466: 

467: 

468: 

469: 

470: 

471: 

472: 

473: 

474: 

475: 

476: 

477: 

478: 

479: 

480: 

481: 

482: 

483: 

484: 

485: 

486: 

487: 

488: 

489: 

490: 

491: 

492: 

493: 

494: 

495: 

496: 

497: 

498: 

499: 

500: 

501: