1: \documentclass[12pt]{amsart}

2: \usepackage{amsfonts}

3: \usepackage{epsfig,graphics}

4: \usepackage{amssymb}

5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

6: \usepackage{amscd}

7: 

8: %TCIDATA{Created=Fri Mar 29 17:07:40 2002}

9: %TCIDATA{LastRevised=Sat Mar 30 20:12:35 2002}

10: 

11: \newtheorem{proposition}{Proposition}[section]

12: \newtheorem{theorem}{Theorem}[section]

13: \newtheorem{lemma}{Lemma}[section]

14: \newtheorem{corollary}{Corollary}[section]

15: \newtheorem{remark}{Remark}[section]

16: \newtheorem{exercise}{Notations}[section]

17: \newtheorem{definition}{Definition}[section]

18: \newtheorem{conjecture}{Conjecture}[section]

19: \newtheorem{example}{Example}[section]

20: \newtheorem{claim}{Claim}[section]

21: 

22: \addtolength{\hoffset}{-0.5cm}

23: \addtolength{\textwidth}{1cm}

24: 

25: \newcommand{\BZ}{{\mathbb{Z}}}

26: \newcommand{\BN}{{\mathbb{N}}} 

27: \newcommand{\BR}{{\mathbb{R}}}

28: \newcommand{\BC}{{\mathbb{C}}}

29: \newcommand{\BF}{{\mathbb{F}}}

30: \newcommand{\BQ}{{\mathbb{Q}}}

31: \newcommand{\BP}{{\mathbb{P}}}

32: \newcommand{\BA}{{\mathbb{A}}}

33: \newcommand{\BG}{{\mathbb{G}}}

34: \newcommand{\BT}{{\mathbb{T}}}

35: \newcommand{\BH}{{\mathbb{H}}}

36: 

37: \newcommand{\gD}{\Delta}

38: \newcommand{\gd}{\delta}

39: \newcommand{\gb}{\beta}

40: \newcommand{\gi}{\iota}

41: \newcommand{\gC}{\Gamma}

42: \newcommand{\gc}{\gamma}

43: \newcommand{\gs}{\sigma}

44: \newcommand{\gS}{\Sigma}

45: \newcommand{\gO}{\Omega}

46: \newcommand{\gep}{\epsilon}

47: \newcommand{\gl}{\lambda}

48: \newcommand{\ga}{\alpha}

49: \newcommand{\gt}{\tau}

50: \newcommand{\gz}{\zeta}

51: \newcommand{\gf}{\phi}

52: \newcommand{\gF}{\Phi}

53: \newcommand{\gth}{\theta}

54: \newcommand{\gTh}{\Theta}

55: \newcommand{\gT}{\Theta}

56: \newcommand{\fT}{{\frak{T}}}

57: \newcommand{\fH}{{\frak{H}}}

58: \newcommand{\ff}{{\frak{f}}}

59: \newcommand{\fg}{{\frak{g}}}

60: \newcommand{\fM}{{\frak{M}}}

61: \newcommand{\fO}{{\frak{O}}}

62: 

63: \def\a{\aleph}

64: \def\w{\omega}

65: \def\k{\kappa}

66: \def\De{\Delta}

67: \def\bbr{{\Bbb R}}

68: \def\vp{\varphi}

69: \def\e{\varepsilon}

70: 

71: \newcommand{\ti}[1]{\tilde{#1}}

72: \newcommand{\p}{\prod}

73: \newcommand{\ol}[1]{\overline{#1}}

74: \newcommand{\Pic}{\text{Pic}}

75: \newcommand{\jac}{\text{Jac}}

76: \newcommand{\prym}{\text{Prym}}

77: \newcommand{\gal}{\text{Gal}}

78: \newcommand{\Sym}{\text{Sym}}

79: \newcommand{\NS}{\text{NS}}

80: \newcommand{\dg}{\text{deg}}

81: \newcommand{\aut}{\text{Aut}}

82: \newcommand{\dm}{\text{dim}}

83: \newcommand{\norm}{\text{Norm}}

84: \newcommand{\odd}{\text{odd}}

85: \newcommand{\even}{\text{even}}

86: \newcommand{\ver}{\text{Ver}}

87: 

88: \newcommand{\rk}{\text{rank}}

89: \newcommand{\rank}{\text{rank}}

90: \newcommand{\vol}{\text{vol}}

91: \newcommand{\diam}{\text{diam}}

92: \newcommand{\diag}{\text{diag}}

93: 

94: \newcommand{\SL}{\text{SL}}

95: \newcommand{\GL}{\text{GL}}

96: \newcommand{\PSL}{\text{PSL}}

97: \newcommand{\PGL}{\text{PGL}}

98: \newcommand{\SO}{\text{SO}}

99: \newcommand{\PSO}{\text{PSO}}

100:    

101: \def\un{\underbar}

102: \def\uhr{\upharpoonright}

103: \def\dist{\text{dist}}

104: \def\Aut{\text{Aut}}

105: \def\N{{\cal N}}

106: 

107: \newcommand{\Cy}[1]{\mathbb{Z}/#1\mathbb{Z}}

108: \newcommand{\hra}{\hookrightarrow}

109: \newcommand{\im}[1]{\text{Im}(#1)}

110: \newcommand{\id}[1]{Id_{#1}}

111: \newcommand{\g}[1]{Gal(\ol{#1}/#1)}

112: \newcommand{\n}[2]{Norm_{#1/#2}}

113: \newcommand{\ord}[2]{\text{ord}_{(\Cy{#1})^{*}}#2}

114: \newcommand{\np}[2]{||#1||_{ #2}}

115: \newcommand{\h}[1]{H^{#1}}

116: \newcommand{\res}{\text{res}}

117: \newcommand{\df}{\stackrel{\text{def}}{=}}

118: \newcommand{\conan}{\ol{\ol{Z}}}

119: 

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

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

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

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

124: \newtheorem{conj}[prop]{Conjecture}

125: \theoremstyle{definition}

126: \newtheorem{nota}[prop]{Notations}

127: \newtheorem{defn}[prop]{Definition}

128: \newtheorem{rem}[prop]{Remark}

129: \newtheorem{exam}[prop]{Example}

130: \newtheorem{prob}[prop]{Problem}

131: \newtheorem{ques}[prop]{Question}

132: \newtheorem{ob}[prop]{Observation}

133: \newtheorem{clm}[prop]{Claim}

134: \newtheorem{Ack}[prop]{Acknowledgments}

135: 

136: %\input tcilatex

137: 

138: \begin{document}

139: \author{Tsachik Gelander}

140: \address{Tsachik Gelander, Institute of Mathematics, Hebrew university, Jerusalem, Israel}

141: \email{tsachik@math.huji.ac.il}

142: %\date{\today}

143: \thanks{Research partially supported by the Clore and the Marie Curie 

144: fellowships.}

145: 

146: \title{Homotopy type and volume of locally symmetric manifolds}

147: \maketitle

148: 

149: 

150: 

151: 

152: 

153: \begin{center}

154: \epsfig{file=b.eps,height=0.5cm}

155: \end{center}

156: 

157: 

158: 

159: 

160: 

161: 

162: \begin{abstract}

163: We consider locally symmetric manifolds with a fixed universal 

164: covering, and construct for each such manifold $M$ a simplicial complex 

165: $\mathcal{R}$ whose size is proportional to the volume of $M$. 

166: When $M$ is non-compact, $\mathcal{R}$ is homotopically equivalent to $M$, 

167: while when $M$ is compact, $\mathcal{R}$ is homotopically equivalent to 

168: $M\setminus N$, where $N$ is a finite union of submanifolds of fairly 

169: smaller dimension.

170: This reflects how the volume controls the topological structure of $M$, and 

171: yields concrete bounds for various finiteness statements which previously had 

172: no quantitative proofs. 

173: For example, it gives an explicit upper bound for the 

174: possible number of locally symmetric manifolds of volume bounded by $v>0$,

175: and it yields an estimate for the size of a minimal presentation for the 

176: fundamental group of a manifold in terms of its volume.

177: It also yields a number of new finiteness results.

178: \end{abstract}

179: 

180: 

181: 

182: 

183: \setcounter{tocdepth}{1}

184: \tableofcontents

185: 

186: 

187: 

188: 

189: %NEW SECTION

190: 

191: 

192: 

193: 

194: 

195: \section{Introduction and statements of the main results}

196: 

197: In this article we study relations between the volume and the topological 

198: structure of locally symmetric manifolds. 

199: We are interested in asymptotic properties, 

200: when the volume tends to infinity. 

201: 

202: \medskip

203: 

204: We shall always fix a symmetric space, $S$, of non-compact type without 

205: Euclidean de-Rham factors, and consider the class of $S$-manifolds, 

206: by which we mean complete Riemannian manifolds locally isometric to $S$, 

207: or equivalently, manifolds of the form

208: $M=\gC\backslash S$ where $\gC$ is a discrete torsion-free group of isometries of $S$.

209: Sometimes we shall restrict our attention to arithmetic manifolds,

210: i.e. to Riemannian manifolds of the form 

211: $M=\gC\backslash S$, where $\gC\leq\textrm{Isom}(S)$ is a torsion-free arithmetic lattice

212: in the Lie group $\textrm{Isom}(S)$ of isometries of $S$.

213: A theorem of Borel and Harish-Chandra \cite{BoHC} says that if $M$ is arithmetic then 

214: $\vol (M)<\infty$. When $\rk (S)\geq 2$, Margulis' arithmeticity theorem \cite{Mar2}

215: gives the converse, i.e. $\vol (M)<\infty$ iff $M$ is arithmetic\footnote{The equivalence between finite

216: volume and arithmeticity is also known for the rank-1 cases $\text{Sp}(n,1)$ and $\text{F}_{4}^{-20}$ 

217: by \cite{GS}, \cite{cor}}. 

218: 

219: The ``complexity'' of the topology of locally symmetric manifolds is

220: controlled by the volume. This is illustrated by a theorem of 

221: Gromov (see \cite{BGS} theorem 2) which asserts  

222: that the Betti numbers are bounded by a constant times the volume, i.e. 

223: $$

224: \sum_{i=1}^{n}b_i(M)\leq c(S)\cdot\vol (M)

225: $$ 

226: for any $S$-manifold $M$.

227: Gromov's theorem applies also to non-locally symmetric manifolds, under appropriate conditions.

228: 

229: We conjecture (and prove in many cases) that, for locally symmetric manifolds,

230: the volume forces stronger topological restrictions. 

231: 

232: \begin{defn}\label{(d,v)-s.c.}

233: A {\bf $(d,v)$-simplicial complex} is a simplicial complex with 

234: at most $v$ vertices, all of them of valence $\leq d$.

235: \end{defn}

236: 

237: \begin{rem}

238: The number of $k$-simplexes in a $(d,v)$-simplicial complex $\mathcal{R}$ is 

239: $\leq\frac{v}{k+1}{n\choose k}$. Thus, the size of $\mathcal{R}$ (the number of its simplexes)

240: is at most $v\cdot\sum_{k=0}^d\frac{1}{k+1}{n\choose k}$, and this depends linearly on 

241: $v$.

242: \end{rem}          

243: 

244: \begin{conj}\label{conjA}

245: For any symmetric space of non-compact type $S$,

246: there are constants $\ga (S),d(S)$, such that any irreducible

247: $S$-manifold $M=\gC\backslash S$ (which is assumed also to be arithmetic in the case

248: $\dim (S)=3$) is homotopically equivalent to a 

249: $\big( d(S),\ga (S)\vol (M)\big)$-simplicial complex.

250: \end{conj}

251: 

252: \begin{rem}

253: The analogous statement is false for non-arithmetic manifolds in dimension $3$.

254: \end{rem}

255: 

256: In this paper we shall establish the following partial answers to Conjecture \ref{conjA}.

257: Since we failed in proving it in full generality, our results cannot be organized in a compact form, and

258: we have to split our statements and formulate strong results under appropriate conditions, and 

259: weaker results for more general cases.

260: 

261: \begin{thm}\label{thmA}

262: Let $S$ be a symmetric space of non-compact type. Then:

263: \begin{enumerate}

264: \item

265: Conjecture \ref{conjA} holds for non-compact arithmetic $S$-manifolds.

266: \item

267: If $S$ is neither isometric to $\SL_3(\BR )/\SO_3(\BR ),~\BH^2\times\BH^2$ nor to

268: $\BH^3$, then for some constants $\ga (S),d(S)$, the fundamental

269: group $\pi_1(M)$ of any 

270: $S$-manifold $M$ is isomorphic to the fundamental group of some

271: $\big( d(S),\ga (S)\vol (M)\big)$-simplicial complex.

272: \item

273: If $S$ is isometric to $\SL_3(\BR )/\SO_3(\BR ),~\BH^2\times\BH^2$ or to

274: $\BH^3$, then the fundamental group $\pi_1(M)$ of any  

275: $S$-manifold $M$ is a quotient of the fundamental group of some

276: $\big( d(S),\ga (S)\vol (M)\big)$-simplicial complex.

277: \end{enumerate}

278: \end{thm}

279: 

280: \begin{rem}

281: For compact arithmetic manifolds, Conjecture \ref{conjA} would follow, 

282: by a straightforward argument, if one could prove that the infimum of the

283: lengths of closed geodesics, taken over all compact arithmetic $S$-manifolds, 

284: is strictly positive. This conjectured phenomenon is strongly related to some 

285: properties of algebraic integers, such as the Lehmer conjecture,

286: which are still a mystery.

287: \end{rem}

288: 

289: Conjecture \ref{conjA} and Theorem \ref{thmA} yield

290: quantitative versions for some classical finiteness 

291: statements. We shall describe our main two applications 

292: in paragraphs \ref{BMP} and \ref{QVWT}. 

293: 

294: 

295: 

296: 

297: 

298: 

299: %subsection

300: 

301: \subsection{A linear bound on the size of a minimal presentation}\label{BMP}

302: 

303: \medskip

304: 

305: It is well known that the fundamental group of a locally symmetric manifold

306: with finite volume, or more generally a lattice in a connected semisimple Lie

307: group, is finitely presented. 

308: The compact case is quite standard, but the non-compact case was proved, step by step, by several authors

309: over several years.

310: Garland and Raghunathan \cite{Ga-Ra} proved it in the rank one case.

311: In the higher rank case, the finite generation was proved by Kazhdan \cite{Kazhdan} 

312: by defining and proving property-$T$ when $\rank (G)>2$,

313: and then by S.P. Wang \cite{S-P.W} when $\rank (G)=2$.

314: Using the finite generation, Margulis \cite{Mar2} proved the classical arithmeticity 

315: theorem for higher rank irreducible lattices\footnote{Margulis proved arithmeticity for higher rank 

316: non-uniform lattices \cite{Mar3}

317: few years before he proved the general case and without using his supper rigidity theorem}.

318: For arithmetic groups the finite presentability follows from the reduction 

319: theory of Borel and Harish-Chandra \cite{BoHC}.  

320: Later, using Morse theory, Gromov (\cite{BGS} theorem 2) gave a geometric proof of the finite 

321: presentability by proving that any locally 

322: symmetric manifold is diffeomorphic to the interior of a compact manifold with 

323: boundary. We remark that a slight modification of the argument of section \ref{6} below yields 

324: a completely elementary proof of this result of Gromov,

325: and hence of finite presentability (see also Remark \ref{finite-presentability}). 

326: However, in order to get a concrete estimate on the minimal possible size of such presentation 

327: in terms of the volume, we shall use arithmeticity (in the higher rank non-compact case).

328: We obtain the following quantitative version of finite presentability:

329: 

330: \begin{thm}\label{thmP}

331: Assume that $S$ is neither isometric to $\BH^3$,

332: $\SL_3(\BR )/\SO_3(\BR )$ nor to $\BH^2\times\BH^2$.

333: Then there is a constant $\eta =\eta (S)$ such that for any irreducible

334: $S$-manifold $M$, the fundamental group $\pi_1(M)$ admits a presentation

335: $$

336:  \pi_1(M)\cong\langle\Sigma :W\rangle

337: $$ 

338: with both $|\Sigma |,|W|\leq\eta\cdot\vol (M)$, in which all the relations 

339: $w\in W$ are of length $\leq 3$. 

340: \end{thm}

341: 

342: \begin{rem}

343: We believe (see also Conjecture \ref{conjA}) that the assumptions that $S$ is not

344: isometric to $\SL_3(\BR )/\SO_3(\BR )$ and to $\BH^2\times\BH^2$, are not really necessary.

345: However, our proof does not work in these cases.

346: It follows, however, from Theorem \ref{thmA}(2) that the analogous statements for non-compact 

347: $S$-manifolds hold also in these cases.

348: \end{rem}

349: 

350: \begin{rem}

351: The analogous statement for non-arithmetic hyperbolic $3$-manifolds is evidently false. 

352: However, if Conjecture \ref{conjA} is true, then the analogue of Theorem \ref{thmP} should

353: hold for arithmetic $3$-manifolds. By Theorem \ref{thmA}, it holds for non-compact arithmetic 

354: $3$-manifolds. Moreover, It was shown in \cite{cooper} that for any hyperbolic $3$-manifold $M$, 

355: the sum of the relations length, in any presentation of $\pi_1(M)$, is at 

356: least $\frac{\vol (M)}{\pi}$. This means that our upper bound is tight in 

357: this case.

358: \end{rem}

359: 

360: In the general case, Theorem \ref{thmA}(3) implies the following (weaker)

361: statement for which even the finiteness is in some sense surprising (in dimension $3$).

362: For a group $\gC$ let $d(\gC )$ denotes the minimal size of a generating set.

363: 

364: \begin{thm}\label{d(gC)}

365: For any $S$, there is a constant $\eta$ (depending on $S$), 

366: such that for any $S$-manifold $M$, $d\big(\pi_1(M)\big)\leq \eta\cdot\vol (M)$.

367: In other words, for any $v>0$, 

368: $$

369:  \sup\{d\big(\pi_1(M)\big) :\vol (M)\leq v\}\leq \eta v.

370: $$

371: Moreover, if $S$ is not isomorphic to $\SL_3(\BR )/\SO_3(\BR )$ then there is

372: a presentation

373: $$

374: \pi_1(M)\cong\langle\Sigma :W\rangle

375: $$ 

376: with $|\Sigma |,|W|\leq\eta\cdot\vol (M)$.

377: \end{thm} 

378: 

379: Note that Theorem \ref{d(gC)} does not give a bound on the length of the relations $w\in W$.

380: 

381: 

382: 

383: 

384: 

385: 

386: 

387: %subsection

388: 

389: \subsection{A quantitative version of Wang's theorem}\label{QVWT}

390: 

391: \medskip

392: 

393: We apply our results in order to estimate the number of $S$-manifolds 

394: with bounded volume (or more generally the number of conjugacy classes of

395: lattices in $G=\text{Isom}(S)$). 

396: This can be considered as a continuous analogue to

397: asymptotic group theory which studies the subgroup growth

398: of discrete groups (where ``covolume of lattices'' extends the notion ``index of subgroups''). 

399: Unlike the situation in the discrete (finitely generated) case, even the 

400: finiteness statements are not clear, and in general not true. 

401: However, a classical theorem of H.C. Wang states that if $S$ is not isometric 

402: to one of the hyperbolic spaces $\BH^2,\BH^3$, then for any $v>0$ there 

403: are only finitely many irreducible $S$-manifolds with total volume $\leq v$, 

404: up to isometries (see \cite{Wa} 8.1, and paragraph \ref{SL_2} below). 

405: We remark that Wang's result and proof do not give explicit estimates.

406: 

407: Denote by $\rho_S(v)$ the number of non-isometric irreducible $S$-manifolds

408: with volume $\leq v$.

409: 

410: By Mostow's rigidity theorem, a locally symmetric manifold of dimension 

411: $\geq 3$ is determined by its fundamental group. Applying 

412: Theorem \ref{thmA}(2), we obtain: 

413: 

414: \begin{thm}

415: If $\dim (S)\geq 4$ and $S$ is neither isometric to $\SL_3(\BR )/\SO_3(\BR )$ 

416: nor to $\BH^2\times\BH^2$ then there is a constant $c$, depending on $S$, with

417: respect to which

418: $$

419: \log\rho_S(v)\leq c\cdot v\log v

420: $$ 

421: for any $v>0$.

422: \end{thm}

423: 

424: This upper bound was first proved for hyperbolic manifolds by M. Burger, A. 

425: Lubotzky, S. Mozes and the author in \cite{BGLM}, where also a lower bound of

426: the same type was established, proving that this estimate is the true 

427: asymptotic behavior in the hyperbolic case. 

428: 

429: \begin{thm}[BGLM]

430: For $n\geq 4$, there are constants $c_n>b_n>0$ and $v_n>0$ such that

431: $$

432:  b_nv\log v\leq \log\rho_{\BH^n}(v)\leq c_nv\log v,

433: $$

434: whenever $v>v_n$.

435: \end{thm}

436: 

437: However, we suspect that in the higher rank case, where all manifolds are

438: arithmetic and conjectured to possess the congruence subgroup property, 

439: this upper bound is far from the true asymptotic behavior.  

440: It seems, in view of \cite{Lub}, that the problem of determining the real

441: asymptotic behavior is closely related to the congruence subgroup problem.

442: It is also related to the analysis of the Galois cohomology of compact 

443: extensions of $G=\text{Isom}(S)$.

444: 

445: For hyperbolic manifolds, the weaker upper bound  

446: $\rho_{\BH^n}(v)\leq v\exp (\exp (\exp (v+n)))$ was proved previously by Gromov \cite{Gr1}.

447: In this paper we establish a first concrete estimate for $\rho_S(v)$ for general $S$.

448: 

449: In dimension $2$ and $3$ the analogue of Wang's finiteness theorem is false.

450: However, as was shown by Borel \cite{Bor}, it remains true when 

451: considering only arithmetic manifolds.

452: The following estimate for the number of non-compact arithmetic $3$-manifolds

453: follows from Theorem \ref{thmA}(1):

454: 

455: \begin{prop}\label{116}

456: For some constant $c>0$, there are at most $v^{cv}$ non-isometric arithmetic 

457: non-compact hyperbolic 3-manifolds with volume $\leq v$.

458: \end{prop}

459: 

460: For compact arithmetic $3$-manifolds we conjecture that the analogous 

461: statement holds, but prove only the following weaker statement:

462: 

463: 

464: \begin{prop}\label{117}

465: Let $M$ be a compact (arithmetic) hyperbolic 3-manifold.

466: Then for some constant $c(M)$, the number of non-isometric hyperbolic

467: manifolds commensurable to $M$, with volume $\leq v$, is at most $v^{c(M)v}$.

468: \end{prop}

469: 

470: We remark that the analogue of Propositions \ref{116} and \ref{117} hold also in 

471: the cases $S=\SL_3(\BR )/\SO_3(\BR )$ and $S=\BH^2\times\BH^2$.

472: 

473: \medskip

474: 

475: In section \ref{complements}, we shall generalize some of our results concerning $S$-manifolds

476: to the larger family of $S$-orbifolds. We shall also indicate how to construct triangulations

477: for rank-1 manifolds which are not necessarily arithmetic.

478: 

479: \medskip

480: 

481: Let us now give a short and not precise explanation of the basic lines of the proofs of our main 

482: results \ref{thmA}.

483: The idea is to construct, inside each $S$-manifold $M$, a submanifold with boundary $M'$, which is

484: similar enough to $M$ and for which we can construct a triangulation of size $\leq c\cdot\vol (M)$.

485: In order to construct such a triangulation we shall require a lower bound on the injectivity radius

486: of $M'$ (independent of $M$) and some bounds on the geometry of the boundary $\partial M'$.

487: We shall construct $M'$ inside the $\gep_s$-thick part, in a way that its pre-image in the 

488: universal cover $S$ will be the complement of some ``locally finite'' union of convex sets, each

489: has a smooth boundary whose curvature is bounded uniformly from below, and the angles at the 

490: corners of $\partial M'$ (where the boundaries of two or more such convex sets meet) are bounded 

491: uniformly from below.

492: 

493: The non-compact arithmetic case is easier to deal with. The reason is that any non-uniform 

494: arithmetic lattice $\gC\leq G=\text{Isom}(S)^0$ comes from a rational structure on $G$ (rather then

495: on a compact extension of $G$ as the case might be when $\gC$ is uniform). 

496: This implies that there is 

497: some constant $\gep_s'$ such that any non-uniform arithmetic $S$-manifold contains no closed 

498: geodesics of length $\leq\gep_s'$ (see section \ref{AML}). In other words, any non-trivial closed 

499: loop which is short enough corresponds to a unipotent element in the fundamental group.

500: We define (the pre-image in $S$ of) $M'$ to be the complement of the union of appropriate

501: sub-level sets for the displacement functions $\{ d_\gc\}$ where $\gc$ runs over all unipotents

502: in $\pi_1(M)$. The injectivity radius in $M'$ is then uniformly bounded from below,

503: and using the fact that unipotents acts nicely on $S$ and on $S(\infty )$ we can 

504: both estimate the 

505: geometry of $\partial M'$, and also construct a deformation retract from $M$ to $M'$.

506: 

507: The situation is more subtle in the compact case. We do not have sufficient information

508: on the thick-thin decomposition. 

509: In this case we construct $M'$ and prove that it is diffeomorphic to $M\setminus N$ where $N$ is a 

510: finite union of submanifolds of codimension $\geq 3$. $N$ will contain the union of all 

511: closed geodesics of length smaller than some fixed constant. 

512: The idea is that at any point outside $N$

513: there is a preferred direction, such that when we move along it, the injectivity radius is

514: increasing most rapidly. This will help us to define a deformation retract from $M\setminus N$ to 

515: $M'$. As $N$ has codimension $\geq 3$, $\pi_1(M)\cong\pi_1(M\setminus N)\cong\pi_1(M ')$.

516: A main difficulty (which arises also in the compact rank one case) is how to control the geometry 

517: of the boundary of $M'$. We shall handle this difficulty in section \ref{bigangles}, where the 

518: main idea is Lemma \ref{bigA} which says that if two isometries commute then the exterior angle 

519: between their sub-level sets is $\geq\frac{\pi}{2}$.

520: 

521: \medskip

522: 

523: The present paper generalizes the work \cite{BGLM} which treated the special case of real 

524: hyperbolic spaces. Although some of the ideas from \cite{BGLM} appear again in the present paper, 

525: the situation in the general case is significantly more complicated than the hyperbolic case. 

526: The proof in \cite{BGLM} uses the explicit description of a hyperbolic compact 

527: thin component of the thick-thin decomposition as a cone over a coaxial Euclidean ellipsoids, 

528: as well as some computations in constant curvature. Hence the

529: argument in \cite{BGLM} does not apply to more general rank-1 symmetric spaces.

530: More crucially, in contrast to the situation in the rank-1 case, in the higher rank case

531: there is currently no good understanding of the structure of the thin components in the 

532: thick-thin decomposition, and hence new ingredients are required also in the skeleton of the proof.

533: 

534: Most of this paper is devoted to the treatment of the higher rank case. 

535: However, whenever it is not required, we shall not make

536: any assumption on the rank. We remark also that many of the arguments in this paper stay valid for

537: general manifolds of non positive curvature. In particular, some parts of this work could be 

538: generalized to non symmetric Hadamard spaces.

539: 

540: 

541: 

542: 

543: 

544: 

545: 

546: 

547: 

548: %NEW SECTION

549: 

550: 

551: 

552: 

553: \section{Notations, definitions and background}

554: In this section we shall fix our notations and

555: summarize some basic facts about semisimple Lie groups, symmetric spaces of 

556: non-compact type, and manifolds of non-positive curvature. For a comprehensive

557: treatment of these subjects we refer the reader to \cite{Rag}, \cite{BGS} and \cite{BH}.  

558: 

559: Let $S$ be a symmetric space of non-compact type. We shall always assume, 

560: that $S$ has no Euclidean 

561: de Rham factors. Let $G=\textrm{Isom}(S)$ be the Lie group of isometries of $S$. $G$ is 

562: center-free, semi-simple without compact factors, and with finitely many connected components. 

563: We denote by $G^0$ the identity component of $G$ with respect to the real 

564: topology. There is an algebraic group $\Bbb{G}$ defined over $\BQ$, such that

565: $G^0$ coincides with the connected component of the group of real points $\Bbb{G}(\BR )^0$.

566: In particular $G^0$ admits, apart from the real topology, a Zariski topology

567: which is defined as the trace in $G^0$ of the Zariski topology in $\Bbb{G}$.

568: $G^0$ acts transitively on $S$, and  we can identify $S$ with $G^0/K$ where $K\leq G^0$ is

569: a maximal compact subgroup. We remark that there is a bijection between

570: symmetric spaces of non-compact type and connected center free semisimple Lie 

571: groups without compact factors.

572: 

573: $S$ is a Riemannian manifold with non-positive

574: curvature such that for each point $p\in S$ there is an isometry $\sigma_p$

575: of $S$ which stabilizes $p$ and whose differential $d_p(\sigma_p)$ at $p$ is $-1$. 

576: The composition of two such isometries $\sigma_p\cdot \sigma_q$ is called a 

577: {\bf transvection}, and is belong to $G^0$.

578: The non-positivity of the sectional curvature means that the distance 

579: function $d:S\times S\to \BR ^+$ is convex and, in fact, its restriction to 

580: a geodesic line $c(t)=\big( c_1(t),c_2(t)\big)$ in $S\times S$ is strictly 

581: convex, unless $c_1$ and $c_2$ are contained in a flat plane in $S$.

582: 

583: For a real valued function $f:X\to\BR$ on a set $X$ we denote by

584: $\{ f<t\}$ the $t$-sub-level set

585: $$

586:  \{ f<t\}=\{ x\in X:f(x)<t\}.

587: $$

588: 

589: For $\gc \in G$ we denote by $d_{\gc}$ the {\bf displacement function}

590: $$

591:   d_{\gc}(x)=d(\gc \cdot x,x).

592: $$

593: This function is convex and smooth outside Fix$(\gc )$. 

594: In particular the sub-level sets $\{ d_{\gc} <t\}$ are convex with smooth boundary.

595: We denote by {\bf min}$(\gc )$ the set 

596: $$

597:  \min (\gc )=\{x\in S:d_{\gc}(x)=\inf d_{\gc}\}.

598: $$

599: 

600: For a set $A\subset S$ we denote by 

601: $$

602:  D_A(x)=d(A,x)

603: $$ 

604: the {\bf distance function}.

605: If $A$ is a convex set then the function $D_A$ is convex and smooth at any

606: point outside the boundary of $A$. We denote its $t$-sub-level set by

607: $$

608:  (A)_t=\{ D_A<t\} 

609: $$

610: and call it the $t$-{\bf neighborhood} of $A$. 

611: Note that $(A)_{t+s}=\big( (A)_t\big)_s$.

612: Similarly, we let $)A(_t$ denote the $t$-{\bf shrinking} of $A$

613: $$

614:  )A(_t=S\setminus (S\setminus A)_t.

615: $$

616: If $A$ is convex then $\big) (A)_t\big(_t= A$ but in general these are 

617: different sets. 

618: 

619: A subset $\mathcal{N}$ of a metric space $X$ is called an {\bf $\gep$-net}, if

620: $(\mathcal{N})_\gep =X$. A subset $\mathcal{N}\subset X$ is called 

621: {\bf $\gep$-discrete} is $d(x,y)>\gep$ for any $x,y\in\mathcal{N},~x\neq y$.

622: If $\mathcal{N}$ is a maximal $\gep$-discrete subset, then it is an $\gep$-net.

623: 

624: If $A$ is closed and convex then for any $x\in S$ there is a unique closest 

625: point $p_A(x)$ in $A$. The map $p_A:S\to A$ is called the {\bf projection} 

626: on $A$, and it is distance decreasing, i.e.

627: $d\big(p_A(x),p_A(y)\big)\leq d(x,y)$ for any $x,y\in S$.

628: 

629: A {\bf flat subspace} is a totally geodesic sub-manifold which is 

630: isometric to a Euclidean space. A {\bf flat} is a maximal flat subspace. 

631: Any flat subspace is contained in a flat. 

632: All flats in $S$ have the same dimension

633: $\textrm{rank}(S)=\textrm{rank}(G)$\footnote{By $\textrm{rank}(G)$ we shall 

634: always mean the real rank of $G$}. A geodesic $c\subset S$ is called 

635: {\bf regular} if it is contained in a unique flat.

636: 

637: An element $\gc \in G$ is unipotent if $\text{Ad}(\gc )-1$ is a nilpotent 

638: endomorphism. A subgroup of $G$ is called a {\bf unipotent subgroup} if all its 

639: elements are unipotent. If $H\leq G^0$ is a unipotent subgroup then its 

640: Zariski closure $\overline{H}^z\leq\Bbb{G}$ is also unipotent, and 

641: $\overline{H}^z_\BR$ is connected. Moreover, a unipotent subgroup $H\leq G^0$ 

642: is connected iff $H=\overline{H}^z_\BR$. 

643: 

644: $S(\infty )$ is the set of equivalence classes of geodesic rays, where two

645: rays are considered equivalent if their traces are of bounded Hausdorff 

646: distance from each other. 

647: There is a structure of a spherical building on $S(\infty )$ - the Tits 

648: building of $S$. The chambers of this building are sometimes called Weyl 

649: chambers. A maximal unipotent subgroup of $G^0$ is the unipotent radical of 

650: some minimal parabolic subgroup. The minimal parabolic subgroups of $G^0$ are 

651: the stabilizers (and the fixator) of the chambers of this building, and there

652: are canonical bijections between the set of minimal parabolic subgroups, the 

653: set of maximal unipotent subgroups and the set of chambers. All maximal 

654: unipotent subgroups are conjugate in $G^0$, or in other words, $G^0$ acts 

655: transitively on the set of chambers $W\subset S(\infty )$. Moreover, each 

656: parabolic subgroup acts transitively on $S$, and hence $G$

657: acts transitively on the set of couples $(W,x)$ of a chamber 

658: $W\subset S(\infty )$ and a point $x\in S$.

659: There is a canonical metric on $S(\infty )$, called the Tits metric, with 

660: respect to which the apartments of the Tits building are isometric to spheres

661: and the chambers are isometric to spherical simplices. The induced action of 

662: $G$ on $S(\infty )$ preserves the Tits metric.

663: A point $p\in S(\infty )$ is called {\bf regular} if it is an interior point

664: of a chamber. For any $p\in S(\infty )$ and $x\in S$ there is a unique geodesic

665: $c$ with $c(0)=x$ and $c(\infty )=p$, and $p$ is regular iff $c$ is regular.

666: 

667: When $\gC$ is a group of isometries of $S$, we say that a subset $A\subset S$ 

668: is $\gC$-{\bf precisely invariant} if $\gc \cdot A=A$

669: whenever $\gc \cdot A \cap A\neq \emptyset$.   

670: If $\gC$ acts freely and discretely (i.e. $\gC \subset G$ is discrete

671: and torsion free) then we denote by $\gC\backslash A$ the image of $A$ in $\gC\backslash S$.

672: If $A$ is a connected simply connected $\gC$-precisely invariant set

673: then 

674: $$

675:  \pi_1(\gC\backslash A )\cong \{ \gc \in \gC :

676:  \gc \cdot A=A\}. 

677: $$

678: 

679: For a subset $M'$ of $M=\gC\backslash S$ we let $\ti M'$ denote its pre-image in $S$, and

680: we shall usually denote by $\ti{M'}^0$ an arbitrary connected component of 

681: $\ti M'$.  

682: 

683: For a subgroup $\gC\leq G$, a constant $\gep >0$ and a point $x\in S$,  

684: we let $\gC_\gep (x)$ denote the group generated by the ``small elements'' 

685: at $x$

686: $$

687:  \gC_\gep (x)=\langle \gc\in\gC :d_\gc (x)\leq\gep\rangle.

688: $$

689: Recall the classical Margulis lemma:

690: 

691: \begin{thm}[{\bf The Margulis lemma}]

692: There are constants $\gep_s>0$ and $n_s\in\BN$, depending only on $S$,

693: such that for any discrete subgroup $\gC\leq G$, and any point $x\in S$,

694: the group $\gC_{\gep_s}(x)$ contains a subgroup 

695: of index $\leq n_s$ which is contained in a connected nilpotent subgroup of 

696: $G$.

697: \end{thm}

698: 

699: The {\bf $\gep$-thick thin} decomposition of an $S$-manifold $M=\gC\backslash S$ reads 

700: $$

701:  M=M_{\geq\gep}\cup M_{\leq\gep}.

702: $$

703: The {\bf $\gep$-thick part} $M_{\geq\gep}$ is defined as the set of all 

704: points $x\in M$ at which the injectivity radius is $\geq\frac{\gep}{2}$, 

705: and the {\bf $\gep$-thin part} $M_{\leq\gep}$

706: is the complement of the interior of the $\gep$-thick part.

707: Note that $M_{\leq\gep}$ is the set of points in $M$ through which there is a

708: non-contractible closed loop of length $\leq\gep$.

709: The pre-image of $M_{\leq\gep}$ in $S$ has a nice description as a 

710: locally finite union of convex sets

711: $$

712:  \ti M_{\leq\gep}=\cup_{\gc\in\gC\setminus\{ 1\}}\{ d_\gc\leq\gep\}.

713: $$

714: The Margulis lemma yields an important piece of information on the structure 

715: of the $\gep$-thick-thin decomposition when $\gep\leq\gep_s$.

716: 

717: If $\gC \subset G$ is a uniform lattice then any element 

718: $\gc \in \gC$ is semisimple, i.e. $\text{Ad}(\gc )$ is diagonalizable over 

719: $\BC$, or equivalently $\min (\gc )\neq \emptyset$. If, in addition, $\gC$ is 

720: torsion free then all its elements are hyperbolic.

721: A semisimple element is hyperbolic iff it has a complex eigenvalue outside the

722: unit disk.

723: A hyperbolic isometry $\gc$ has an axis, i.e. a geodesic line on which

724: $\gc$ acts by translation. Any two axes  are parallel, and 

725: $x\in \min (\gc )$ iff the geodesic line $\overline{x,\gc \cdot x}$

726: is an axis of $\gc$. 

727: We define a {\bf regular isometry} to be a hyperbolic isometry whose axes are regular 

728: geodesics (if one axis is a regular geodesic, then so are all of them, since axes are parallel). 

729: If $g\in G$ acts

730: as a regular isometry then $g$ is a regular element in $G$ in the usual sense,

731: but not necessarily vice versa.

732: If $\gc\in G$ acts as a regular isometry, and $F$ is the unique flat containing

733: an axis of $\gc$ then $\gc$ preserves $F$ and acts by translation on it, and

734: in particular $\min (\gc )=F$.

735: (This follows for example from the facts that the Weyl group 

736: $N_{G^0}(A)/A$, for $A=C_{G^0}(\gc )$, acts simply transitively on the set of Weyl chambers 

737: in $F(\infty )$, and that a

738: regular point $p\in F(\infty )$ determines its ambient Weyl chamber.)

739: 

740: If $A\subset S$ is a closed convex set, and $\ga\in G$ preserves $A$, i.e.

741: $\ga\cdot A=A$, then the projection $P_A:S\to A$ commutes with $\ga$, and 

742: since $P_A$ does not increase distances, we have

743: $d_\ga \big( P_A(x)\big)\leq d_\ga(x)$. In particular, $\ga$ is semisimple iff

744: $\min (\ga )\cap A\neq\emptyset$. Since for any isometry $i$ of a flat 

745: subspace $F$, $\min (i)\neq\emptyset$, it follows that if $\ga\in G$ preserves

746: a flat subspace then $\ga$ is semisimple. If $g\in G^0$ is semisimple, then

747: its centralizer group $C_{G^0}(g)$ is a closed reductive group.

748: If $g\in G$ is a transvection with axis $c$, then

749: $$

750:  C_{G^0}(g)=\{ h\in G^0:h\cdot c \text{ is parallel to } c\}.

751: $$

752:    

753: If $\gC \subset G$ is a non-uniform lattice, then it is almost generated by

754: unipotents. $\gc$ is unipotent iff $\text{inf}(d_\gc )=0$ while $d_\gc$ never 

755: vanishes on $S$. In particular a unipotent element stabilizes a point at infinity but not in $S$. 

756: 

757: 

758: If $f:G^0\to H$ is a surjective Lie homomorphism, and $g\in G$ is semisimple

759: (resp. unipotent) then $f(g)$ is semisimple (resp. unipotent).

760: 

761: 

762: 

763: 

764: 

765: 

766: 

767: %NEW SECTION

768: 

769: 

770: 

771: \section{Some deformation retracts}\label{3}

772: 

773: In several places in this paper we will need to produce a deformation 

774: retract from a submanifold (maybe with boundary) to a subset of it which

775: is defined by the condition that all the functions from some given family

776: are larger than a given constant. A typical example of such a situation is 

777: given by the manifold itself and the $\gep$-thick part.

778: The family of functions in this example is indexed by the fundamental group.

779: To each homotopy class of closed loops corresponds the function for which 

780: the value at any point  $x$ is the minimal length of a loop from the class, 

781: which passes through $x$. The $\gep$-thick part is exactly the subset where 

782: all these functions are $\geq\gep$. The Margulis lemma yields information 

783: about the set of functions which are $<\gep$ at $x$, which can sometimes be exploited in 

784: order to define a deformation retract to the $\gep$-thick part.

785: Another example is given by a subset $M'$

786: of the manifold $M$ and its $\gep$-shrinking $)M'(_\gep$. If the complement of $M'$ is 

787: given by the union of subsets satisfying certain properties,

788: it is sometimes convenient to use the distance functions from these subsets

789: in order to define a deformation retract to $)M'(_\gep$.

790: 

791: The idea is to construct a continuous vector

792: field which makes an acute angle at any point $x$ with the gradients of all the

793: functions from the given family which are small at $x$, and then to let points

794: flow along its integral curves.

795: This idea has been used in an elegant way in other places (c.f. \cite{BGS}). The situation concerned here, however,

796: is more subtle since we consider manifolds with boundary, and have to make sure that on the 

797: boundary, the vector field is pointing towards the interior.

798: 

799: \begin{defn}

800: For a family of real valued continuous functions 

801: $\mathcal{F} =\{\phi_i\}_{i\in I}$ on

802: a manifold $Y$ (with or without boundary) and $\gep >0$, we define the {\bf  

803: $(\mathcal{F},\gep )$-thick part} (or simply the {\bf $\mathcal{F}$-thick part) 

804: $\mathcal{F}_{\geq\gep}$} as follows

805: $$

806:  \mathcal{F}_{\geq\gep}=\cap_{\phi\in\mathcal{F}}\overline{\{\phi >\gep\}}.

807: $$

808: When $\mathcal{F}$ is given, we denote by 

809: $\Psi_{x,\gt }~(x\in Y,\gt\in\BR )$ the subset

810: $$

811:  \Psi_{x,\gt } =\{\phi\in\mathcal{F} : \phi (x)\leq\gt\}.

812: $$

813: Abusing notations, we shall sometimes write $\Psi_{x,\gt }$ for the 

814: corresponding set of indices

815: $\Psi_{x,\gt } =\{i\in I: \phi_i(x)\leq\gt\}$.

816: We say that the family $\mathcal{F}$ is {\bf locally finite} if

817: $\Psi_{x,\gt }$ is finite for any $x\in Y$ and $\gt\in\BR$. 

818: We say that $\gt\in\BR$ is a {\bf critical value} of a continuous function

819: $\phi$ if $\overline{\{\phi >\gt\}}\neq \{\phi\geq\gt\}$, i.e. if $\gt$ is 

820: a value of a local maximum of $\phi$.

821: \end{defn}

822: 

823: If $Z\subset Y$ is a submanifold with smooth boundary, and $\phi$ a real 

824: valued function on $Z$. The gradient $\nabla\phi$ is well defined on $Z$ 

825: wherever it is unique (on 

826: the boundary it is the tangent vector with length and direction equal to the 

827: value of the maximal directional 

828: derivative and the direction at which it occurs). The function $\phi$ is said 

829: to be smooth (or $C^1$) if its gradient $\nabla\phi$ is a continuous function on the 

830: whole of $Z$. When $\phi$ is smooth, its directional derivative with respect to

831: $v\in T_x(Y)$ is given by the usual formula $v\cdot\nabla\phi (x)$.

832:  

833: In later sections, we will consider families $\mathcal{F}$ which consist of functions of the 

834: following 2 types:

835: \begin{itemize}

836: \item 

837: The displacement function $d_\gc (x)=d(x,\gc\cdot x)$ associated with

838: a non-trivial isometry $\gc$ of $S$.

839: \item 

840: The distance function $D_A(x)=d(x,A)$ from a given closed convex set with

841: smooth boundary.

842: \end{itemize}

843: In both cases the function is convex, non-negative, and without positive 

844: critical values. The function $d_\gc$ is smooth on $\{ d_\gc >0\}$, and the 

845: function $D_A$ is smooth on $\overline{\{ D_A>0\}}$.

846: 

847: \medskip

848: 

849: %Let us now formulate a precise statement.

850: 

851: \begin{lem}[\bf A deformation retract which increases functions]

852: \label{deformation-retract}

853: Consider a submanifold without boundary $Y\subset S$.

854: \begin{itemize}

855: \item 

856: Let $\mathcal{F} =\{\phi_i\}_{i\in I}$ be a locally finite

857: family of non-negative continuous functions on $Y$.

858: \item

859: Assume that for each $\phi\in\mathcal{F}$ with $\{\phi =0\}\neq\emptyset$, the set 

860: $\overline{\{\phi >0\}}$ is a submanifold with smooth boundary.

861: \item

862: Assume that each $\phi\in\mathcal{F}$ is $C^1$ on $\overline{\{\phi >0\}}$.

863: \item 

864: Let $L=\mathcal{F}_{\geq 0}=\cap_{\phi\in\mathcal{F}}\overline{\{\phi >0\}}$.

865: \item

866: Let $\gb :[0,3\gep ]\to\BR^{>0}$ be a given continuous positive function 

867: (in case all $\phi\in\mathcal{F}$ are strictly positive we allow $\gb$ to

868: be defined only on $(0,3\gep ]$).

869: \item

870: Assume that $\gep$ is not a critical value for any $\phi\in\mathcal{F}$.

871: \item

872: Assume that for any $x\in Y$    

873: there is a unit tangent vector $\hat n(x)\in T_x(S)$ such that  

874: $$

875:  \hat n(x)\cdot \nabla \phi (x)\geq\gb\big( \phi (x)\big)

876: $$

877: for any $\phi\in\Psi_{x,3\gep}$. 

878: \end{itemize}

879: Then there is a deformation retract from $L$ to the $\mathcal{F}$-thick 

880: part $\mathcal{F}_{\geq\gep}$. 

881: (In particular it follows that $\mathcal{F}_{\geq\gep}\neq\emptyset$.)

882: 

883: If $\gC\leq\text{Isom}(S)$ is a discrete subgroup, and if $L$ and the family 

884: $\mathcal{F}$ are $\gC$-invariant, in the sense that $\gc\cdot L=L$, and

885: the function $\gc\cdot\phi (x):=\phi(\gc^{-1}\cdot x)$ belongs to $\mathcal{F}$ for any

886: $\phi\in\mathcal{F},~\gc\in\gC$,

887: then there exists such a deformation retract

888: which is $\gC$-invariant (in the obvious sense)

889: and hence induces a deformation retract between 

890: the corresponding subsets $\gC\backslash L$ and 

891: $\gC\backslash\mathcal{F}_{\geq\gep}$ of $M=\gC\backslash S$.

892: \end{lem}

893: 

894: \begin{proof}

895: We will define an appropriate continuous vector field on $L$. 

896: The desired deformation retract will be the flow along this vector field.

897: 

898: Let $\gd (x)$ denote 

899: $$

900:  \gd (x)=\min_{\phi\in \mathcal{F}}\phi (x).

901: $$

902: 

903: For any non-empty subset $\Psi\subset\mathcal{F}$ for which 

904: $\cap_{\phi\in\Psi}\{\phi\leq 3\gep\}\neq\emptyset$ and any point $x$ of this intersection, let 

905: $\hat f(x,\Psi )\in T_x(S)$ be a unit tangent vector which maximizes the 

906: expression

907: $$

908:  \min\{\hat f\cdot \nabla\phi (x):\phi\in\Psi\},

909: $$

910: and let 

911: $$

912:  \gb_\Psi (x)=\min_{\phi\in\Psi}\gb\big(\phi (x)\big).

913: $$

914: Then for any $\phi\in\Psi$

915: \begin{eqnarray*}

916:  \hat f(x,\Psi )\cdot \nabla \phi (x)\geq

917:  \min_{\phi\in\Psi}\hat f(x,\Psi )\cdot \nabla \phi (x)\geq

918:  \min_{\phi\in\Psi}\hat n(x)\cdot \nabla \phi (x)\geq

919:  \min_{\phi\in\Psi}\gb\big(\phi (x)\big)=

920:  \gb_\Psi (x).

921: \end{eqnarray*} 

922: Moreover, it follows from the strict convexity of the Euclidean unit disk that

923: $\hat f(x,\Psi )$ is uniquely determined, and consequently, that for a fixed

924: $\Psi$, the vector field $\hat f(x,\Psi )$ is continuous on

925: $\cap_{\phi\in\Psi}\{\phi\leq 3\gep\}$.

926: 

927: The desired vector field is defined as follows:

928: \begin{eqnarray*}

929:  &&\!\!\!\!\!\!\! \overrightarrow V(x)=

930:  \sqrt{2\big(\gep -\gd(x)\big)\lor 0}\cdot\\

931: &\phantom{\leq}&

932:  \cdot\sum_\Psi

933:  \frac{\big( 3\gep -\max_{\phi\in\Psi}\phi (x) \big)\lor 0}{\gep}\cdot

934:  \frac{\big(\min_{\phi\notin\Psi}\phi (x)-\gep \big)\lor 0}{\gep}\cdot

935:  \frac{1}{\gb_\Psi(x)}\hat f(x,\Psi ),

936: \end{eqnarray*}

937: where the sum is taken over all non-empty finite subsets 

938: $\Psi \subset\mathcal{F}$.

939: 

940: Notice that $\gb_\Psi(x)$ is strictly positive, and 

941: all the terms in each summand are continuous, and $\overrightarrow V\equiv 0$

942: on the $\mathcal{F}$-thick part

943: $$

944:  \mathcal{F}_{\geq\gep}=

945:  \cap_{\phi\in\mathcal{F}}\overline{\{\phi >\gep\}}=

946:  \cap_{\phi\in\mathcal{F}}\{\phi\geq\gep\}=

947:  \{\gd\geq\gep\}.

948: $$

949: The term $\sqrt{2\big(\gep -\gd(x)\big)\lor 0}$ takes care of the continuity

950: on the boundary $\{\gd =\gep\} =\partial\{\gd \leq\gep\}$ of the 

951: $\mathcal{F}$-thin part. 

952: The terms $\frac{\big( 3\gep -\max_{\phi\in\Psi}\phi (x) \big)\lor 0}{\gep}$ 

953: guarantee that all the non-zero summands correspond to sets which are 

954: contained in the finite set $\Psi_{x,3\gep}$. In particular the summation is 

955: finite for any $x\in L$, and $\hat f(x,\Psi )$ is well defined for any 

956: non-zero summand.

957: The terms $\frac{\big(\min_{\phi\notin\Psi}\phi (x)-\gep \big)\lor 0}{\gep}$ 

958: guarantee that all the non-zero summands correspond to $\Psi$'s which contains

959: $\Psi_{x,\gep}$.   

960: 

961: If $\gd (x)<\gep$ then $\Psi_{x,\gep}\neq\emptyset$, and, when looking only on the summand 

962: corresponding to $\Psi_{x,2\gep}$, we see that

963: $$

964:  \nabla \phi (x)\cdot\overrightarrow V(x)\geq\sqrt{2\big(\gep -\gd(x)\big)}

965: $$

966: for any $\phi\in\Psi_{x,\gep}\subset\Psi_{x,2\gep}$, and in particular for any 

967: $\phi$ with $\phi (x)=\gd (x)$.

968: 

969: It follows that if $x(t)$ is an integral curve of $\overrightarrow V$ with

970: $\gd\big( x(0)\big) <\gep$, then

971: $$

972:  \frac{d}{dt}\Big(\gd\big( x(t)\big)\Big)\geq 

973:  \sqrt{2\Big(\gep -\gd\big( x(t)\big)\Big)}.

974: $$

975: (To be more precise, since $\gd\big( x(t)\big)$ is not necessarily 

976: differentiable, we should write 

977: $\liminf_{\gt\to 0}\frac{\gd\big( x(t+\gt)\big)-\gd\big( x(t)\big)}{\gt}$ 

978: instead of $\frac{d}{dt}\Big(\gd\big( x(t)\big)\Big)$ in the last inequality.)

979: Thus, for $t=\sqrt{2\Big(\gep -\gd\big( x(0)\big)\Big)}$ we have

980: $\gd\big( x(t)\big) =\gep$. 

981: 

982: Since $x\in L$ belongs to $\partial L$ iff $\gd (x)=0$, and hence, 

983: the vector field 

984: $\overrightarrow V$ points everywhere towards the interior $\textrm{int}(L)$,

985: it follows from the Peano existence theorem of solution for ordinary 

986: differential equations, that for any $x\in L$ there is an integral curve 

987: $c_x(t)$ of $\overrightarrow V$, defined for all $t\geq 0$ with $c_x(0)=x$ and 

988: with $c_x(t)\in\textrm{int} L$ for $t>0$. Stability of the solutions implies

989: that $c_x(t)$ depends continuously on $x$ and $t\geq 0$.

990: 

991: As a conclusion, we get that the flow along 

992: $\overrightarrow V$ for

993: $\sqrt{2\gep}$ time units defines a deformation retract from $L$ to 

994: $\mathcal{F}_{\geq\gep}=\{\gd\geq\gep\}.$

995: 

996: If $L$ and $\mathcal{F}$ are

997: $\gC$-invariant, then $\overrightarrow V(x)$, as it is defined above, is also 

998: $\gC$-invariant. Hence it induces a vector field on $\gC\backslash L$, 

999: and a deformation retract from $\gC\backslash L$ to $\gC\backslash\mathcal{F}_{\geq\gep}$.

1000: \end{proof}

1001:  

1002: A second kind of deformation retracts that we shall often use is the following.

1003: Let $A\subset S$ be a closed convex set, and $B\subset S$ a set containing $A$.

1004: We say that $B$ is {\bf star-shaped} 

1005: with respect to $A$ if for any $b\in B$ the geodesic segment 

1006: connecting $b$ to its closest point (the projection) in $A$ is contained 

1007: in $B$. In that case we can define a deformation retract from $B$ to $A$ by 

1008: moving $b$ at a constant rate (equal to the initial distance) 

1009: towards its projection in $A$. We call this the {\bf star-contraction} 

1010: from $B$ to 

1011: $A$. From the convexity of the distance function we conclude that

1012: the star contraction from $B$ to $A$

1013: is distance decreasing, and hence continuous, and if 

1014: the Hausdorff distance $\textrm{Hd}(A,B)$ is finite, then the star-contraction 

1015: gives a homotopy equivalence between $(B,\partial B)$ and $(A,\partial A)$.

1016: 

1017: More generally, if $A\subset B_1\subset B_2\subset S$, and if the $B_i$'s are closed and

1018: star-shaped with respect to $A$, we can define a deformation retract from $B_2$

1019: to $B_1$ by letting any $b\in B_2\setminus B_1$ flow in the direction of

1020: its projection $P_A(b)$ in $A$ at constant speed $s$, where $s$ equals to

1021: the length of the segment $[b,P_A(b)]\cap (B_2\setminus B_1)$. 

1022: By a similar procedure one can show that there is a deformation retract from $S\setminus B_1$

1023: to $\overline{S\setminus B_2}$.

1024: In this way we obtain:

1025: 

1026: \begin{lem}[{\bf A generalized star-contraction}]\label{GSC}

1027: Assume that 

1028: \begin{itemize}

1029: \item

1030: $A$ is convex and closed,

1031: \item

1032: $A\subset B_1\subset B_2$ where $B_i$ are closed and star-shaped with respect 

1033: to $A$.

1034: \item

1035: $\textrm{Hd}(A,B_2)<\infty$.

1036: \end{itemize}

1037: Then there is a deformation retract from $(B_2,\partial B_2)$ to $(B_1,\partial B_1)$.

1038: Similarly, there is a deformation retract from $S\setminus B_1$

1039: to $\overline{S\setminus B_2}$.

1040: \end{lem}

1041: 

1042: 

1043: 

1044: 

1045: 

1046: 

1047: %NEW SECTION

1048: 

1049: 

1050: 

1051: 

1052: \section{Constructing a simplicial complex in a thick submanifold with 

1053: nice boundary}\label{4}

1054: 

1055: Throughout this section, $M=\gC\backslash S$ is a fixed $S$-manifold with finite volume,

1056: $M'\subset M$ a connected submanifold with boundary, and $\gep >0$ is fixed. 

1057: The main result of this section is Lemma \ref{simplicial-complex}. 

1058: 

1059: We wish to formulate some convenient conditions on $M'$, under which $M'$ is 

1060: homotopically equivalent to a simplicial complex $\mathcal{R}$ whose 

1061: combinatorics is bounded in terms of $\vol (M)$. 

1062: More precisely, we would like $\mathcal{R}$ to be 

1063: a $\big( d,\ga\cdot\vol (M')\big)$-simplicial complex, where $d$ and $\ga$

1064: are some constants depending only on $S$ and $\gep$ (see Definition \ref{(d,v)-s.c.}).

1065: 

1066: To construct $\mathcal{R}$ we will use a ``good covering'' argument.

1067: Recall that a cover of a topological space $T$ is called a {\bf good cover} 

1068: if any non empty intersection of sets of the cover is contractible.

1069: In this case the simplicial complex $\mathcal{R}$ which corresponds to

1070: the nerve of the cover is homotopically equivalent to $T$. By definition, the

1071: vertices of $\mathcal{R}$ correspond to the sets of the cover, and a 

1072: collection of vertices form a simplex when the intersection of the 

1073: corresponding sets is non-empty (see \cite{BT}, theorem 13.4). 

1074: 

1075: In a manifold with injectivity radius bounded uniformly 

1076: from below by $\epsilon$, such a good cover is achieved by taking 

1077: $\epsilon$-balls for which the set of centers form an

1078: $\epsilon /2$ (say) discrete net. In our case, in order to be able to use 

1079: $\gep$-balls (or more precisely $\gep /c$-balls for some constant $c$) 

1080: we shall require that $M'$ lies inside $M_{\geq\gep}$. 

1081: However, $M'$ is not a manifold but a manifold with boundary, and balls may 

1082: not be ``nice'' subsets - they may not be convex or even contractible, and an intersection of

1083: balls may not be connected.

1084: Therefore we should be more careful.

1085: Our problems arise only near the boundary (far away from 

1086: the boundary balls are nice). So we need some control on the geometry of the boundary.

1087: 

1088: \begin{lem}\label{simplicial-complex}

1089: Let $M=\gC\backslash S$ be a fixed $S$-manifold with finite volume,

1090: let $M'\subset M$ be a connected submanifold with boundary, 

1091: and let $\gep >0$ be fixed. 

1092: \begin{itemize}

1093: \item

1094: Assume that $M'$ is contained in the $\gep$-thick part $M_{\geq \gep}$.

1095: \item

1096: Write $X=M\setminus M'$, and assume that its pre-image $\tilde X\subset S$ 

1097: under the universal covering map is a locally finite union of convex open sets 

1098: with smooth boundary $\ti X=\cup_{i\in I}X_i$

1099: (by locally finite we mean that every compact set in $S$ intersects only finitely many $X_i$'s).

1100: \item

1101: Assume that $M'$ is homotopically equivalent to its $\frac{\gep}{2}$-shrinking 

1102: $)M'(_{\frac{\gep}{2}}$.

1103: \item

1104: Let $b>1$ be a constant which depends only on $S$ and on $\gep$.

1105: \item

1106: Assume that for any point $x\in S\setminus\ti X$ with $d(x,\ti X)\leq\gep$, 

1107: there is a unit tangent vector $\hat n(x)\in T_x(S)$,

1108: such that for each $i\in I$ with $d(x,X_i)=d(x,\ti X)$, 

1109: the inner product of $\hat n(x)$ with the gradient $\nabla D_{X_i}(x)$ satisfies 

1110: $$

1111:  \hat n(x)\cdot\nabla D_{X_i}(x)>\frac{1}{b}.

1112: $$

1113: \end{itemize}

1114: Then there are constants $\ga$ and $d$, depending only on $S$ and on $\gep$, 

1115: such that $M'$ is homotopically equivalent to some 

1116: $\big( d,\ga\cdot\vol (M')\big)$-simplicial complex.

1117: \end{lem}

1118: 

1119: Throughout this section we use the notation of the statement of 

1120: Lemma \ref{simplicial-complex}.\\

1121: 

1122: The proof of Lemma \ref{simplicial-complex} relies on a uniform estimate on the distance

1123: between shrinkings of $M'$.

1124: 

1125: \begin{prop}\label{closeN}

1126: For any $\gt$ and $\gd$ with $\gep\geq \gt+\gd\geq\gt>0$ we have

1127: $$

1128:  \big( )M'(_{\gt+\gd}\big)_{b\gd}\supset )M'(_\gt.

1129: $$

1130: \end{prop}

1131: 

1132: The proposition can also be stated as follows: For any such $\gt$ and $\gd$, 

1133: the Hausdorff distance between the corresponding sets satisfies

1134: $$

1135:  \textrm{Hd}\big()M'(_{\gt+\gd}~,~)M(_{\gt}\big)= \leq b\gd.

1136: $$

1137: In other words, in order to prove the proposition, we need to show that for any

1138: $x\in )M(_\gt\setminus)M(_{\gt+\gd}=(X)_{\gt+\gd}\setminus (X)_{\gt}$ there is 

1139: $y\notin(X)_{\gt+\gd}$ with 

1140: $d(x,y)\leq b\gd$.

1141: 

1142: \begin{proof}

1143: Let $x\in(X)_{\gt+\gd}\setminus (X)_{\gt}$. Choose a lifting

1144: $\ti x\in(\ti X)_{\gt+\gd}\setminus (\ti X)_{\gt}$ of $x$.

1145: As $\ti x\in(\ti X)_{\gt+\gd}\setminus (\ti X)_{\gt}$, there is $\gt_1~(\gt+\gd\geq\gt_1\geq\gt)$

1146: such that $\ti x\in\partial(\ti X)_{\gt_1}$.

1147: 

1148: Let $c(t)$ be the piecewise

1149: geodesic curve of constant speed $b$, passing through $\ti x$, which is defined 

1150: inductively (for $t\geq \gt_1$) as follows:

1151: At time $t_1=\gt_1$ set $c(t_1)=\ti x$ and define its one sided derivative by 

1152: $$

1153:  \frac{d}{dt}_+ c(t_1)=b\hat n(\ti x),

1154: $$ 

1155: and define 

1156: $$

1157:  I_c(t_1)=\{ i\in I:x\in\partial (X_i)_{t_1}\}.

1158: $$

1159: Identify $c$ with the constant speed geodesics determined by this condition, as

1160: long as $t >t_1$ and $c(t )\notin \overline{(X_i)_\gt}$ for any $i\notin 

1161: I_c(t_1)$. Let $t_2$ be the first time (if such exist) $t_1\leq t<\gt+\gd$ at 

1162: which $c(t)$ hits some $\overline{(X_i)_t}$ for $i\notin I_c(t_1)$. 

1163: As the collection $\{ X_i\}_{i\in I}$ is locally finite, $t_2$ is well defined and strictly 

1164: bigger than $t_1$. We claim that $c(t_2)\in\partial (\tilde X)_{t_2}$, and 

1165: that $c(t)\notin (\ti X)_t$ for $t_1<t<t_2$. To prove this, we

1166: need to show that $c(t )\notin (\ti X_i)_t$ for any 

1167: $t_1\leq t\leq t_2$ and $i\in I_c(t_1)$. Fix such $i$ and observe that

1168: $$

1169:  \frac{d}{dt}|_{t=t_1} D_{X_i}\big(c(t)\big)=

1170:  b\hat n(x)\cdot \nabla D_{X_i}(x)>b\frac{1}{b}=1.

1171: $$ 

1172: Since the convex function $D_{X_i}\big( c(t )\big)$ has 

1173: non-decreasing derivative we get that $\frac{d}{dt}\Big( D_{X_i}\big( c(t)\big)\Big) >1$ 

1174: for any $t_2>t>t_1$, and hence, 

1175: $D_{X_i}\big( c(t)\big)>D_{X_i}\big( c(t_1)\big)+(t-t_1)=t$ 

1176: for any such $t$. 

1177: Thus, the point $c(t)$ is outside $\big( (X_i)_{t_1}\big)_{t -t_1}=(X_i)_t$.

1178: 

1179: Then we define the second piece of $c$ by the condition that its one sided

1180: derivative $\frac{d}{dt}_+ c(t_2)$ satisfies 

1181: $\frac{d}{dt}_+ c(t_2)=b\hat n\big( c(t_2)\big)$, 

1182: and we define 

1183: $$ 

1184: I_c(t_2)=\{ i\in I:c(t_2)\in \partial (X_i)_{t_2}\}.

1185: $$

1186: 

1187: Note that since $t_2$ is smaller than $\gt+\gd <\gep$ and 

1188: $c(t_2)\in\partial (X_2)$, the  direction $\hat n\big( c(t_2)\big)$ is well 

1189: defined. In this way, we continue to define $c$ inductively.

1190: 

1191: If $t_i$ converge to some $t_{\omega_0}<\gt+\gd$, then we define 

1192: $c(t_{\omega_0})$ to be the limit of $c(t_i)$ and 

1193: $\frac{d}{dt_+}c(t_{\omega_0})=b\hat n\big( c(t_{\omega_0})\big)$

1194: and $I_c(t_{\omega_0})=\{i\in I:c(t_{\omega_0})\in\partial 

1195: (X_i)_{t_{\omega_0}}\}$,

1196: and denote the next index by $\omega_0+1$, and so on.

1197: By this way we obtain a piecewise

1198: geodesic curve with at most countably many pieces, connecting 

1199: $\ti x=c(\gt_1)$ to $\ti y=c(\gt +\gd)$, 

1200: with $\ti y=c(\gt +\gd )\notin (\tilde X)_{\gt +\gd}$. 

1201: Since the length of 

1202: $c\big( [\gt_1,\gt +\gd ]\big)$ is $b(\gt +\gd-\gt_1)\leq b\gd$ 

1203: this proves the proposition.

1204: \end{proof}

1205: 

1206: \begin{cor}\label{cover}

1207: Let $\gt$ and $\gd$ be as in Proposition \ref{closeN}. Let $\mathcal{C}$ be a

1208: collection of balls of radius $(b+1)\gd$ for which the set of centers

1209: $\mathcal{C}'$ form a maximal $\gd$ discrete subset of $)M(_{\gt+\gd}$.

1210: Then $\mathcal{C}$ is a cover of $)M(_\gt$.

1211: \end{cor}

1212: 

1213: \begin{proof}

1214: As $\mathcal{C}'$ is maximal $(\mathcal{C}')_\gd\supset )M(_{\gt+\gd}$.

1215: By Proposition \ref{closeN} we have 

1216: $$

1217:  \big( )M'(_{\gt+\gd}\big)_{b\gd}\supset )M'(_\gt.

1218: $$

1219: Thus

1220: $$ 

1221:  \cup_{C\in \mathcal{C}}C=(\mathcal{C}')_{(b+1)\gd}=

1222:  \big( (\mathcal{C}')_\gd\big)_{b\gd}\supset

1223:  \big( )M'(_{(\gt +\gd)}\big)_{b\gd}\supset )M'(_\gt.

1224: $$

1225: \end{proof}

1226: 

1227: In the sequel we will take $\gt=\frac{\gep}{2}$.

1228: The next proposition provides a uniform bound on the curvature of the smooth pieces

1229: of $\partial )M'(_{\frac{\gep}{2}}$.

1230: 

1231: \begin{prop}\label{smallC}

1232: For any $X_i$ and any point $x \in \partial (X_i)_{\frac{\epsilon}{2}}$,

1233: the $\frac{\gep}{2}$-ball, whose boundary sphere is tangent at $x$ to  

1234: $\partial (X_i)_{\frac{\epsilon}{2}}$, which lies on the same side of 

1235: $\partial (X_i)_{\frac{\gep}{2}}$ as $(X_i)_{\frac{\gep}{2}}$,

1236: is contained in $(X_i)_{\frac{\epsilon}{2}}$.

1237: \end{prop}

1238: 

1239: \begin{proof}

1240: The distance between $x$ and its closest point $p_{X_i}(x)$ in the closed 

1241: convex set $X_i$ is easily seen to be $\frac{\gep}{2}$, and the 

1242: $\frac{\gep}{2}$-ball 

1243: centered at $p_{X_i}(x)$ is the required one.

1244: \end{proof}

1245: 

1246: The following proposition follows directly from the definition of a deformation 

1247: retract.

1248: 

1249: \begin{prop}\label{B-B'}

1250: Let $B'\subset B$ be topological spaces, and let $F_t~\big( t\in [0,1]\big)$ 

1251: be a deformation retract of $B$ such that $F_t(b)\in B'$ for any 

1252: $b\in B',t\in [0,1]$. 

1253: Then $F_t|_{B'}$ is a deformation retract of $B'$.

1254: \end{prop}

1255: 

1256: Let $B_r$ denote an arbitrary ball in $S$ of radius $r$, and $B_r(x)$ the ball 

1257: of radius $r$ centered at $x$. The following proposition follows from the fact 

1258: that the volume of a ball of radius $r$ is a convex function of $r$ (because 

1259: the surface area of the $r$-sphere is an increasing function of $r$).

1260: 

1261: \begin{prop}

1262: There is a constant $m$ such that for any $\delta <1$, 

1263: $$

1264:  m\cdot \vol (B_{\delta /2})

1265:  \geq \vol (B_{(b+1.5)\delta}).

1266: $$ 

1267: Thus any $\delta$-discrete subset of 

1268: $B_{(b+1)\delta}$ consists of at most $m$ elements.

1269: \end{prop}

1270: 

1271: For a finite set

1272: $\{ y_1,...,y_t\} \subset S$ we denote by $\sigma (y_1,...,y_t)$

1273: its Chebyshev center, i.e. the unique point $x$ which minimizes the function

1274: $\max_{1\leq i\leq t}d(x,y_i)$.

1275: 

1276: We will soon take $B$ to be an intersection of balls, and $B'\subset B$ to be

1277: the intersection of $B$ with $)M'(_{\frac{\gep}{2}}$. 

1278: We intend to use \ref{B-B'} in order to show that under some certain 

1279: conditions, $B'$ is contractible.

1280: It will be natural to use the star contraction to the Chebyshev center 

1281: of the centers of the associated balls - The deformation retract which 

1282: makes any point of $B$ flow along the geodesic segment which connects it to 

1283: the required Chebyshev center, at constant velocity $s$, where $s$ 

1284: equals the initial distance. In order to do this, we need the following:

1285: 

1286: \begin{prop}[Defining the constant $\gd$] \label{positiveD}

1287: There exists $\gd~(0<\delta < \frac{\epsilon}{2(b+1)})$ 

1288: such that for any point $x\in S$, any $\frac{\gep}{2}$-ball

1289: $C$ which contains $x$ on its boundary sphere, and  

1290: any $m$ points $y_1,\ldots ,y_m\in B_{(b+1)\delta }(x)\setminus 

1291: (C)_{\delta}$, the inner product of the external normal vector of $C$ at $x$ 

1292: with the tangent at $x$ to the geodesic segment $[x,\sigma (y_1,\ldots ,y_m)]$ 

1293: is positive.

1294: \end{prop}

1295: 

1296: \begin{proof}

1297: Assume the contrary. Then there is a sequence $\delta_n \to 0$,

1298: a corresponding sequence $(C_n)$ of $\frac{\epsilon}{2}$-balls tangent to some 

1299: $x_0\in S$ (which we may assume converge to some fixed such ball), and a 

1300: corresponding sequence of

1301: $m$-tuples of points 

1302: $y_1^n,\ldots ,y_m^n\in B_{(b+1)\delta_n}(x_0)\setminus ( C_n)_{\gd_n}$, 

1303: such that the inner product of the

1304: tangent at $x_0$ to the geodesic segment $[x_0,\sigma (y_1^n,...,y_m^n)]$ with

1305: the external normal vector of the corresponding sphere $\partial C_n$ is non-positive (we may fix $x=x_0$ 

1306: since $G$ acts transitively).

1307: 

1308: Rescaling the metric each time we can

1309: assume that $\delta_n$ is fixed and equals $1$. We then get a sequence of 

1310: Riemannian metrics converging, on the ball of radius $b+1$ around $x_0$,

1311: to the Euclidean metric on the ball of radius $b+1$. 

1312: More precisely, we look at the ball of radius $b+1$ in the tangent space

1313: $T_{x_0}S$. We identify it

1314: each time with the ball of radius $(b+1)\delta_n$ around $x_0$ in $S$ via the 

1315: map $X\to \exp_{x_0}(\delta_n X)$, and we rescale the metric there to 

1316: $$

1317:  d_n(X,Y)=\frac{1}{\delta_n}\cdot d\big(\exp_{x_0}(\delta_nX),\exp_{x_0}(\delta_nY)\big).

1318: $$

1319: ( All these metrics induce the same topology.)

1320: 

1321: Now, in the rescaled metrics our tangent balls tend to a half space 

1322: (since $\epsilon /\delta_n \to \infty$), and we may as well assume that our 

1323: $m$-tuples also converge. 

1324: In the limit, we get an $m$-tuple of points

1325: in a Euclidean space at distance at least $1$ (and at most $b+1$)

1326: from a half space, for which

1327: the inner product of the external normal vector to this half space with the 

1328: vector $\overrightarrow v$, pointing from \textrm{so}me $x_0$ on the boundary 

1329: hyper-plane, towards the Chebyshev center of this $m$-tuple is non-positive. This is an absurd.

1330: \end{proof}  

1331: 

1332: Finally, we claim

1333: 

1334: \begin{prop}

1335: Let $\mathcal{C}$ be a collection of balls of radius $(b+1)\delta$, for which 

1336: the set of centers $\mathcal{C}'$ form a maximal $\delta$-discrete subset of

1337: $)M'(_{\frac{\gep}{2} +\gd}$. Then $\mathcal{C}$, i.e. the restrictions of

1338: its sets to $)M'(_{\frac{\gep}{2}}$, is a good cover of 

1339: $)M'(_{\frac{\gep}{2}}$.

1340: \end{prop}

1341: 

1342: \begin{proof}

1343: By corollary \ref{cover}, $\mathcal{C}$ is a cover of $)M'(_{\frac{\gep}{2}}$.

1344: 

1345: Let $B$ be the intersection of $m$ (not necessarily different) balls of 

1346: $\mathcal{C}$, with centers $y_1,\ldots,y_m\in\mathcal{C}$. 

1347: 

1348: Proposition \ref{smallC} implies that for any 

1349: $\tilde x\in \partial )\tilde M'(_{\frac{\gep}{2}}$ 

1350: and for any  $X_i$ with

1351: $\tilde x\in \partial  (X_i)_{\frac{\epsilon}{2}}$

1352: there is an $\frac{\epsilon}{2}$-ball,

1353: tangent to $\partial (X_i)_{\frac{\gep}{2}}$ at $\tilde x$,

1354: which is contained in $(X_i)_{\frac{\gep}{2}} \subset 

1355: (\tilde X)_{\frac{\gep}{2}}$.

1356: Thus, if in addition the image $x$ of $\tilde x$ belongs to $B$, then 

1357: Proposition \ref{positiveD} implies that the geodesic segment 

1358: $[x,\sigma (y_1,\ldots ,y_m)]$ lies inside $B'=B\cap )M'(_{\frac{\gep}{2}}$.

1359: Let us explain this point as follows. Let $c(t),~t\in [0,1]$ be a 

1360: parameterization of the geodesic segment $[x,\sigma (y_1,\ldots ,y_m)]$. 

1361: Propositions \ref{smallC} and \ref{positiveD} imply that 

1362: $c(t)\notin \overline{(X)}_{\frac{\gep}{2}}$ for all sufficiently small $t>0$.

1363: Assume that $c(t)\in\partial (X)_{\frac{\gep}{2}}$ for some $t_0$, $0<t_0<1$, 

1364: (and let $t_0$ be the first such  time). Then Propositions \ref{smallC} and 

1365: \ref{positiveD} applied to $c(t_0)$ imply that 

1366: $c(t_0-\gD t)\in(X)_{\frac{\gep}{2}}$ for small $\gD t$. But this is a 

1367: contradiction.

1368: We conclude that if $B$ is not empty 

1369: then $\sigma (y_1,\ldots ,y_m)\in B'$ and the star-contraction from $B$ to

1370: $\sigma (y_1,\ldots ,y_m)$ induces a contraction of $B'$. Hence the 

1371: set $B'$ is non-empty and contractible.

1372: This means that $\{ C':C\in \mathcal{C} \}$ is a good cover of 

1373: $)M'(_{\frac{\gep}{2}}$, where $C':=C\cap )M'(_{\frac{\gep}{2}}$.

1374: \end{proof}

1375: 

1376: We conclude that $)M'(_{\frac{\gep}{2}}$, and therefore $M'$, is homotopically 

1377: equivalent to the simplicial complex $\mathcal{R}$ which corresponds to the 

1378: nerve of $\mathcal{C}$. Since the collection of centers $\mathcal{C}'$ is 

1379: $\gd$-discrete, and therefore the $\frac{\gd}{2}$-balls with the same centers

1380: are disjoint, we get that $|\mathcal{C}'|$,  

1381: the number of the vertices of $\mathcal{R}$, is 

1382: $\leq\frac{\vol (M)}{\vol (B_{\gd /2})}$. 

1383: Since the sets of $\mathcal{C}$ are subsets of $(b+1)\gd$-balls, each of them 

1384: intersects at most 

1385: $\frac{\vol ( B_{2(b+1)\gd+\gd /2})}{\vol (B_{\gd /2})}$ of the 

1386: others. Hence, the degree of any vertex in $\mathcal{R}$ is 

1387: $\leq d:=\frac{\vol ( B_{2(b+1.25)\gd})}{\vol (B_{\gd /2})}$.

1388: This completes the proof of Lemma \ref{simplicial-complex}.

1389: 

1390: 

1391: 

1392: 

1393: 

1394: 

1395: %NEW SECTION

1396: 

1397: 

1398: 

1399: 

1400: 

1401: \section{An arithmetic variant of the Margulis lemma}\label{AML}

1402: 

1403: The classical Margulis lemma yields information on the algebraic structure

1404: of a discrete group of isometries which is generated by ``small elements''. If,

1405: in addition, this discrete group lies inside an arithmetic group then we can 

1406: say a little more. In this section we shall explain this idea in the

1407: non-uniform case. 

1408: 

1409: The Lie group of isometries $G=\textrm{Isom}(S)$

1410: is center-free, semi-simple, without compact factors and with 

1411: finitely many connected components. Its identity component $G^0$ coincides 

1412: with the connected component of the group of real points $\Bbb{G}(\BR )^0$ of some

1413: $\BQ$-algebraic group $\Bbb{G}$. 

1414: We shall identify $G^0$ with its adjoint group

1415: $\text{Ad}(G^0)\leq\text{GL}(\mathfrak{g})$ and think of it as a group of 

1416: matrices. We will denote by $\mu$ a fixed Haar measure on $G$.

1417: 

1418: \begin{lem}[\bf An arithmetic variant of the Margulis lemma]\label{arith-M-L}

1419: There are constants $\gep =\gep (S)>0$ and $m=m(S)\in\BN$, such that 

1420: if $\gC\leq G$ is a non-uniform torsion-free arithmetic lattice, then 

1421: for any $x\in S$, the group of real points of the Zariski closure

1422: $\big(\overline{\gC_\gep (x)}^z\big)_\BR$ of the group 

1423: $$

1424:  \gC_\gep (x)=\langle\gc\in\gC:d_\gc (x)\leq\gep\rangle

1425: $$

1426: has at most $m$ connected components, and its identity component is a 

1427: unipotent subgroup.

1428: \end{lem}

1429: 

1430: The following claims (\ref{Q-structure}, \ref{5.3}, \ref{5.4}) 

1431: are well known (c.f. \cite{Mar1}, chapter $IX$ section 4). 

1432: 

1433: \begin{lem}\label{Q-structure}\label{5.2}

1434: For any non-uniform arithmetic lattice $\gD\leq G^0$, there is a rational 

1435: structure on $G^0$ (coming from a rational structure on the vector space $\mathfrak {g}$)

1436: with respect to which $\gD$ is contained in $G^0(\BQ )$ and

1437: commensurable to $G^0(\BZ )$. Conjugating by an element $g\in G^0(\BQ )$ we can

1438: assume that $\gD\subset\ G^0(\BZ )$.

1439: \end{lem}

1440: 

1441: \begin{proof}[Explanation]

1442: In general, if $\gD$ is an arithmetic lattice in $G$, there is 

1443: a compact extension $H$ of $G$, a rational structure on $H$, and a subgroup

1444: $\gD'\leq H_\BQ$ commensurable to $H_\BZ$ whose projection to $G$ is $\gD$.   

1445: However, when $\gD$ is non-uniform we can always take $H=G$.

1446: This could be deduced, for example, from the fact that $\gD$ is almost 

1447: generated by unipotent elements, and that compact groups contain no non-trivial 

1448: unipotent element.

1449: 

1450: For a given rational structure, the group $G_\BZ$ is defined only up to 

1451: commensurability. A subgroup $\gD\leq G_\BQ$ which is commensurable to 

1452: $G_\BZ$ is always conjugate, by an element of $G_\BQ$, to a subgroup of $G_\BZ$.

1453: \end{proof}

1454: 

1455: If $g\in G$ is close to $1\in G$, then its characteristic polynomial (i.e. the characteristic 

1456: polynomial of the endomorphism $\text{Ad}(g)$) is close to $(\gl -1)^n$ where 

1457: $n=\dim (\mathfrak {g})$. In particular:

1458: 

1459: \begin{prop}\label{5.3}

1460: There is an identity neighborhood $\gO_1\subset G$ such that, if $g\in \gO_1$,

1461: and the characteristic polynomial of $g$ has integral coefficients, 

1462: then $g$ is unipotent.

1463: \end{prop}

1464: 

1465: \ref{5.2} and \ref{5.3} implies:

1466: 

1467: \begin{cor}[\cite{Mar1} 4.21]\label{5.4}

1468: For any non-uniform arithmetic lattice $\gD\leq G$, the intersection 

1469: $\gO_1\cap\gD$ consists of unipotent elements only.

1470: \end{cor}

1471: 

1472: Recall also the following theorem of Zassenhaus and Kazhdan-Margulis (see \cite{Rag} 

1473: theorem 8.16):   

1474: 

1475: \begin{thm}[Zassenhaus, Kazhdan-Margulis]

1476: There exists an identity neighborhood $\gO_2\subset G$, called a

1477: Zassenhaus neighborhood, so that for any discrete subgroup $\gS\leq G$,

1478: the intersection $\gS\cap\gO_2$ is contained in a connected nilpotent Lie

1479: subgroup of $G$.

1480: \end{thm}

1481: 

1482: \begin{proof}[Proof of Lemma \ref{arith-M-L}]

1483: Let $\gO\subset G$ be a relatively compact symmetric identity neighborhood 

1484: which satisfies $\gO^2\subset\gO_1\cap\gO_2$.

1485: 

1486: Fix an integer $m$,

1487: $$

1488:  m> \inf_{h\in G}\frac{\mu \big(\{g\in G:d_g(x)\leq 1\}\cdot h\gO h^{-1} \big)}

1489:  {\mu (\gO )}, 

1490: $$ 

1491: and

1492: $$

1493:  \gep =\frac{1}{m}.

1494: $$ 

1495: As $G$ is unimodular, $m$ can be chosen independently of $x$. 

1496: Replacing $\gO$ by some conjugate $h\gO h^{-1}$, if needed, we can assume that

1497: $$

1498:  m>\frac{\mu \big(\{g\in G:d_g(x)\leq 1\}\cdot\gO \big)}{\mu (\gO )}.

1499: $$ 

1500: Let

1501: $$

1502:  \gC_\gO=\langle\gC\cap\gO^2\rangle.

1503: $$

1504: Then 

1505: $$

1506:  [\gC_\gep (x): \gC_\gep (x)\cap\gC_\gO ]\leq m.

1507: $$

1508: To see this, assume for a moment that this index was $\geq m+1$. Then we 

1509: could find $m+1$ representatives $\gc_1,\gc_2,\ldots,\gc_{m+1}\in\gC_\gep (x)$

1510: for different cosets of $\gC_\gep (x)\cap\gC_\gO$ in the ball of radius $m$

1511: in $\gC_\gep (x)$ according to the word metric with respect to the generating

1512: set $\{\gc\in\gC_\gep (x): d_\gc (x)\leq\gep\}$. As they belong to different

1513: cosets, $\gc_i\gO\cap\gc_j\gO =\emptyset$ for any $1\leq i<j\leq m+1$.

1514: Since $d_{\gc_i}(x)\leq m\cdot\gep =1$ these $\gc_i$'s

1515: are all inside $\{ g\in G:d_g(x)\leq 1\}$. This contradicts the assumption 

1516: $m\cdot\mu (\gO )>\mu \big(\{ g\in G:d_g(x)\leq 1\}\cdot\gO\big)$.

1517: 

1518: It follows from the Zassenhaus-Kazhdan-Margulis theorem that $\gC_\gO$ is contained in a 

1519: connected nilpotent Lie subgroup of $G$, and therefore, by Lie's theorem,

1520: $\gC_\gO$ is triangulable over $\BC$. As $\gC_\gO$ is generated by unipotent

1521: elements, it follows that $\gC_\gO$ is a group of unipotent elements.

1522: Thus the Zariski closure $\overline{\gC_\gO}^z$ is a unipotent algebraic group,

1523: and hence, the group of its real points $(\overline{\gC_\gO}^z)_\BR$ is 

1524: connected in the real topology. Similarly, its subgroup

1525: $(\overline{\gC_\gep (x)\cap\gC_\gO}^z)_\BR$ is connected.

1526: Clearly, $(\overline{\gC_\gep (x)\cap\gC_\gO}^z)_\BR$ is the identity 

1527: connected component of $(\overline{\gC_\gep (x)}^z)_\BR$, 

1528: and its index is at most $m$.

1529: 

1530: \end{proof}

1531: 

1532: \begin{rem}

1533: Although $\gep$ could be taken to be $1/m$, we use different letters for them

1534: because they play different roles.

1535: In the sequel we will assume that the above lemma is satisfied with $\gep$

1536: replaced by $10\gep$.

1537: \end{rem}

1538: 

1539: \begin{rem}\label{5.7}

1540: It follows from Lemma \ref{arith-M-L} that if $\gc$

1541: is an element of a non-uniform arithmetic lattice of $G$ and 

1542: $\inf d_\gc < \gep$, then $\gc^j$ is unipotent for some $j\leq m$.

1543: This implies that in a non-compact arithmetic $S$-manifold there are no closed geodesic of length 

1544: $\leq\gep$, and that in the $\gep$-thick thin decomposition, the thin part has no

1545: compact connected component. For example, the $\gep$-thin part of any 

1546: non-compact arithmetic hyperbolic surface (or more generally of any rank-1

1547: manifold) consists only of cusps.

1548: \end{rem}

1549: 

1550: 

1551: 

1552: 

1553: 

1554: %new section

1555: 

1556: 

1557: 

1558: 

1559: 

1560: \section{The proof of \ref{thmA}(1)}\label{6}

1561: 

1562: In this section we shall prove:

1563: 

1564: \begin{thm}[{\bf Theorem \ref{thmA}(1) of the introduction}]\label{thm1}

1565: Let $S$ be a symmetric space of non-compact type. Then there are constants

1566: $\ga$ and $d$ (depending only on $S$) such that any non-compact arithmetic

1567: $S$-manifold $M$ is homotopically equivalent some 

1568: $\big( d,\ga\cdot\vol (M)\big)$-simplicial complex.

1569: \end{thm}

1570: 

1571: Fix $\gep' =\gep' (S),m=m(S)$, such that Lemma \ref{arith-M-L} is satisfied 

1572: with $10\gep'$, and $m$. Assume that $M=\gC\backslash S$ is a given non-compact 

1573: arithmetic $S$-manifold. Denote by $\gC^u$ the set of unipotent elements in 

1574: $\gC$,

1575: $$

1576:  \gC^u=\{ \gc\in\gC :\gc \textrm{ is unipotent}\},

1577: $$

1578: and define

1579: $$

1580:  \tilde X=\cup_{\gc\in\gC^u\setminus\{ 1\}}\{ d_\gc<\gep'\},

1581: $$

1582: and

1583: $$

1584:  \tilde M'=\tilde M \setminus \tilde X,

1585: $$

1586: and let $X\subset M$ and $M'\subset M$ be the images of $\ti X$ and 

1587: $\ti M'$ under the universal covering map.

1588: 

1589: In order to prove Theorem \ref{thm1} we shall show:

1590: 

1591: \begin{enumerate}

1592: \item\label{1}

1593: There is a deformation retract from $M$ to $M'$. In particular $M$ is 

1594: homotopically equivalent to $M'$.

1595: \item\label{2}

1596: $M'$ satisfies the conditions of Lemma \ref{simplicial-complex} with 

1597: $\gep =\gep'/m$.

1598: \end{enumerate}

1599: 

1600: Let us start with some preliminaries. Let $W\subset S(\infty )$ be a Weyl

1601: chamber of the Tits spherical building. Then $W$ is isometric to a spherical

1602: simplex.

1603: We define its center of mass $z$ to be

1604: $$

1605:  z=\frac{\int_W xd\mu (x)}{\|\int_W xd\mu (x)\|}

1606: $$

1607: where $\mu$ is the Lebesgue measure on the sphere. Since $W$ is 

1608: contained in a half 

1609: sphere (see \cite{BGS} appendix 3), we get $\|\int_W xd\mu(x)\|\neq 0$. Since $W$

1610: is convex (in the spherical metric), and since its interior is non-empty and is 

1611: exactly the set of regular points in $W$, it follows that $z$ is regular

1612: and contained in $W$.

1613: 

1614: We shall also use:

1615: 

1616: \begin{lem}\label{321}

1617: Assume that $g$ is a parabolic isometry (i.e. $\min (g)=\emptyset$), $c(t)$ is a regular geodesic, and

1618: $g$ stabilizes $c(-\infty )$. Then $\frac{d}{dt}d_g\big( c(t)\big)\neq0$

1619: for any $t\in\BR$.

1620: \end{lem}

1621: 

1622: \begin{proof}

1623: Let $c(t)$ be a regular geodesic in $S$, and let $g$ be an isometry which 

1624: stabilizes $c(-\infty )$. Assume that 

1625: $\frac{d}{dt}|_{t=t_o}d_g\big( c(t)\big) =0$, then the function 

1626: $d_g\big( c(t)\big)$, being analytic and convex 

1627: must be constant, since it doesn't tend to $\infty$ as 

1628: $t\to -\infty$. This means that $g\cdot c$ is a geodesic parallel to $c$.

1629: We conclude that $g$ preserves the unique flat which contains $c$, since this

1630: flat is exactly the set of points through which there is a geodesic parallel

1631: to $c$. But this implies that $g$ is semisimple, a contradiction.

1632: \end{proof}

1633: 

1634: Furthermore, if $g$ is unipotent, then $d_g\big( c(t)\big)\to 0$ as $t\to -\infty$

1635: (see \cite{BGS} appendix 5, section 4).

1636: 

1637: \begin{proof}[{\bf The proof of (\ref{1})}]

1638: We will show that the conditions of Lemma \ref{deformation-retract} are 

1639: satisfied with $L=Y=S$ and the $\gC$-invariant family of functions

1640: $\mathcal{F}=\{d_\gc\}_{\gc\in\gC^u\setminus\{ 1\}}$. 

1641: Then $\ti M'=\mathcal{F}_{\geq\gep'}$.

1642: 

1643: The finiteness of the sets 

1644: $$

1645:  \Psi_{x,\gt }=\{\gc\in\gC^u\setminus\{ 1\}:d_\gc (x)\leq\gt\}

1646: $$ 

1647: follows from the compactness of $\{ g\in G:d_g(x)\leq\gt\}$

1648: together with the discreteness of $\gC$. All the functions 

1649: $\{ d_\gc\}_{\gc\in\gC^u\setminus\{ 1\}}$ are convex, strictly positive, 

1650: without critical values. 

1651: We need to define the continuous function $\gb :(0,3\gep )\to\BR^{>0}$, 

1652: and the appropriate direction $\hat n(x)\in T_x(S)$ for any $x\in S$ with

1653: $\Psi_{x,3\gep}\neq\emptyset$.

1654: 

1655: Fix $x\in S$ with $\Psi_{x,3\gep}\neq\emptyset$ and let 

1656: $\Psi_{x,3\gep}=\{\gc_1,\gc_2,\ldots,\gc_k\}$.

1657: By Lemma \ref{arith-M-L} the Zariski closure $\overline\gD^z$ of

1658: the group $\gD =\langle \gc_1,\gc_2,\ldots,\gc_k \rangle$ 

1659: has a unipotent identity component, and $\leq m$ connected components. 

1660: Since $\gc_i$ is unipotent,

1661: it is contained in the Zariski closure of the cyclic group generated by any 

1662: power of it. 

1663: As $\gc_i^j$ belongs to the identity component $(\overline{\gD}^z)^0$ for some

1664: $j\leq m$, also $\gc_i\in(\overline{\gD}^z)^0$. Since this holds for any

1665: generator $\gc_i$, we get that

1666: $\overline{\gD}^z=(\overline{\gD}^z)^0$ and hence the group of its

1667: real points $(\overline{\gD}^z)_\BR$ is a connected unipotent group.

1668: Hence $\gc_1,\gc_2,\ldots,\gc_k$ are contained in a connected unipotent group.

1669: 

1670: Let $N\leq G$ be a maximal 

1671: connected unipotent subgroup which contains $\gc_1,\gc_2,\ldots ,\gc_k$.

1672: Let $W\leq S(\infty )$ be the Weyl chamber of the Tits boundary of $S$ 

1673: which corresponds to $N$. Let $c(t)=c_x(t)$ be the

1674: geodesic line with $c(0)=x$ for which $c(-\infty )$ is the center of mass of

1675: $W$.

1676: 

1677: By Lemma \ref{321}

1678: $$

1679:  \frac{d}{dt}|_{t=0}\Big( d_g\big( c(t)\big)\Big)>0

1680: $$ 

1681: for any $g\in N\setminus\{ 1\}$. In addition, the continuous function

1682: $$

1683:  h(g)=\dot c(0)\cdot\nabla d_g(x)

1684: $$ 

1685: attains a minimum on the compact set

1686: $$

1687:  \{g\in N:d_g(x)=\gt\}.

1688: $$

1689: We define $\gb (\gt )$ to be this minimum (then $\gb (\gt )$ is defined for

1690: any $\gt >0$). By definition, 

1691: $\dot c(0)\cdot\nabla d_{\gc_i}(x)\geq\gb\big(d_{\gc_i}(x)\big)$.  

1692: Moreover, it is easy to see that $\gb$ is a 

1693: continuous positive function. Since $G$ acts transitively on the

1694: set of couples $(W,x)$ of a Weyl chamber $W\subset S(\infty )$ and a point $x\in S$,

1695: $\gb$ is independent of $N$ and of $x$.

1696: The conditions of Lemma \ref{deformation-retract} are satisfied with the 

1697: tangent vector $\hat n(x)=\dot c_x(0)$.

1698: \end{proof}

1699: 

1700: \begin{rem}

1701: We used the existence and uniqueness of the center of mass

1702: of $W$ in order to define 

1703: $\gb$ in a canonical way. However, as any two Weyl chambers are isometric,

1704: we could choose arbitrarily a regular point in one Weyl chambers, and translate

1705: it to any other Weyl chambers, and to use these points when defining 

1706: $c(-\infty )$ and then $c, \hat n(x)$ and

1707: $\gb$ in our proof.

1708: \end{rem}

1709: 

1710: \begin{rem}\label{finite-presentability}

1711: A slight modification of the argument above, yields an elementary proof of the 

1712: fact that $M=\gC\backslash S$ is homotopic to a compact manifold with boundary, and 

1713: hence that $\gC$ is finitely presented, when $\gC\leq G$ is any torsion free (non-uniform)

1714: lattice. We used the arithmeticity of $\gC$ in order to get a uniform estimate for all $\gC$'s.

1715: However, for a fixed $\gC$, we could have used corollary 11.18 from \cite{Rag} instead of 

1716: corollary \ref{5.4} above, and by this to avoid the assumption that $\gC$ is arithmetic. Moreover, 

1717: we don't really have to assume that $\gC$ is torsion free (when proving just the finiteness without

1718: explicit estimate). The same argument shows that in general, any $S$-orbifold 

1719: $\gC\backslash S$ is homotopic to a compact orbifold with boundary, and it is not hard to show by the same 

1720: means that $\gC\backslash G$ is homotopic to a compact manifold with boundary. In particular, our method 

1721: gives a quite elementary proof of the finite presentability of lattices.

1722: Furthermore, changing appropriately the factor $\sqrt{\cdot}$ in the vector field which induces

1723: the deformation retract, so that it will decay more slowly, we can get a simple proof that 

1724: $\gC\backslash G$ is diffeomorphic to the interior of a compact manifold with boundary.

1725: \end{rem}

1726: 

1727: \begin{proof}[{\bf The proof of (\ref{2})}]

1728: In order to check that the conditions of Lemma \ref{simplicial-complex}

1729: are satisfied, we will show that

1730: 

1731: \begin{itemize}

1732: \item $M'\subset M_{\geq\gep}$ (where $\gep =\frac{\gep '}{m}$).

1733: \item $\ti X$ is a locally finite union of convex open sets with smooth boundary.

1734: \item $M'$ is homotopically equivalent to its $\frac{\gep'}{2m}$-shrinking 

1735: $)M'(_{\frac{\gep'}{2m}}$.

1736: \end{itemize}

1737: And that there are

1738: \begin{itemize}

1739: \item $b=b(S)>0$ and

1740: \item $\hat n(x)\in T_x(S)$ for any 

1741: $x\in (\ti X)_{\frac{\gep'}{2m}}\setminus (\ti X)$

1742: \end{itemize}

1743: such that

1744: $$

1745: \hat n(x)\cdot\nabla D_{\{\gc\leq\gep'\}}(x)>\frac{1}{b}

1746: $$

1747: for any $\gc\in\gC^u\setminus\{ 1\}$ with 

1748: $D_{\{\gc\leq\gep'\}}(x)\leq\frac{\gep'}{2m}$ (this is stronger than the condition on the

1749: inner products required in Lemma \ref{simplicial-complex}).\\

1750: 

1751: If $x\in S$ and $d_\gc (x)\leq\frac{\gep'}{m}\leq\gep'$ for some $\gc\in\gC$, 

1752: then by Lemma \ref{arith-M-L} $\gc^j$ is unipotent for some $j\leq m$,

1753: and since $d_{\gc^j} (x)\leq\frac{j\cdot\gep'}{m}\leq\gep'$, we obtain that

1754: $x\in\ti X$. This proves that $M'$ is contained in the $\frac{\gep'}{m}$-thick

1755: part $M_{\geq\frac{\gep'}{m}}$.\\

1756: 

1757: Since for $\gc\in\gC^u\setminus\{ 1\}$ the displacement function is convex and

1758: analytic and $\inf d_\gc=0$, and since $\gC$ is discrete, we have that 

1759: $\ti X=\cup_{\gc\in\gC^u\setminus\{ 1\}}\{d_\gc<\gep\}$ is a locally finite 

1760: union of convex open sets with smooth boundary.

1761: \\

1762: 

1763: We shall prove that there is a deformation retract from $M'$ to 

1764: its $\frac{\gep'}{2m}$-shrinking $)M'(_{\frac{\gep'}{2m}}$

1765: by showing that the conditions of Lemma \ref{deformation-retract}

1766: are satisfied with 

1767: ${\mathcal F}=\{ D_{\{ d_\gc\leq\gep'\}}\}_{\gc\in\gC^u\setminus\{ 1\}}$,

1768: $Y=S$, $L={\mathcal F}_{\geq 0}=\ti M'$, and 

1769: ${\mathcal F}_{\geq\frac{\gep'}{2m}}=)\ti M'(_{\geq\frac{\gep'}{2m}}$.

1770: (In this proof $\frac{\gep'}{2m}$ plays the rule of $\gep$ in Lemma

1771: \ref{deformation-retract}.)

1772: 

1773: The finiteness of the sets 

1774: $$

1775:  \Psi_{x,\gt}=\{\gc\in\gC^u\setminus\{ 1\}:

1776:  D_{\{d_\gc\leq\gep'\}}(x)\leq\gt\}

1777: $$ 

1778: follows from the discreteness of $\gC$

1779: together with the compactness of 

1780: $$

1781: \{g\in G:D_{\{d_g\leq\gep\}}(x)\leq\gt\}.

1782: $$ 

1783: 

1784: We shall define the direction $\hat n(x)\in T_x(S)$ analogously 

1785: to the way it is done in the proof of (\ref{1}), 

1786: and shall show that the conditions of Lemma \ref{deformation-retract} are 

1787: satisfied with the constant function $\frac{\gb (\gep' )}{2}$, where $\gb$ 

1788: is the function defined in the proof of (\ref{1}).

1789: 

1790: Fix $x\in L=\mathcal{F}_{\geq 0}$, with $\Psi_{x,\frac{3\gep'}{2m}}\neq\emptyset$, and denote

1791: $\Psi_{x,\frac{3\gep'}{2m}}=\{\gc_1,\gc_2,\ldots,\gc_k\}$.

1792: Since $\gep'+2\frac{3\gep'}{2m}=4\gep'$ we have

1793: $(\{ d_{\gc_i}\leq\gep'\} )_{\frac{3\gep'}{2m}}\subset\{d_{\gc_i}\leq 4\gep'\}$

1794: which implies that $\cap_1^k\{d_{\gc_i}\leq 4\gep'\}\neq\emptyset$ and hence

1795: $\gc_1,\gc_2,\ldots,\gc_k$ are contained in a connected 

1796: unipotent group. As in the proof of (\ref{1}), let $N$ be a 

1797: maximal connected unipotent group which contains $\gc_1,\gc_2,\ldots,\gc_k$, 

1798: let $W\subset S(\infty )$ be the Weyl chamber which corresponds to $N$ in the 

1799: Tits boundary. Let $c(t)=c_x(t)$ be the geodesic line with $c(0)=x$ and $c(-\infty)=$ the center 

1800: of mass of $W$. Since $\gc_i\in N$,

1801: $d\big( c(t),\gc_i\cdot c(t)\big)$ tends to $0$ as $t\to -\infty$, and since

1802: $d_{\gc_i}(x)>\gep'$ we have

1803: $d_{\gc_i}\big( c(t_0)\big)=\gep'$ for some negative $t_0$. 

1804: 

1805: The function $d_{\gc_i}\big( c(t)\big)$ is convex and hence has a non-decreasing derivative,

1806: thus, taking $\hat n(x)=\dot c(0)$, we have:

1807: 

1808: \begin{eqnarray*}

1809:   \hat n(x)\cdot\nabla D_{\{d_{\gc_i}\leq\gep'\}}

1810:  (x)

1811: %&=&\\

1812: &=& \dot c(0)\cdot\nabla D_{\{d_{\gc_i}\leq\gep'\}}(x)\\ 

1813: & =& 

1814:  \dot c(t_0)\cdot\frac{\nabla d_{\gc_i}\big( c(0)\big)}

1815:  {\|\nabla d_{\gc_i}\big( c(0)\big)\|}\\

1816: &\geq &

1817:  \dot c(0)\cdot\frac{\nabla d_{\gc_i}\big( c(0)\big)}{2}\\

1818: &=& 

1819: \frac{1}{2}\cdot\frac{d}{dt}|_{t=0}d_{\gc_i}\big( c(t)\big)\\

1820: & \geq& 

1821: \frac{1}{2}\cdot\frac{d}{dt}|_{t=t_0}d_{\gc_i}\big( c(t)\big)\\

1822: &=&

1823: \frac{1}{2}\cdot\dot c(t_0)\cdot\nabla d_{\gc_i}\big( c(t_0)\big)

1824:  \geq\frac{\gb(\gep')}{2},

1825: \end{eqnarray*}

1826: where $\gb$ is the function defined in the proof of 

1827: \ref{1}.

1828: In the above computation we made use of the facts that 

1829: $\|\nabla  D_{\{d_{\gc_i}\leq\gep'\}}\| =1$ everywhere outside 

1830: $\{d_{\gc_i}<\gep '\}$, 

1831: and that $d_{\gc_i}$ is 2-Lipschitz and hence $\|\nabla d_{\gc_i}\|\leq 2$.

1832: 

1833: We completed the verification that the conditions of Lemma 

1834: \ref{deformation-retract} are satisfied, and hence proved that $M'$ is 

1835: homotopic to its $\frac{\gep'}{2m}$-shrinking.\\

1836: 

1837: Finally, observe that we can take also the constant $b$

1838: to be $b=\frac{2}{\gb(\gep')}$, and the unit tangent vector $\hat n(x)$ from 

1839: \ref{simplicial-complex} to be the same $\hat n(x)$ used above. Then all the conditions of Lemma

1840: \ref{simplicial-complex} are satisfied, and the proof of (\ref{2}) is 

1841: completed. 

1842: \end{proof}

1843: 

1844: 

1845: 

1846: 

1847: 

1848: 

1849: 

1850: %NEW SECTION

1851: 

1852: 

1853: 

1854: 

1855: 

1856: 

1857: 

1858: \section{Estimating angles at corners of the boundary}\label{bigangles}\label{7}

1859: 

1860: The next 3 sections are devoted to the remaining cases (2), (3) of Theorem \ref{thmA}.

1861: The main result on this section is Theorem \ref{nice-boundary}.

1862: 

1863: Our method is to replace a manifold $M$ by a submanifold with boundary $M'$, 

1864: in which the injectivity radius is uniformly bounded from below, and then  

1865: to apply Lemma \ref{simplicial-complex}. This requires some

1866: information on the boundary. The pre-image $\ti X$ of the complement 

1867: $X=M\setminus M'$ in the universal covering is required to be

1868: a locally finite union of 

1869: convex open sets with smooth boundary, and what we need is a control on the 

1870: angles of the corners - where the boundaries of two or more such sets intersect.

1871: In the non-uniform case we used the presence of many unipotents, and the nice

1872: actions of unipotent groups on the boundary at infinity. In the compact case

1873: there are no unipotents, so different tools are required.

1874: 

1875: Let $A,B\subset S$ be convex bodies with smooth boundary and with a common

1876: interior point in their intersection. The angle between the boundaries at a 

1877: common point

1878: $x\in \partial A \cap \partial B$ is measured by $\pi$ minus the angle between

1879: the external normal vectors $\hat n_A(x), \hat n_B(x)$

1880: $$

1881:  \phi_x(\partial A,\partial B)=

1882:  \pi -\angle \big(\hat n_A(x),\hat n_B(x)\big).

1883: $$ 

1884: Thus, a big angle between the boundaries corresponds to a small angle between 

1885: the normals.

1886: The following lemma states that when $A$ and $B$ are sub-level sets 

1887: of commuting isometries, these angles are $\geq\frac{\pi}{2}$. A similar statement appeared in \cite{BGS}.

1888: 

1889: For an isometry $\gc$, and for $x\in S$ with 

1890: $d_{\gc}(x)=a>\min d_\gc$ 

1891: we denote by $\hat n_{\gc}(x)$ the external (with respect to 

1892: $\{ d_{\gc} \leq a \}$)

1893: normal to 

1894: $$

1895:  \{ d_{\gc} =a \}=\partial\{ d_{\gc}\leq a \}.

1896: $$

1897: 

1898: \begin{lem}\label{bigA}{\bf (Commutativity implies big angles):}

1899: If the isometries $\alpha$ and

1900: $\beta$ commute then $\hat n_{\alpha}(x)\cdot \hat n_{\beta}(x)\geq 0$.

1901: \end{lem}

1902: 

1903: \begin{proof}

1904: Let $a_\alpha =d_\alpha (x),a_\beta =d_\beta (x)$.

1905: Since $\{ d_\alpha =a_\alpha \}$ is a level set for $d_\alpha$, we see that

1906: $\hat n_{\alpha}(x)$ is the direction of the gradient

1907: $(\nabla \cdot d_\alpha) (x)$.

1908: Thus if $c(t)$ is the geodesic line through $x$ with 

1909: $\dot c(0)=\hat n_{\beta}(x)$

1910: then it is enough to show that

1911: $$

1912:  (\nabla \cdot d_\alpha )(x) \cdot \hat n_{\beta}(x)=

1913:  \frac{d}{dt}|_{t=0}\{ d_\alpha \big( c(t)\big)\} \geq 0.

1914: $$

1915: Let $p$ denote the projection on the convex set 

1916: $\{ d_{\beta} \leq a_\beta \}$. Since $\alpha$ and $\beta$ commute, the 

1917: set $\{ d_{\beta} \leq a_\beta \}$ is $\alpha$-invariant and therefore

1918: $\alpha$ commute with $p$. Since $p$ decreases distances and since the 

1919: geodesic lines $c(t),\ga\cdot c(t)$ are both orthogonal to 

1920: $\partial\{ d_\gb\leq a_\gb\}$ one sees that $d\big( c(t),\ga\cdot c(t)\big)$ 

1921: is a non-decreasing function of $t$. Thus 

1922: $$

1923:  (\nabla \cdot d_\alpha )(x)\cdot \hat n_{\beta}(x)=

1924:  \frac{d}{dt}|_{t=0}\{ d_\alpha \big( c(t)\big)\} =

1925:  \frac{d}{dt}|_{t=0}\{d\big( c(t),\alpha \cdot c(t)\big)\} \geq 0.

1926: $$

1927: \end{proof}

1928: 

1929: We shall need similar information when $A,B$ are replaced by their 

1930: $\gep$-neighborhoods $(A)_\gep ,(B)_\gep$. The following lemma explains that

1931: the situation then only improves.

1932:  

1933: Define

1934: $$

1935:  \varphi_t= \sup_{x\in \partial (A)_t \cap \partial (B)_t}

1936:  \angle\big(\hat n_{(A)_t}(x),\hat n_{(B)_t}(x)\big),

1937: $$

1938: then we have

1939: 

1940: \begin{lem}\label{monoton}{\bf (Monotonicity of angles):}

1941: $\varphi_t$ is a non-increasing function of $t$.

1942: \end{lem}

1943: 

1944: \begin{proof}

1945: We need to show that if $t_1>t_2\geq 0$ then $\varphi_{t_1}\leq 

1946: \varphi_{t_2}$. If $\varphi_t$ vanishes at some point $t=t_0$ then, as it is

1947: easy to verify, the union $A_{t_0}\cup B_{t_0}$ is convex, and $\varphi_t=0$

1948: for any $t>t_0$. We may therefore assume that $\varphi_t$ is strictly positive

1949: in our segment $[t_2,t_1]$. 

1950: 

1951: Fix a common interior point 

1952: $$

1953:  y\in\textrm{int}(A)\cap\textrm{int}(B).

1954: $$

1955: Now 

1956: $$

1957:  \varphi_t=\sup_{x\in \partial (A)_t \cap \partial (B)_t}

1958:  \angle \big(\hat n_{(A)_t}(x),\hat n_{(B)_t}(x)\big) =

1959:  \sup_{R\subset S}\{\max_{x\in R\cap\partial (A)_t \cap \partial (B)_t} 

1960:  \angle \big(\hat n_{(A)_t}(x),\hat n_{(B)_t}(x)\big)\}

1961: $$

1962: where $R$ runs over the (compact) balls centered at $y$.

1963: It is therefore enough to show that 

1964: $$

1965:  \varphi_{t_2}\geq\max_{x\in R\cap\partial (A)_{t_1} \cap \partial (B)_{t_1}} 

1966:  \angle \big(\hat n_{(A)_{t_1}}(x),\hat n_{(B)_{t_1}}(x)\big)

1967: $$

1968: for any such $R$.

1969: 

1970: Fix $R$ large enough (so that the intersection 

1971: $R\cap\partial (A)_{t_1} \cap \partial (B)_{t_1}$ is not empty)

1972: and let 

1973: $$

1974:  \tilde\varphi_t=\max_{x\in R\cap\partial (A)_t \cap 

1975:  \partial (B)_t} \angle \big(\hat n_{(A)_t}(x),\hat n_{(B)_t}(x)\big).

1976: $$

1977: Since $\tilde\varphi_t$ is obviously continuous,

1978: it is enough to show that for any $t_2 < t \leq t_1$ we have 

1979: $\tilde\varphi_{t-\gD t}\geq\tilde\varphi_t$ whenever $\gD t$ is small enough.

1980: Fix $t$, and let $\gD t$ be sufficiently small so that the argument below holds.

1981: 

1982: To simplify notation we replace $A$ (resp. $B$) by $(A)_{t-\gD t}$ 

1983: (resp. $(B)_{t-\gD t}$) and assume that $t-\gD t=0$ (this is just a matter of

1984: changing names after $t$ and $\gD t$ are fixed). 

1985: Let $x_t\in R\cap\partial (A)_t \cap\partial (B)_t$ be such that 

1986: $$

1987:  \angle \big(\hat n_{(A)_t}(x_t),\hat n_{(B)_t}(x_t)\big) =\tilde\varphi_t.

1988: $$

1989: It is enough to show that there is $x_0\in R\cap\partial A \cap \partial B$

1990: with

1991: $$

1992:  \angle \big(\hat n_A(x_0),\hat n_B(x_0)\big)\geq

1993:  \angle \big(\hat n_{(A)_t}(x_t),\hat n_{(B)_t}(x_t)\big).

1994: $$

1995: 

1996: Let $x_0\in \partial A \cap \partial B$ be a point at the minimal possible distance from $x_t$.

1997: Denote by $\hat u_0$ (resp. $\hat u_t$) the tangent to the geodesic line

1998: $\overline{x_0,x_t}$ at $x_0$ (resp. at $x_t$).

1999: 

2000: Since $\gD t$ is arbitrarily small, $\hat u_t$ is roughly in the direction of

2001: the bisector of the angle between $\hat n_{(A)_t}(x_t)$ and 

2002: $\hat n_{(B)_t}(x_t)$, and $x_0$ is closer to $y$ than $x_t$. In particular 

2003: $x_0\in R$.

2004: 

2005: By the Lagrange multipliers theorem $\hat u_0$ is a linear combination of

2006: $\hat n_A(x_0)$ and $\hat n_B(x_0)$ and since $\gD t$ is arbitrarily small, we

2007: can assume that $\hat u_0$ is in the convex cone spanned by   

2008: $\hat n_A(x_0)$ and $\hat n_B(x_0)$ (again, in the limit case $\hat u_0$

2009: is the direction of the bisector of the angle between

2010: $\hat n_A(x_0),\hat n_B(x_0)$). Thus 

2011: $$ 

2012:  \angle \big(\hat n_A(x_0),\hat n_B(x_0)\big) =

2013:  \angle \big(\hat n_A(x_0),\hat u_0\big) +

2014:  \angle \big(\hat u_0,\hat n_B(x_0)\big),

2015: $$

2016: and since (by the triangle inequality for angles) 

2017: $$

2018:  \angle \big(\hat n_{(A)_t}(x_t),\hat n_{(B)_t}(x_t)\big)\leq

2019:  \angle \big(\hat n_{(A)_t}(x_t),\hat u_t\big)+

2020:  \angle \big(\hat u_t,\hat n_{(B)_t}(x_t)\big),

2021: $$

2022: it is enough to show that 

2023: $$

2024:  \angle \big(\hat n_A(x_0),\hat u_0\big)\geq 

2025:  \angle \big(\hat n_{(A)_t}(x_t),\hat u_t\big)

2026: $$

2027: (and the analogous inequality for $B$ instead of $A$ whose proof is the same).

2028: 

2029: Let $c(s)$ be the geodesic line of unit speed 

2030: with $c(0)=x_0,\dot c(0)=\hat u_0$, and let

2031: $c(s_0)=x_t$ (i.e. $s_0=d(x_0,x_t)$). Then the above inequality follows from

2032: the following

2033: $$

2034:  \hat u_0\cdot \hat n_A(x_0)=\frac{d}{ds}|_{s=0}D_A\big( c(s)\big)\leq 

2035:  \frac{d}{ds}|_{s=s_0}D_A\big( c(s)\big)=

2036:  \hat u_t \cdot \hat n_{(A)_t}(x_t)

2037: $$

2038: and this follows from the convexity of the function $D_A$.

2039: \end{proof}

2040: 

2041: The following lemma, explains how we can use the information obtained above,

2042: in order to find a direction with respect to which the directional derivatives

2043: of all the corresponding distance functions are large.

2044: 

2045: \begin{lem}{\bf (Existence of a good direction):}\label{b(d)}

2046: There is a constant $b(d)>0$ such that for any set of unit vectors

2047: $\{ \hat n_i\}_{i\in I}\subset \BR^d$, that satisfy the condition 

2048: $\hat n_i\cdot \hat n_j \geq 0$ for any $i,j\in I$, there is a unit vector

2049: $\hat f$ such that $\hat f\cdot \hat n_i > \frac{1}{b(d)}$ for any $i\in I$.

2050: \end{lem}

2051: 

2052: \begin{proof}

2053: Let $\gD$ be a maximal $1$-discrete subset of $\{ \hat n_i\}_{i\in I}$.

2054: Then there are at most $\frac{b(d)}{2}$ elements in $\gD$, where $b(d)$ is 

2055: some constant.

2056: For any $\hat n_i \in \{ \hat n_i\}_{i\in I}$ there is 

2057: $\hat n_{i_0}\in \gD$ with $\| \hat n_i-\hat n_{i_0}\| <1$, which implies

2058: $\hat n_i \cdot \hat n_{i_0} >1/2$.

2059: Let 

2060: $$

2061:  \hat f=\frac{\Sigma_{\hat n_j\in \gD} \hat n_j}{\|\Sigma_{\hat n_j\in \gD} 

2062:  \hat n_j\|},

2063: $$

2064: then 

2065: $$

2066:  \hat n_i \cdot \hat f >\hat n_i \cdot \frac{\Sigma_{\hat n_j\in \gD} 

2067:  \hat n_j}{\frac{b(d)}{2}}\geq

2068:  \frac{2}{b(d)}\hat n_i \cdot \hat n_{i_0}\geq \frac{1}{b(d)}.

2069: $$

2070: \end{proof}

2071: 

2072: Summarizing the above discussion together with the arguments of the previous 

2073: section, we conclude the following less general but more practical version of 

2074: Lemma \ref{simplicial-complex}:

2075: 

2076: \begin{thm}\label{nice-boundary}

2077: Let $S$ be a symmetric space of non-compact type, $M=\gC\backslash S$ an $S$-manifold, and

2078: $\gep'\geq\gep >0$. Assume that: 

2079: \begin{itemize}

2080: \item 

2081: $M'\subset M_{\geq\gep}$ is a submanifold with boundary

2082: for which the complement of the pre-image in the universal covering is given

2083: by 

2084: $$

2085:  \ti X=S\setminus\ti M'=\cup_{\gc\in\gC'}\{d_\gc <\gep'\},

2086: $$

2087: where $\gC'$ is a subset of $\gC$ which is invariant under conjugation by elements of $\gC$.

2088: \item

2089: $M'$ is homotopically equivalent to its $\frac{\gep}{2}$-shrinking

2090: $)M'(_\frac{\gep}{2}$.

2091: \item 

2092: For any $x\in S$ the group 

2093: $$

2094:  \langle \gc\in\gC':d_\gc (x) \leq 3\gep'\rangle

2095: $$

2096: is either unipotent or commutative.

2097: \end{itemize} 

2098: Then there are positive constants $d,\ga$ which depend on $S,~\gep,\gep'$, 

2099: such that $M'$ is homotopically equivalent to some 

2100: $\big( d,\ga\cdot\vol (M')\big)$-simplicial complex.

2101: \end{thm}

2102: 

2103: \begin{proof}

2104: We shall show that the conditions of Lemma \ref{simplicial-complex} are 

2105: satisfied. 

2106: The sets $\{d_\gc <\gep'\}$ are convex and open with smooth boundary and 

2107: $X$ is the locally finite union of them.\\

2108: 

2109: It is given that $M'$ is contained in the $\gep$-thick part and that it is

2110: homotopic to its $\frac{\gep}{2}$-shrinking.\\

2111: 

2112: We should indicate how to define the constant $b$ and the direction 

2113: $\hat n(x)$, from the conditions of Lemma \ref{simplicial-complex}.

2114: 

2115: We define

2116: $$

2117:  b=\max \{\frac{2}{\gb (\gep')},b(d)\}

2118: $$ 

2119: where $\gb$ is the function defined in the proof of Theorem \ref{thm1} in the 

2120: previous section, and $b(d)$ is the

2121: constant defined in Lemma \ref{b(d)} for $d=\dim S$.

2122: 

2123: Next we define the directions $\hat n(x)$. Let $x\in S\setminus\ti X$ be a 

2124: point with $d(x,\ti X)\leq\gep$. If the group 

2125: $\langle \gc\in\gC':d_\gc (x) \leq 3\gep'\rangle$ is unipotent, then we define

2126: the direction $\hat n(x)$ in the same way as it is done in the previous 

2127: section.

2128: (It is shown in the previous section that in this case the conditions of 

2129: Lemma \ref{simplicial-complex} are satisfied.)

2130: Assume that the group $\langle \gc\in\gC':d_\gc (x) \leq 3\gep'\rangle$ is 

2131: abelian. Let $A(x)\subset \gC'$ be the set of elements $\gc\in \gC'$

2132: with $D_{\{d_\gc <\gep'\}}(x)=d(x,\ti X)$. Then for $\gc\in A(x)$

2133: $$

2134:  d_\gc (x)\leq 2d(x,\ti X)+\gep'\leq 3\gep',

2135: $$

2136: and hence $A(x)$ is contained in 

2137: $\langle \gc\in\gC':d_\gc (x) \leq 3\gep'\rangle$. Thus $A(x)$ is abelian.

2138: By Lemma \ref{bigA}, for any $\ga ,\gb\in A(x)$ and any 

2139: $y\in\partial \{d_\ga <\gep'\}\cap\partial \{d_\gb <\gep'\}$, the inner 

2140: product of the corresponding external normal vectors at $y$ is non-negative. 

2141: By Lemma \ref{monoton},

2142: the inner product at $T_x(S)$ of the external normal vectors to

2143: $\partial (\{d_\ga <\gep'\})_{d(x,\ti X)}$ and

2144: $\partial (\{d_\gb <\gep'\})_{d(x,\ti X)}$ at $x$, is also non-negative. 

2145: Therefore it follows from Lemma \ref{b(d)} there is a direction $\hat f\in T_x(S)$ 

2146: for which the inner product of $\hat f$ with the external normal vector at $x$ to

2147: $\partial (\{d_\gc <\gep'\})_{d(x,\ti X)}$ is $\geq\frac{1}{b(d)}$ for any 

2148: $\gc\in A(x)$.

2149: Thus, we can take $\hat n (x)=\hat f$.

2150: \end{proof}

2151: 

2152: 

2153: 

2154: 

2155: 

2156: 

2157: %NEW SECTION

2158: 

2159: 

2160: 

2161: 

2162: 

2163: \section{The proof of \ref{thmA}(2) in the rank-1 case}\label{rk-1}

2164: 

2165: Theorem \ref{thmA}(2) can be proved, independently of the rank, using the argument

2166: that we shall present in section \ref{pfBC}. 

2167: For two reasons we chose to do the rank-1

2168: case separately. The first reason is that in the remaining higher rank case all

2169: locally symmetric manifolds of finite volume are arithmetic. Since Theorem

2170: \ref{thmA} already took care of the non-compact arithmetic case, we can assume compactness when 

2171: considering higher rank manifolds. This would make the argument in section \ref{9} simpler.

2172: The second reason is that in the rank-1 case there is another proof, which is

2173: in some sense more simple. This proof, which we shall present below, is 

2174: basically the same as the one given in \cite{BGLM} for the hyperbolic case, and with the tools

2175: developed in sections 3,4,6 and 7, it could be applied to general rank-1 symmetric spaces which

2176: are not necessarily of constant curvature.

2177: 

2178: \begin{rem}

2179: For $S=\BH^2$ the statement of Conjecture \ref{conjA} follows 

2180: from the Gauss Bonnet theorem without the arithmeticity assumption.

2181: This is because the volume determines the genus and bound the possible number of

2182: cusps. For this reason we allow 

2183: ourselves to ignore this case in this and in the following section.

2184: \end{rem}

2185: 

2186: Recall what we intend to prove:

2187: 

2188: \begin{thm}[{\bf Theorem \ref{thmA}(2) for the rank-1 case}]\label{rk-1thm}

2189: If $S$ is a rank-$1$ symmetric space of dimension $\geq 4$, then for some 

2190: constants $\ga ,d$, the fundamental group $\pi_1(M)$ of any 

2191: $S$-manifold $M=\gC\backslash S$ is isomorphic to the fundamental group of some

2192: $\big( d,\ga\cdot\vol (M)\big)$-simplicial complex.

2193: \end{thm}

2194: 

2195: We shall use the ordinary thick-thin decomposition. 

2196: Let $\gep_s$ be the constant from the Margulis lemma, and let 

2197: $\gep =\frac{\gep_s}{2}$.

2198: 

2199: \begin{thm}[{\bf Thick-thin decomposition in rank-1}](see \cite{Th} section 4.5, and \cite{BGS} 

2200: section 10)\label{thick-thin} 

2201: Assume that $\rk (S)=1$. Let $M$ be an $S$-manifold of finite volume and $M_{\leq \epsilon}^0$ an 

2202: arbitrary connected component of $M_{\leq \epsilon}$. Then $M_{\leq \epsilon}^0$ belongs to either 

2203: one of the following two types:

2204: \begin{enumerate}

2205: \item 

2206: A tube, i.e. a tubular neighborhood of a short geodesic. In this case 

2207: $M_{\leq \epsilon}^0$ is topologically a ball-bundle 

2208: over the circle. Its fundamental group $\pi_1(M_{\leq \epsilon}^0)$ is 

2209: infinite cyclic, and in particular abelian.

2210: 

2211: or

2212:  

2213: \item  

2214: A cusp. In which case $M_{\leq \epsilon}^0$ is homeomorphic to 

2215: $\BR^{\geq 0}\times \partial M_{\leq \epsilon}^0$, where 

2216: $\partial M_{\leq \epsilon}^0$ is topologically a sub-manifold of 

2217: codimension $1$.

2218: A connected component of its pre-image  $\tilde  M_{\leq \epsilon}^0$ is

2219: given by $\tilde  M_{\leq \epsilon}^0=

2220: \cup_{\gc \in \gC_0}\{ d_\gc \leq \epsilon \}$ where $\gC_0\leq\gC$ is a 

2221: subgroup isomorphic to $\pi_1(M_{\leq\gep}^0)$, and it is a ``star-shaped''

2222: neighborhood of some point $z\in S(\infty )$ (i.e. each geodesic line with

2223: $c(\infty )=z$ enters once into $\tilde M_{\leq \epsilon}^0$ and stays in it).

2224: $\gC_0$ consists of unipotent elements only,

2225: and each $\gc \in \gC_0$

2226: fixes $z$ and leaves the horospheres around $z$ invariant.

2227: Moreover $\gC_0$ is metabelian.

2228: \end{enumerate}

2229: \end{thm}

2230: 

2231: There are only finitely many connected components of $M_{\leq\gep}$.

2232: Moreover, as $\dim (M)\geq 3$ the boundary of each 

2233: connected component of $M_{\leq\gep}$ is connected, and hence the thick 

2234: part  $M_{\geq\gep}$ is connected. Each cusp is homotopically equivalent to its

2235: boundary. As $\dim (M)\geq 4$ each tube is a ball bundle over the circle,

2236: with fibers of dimension $\geq 3$, and hence, its boundary is a sphere bundle

2237: for a sphere of dimension $\geq 2$. Therefore, for each connected component of

2238: the thin part, the injection of the boundary into the component induces an

2239: isomorphism between the fundamental groups (which are both $\cong\BZ$). 

2240: Thus by Van-Kampen's theorem we obtain:

2241: 

2242: \begin{cor}

2243: $$

2244:  \pi_1(M)\cong\pi_1(M_{\geq\gep}).

2245: $$

2246: \end{cor}

2247: 

2248: Take $M'=M_{\geq\gep}$. Then the following claim 

2249: finishes the proof of Theorem \ref{rk-1thm}.

2250: 

2251: \begin{clm}

2252: $M'$ satisfies the condition of Theorem \ref{nice-boundary}, with respect 

2253: to $\gep'=\gep$.

2254: \end{clm}

2255: 

2256: \begin{proof}

2257: We only need to explain why $M'$ is homotopic to its 

2258: $\frac{\gep}{2}$-shrinking, since the other conditions of Theorem

2259: \ref{nice-boundary} follow directly from the definition of $M'$ and from Theorem \ref{thick-thin}. 

2260: 

2261: Since $\gep =\frac{\gep_s}{2}$, the $\frac{\gep}{2}$-neighborhoods of different 

2262: connected components of

2263: $M_{\leq\gep}$ are still disjoint, and hence, we can prove the homotopy 

2264: equivalence by

2265: showing that there is a deformation retract from 

2266: $\overline{(M_{\leq\gep}^0)_\frac{\gep}{2}}\setminus M_{\leq\gep}^0$ to

2267: $\partial (M_{\leq\gep}^0)_\frac{\gep}{2}$, for each connected component 

2268: $M_{\leq\gep}^0$ of $M_{\leq\gep}$.

2269: 

2270: If $M_{\leq\gep}^0$ is a cusp, then one can define such a deformation retract 

2271: by letting each point flow (at constant speed = the initial distance) along the 

2272: unique geodesic line which connects it to the unique end of the cusp, i.e. along

2273: the geodesic whose lifting in the universal covering converges, when $t\to -\infty$, to the unique

2274: limit point in $S(\infty )$ of a lifting of the cusp.

2275: 

2276: If $M_{\leq\gep}^0$ is a tube, then the lifting of $M_{\leq\gep}^0$ and 

2277: its $\frac{\gep}{2}$-neighborhood in $S=\ti M$ are both star-shaped with 

2278: respect to the lifting $c$ of the short closed geodesic which lies inside $M_{\leq\gep}^0$. 

2279: The generalized star-contraction (see \ref{GSC}) from 

2280: $\overline{(\ti M_{\leq\gep}^0)_\frac{\gep}{2}}\setminus \ti M_{\leq\gep}^0$ to

2281: $\partial (\ti M_{\leq\gep}^0)_\frac{\gep}{2}$ 

2282: projects to a deformation retract of the corresponding subsets of $M$. 

2283: \end{proof}

2284: 

2285: 

2286: 

2287: 

2288: 

2289: 

2290: 

2291: %NEW SECTION

2292: 

2293: 

2294: 

2295: 

2296: \section{The proofs of \ref{thmA}(2) and \ref{thmA}(3)}\label{pfBC}

2297: 

2298: We shall use the fact that when a low dimensional submanifold is 

2299: removed from a high dimensional manifold, it doesn't change the low dimensional

2300: homotopy groups. More precisely:

2301: 

2302: \begin{lem}\label{M-N}

2303: Let $M$ be a connected manifold and $N\subset M$ a closed (not 

2304: necessarily connected) sub-manifold.

2305: \begin{itemize}

2306: \item If $\text{codim}_M(N)\geq 2$ then there is a surjective homomorphism

2307: $$

2308:  \pi_1(M\setminus N)\to\pi_1(M).

2309: $$

2310: \item If $\text{codim}_M(N)\geq 3$ then $\pi_1(M\setminus N)\cong\pi_1(M)$.

2311: \end{itemize}

2312: \end{lem}

2313: 

2314: \begin{proof}

2315: If $\text{codim}_M(N)\geq 2$ then any closed loop can be pushed to one which 

2316: dose not intersect $N$. Similarly, if $\text{codim}_M(N)\geq 3$ then any 

2317: homotopy of loops can be pushed to $M\setminus N$.

2318: \end{proof}

2319: 

2320: The submanifold we are about to remove consists

2321: of the union of all short closed geodesics of some certain type. Observe that

2322: the lifting to $S=\ti M$ of a closed geodesic in $M=\gC\backslash S$ which corresponds to

2323: $\gc\in\gC$ is an axis of $\gc$. In particular, any two closed geodesics from

2324: the same homotopy class are parallel.

2325: 

2326: \begin{lem}\label{codimN}

2327: Let $\gc\in G^0$ be a hyperbolic isometry which projects non-trivially to each 

2328: simple factor of $G^0$, and let $\ti N\subset S$ be the union of all geodesics 

2329: which are axes of $\gc$.

2330: 

2331: \begin{itemize}

2332: \item If $S$ is not isometric to $\BH^2$ then $\text{codim}_S(\ti N)\geq 2$.

2333: \item If additionally $S$ is neither isometric to $\BH^3, \BH^2\times\BH^2$ nor to

2334: $\PSL_3(\BR )/\PSO_3(\BR )$, then $\text{codim}_S(\ti N)\geq 3$.

2335: \end{itemize}

2336: \end{lem}

2337: 

2338: \begin{proof}

2339: Let $S^*$ be an irreducible factor of $S$. Denote by $\gc^*$ the projection of 

2340: $\gc$ to the corresponding simple factor $G^*$ of $G^0$.

2341: We need to estimate $\text{codim}_{S^*}\big(\min (\gc^*)\big)$. 

2342: 

2343: If $\gc^*$ fixes a point then we just remark that since the projection $\gc^*$

2344: is non-trivial, 

2345: $\text{codim}_{S^*}\big(\min (\gc^*)\big) >0$. Otherwise let $c$ be an axis

2346: of $\gc^*$ in $S^*$ and let

2347: $$

2348:  G^*(c) = \{ g\in G^*:g\cdot c \text{~is parallel to~} c\}.

2349: $$

2350: An alternative way to define $G^*(c)$ is as follows. Pick two points

2351: $p,q\in c(\BR )$ and let $g_{p,q}=\gs_p\cdot\gs_q$ be the corresponding 

2352: transvection,  then $G^*(c)$ coincides with the centralizer group 

2353: $C_{G^*}(g_{p,q})\leq G^*$ of $g_{p,q}$.

2354: $G^*(c)$ is a closed reductive Lie subgroup of $G^*$ which acts 

2355: transitively on $N^*=$ the union of all geodesics parallel to $c$ in $S^*$. 

2356: Observe that $N^*$ contains the projection of $\ti N$ to $S^*$.

2357: 

2358: Choose an Iwasawa decomposition of $G^*$ which induces an Iwasawa 

2359: decomposition of $G^*(c)^0$, i.e. write the Iwasawa decompositions of

2360: $G^*$ and $G^*(c)$ simultaneously 

2361: $$

2362:  G^*=K\cdot A\cdot U, \text{~and~} G^*(c)^0=K(c)\cdot A\cdot U(c),

2363: $$

2364: such that $K(c)=K\cap G^*(c)^0$ and $U(c)=U\cap G^*(c)^0$.

2365: (This could be done by choosing a flat $F\supset c$ and defining the torus $A$

2366: to be the subgroup of $G^*$ which acts on $F$ by translations, and then 

2367: choosing an order on the dual of $\mathfrak{a}=\text{Lie}(A)$ such that if 

2368: $c(t)$ is given by $c(t)=e^{tH}$ for $H\in\mathfrak{a}$, then $\ga (H)\geq 0$ 

2369: for any simple 

2370: root $\ga$, and then taking the Iwasawa decomposition which corresponds to 

2371: $A$ with this ordering.) 

2372: 

2373: Then $\dim (S^*)=\dim (A\cdot U)$ while $\dim (N^*)=\dim \big( A\cdot U(c)\big)$,

2374: and hence

2375: $$

2376:  \text{codim}_{S^*}(N^*)=\text{codim}_U\big( U(c)\big).

2377: $$

2378: Let $P_{c(-\infty )}$ be the parabolic subgroup of $G^*$ which corresponds to

2379: the point $c(-\infty )\in S^*(\infty )$, and let $P^-$ be the minimal parabolic

2380: opposite to $P^+=AU$. Then, as it is easy to verify, the Lie algebra of $G^*$

2381: is the direct sum

2382: $$

2383:  \text{Lie}(G^*)=\text{Lie}(U)+\text{Lie}(P^-),

2384: $$

2385: while the Lie algebra of the parabolic $P_{c(-\infty )}$ is the direct sum

2386: $$

2387:  \text{Lie}\big( P_{c(-\infty )}\big)=

2388:  \text{Lie}\big( U(c)\big)+\text{Lie}(P^-).

2389: $$

2390: Hence $\text{codim}_U\big( U(c)\big)=\text{codim}_{G^*}(P_{c(-\infty )})$.

2391: The lemma follows from the following proposition by a standard

2392: case analysis.

2393:   

2394: \begin{prop}[see \cite{BaNe} lemma 3.4 and corollary 8]

2395: Let $H$ be a connected simple Lie group and $P\leq H$ a proper closed connected 

2396: subgroup. Then $\text{codim}_H(P)\geq\rk (H)$, and the equality can hold only

2397: if $H$ is locally isomorphic to $\SL_n(\BR )$.

2398: \end{prop}

2399: \end{proof}

2400: 

2401: We shall now prove: 

2402: 

2403: \begin{thm}[{\bf Theorem \ref{thmA}(2) from the introduction

2404: in the general case}]\label{thm2}

2405: Let $S$ be a symmetric space of non-compact type, not isometric to

2406: $\BH^2,\BH^3, \BH^2\times\BH^2, \PSL_3(\BR )/\PSO_3(\BR )$.

2407: Then there are constants $\ga ,d$ such that the fundamental group of any

2408: irreducible $S$-manifold of finite volume $M$ is isomorphic to the fundamental 

2409: group of some $\big( d,\ga\cdot\vol (M)\big)$-simplicial complex.

2410: \end{thm}

2411: 

2412: \begin{proof}[Proof of Theorem \ref{thm2}] 

2413: Let $G=\text{Isom}(S)$, let $r$ be the rank of $S$, i.e. the real rank of $G$, let $n_s$ be the

2414: index from the Margulis lemma and let $n=n_s!$. Fix $\gep$ to be one third of 

2415: the $\gep_s$ from the Margulis lemma.

2416: 

2417: Since we already proved the theorem in the rank one case, we can assume that $\rk (S)\geq 2$.

2418: Then any $S$-manifold is arithmetic. Moreover, as Theorem \ref{thm1} implies Theorem \ref{thm2}

2419: for non-compact arithmetic manifold, we may assume that $M=\gC\backslash S$ is compact.

2420: In this case, we have the following strengthening of the Margulis lemma.

2421: 

2422: \begin{lem}

2423: For any $x\in S$, the group $\gC_{\gep_s} (x)$ contains an abelian subgroup of

2424: index $n_s$.

2425: \end{lem}

2426: 

2427: \begin{proof}

2428: By the Margulis lemma $\gC_{\gep_s} (x)$ contains a subgroup 

2429: $\gC_{\gep_s}^0 (x)$ of index $n_s$ which is contained in a connected nilpotent

2430: Lie subgroup of $G$. By Lie's theorem $\gC_{\gep_s}^0 (x)$ is triangulable

2431: over $\BC$. Therefore its commutator $[\gC_{\gep_s}^0 (x),\gC_{\gep_s}^0 (x)]$

2432: contains only unipotents. As $\gC$ is cocompact it has no 

2433: non-trivial unipotents. Thus $\gC_{\gep_s}^0 (x)$ is abelian.

2434: \end{proof}

2435: 

2436: It follows that when $\gc_1 ,\gc_2\in\gC$ satisfy 

2437: $\{ d_{\gc_1}\leq\gep\}\cap\{ d_{\gc_2}\leq\gep\}\neq\emptyset$ then $\gc_1^n$

2438: and $\gc_2^n$ commute.

2439: 

2440: Consider an element $\gc\in\gC$. Replacing $\gc$ by $\gc^{[G:G^0]}$, we may

2441: assume that $\gc\in G^0$. As $\gC\cap G^0$ is a uniform 

2442: lattice in $G^0$, $\gc$ is semisimple.

2443: Consider the sequence of centralizers

2444: $$

2445:  C_G(\gc)\leq C_G(\gc^n)\leq C_G(\gc^{n^2})\leq C_G(\gc^{n^3})\leq\ldots

2446:  \leq C_G(\gc^{n^r}).

2447: $$

2448: As the centralizer of a semisimple element is determined by the type of the 

2449: singularity of the element and by a torus which contains it, two consecutive

2450: terms $C_G(\gc^{n^i})$ and $C_G(\gc^{n^{i+1}})$ in this sequence 

2451: must coincide. Take $i$ to be the first time at which this happens and write 

2452: $\gc'=\gc^{n^i}$. In this way we attach $\gc'$ to each $\gc\in\gC$.

2453: Notice that if $\gc\in G^0$ then $\gc'=\gc^j$ for some $j\leq n^r$, 

2454: and in general $\gc'=\gc^j$ for some $j\leq [G:G^0]n^r$. Set $m=[G:G^0]n^r$.

2455: 

2456: The reason we prefer to work with the $\gc'$'s is the following. If 

2457: $$

2458: \{d_{\gc_1'}\leq 3\gep\}\cap\{d_{\gc_2'}\leq 3\gep\}\neq\emptyset

2459: $$ 

2460: then $\gc_1'$ and $\gc_2'$ commute. Of course, by the above lemma, the 

2461: non-empty 

2462: intersection implies that $\gc_1'^n,\gc_2'^n$ commute. Since $\gc_1'$ has the 

2463: same centralizer as $\gc_1'^n$, this implies that $\gc_1',\gc_2'^n$ commute, 

2464: and since $\gc_2'$ has the same centralizer as $\gc_2'^n$,  

2465: this implies that $\gc_1',\gc_2'$ commute.

2466: 

2467: Let $N\subset M$ be the subset which consists of the union of all

2468: closed geodesics of length $\leq\gep$ which correspond to elements of the

2469: form $\gc'$ for $\gc\neq 1$. In other words, its pre-image in $S$ is given by 

2470: $$

2471:  \ti N=\cup\{\min (\gc' ):\gc\in\gC\setminus\{ 1\} ,\min (d_{\gc'})\leq\gep\}.

2472: $$ 

2473: Then $N$ is a finite union of totally geodesic closed submanifolds. 

2474: Since $\gC$ is irreducible and $G$ is center free, any $\gc\in\gC\setminus\{ 1\}$ 

2475: projects non-trivially to each factor of $G$. So it follows from

2476: Lemma \ref{codimN} that $\text{codim}_M(N)\geq 3$, and hence, by Lemma 

2477: \ref{M-N}

2478: $$

2479:  \gC\cong\pi_1(M)\cong\pi_1(M\setminus N).

2480: $$

2481: 

2482: Define

2483: $$

2484:  \ti X=\cup_{\gc\in\gC\setminus\{ 1\}}\{ d_{\gc'}<\gep\},~~

2485:  \ti M'=S\setminus \ti X,

2486: $$

2487: and 

2488: $$

2489:  X=\gC\backslash\ti X,~~M'=\gC\backslash\ti M'=M\setminus X.

2490: $$

2491: 

2492: Clearly $N\subset X$.

2493: In order to prove the theorem we shall show:

2494: \begin{enumerate}

2495: \item 

2496: $M\setminus N$ and $M'$ are homotopically equivalent.

2497: \item 

2498: $M'$ satisfies the conditions of Theorem \ref{nice-boundary}

2499: with respect to $\gep$ and $\frac{\gep}{m}$ (corresponding to $\gep'$ and 

2500: $\gep$ respectively in \ref{nice-boundary} ).

2501: \end{enumerate}

2502: 

2503: {\bf Proof of (1):}

2504: We shall construct the desired homotopy in two steps.

2505: Since $M$ and $N$ are compact, and each connected component of $N$ is a 

2506: finite union of totally geodesic submanifolds, there is small positive number 

2507: $\eta>0$, such that the 

2508: $\eta$-neighborhoods of the components of $N$ are still disjoint and contained 

2509: in $X$. It is easy to verify that if $\eta$ is small,

2510: $M\setminus N$ and $M\setminus \overline{(N)_\eta}$ are diffeomorphic.

2511: 

2512: Next, we claim that there is a deformation retract from 

2513: $M\setminus \overline{(N)_\eta}$ to $M'$. In order to show this, we shall

2514: apply Lemma \ref{deformation-retract} to the 

2515: set of functions 

2516: $$

2517:  \mathcal{F}=\{d_{\gc'}:\gc\in\gC\setminus\{1\},\min (d_{\gc'})<\gep\}

2518: $$ 

2519: on $Y=L=S\setminus\overline{(\ti N)_\eta}$, and $M'=\mathcal{F}_{\geq\gep}$. 

2520: We have to indicate what are the directions 

2521: $\hat n(x)$, and what is the continuous function $\gb$.

2522: 

2523: Let $x\in \ti X\setminus\overline{(\ti N)_\eta}$, and let 

2524: $\{ d_{\gc'_1},\ldots ,d_{\gc'_k}\}\subset \mathcal{F}$ be the subset of

2525: functions satisfying $d_{\gc'_i}(x)\leq 3\gep$. Then the group

2526: $\langle \gc'_1,\ldots ,\gc'_k\rangle$ is abelian. Thus for each $j\leq k$

2527: the convex set $\cap_{i=1}^{j-1}\min (\gc'_i)$ is $\gc'_j$ invariant, 

2528: and hence, by induction 

2529: $$

2530: \cap_{i=1}^j\min (\gc'_i)\neq\emptyset.

2531: $$

2532: Pick arbitrarily $y\in\cap_{i=1}^k\min ({\gc'_i})$ and define 

2533: $\hat n(x)$ to be the tangent at $x$ 

2534: to the geodesic line which goes from $y$ to $x$.

2535: 

2536: For $t\leq 3\gep$ we define $\gb (t)$ to be the minimum of the directional 

2537: derivative of $d_{\gc'}$ at $x$ with respect to the tangent to the geodesics 

2538: $\overline{z,x}$, where this minimum is taken over all 

2539: $x\in\ti X\setminus (\ti N)_\eta$, all $d_{\gc'}\in \mathcal{F}$ with 

2540: $d_{\gc'}(x)=t$, and all

2541: $z\in\min (\gc' )$.

2542: Since 

2543: \begin{itemize}

2544: \item

2545: $N$ and $M\setminus (N)_\eta$ are compact, 

2546: \item

2547: up to conjugations there are only finitely many $\gc'$'s with 

2548: $\min (d_{\gc'})<\gep'$, and 

2549: \item 

2550: for any selection

2551: of $x,z,\gc'$ as above, the corresponding directional derivative is positive,

2552: \end{itemize}

2553: it follows that $\gb$ is a well defined

2554: continuous positive function.\\

2555: 

2556: {\bf Proof of (2):}

2557: We shall check that the conditions of Theorem

2558: \ref{nice-boundary} are satisfied.

2559: 

2560: Let $x\in S$ and assume $d_\gc (x)\leq\frac{\gep}{m}$

2561: for some $\gc\neq 1$ in $\gC$. As $\gc'=\gc^j$ for some $j\leq m$, it follows

2562: that 

2563: $$

2564: d_{\gc'}(x)=d_{\gc^j}(x)\leq j\cdot d_\gc (x)\leq\frac{j\cdot\gep}{m}\leq\gep.

2565: $$

2566: Thus $x\in\ti X$. This shows that $M'$ is contained in 

2567: $M_{\geq\frac{\gep}{m}}$.

2568: 

2569: If $\{ d_{\gc_1'}\leq 3\gep\}\cap\{ d_{\gc_2'}\leq 3\gep\}$ then 

2570: $\gc'_1$ commutes with $\gc'_2$. Thus, the last condition of Theorem

2571: \ref{nice-boundary} is also satisfied.

2572: 

2573: We shall show that there is a deformation retract from $M'$ to its 

2574: $\frac{\gep}{2m}$-shrinking, by an analogous way to the second step in the 

2575: proof of (1) above. 

2576: This time take

2577: $$

2578:  \mathcal{F}=\{D_{\{ d_{\gc'}\leq\gep\}}:\gc\in\gC\setminus\{1\},\min (d_{\gc'})<\gep\}

2579: $$ 

2580: while $\frac{\gep}{2m}$ plays the role of $\gep$ in Lemma 

2581: \ref{deformation-retract}.

2582: We define the directions $\hat n(x)\in T_x(S)$ 

2583: to be the tangent to the geodesic $\overline{y,x}$ for arbitrary

2584: $y\in\cap\{\min (\gc' ):d_{\gc'}\in\Psi_{x,3\gep}\}$. 

2585: Additionally, we define

2586: $\gb (t)$ to be the minimum of the directional 

2587: derivative of $D_{\{d_{\gc'}\leq\gep\}}$ at $x$ with respect to the tangent 

2588: of the geodesics $\overline{z,x}$, where the minimum is taken over all 

2589: $x\in S$, all $D_{\{d_{\gc'}\leq\gep\}}\in \mathcal{F}$ with 

2590: $D_{\{d_{\gc'}\leq\gep\}}(x)=t$, and all

2591: $z\in\min (\gc' )$. Since for $t\geq 0$ and for a relevant $\gc'$, the set

2592: $\{ D_{\{d_{\gc'}\leq\gep\}}\leq t\}$ is a convex body with smooth boundary and since

2593: any $z$ as above belongs to the interior of this set, the directional

2594: derivatives mentioned above are always positive. By compactness we get that

2595: $\gb$ is a continuous positive function defined for any $0\leq t\leq\frac{3\gep}{2m}$.

2596: This finishes the proof of Theorem \ref{thm2} 

2597: \end{proof}

2598: 

2599: Next we prove:

2600: 

2601: \begin{thm}[{\bf Theorem \ref{thmA}(3) of the introduction}]\label{thm3}

2602: For any $S$, there are constants $\ga ,d$, such that the fundamental group

2603: of any irreducible $S$-manifold $M$ is isomorphic to a quotient of the fundamental group 

2604: of some $\big( d,\ga\cdot\vol (M)\big)$-simplicial complex. 

2605: \end{thm}

2606: 

2607: \begin{proof}

2608: There are only $3$ cases left to deal with. $\PSL_3(\BR )/\PSO_3(\BR ),

2609: \BH^2\times\BH^2$ and $\BH^3$.

2610: 

2611: The proof for the first two cases goes verbatim as the proof of Theorem 

2612: \ref{thm2}, since these cases are of higher rank and we can assume compactness.

2613: The only difference is that in these cases, we have only $\text{codim}_M(N)\geq 2$ (instead of $\geq 3$) by 

2614: the first part of Lemma \ref{codimN}, 

2615: so the result follows from the first part of Lemma \ref{M-N}.

2616: 

2617: For hyperbolic $3$-manifolds, one should only throw out finitely many circles

2618: which are all the closed geodesics of length $\leq\gep$ and then prove that

2619: what is left is homotopically equivalent to the $\gep$-thick part. This is 

2620: done by deforming each cusp to its boundary as in section 

2621: \ref{rk-1}, and deforming each compact component minus a circle to its 

2622: boundary, as it is done in the proof of \ref{thm2}. 

2623: Then, again, Theorem \ref{nice-boundary} finishes the proof. 

2624: \end{proof}

2625: 

2626: 

2627: 

2628: 

2629: 

2630: 

2631: %new section

2632: 

2633: 

2634: 

2635: 

2636: 

2637: 

2638: 

2639: \section{Some remarks on Conjecture \ref{conjA}, Theorem \ref{thmA} and their relations to algebraic number 

2640: theory}\label{9} 

2641: 

2642: Let $p(x)\in\BZ [x]$ be an integral monic polynomial, and let

2643: $$

2644:  p(x)=\prod_{i=1}^k(x-\ga_i)

2645: $$ 

2646: be its factorization into linear factors over $\BC$. Denote by $m(p)$ its

2647: {\bf exponential Mahler measure}

2648: $$

2649:  m(p)=\prod_{|\ga_i|>1}|\ga_i|.

2650: $$

2651: The following is known as Lehmer's conjecture. 

2652: 

2653: \begin{conj}

2654: There exists a constant $\ell >0$ such that if $p(x)$ is an integral 

2655: monic polynomial with $m(P)\neq 1$, then $m(p)>1+\ell$.

2656: \end{conj}   

2657: 

2658: Denote by $d(p)$ the number of roots $\ga_i$ with absolute value $>1$

2659: $$

2660:  d(p)=\#\{ \ga_i:|\ga_i|>1\}.

2661: $$

2662: 

2663: The following conjecture of Margulis 

2664: is weaker than Lehmer's conjecture.

2665: 

2666: \begin{conj}[see \cite{Mar1} $IX$ 4.21]\label{MargulisConj}

2667: There is a function $\ell :\BN\to\BR^{>0}$ such that 

2668: $m(p)\geq 1+\ell \big( d(p)\big)$ for any non-cyclotomic monic

2669: polynomial $p(x)\in\BZ [x]$.

2670: \end{conj}

2671: 

2672: If Conjecture \ref{MargulisConj} is true, then for any 

2673: symmetric space of  non-compact type $S$, the minimal injectivity radius of any compact 

2674: arithmetic $S$-manifold is bounded from below by some positive constant $r=r(S)$ (see also 

2675: \cite{Mar1}, page 322 for a similar statement). 

2676: To see this, we argue as follows: 

2677: Let $G^0$ be the identity component of $G=\textrm{Isom}(S)$. As $G^0$

2678: is center-free we can identify it with its adjoint group 

2679: $\textrm{Ad}(G^0)\leq\textrm{GL}(\mathfrak{g})$.

2680: Let $\gC\leq G^0$ be a torsion-free uniform arithmetic lattice in $G^0$. 

2681: We think of $\gC$ as the intersection of the fundamental group of some compact

2682: arithmetic $S$-manifold with $G^0$.

2683: Since $\gC$ is arithmetic, there is a compact extension 

2684: $G^0\times O$ of $G^0$ and a $\BQ$-rational structure on the Lie algebra 

2685: $\mathfrak{g}\times \mathfrak{o}$ of $G^0\times O$,

2686: such that $\gC$ is the projection to $G^0$ of a lattice $\ti{\gC}$, which is 

2687: contained in $(G^0\times O)_{\BQ}$ and commensurable to the group of integral

2688: points $(G^0\times O)_{\BZ}$ with respect to some $\BQ$-base of

2689: $(\mathfrak{g}\times \mathfrak{o})_{\BQ}$. By changing this $\BQ$-base, we 

2690: can assume that $\ti{\gC}$ is in fact contained in $(G^0\times O)_{\BZ}$.

2691: This means that the characteristic polynomial $p_{\ti{\gc}}$ of any 

2692: $\ti{\gc}\in\ti{\gC}$ is a monic integral polynomial. 

2693: As $\gC$ is discrete and

2694: torsion-free, $m(p_{\ti{\gc}})>1$ for any $\ti{\gc}\in\ti{\gC}$

2695: which projects to a non-trivial element in $\gC$.

2696: Since $O$ is compact, any eigenvalue of $\tilde\gc$ with absolute value 

2697: different from $1$ is also an eigenvalue of its projection $\gc\in G^0$. 

2698: In particular

2699: $$

2700:  m(p_{\gc})= m(p_{\ti\gc})\geq 1+\min_{i\leq\dim G}\ell (i).

2701: $$ 

2702: We conclude that $\gc$ is outside the open set 

2703: $$

2704:  U=\{ g\in G^0: m(p_g)<1+\min_{i\leq\dim G}\ell (i)\}.

2705: $$ 

2706: Clearly $U$ contains any

2707: compact subgroup of $G$, and contains a subset of the form

2708: $$

2709:  \{ g\in G^0:g \textrm{~is semisimple and~} \min d_g < \ti r(S)\}

2710: $$ 

2711: for some positive constant $\ti r(S)$.

2712: Hence the minimal injectivity radius of $\gC\backslash S$ is $\geq \ti r(S)$.

2713: Finally, if $\gC\leq G$ is a torsion free lattice which is not necessarily 

2714: contained in $G^0$ then the minimal injectivity radius of $\gC\backslash S$ is at 

2715: least $\frac{\ti r(S)}{[G:G^0]}=r(S)$.

2716: This implies: 

2717: 

2718: \begin{cor}

2719: Conjecture \ref{MargulisConj} implies Conjecture \ref{conjA} (for compact 

2720: arithmetic manifolds).

2721: \end{cor}

2722: 

2723: \begin{proof}

2724: For a given $M$, choose a maximal $r$-discrete net $\mathcal{C}$, and take $\mathcal{R}$ to be the

2725: simplicial complex which corresponds to the nerve of the cover of $M$ by the $r$-balls whose centers form $\mathcal{C}$.

2726: Then $M$ is homotopic to $\mathcal{R}$ which is a $\big(\frac{\vol (B_{2.5r})}{\vol (B_{r/2})},

2727: \frac{\vol (M)}{\vol (B_r)}\big)$-simplicial complex.

2728: \end{proof}

2729: 

2730: For compact locally symmetric manifolds, the minimal injectivity radius  

2731: equals half of the length of the shortest close geodesic. 

2732: For non compact arithmetic $S$-manifolds we already know the absence of short 

2733: closed geodesics (see Remark \ref{5.7}). Therefore

2734: Conjecture \ref{MargulisConj} implies 

2735: 

2736: \begin{conj}\label{injrarCong}

2737: For any $S$, there exists a constant $l=l(S)$ such that no arithmetic 

2738: $S$-manifold contains a closed geodesic of length $\leq l$.

2739: \end{conj}

2740: 

2741: A simple argument, which uses the fact that for non-cyclotomic monic integral

2742: polynomials $F(x)$ of a fixed degree $k=\dim (G)+\dim (O)$, $m(F)$ is bounded

2743: away from $1$, shows that 

2744: for all the compact arithmetic $S$-manifolds, which

2745: arise by constructions in which the compact extending group $O$ is fixed, 

2746: the infimum on the minimal injectivity radius is positive, and the statement

2747: of Conjecture \ref{injrarCong} holds, independently of the rational structure 

2748: and of the specific choice of a manifold within a commensurability class. 

2749: More precisely:

2750: 

2751: \begin{prop}\label{fixedO}

2752: Given a compact semi-simple Lie group $O$, there is a positive constant

2753: $r=r(S,O)$, such that the minimal injectivity radius is $\geq r$,

2754: for any compact manifold $M=\gC\backslash S$ such that $\gC\cap G^0$ is commensurable to the group 

2755: $\pi_{G^0}\big( (G^0\times O)_\BZ\big)$ for some $\BZ$-structure on 

2756: $\mathfrak{g}\times \mathfrak{o}$.

2757: In particular, any such manifold is homotopically equivalent to a 

2758: $\big(\frac{\vol (B_{2.5r})}{\vol (B_{r/2})},

2759: \frac{\vol (M)}{\vol (B_r)}\big)$-simplicial complex.

2760: \end{prop}

2761:  

2762: A real algebraic integer $\gt >1$ is called a {\bf Salem number} if all its 

2763: conjugates in $\BC$ have absolute value $\leq 1$.

2764: In Conjecture \ref{MargulisConj} it is not even known whether $\ell (1)>0$, i.e.

2765: whether there is a positive gap between $1$ and the set of Salem 

2766: numbers. Sury \cite{sury} showed that the existence of such a gap is equivalent to the 

2767: existence of an identity neighborhood in $\SL_2(\BR )$ which intersects 

2768: trivially any uniform arithmetic lattice. 

2769: Therefore the existence of such a gap implies

2770: a positive infimum on the length of closed geodesics, when considering all the 

2771: arithmetic surfaces.

2772: We shall now discuss the relation between this gap and 

2773: manifolds locally isometric to $S=\textrm{SL}_3(\BR )/\textrm{SO}_3(\BR )$ or to 

2774: $\BH^2\times\BH^2$ (the two cases for which we could not prove the analog of Theorem \ref{thmA}(2)).

2775: 

2776: \begin{clm}

2777: Assume there is a positive gap between $1$ and the set of Salem numbers (i.e. $\ell (1)>0$).

2778: Then the analog of Theorem \ref{thmA}(2) holds also for the symmetric space

2779: $S=\textrm{SL}_3(\BR )/\textrm{SO}_3(\BR )$. 

2780: \end{clm}

2781: 

2782: \begin{proof}

2783: Let $\gt >1$, and assume that $\gc =\textrm{diag}(\gt ,\gt ,\gt^{-2})$

2784: is an element of a lattice $\gC\leq\textrm{SL}_3(\BR )$. One 

2785: can easily compute the eigenvalues of $\textrm{Ad}(\gc )$ : they are 

2786: $\gt^3,\gt^{-3}$ and $1$. As $\gC$ is arithmetic, $\gt$ is a Salem number, 

2787: and thus, by

2788: our assumption, is bounded away from $1$. We conclude that the displacement 

2789: functions $d_\gc$ of such elements are uniformly bounded away from $0$.

2790: Take 

2791: $$

2792:  \gep^* < \text{inf}\{\min d_\gc :\gc =\textrm{diag}(\gt ,\gt ,\gt^{-2}), \gt 

2793:  \text{~is a Salem number}\}

2794: $$

2795: which is also smaller than $\gep_s$ from the Margulis lemma.

2796: 

2797: If $g$ is a hyperbolic element of an arithmetic lattice in $\SL_3(\BR )$ with $\min (d_g )\leq\gep^*$, then

2798: $g$ has no real eigenvalues of multiplicity $2$. It is easy to see that $\min (g)$ is then a flat or a single

2799: geodesic. In particular $\dim\big(\min (g)\big)\leq 2$.

2800: As $\dim (S)=5$, the submanifold $N$ which consists of the union of all closed

2801: geodesics of length $\leq\gep$ has codimension $\geq 3$. 

2802: Therefore we can apply word by word the argument of

2803: the proof of Theorem \ref{thm2} for $S$ with $\gep^*$.

2804: \end{proof}     

2805: 

2806: If moreover $\ell (2)>0$, then the same is true also for the 

2807: second remaining case $S=\BH^2\times\BH^2$. 

2808: 

2809: \begin{clm}

2810: Assume $\ell (1)\cdot\ell (2)>0$. Then the analog of \ref{thmA}(2) holds also for $S=\BH^2\times\BH^2$.

2811: \end{clm}

2812: 

2813: \begin{proof}

2814: In contrast with the situation for $\SL_3(\BR )$, our problem here arises only for {\it regular}

2815: elements $\gc\in\SL_2(\BR )\times\SL_2(\BR )$ (for otherwise

2816: $\min (\gc )$ is a single geodesic and its codimension is $3$). 

2817: But such an element has always

2818: the form $\gc =(\gc_1,\gc_2)$ where $\gc_i\in\SL_2(\BR )$ are diagonalizable

2819: over $\BR$. If $\gc_i$ is conjugate to 

2820: $\text{diag}(\gt_i,\gt_i^{-1}),~\gt_i >1$

2821: then the only possible non-trivial conjugate of $\gt_1$ outside the unit disk

2822: is $\gt_2$.

2823: Therefore, assuming $\ell (1),\ell (2)>0$ we can choose $\gep^*$ small enough, so that all 

2824: arithmetic elements with minimal displacement $\leq\gep^*$ are singular, and each has a unique axis.

2825: Then we can apply the argument of the proof of Theorem \ref{thm2} to this case. 

2826: \end{proof}

2827: 

2828: \begin{rem}

2829: Similarly, $\ell (1)>0$ implies also the analog of  

2830: \ref{thmA}(2) for compact arithmetic $3$-manifolds.

2831: \end{rem} 

2832: 

2833: 

2834: 

2835: 

2836: %new section

2837: 

2838: 

2839: 

2840: 

2841: 

2842: \section{Estimating the size of a minimal presentation}

2843: 

2844: If Conjecture \ref{conjA} is true, 

2845: then, given a symmetric space of non-compact type $S$, and an $S$-manifold $M$, 

2846: such that either

2847: \begin{itemize}

2848: \item $S$ is not $\BH^3$, or

2849: \item $M$ is arithmetic,

2850: \end{itemize}

2851: the minimal size of a presentation for the fundamental group should be bounded

2852: linearly by the volume. 

2853: From Theorem \ref{thmA} we deduce this for most cases: 

2854: 

2855: \begin{defn}

2856: We say that a presentation of a group $\langle \gS : W\rangle$ is 

2857: {\bf standard} if the length of each $w\in W$ is $\leq 3$.

2858: \end{defn}

2859: 

2860: \begin{thm}\label{presentation-thm}

2861: Assume that either

2862: \begin{itemize}

2863:  \item $S$ is not isomorphic to $\BH^3, \BH^2\times\BH^2, \PSL_3(\BR )/\PSO_3(\BR )$, or

2864:  \item $M$ is non-compact arithmetic.

2865: \end{itemize}

2866: Then for some constant $\eta =\eta (S)$, independent of $M$,

2867: the fundamental group $\pi_1(M)$ admits a standard presentation 

2868: $$

2869:  \pi_1(M)\cong\langle \gS : W\rangle

2870: $$

2871: with $|\gS |,|W|\leq\eta\cdot\vol (M)$.

2872: 

2873: \end{thm}

2874: 

2875: \begin{proof}

2876: Let $\mathcal{R}$ be the $\big( d,\ga\cdot\vol (M)\big)$-simplicial complex which corresponds to $M$ by Theorem

2877: \ref{thmA} (1) or (2). Fix a spanning tree $T$ for 

2878: $\mathcal{R}$, and take the generating set $\gS$ for $\pi_1(\mathcal{R})\cong\pi_1(M)$ 

2879: which consists those closed loops which contain exactly one edge outside $T$.

2880: We thus obtain a generating set of size less then the number of edges of the 

2881: 1-skeleton $\mathcal{R}^1$ which is at most $\frac{\ga (S)\vol (M)d(S)}{2}$.

2882: In other words, we take for each edge of $\mathcal{R}^1\setminus T$ the element

2883: of $\pi_1(\mathcal{R}^1)$ which corresponds to the unique cycle (with 

2884: arbitrarily chosen orientation) which is obtained by adding this edge to $T$.

2885: Additionally, let the set of relations $W$ consist exactly 

2886: those words which are induced from 2-simplexes of $\mathcal{R}^2$ (we take

2887: one such relation for each 2-simplex). In this way we obtain

2888: a set of relations of size $\leq\ga (S)\vol (M) d^2(S)$ which is a bound for

2889: the number of triangles in $\mathcal{R}^1$.

2890: Thus, 

2891: $$

2892:  \eta =\ga (S)d^2(S)=\max\{ \ga (S)d^2(S),~\ga (S)d(S)/2\}

2893: $$

2894: will do.

2895: Finally, the length of each $w\in W$ is exactly the number of edges in the 

2896: corresponding 2-simplex which lie outside $T$, and thus the 

2897: presentation $\langle \gS : W\rangle$ is standard.

2898: \end{proof} 

2899: 

2900: \begin{rem}

2901: Lower bounds for the size of any presentation, are known for hyperbolic 

2902: $3$-manifolds (see \cite{cooper}). In this case, the upper bound obtained 

2903: above (for non-compact arithmetic $3$-manifolds) is tight. 

2904: \end{rem}

2905: 

2906: For non-arithmetic hyperbolic $3$-manifolds the analogous statement is 

2907: evidently false (see Remark \ref{55}). Surprisingly the following result holds:

2908: 

2909: \begin{thm}\label{Pre-Hyp}

2910: There is a constant $\eta$ such that the fundamental group of 

2911: any complete hyperbolic $3$-manifold $M$ admits a presentation

2912: $$

2913:  \pi_1(M)\cong\langle \gS : W\rangle

2914: $$

2915: for which both $|\gS |$ and $|W|$ are $\leq\eta\cdot\vol (M)$.

2916: \end{thm}

2917: 

2918: \begin{lem}

2919: Let $S$ be a rank one symmetric space, and let $\gep =\frac{\gep_s}{3}$

2920: where $\gep_s$ is the constant from the Margulis lemma. 

2921: There is a constant $c=c(S)$ such that for every $S$-manifold $M$, the number of closed geodesics of length $\leq\gep$ in $M$ is at most

2922: $c\cdot\vol (M)$.

2923: \end{lem}

2924: 

2925: \begin{proof}

2926: Write $M=\gC\backslash S$ and let $\ga, \gb\in\gC$ be elements which correspond to two

2927: different closed geodesics in $M$ of length $\leq\gep$. Then the axes of $\ga$

2928: and $\gb$ are bounded away from each other, and hence, for large enough $m$, $\langle\ga^m,\gb^m\rangle$ is a non-abelian

2929: free group. It follows that $\{ d_\ga <\gep_s\}\cap\{ d_\gb <\gep_s\}=\emptyset$.

2930: 

2931: Since $d_\gc (x)\leq \gep+2D_{\{ d_\gc <\gep\}}(x)$, we see that

2932: $$

2933:  \{ D_{\{ d_\ga <\gep\}}<\gep\}\cap\{ D_{\{ d_\gb <\gep\}}<\gep\}=\emptyset.

2934: $$

2935: This implies that if we take, for each connected component $M_{\leq\gep}^0$

2936: of $M_{\leq\gep}$, an $\gep$-ball $B_{\gep}$, whose center lies on the

2937: boundary of $M_{\leq\gep}^0$, then these balls are disjoint and injected.

2938: Thus the number of geodesics of length $\leq\gep$ (which coincides with

2939: the number of connected components of $M_{\leq\gep}$ and hence with the number 

2940: of these ${\gep}$-balls) is $\leq\vol (M)/\vol (B_{\gep})$.

2941: \end{proof}

2942: 

2943: \begin{proof}[Proof of Theorem \ref{Pre-Hyp}]

2944: Let $\gep =\frac{\gep_s}{3}$, let $M$ be a complete hyperbolic 

2945: $3$-manifold, and let $N\subset M$ be the union of all closed geodesics in $M$

2946: of length $\leq\gep$. Then $N$ consists of 

2947: $\leq\frac{\vol (M)}{\vol (B_{\gep})}$

2948: circles. As in the proof of \ref{thm3}, $\pi_1(M\setminus N)$ is

2949: isomorphic to $\pi_1(\mathcal{R})$ for some 

2950: $\big(\ga,d\cdot\vol(M)\big)$-simplicial complex $\mathcal{R}$.

2951: It follows that $\pi_1(M\setminus N)$ admits a presentation with 

2952: $\leq\frac{\ga d}{2}\vol (M)$ generators, and $\leq \ga d^2\vol (M)$ relations.

2953: Van-Kampen's theorem implies that when adding these circles one by one

2954: to $M\setminus N$, we should add one relation for each (note that each circle

2955: has a neighborhood homeomorphic to a solid torus or a solid Klein bottle) .

2956: Hence, we get a presentation of $\pi_1(M)$

2957: with the same number of generators and with at most 

2958: $\frac{\vol (M)}{\vol (B_\gep)}$ additional relations.

2959: \end{proof}

2960: 

2961: \begin{rem}\label{55}

2962: In contrast with Theorem \ref{presentation-thm}, Theorem \ref{Pre-Hyp} does 

2963: not yield bounds for the length of the relations. Since for $v$ large enough, 

2964: there are infinitely many complete hyperbolic $3$-manifolds with volume 

2965: $\leq v$, the length of the relations in the above presentations can not be 

2966: bound in terms of $\vol (M)$.

2967: \end{rem}

2968: 

2969: \begin{rem}

2970: For $S=\BH^2\times\BH^2$ the analog of Theorem \ref{Pre-Hyp} holds. 

2971: The proof is almost the same, except

2972: that in this case, each of the $\leq\frac{\vol (M)}{\vol (B_{\gep /2} )}$ 

2973: connected components of the union $N$ of all closed 

2974: geodesics of length $\leq\gep$ is either a circle or a two dimensional torus

2975: or a Klein bottle.

2976: \end{rem}

2977: 

2978: However for the symmetric space $S=\SL_3(\BR )/\SO_3(\BR )$ we have only the 

2979: following result which follows directly from Theorem \ref{thm3}.

2980: 

2981: \begin{prop}

2982: For any $S$, there is a constant $\eta (S)$ such that the fundamental group

2983: of any $S$-manifold $M$ has a generating set of size $\eta (S)\cdot\vol (M)$.

2984: \end{prop}

2985: 

2986: 

2987: 

2988: 

2989: 

2990: 

2991: 

2992: 

2993: 

2994: %NEW SECTION

2995: 

2996: 

2997: 

2998: 

2999: \section{A quantitative version of Wang's theorem}\label{QuaWang}

3000: 

3001: Denote by $\rho_S(v)$ the number of isometric classes of irreducible 

3002: $S$-manifolds of volume $\leq v$. 

3003: 

3004: If Conjecture \ref{conjA} is true then for any

3005: symmetric space, $S$, of non-compact type of dimension $\geq 4$, there is some constant $c=c(S)$,

3006: such that

3007: $$

3008:  \rho_S(v)\leq v^{c\cdot v}

3009: $$

3010: for any $v>0$.

3011: While in dimension $3$, the validity of Conjecture \ref{conjA} would yield an analogous upper bound

3012: for the growth of {\it arithmetic} $3$-manifolds.

3013: 

3014: Theorems \ref{thmA}(2) implies: 

3015: 

3016: \begin{thm}\label{QWthm}

3017: Assume that $S$ is neither isometric to 

3018: $\BH^2,\BH^3,\SL_3(\BR )/\SO_3(\BR )$ nor to $\BH^2\times\BH^2$.

3019: Then we have

3020: $$

3021:  \rho_S(v)\leq v^{c(S)\cdot v}.

3022: $$

3023: \end{thm}

3024: 

3025: And \ref{thmA}(1) implies:

3026: 

3027: \begin{prop}\label{12.2}

3028: The number of non-compact arithmetic hyperbolic $3$-manifolds of volume

3029: $\leq v$ is at most $v^{cv}$ for some constant $c$.

3030: \end{prop}

3031: 

3032: \begin{rem}\label{12.3}

3033: Similarly, for $S=\SL_3(\BR )/\SO_3(\BR )$ or $\BH^2\times\BH^2$, the

3034: number of non-compact irreducible manifolds with volume $\leq v$ is bounded

3035: by $v^{cv}$ for some $c$.

3036: \end{rem}

3037: 

3038: 

3039: \begin{proof}[Proof of \ref{QWthm}, \ref{12.2} and \ref{12.3}]

3040: Since $\dim (S)\geq 3$, it follows from Mostow's rigidity theorem that an 

3041: irreducible $S$-manifold $M$ is characterized by its fundamental 

3042: group, which, by Theorem \ref{presentation-thm}, has a presentation 

3043: $\pi_1(M)\cong\langle \gS , W\rangle$ with $|\gS |,|W|\leq\eta (S)\vol (M)$ in which all the relations has length $\leq 3$.

3044: A rough estimate of the number

3045: of groups admitting such a presentation yields \ref{QWthm}. 

3046: \end{proof} 

3047: 

3048: \begin{rem}

3049: It was shown in \cite{BGLM} that for $\BH^n$ when $n\geq 4$ this estimate

3050: is tight. However in the higher rank case it is very likely that $\rho_S(v)$ grows

3051: much slower. It was guessed in \cite{BGLM} that when $\rk (S)\geq 2$

3052: $$

3053: \log\rho_S(v)\approx {c(S)\frac{(\log V)^2}{\log\log V}}.

3054: $$

3055: \end{rem}

3056: 

3057: \begin{rem}

3058: Since Mostow rigidity does not hold for 

3059: surfaces, our method does not yield a quantitative version for Borel's 

3060: finiteness theorem for arithmetic hyperbolic surfaces of a given genus.

3061: \end{rem}

3062: 

3063: 

3064: 

3065: 

3066: 

3067: %NEW SECTION

3068: 

3069: 

3070: 

3071: 

3072: 

3073: \section{Some complements}\label{complements}

3074: 

3075: 

3076: \subsection{Extending some of the results to non-compact orbifolds} 

3077: 

3078: In the previous sections we have considered $S$-manifolds of finite volume. It is natural to try 

3079: to generalize the results obtained, to the larger family of $S$-orbifolds of finite volume. 

3080: This amounts to consider general lattices in $G$ instead of just torsion free lattices.

3081: 

3082: It turns out that some of the main statements could be generalized to  

3083: non-compact $S$-orbifolds, i.e. to general non-uniform lattices 

3084: $\gC\leq G$. We remark that we do not know how to deal with general compact orbifolds.

3085: In the non-compact case, our generalizations rely on the following effective version of Selberg's lemma: 

3086: 

3087: \begin{lem}\label{i(t.f.)}

3088: There is a constant $i=i(G)\in\BN$ such that any non-uniform arithmetic lattice

3089: $\gC\leq G$ has a torsion free normal subgroup of index $\leq i$.

3090: \end{lem}

3091: 

3092: \begin{proof}

3093: Let $\mathfrak{g}$ denote the Lie algebra of $G$ and let $n$ be its dimension.

3094: Replacing $G$ by its identity component we can assume it is connected and 

3095: center free, and therefore may be identified with its image under the adjoint 

3096: representation $\textrm{Ad}(G)\leq\textrm{GL}(\mathfrak{g})\cong GL_n(\BR )$. 

3097: As follows from the proof of Margulis' arithmeticity theorem, for any 

3098: non-uniform arithmetic lattice $\gC\leq G$ there is a base $B$ for the

3099: vector space $\mathfrak{g}\cong \BR^n$ with respect to which 

3100: $\gC\leq \textrm{GL}_n(\BQ )$ and is commensurable to $\textrm{GL}_n(\BZ )$.

3101: Replacing this base by a $\BZ$-base for the $\BZ$-span of $\gC\cdot B$ (which

3102: is easily seen to be a $\BZ$-lattice in $\BR^n$), we can assume that $\gC$ is 

3103: contained in $\textrm{GL}_n(\BZ )$.

3104: 

3105: Let $T\leq\textrm{GL}_n(\BZ )$ be a fixed torsion-free congruence subgroup  

3106: (which exists, for instance, by Selberg's lemma) and let $i=i(G)$ be its index

3107: $$

3108:  i=[\textrm{GL}_n(\BZ ):T].

3109: $$

3110: Clearly, $\gC\cap T$ is torsion free and $[\gC :\gC\cap T]\leq i$.

3111: \end{proof}

3112: 

3113: The following generalization of Theorem \ref{presentation-thm} follows 

3114: immediately:

3115: 

3116: \begin{thm}\label{Gpresentation}

3117: There is a constant $\eta (G)$ such that any non-uniform arithmetic 

3118: lattice $\gC\leq G$ has a presentation 

3119: $\gC\cong \langle \Sigma :W\rangle$

3120: with 

3121: $$

3122:  |\Sigma |,|W|\leq\eta (G)\cdot\vol (G/\gC ).

3123: $$

3124: \end{thm}

3125: 

3126: \begin{proof}

3127: Take a torsion free normal subgroup $\gC_1$ of index $\leq i$ in $\gC$.

3128: Then $\vol (G/\gC_1)=\vol (G/\gC )|\gC/\gC_1|\leq\vol (G/\gC )\cdot i$, and 

3129: $\gC_1$, being torsion free, has a presentation with $\leq\eta'\vol (G/\gC_1)$ generators and 

3130: relations by \ref{presentation-thm}. We should add at most $|\gC/\gC_1|\leq i$ generators and 

3131: $|\gC/\gC_1|^{|\gC/\gC_1|}\leq i^i$ relations to get a presentation for $\gC$.

3132: \end{proof}

3133: 

3134: However, unlike the case of \ref{presentation-thm}, we do not know how to

3135: bound the lengths of the relations in $W$.

3136: 

3137: \medskip

3138: 

3139: The following extends the results of section \ref{QuaWang}.

3140: 

3141: \begin{thm}\label{countingL}

3142: There is a constant $c=c(G)$ such that for any $v>0$, the number of conjugacy 

3143: classes of non-uniform arithmetic lattices of covolume $\leq v$ is at most

3144: $v^{c\cdot v}$.

3145: \end{thm}

3146: 

3147: \begin{proof}

3148: We already know (by section \ref{QuaWang}) that for any $v>0$ there are at most 

3149: $v^{c'\cdot v}$ 

3150: conjugacy classes of torsion free lattices of covolume $\leq v$, and therefore 

3151: at most $v^{c''\cdot v}$ conjugacy classes of torsion free lattices of covolume

3152: $\leq i\cdot v$ where $i=i(G)$ is the constant from Lemma \ref{i(t.f.)}.

3153: 

3154: Let $v_0$ be the minimal covolume of a lattice $\gC\leq G$. Let $\gC\leq G$ be 

3155: a lattice of covolume $\leq v$. By Lemma \ref{i(t.f.)}, $\gC$ contains a 

3156: torsion free normal subgroup $\gC '$ of index $\leq i$. Let $N_G(\gC ')$ be 

3157: the normalizer of $\gC '$ in $G$. Then $N_G(\gC ')$ is a lattice containing 

3158: $\gC$ whose covolume satisfies

3159: $$

3160:  v_0\leq \vol\big( G/N_G(\gC ')\big)\leq v.

3161: $$ 

3162: 

3163: It follows from Theorem \ref{Gpresentation} that $N_G(\gC ')$ has a generating 

3164: set of size $\leq [\eta\cdot v]$. Thus, for any $j$, the number of subgroups 

3165: of $N_G(\gC ')$ of index $j$ is no more than 

3166: $\big(j+[\eta\cdot v]\big) !^2$ 

3167: (which is a trivial upper bound for the number of index $j$ subgroups of the 

3168: free group of rank $[\eta\cdot v]$). The index $[N_G(\gC '):\gC ]$ is at most 

3169: $v/v_0$, so there are at most

3170: $\big( [v/v_0]+[\eta\cdot v]\big) !^2$ choices for $\gC$ as a

3171: subgroup of index

3172: $[N_G(\gC '):\gC ]$ of $N_G(\gC ')$. Summing over all indexes $\leq v/v_0$, 

3173: we get that there are at most

3174: $[v/v_0]\cdot \big( [v/v_0]+[\eta\cdot v]\big) !^2\leq v^{c'''\cdot v}$ 

3175: choices for $\gC$ as a

3176: subgroup of index $\leq [N_G(\gC '):\gC ]$ of $N_G(\gC ')$. 

3177: 

3178: Thus, the number of lattices $\gC$ which contain the same $\gC '$ as a 

3179: normal subgroup of index $\leq i$ is at most $v^{c'''\cdot v}$. 

3180: Since there are at most $v^{c''v}$ possibilities for $\gC'$,

3181: it follows that the number

3182: of conjugacy classes of lattices of covolume $\leq v$ is at most 

3183: $v^{c''\cdot v}\cdot v^{c'''\cdot v}\leq v^{c\cdot v}$.

3184: \end{proof}

3185: 

3186: 

3187: 

3188: 

3189: 

3190: %NEW subSECTION

3191: 

3192: 

3193: 

3194: \subsection{Commensurable growth}

3195: Let us now restrict our attention to a fixed commensurability class. 

3196: 

3197: \begin{defn}

3198: Two $S$-manifolds $M,N$ are called {\bf commensurable} if they have a common

3199: finite cover. I.e. $\gC_1\backslash S$ is commensurable to $\gC_2\backslash S$ iff $\gC_1$ is

3200: commensurable to some conjugate of $\gC_2$ in $G=\text{Isom}(S)$.

3201: \end{defn}

3202:    

3203: The following definition is natural.

3204: 

3205: \begin{defn}

3206: The {\bf commensurable growth} $\kappa_M(v)$ of a locally symmetric manifold 

3207: $M$, is the number of non-isometric manifolds commensurable to $M$  

3208: with volume $\leq v$.

3209: The {\bf commensurable growth} $\kappa_\gC (v)$ of a lattice $\gC\leq G$ is the

3210: number of conjugacy classes of lattices commensurable to $\gC$ with

3211: covolume $\leq v$.

3212: \end{defn}

3213: 

3214: %{\bf Some remarks and questions}

3215: 

3216: %\begin{itemize}

3217: %\item 

3218: One can define the notion of commensurable growth for arbitrary subgroup

3219: $\gC\leq G$, not necessarily a lattice, as follows: define the 

3220: generalized index between commensurable subgroups

3221: $\gC ,\gC'\leq G$ to be the rational number 

3222: $$

3223:  [\gC :\gC']=\frac{[\gC :\gC\cap\gC']}{[\gC' :\gC\cap\gC']},

3224: $$ 

3225: and use this concept instead of ``covolume'' in the above definition.

3226: %\item 

3227: 

3228: Clearly, for a locally symmetric manifold $M=\gC\backslash S$, we have

3229: $\kappa_M(v)\leq\kappa_\gC (v)$. It is natural to ask what is the relation

3230: between these functions. In particular, 

3231: do they have the same asymptotic behavior?    

3232: %\item 

3233: 

3234: Another interesting question is what is the relation between the commensurable growth  

3235: and the congruence subgroup problem.

3236: %\end{itemize}

3237: 

3238: \medskip

3239: 

3240: We shall now give upper bounds for the commensurable growth of locally 

3241: symmetric manifolds and for its fundamental group when the dimension is $>2$.

3242: 

3243: \begin{clm}

3244: Let $M=\gC\backslash S$ be an irreducible locally symmetric manifold of dimension $>2$ 

3245: with finite volume. Then there is a constant $c=c(M)$ such that 

3246: $\kappa_M(v)\leq v^{cv}$.

3247: \end{clm}

3248: 

3249: \begin{proof}

3250: If $M=\gC\backslash S$ is not arithmetic then, by Margulis' criterion for arithmeticity,

3251: the commensurability class of $\gC$ admits a unique maximal element which 

3252: contains all the others, and the result follows by considering the subgroup 

3253: growth of this maximal element. If $M$ is arithmetic, this follows from \ref{fixedO} and from 

3254: \ref{thm1}.

3255: \end{proof}

3256: 

3257: When $\gC\leq G$ is a non-arithmetic lattice, the above proof applies also to $\kappa_\gC$ and gives 

3258: $\kappa_\gC (v)\leq v^{cv}$ as well. In the arithmetic case, as in 

3259: \ref{countingL}, we obtain similar upper bounds by using the following 

3260: lemma, which can be proved in the same way as Lemma \ref{i(t.f.)}.

3261: 

3262: \begin{lem}\label{180}

3263: For any commensurability class $\mathfrak{N}$ of arithmetic lattices in $G$,

3264: there is a constant $i=i(\mathfrak{N})$ such that any $\gC\in\mathfrak{N}$ 

3265: contains a torsion free subgroup of index $\leq i$.

3266: \end{lem}

3267: 

3268: The following is immediate from \ref{fixedO} and \ref{180}.

3269: 

3270: \begin{lem}

3271: Given a commensurability class $\mathfrak{N}$ of arithmetic lattices in $G$,

3272: there is a constant $\eta =\eta (\mathfrak{N})$ such that any 

3273: $\gC\in\mathfrak{N}$

3274: has a presentation $\gC\cong \langle \Sigma :W\rangle$

3275: with $|\Sigma |,|W|\leq\eta (G)\cdot\vol (G/\gC )$.

3276: \end{lem}

3277: 

3278: As in the proof of \ref{countingL}, these two lemmas imply:

3279: 

3280: \begin{prop}

3281: For any lattice $\gC\leq G$, $\kappa_\gC (v)\leq v^{c(\gC )v}$. 

3282: \end{prop}

3283: 

3284: 

3285: 

3286: 

3287: 

3288: 

3289: %NEW subSECTION

3290: 

3291: 

3292: 

3293: 

3294: 

3295: \subsection{How to construct a simplicial complex for non-arithmetic manifolds}

3296: 

3297: We shall now explain how to attach simplicial complexes to

3298: non-arithmetic, or more generally to rank-$1$ manifolds, of dimension $\geq 4$. We are not

3299: trying to do it in the most economical way, but just to explain an idea of how this could be done.

3300: 

3301: The following proposition follows from a rough estimate for the diameter and 

3302: the minimal injectivity radius of compact connected components of the thin 

3303: part.

3304: 

3305: \begin{prop}\label{rank-1prop}

3306: There are positive constants $\ga =\ga (n)$ and $d=d(n)$ such that 

3307: any compact rank-$1$ locally symmetric manifold $M$ of dimension $n\geq 4$ is 

3308: homotopically equivalent to a 

3309: $\big(d,\ga\cdot\vol (M)^{3n^2+1}\big)$-simplicial complex.

3310: \end{prop}

3311: 

3312: \begin{proof}

3313: Fix $n$, let $\gep (n)$ be the constant of the Margulis lemma, and let 

3314: $M_{\leq\gep (n)}$ be the thin part of the ordinary $\gep (n)$ thick-thin 

3315: decomposition.

3316: 

3317: It follows from \cite{BS} (see proposition 3.2 there) that for some constant

3318: $c$, the diameter of any compact connected component of $M_{\leq\gep (n)}$ is

3319: at most $3\log \big( c\cdot\vol (M)\big)$. 

3320: Applying formula 8.5 from \cite{Gr} (page 381), which implies that the 

3321: injectivity radius decreases at most exponentially as one moves along the 

3322: manifold, we get that the injectivity radius at any point belonging to

3323: $M_{\leq\gep (n)}$, and therefore at any point of $M$, 

3324: is at least $c'\cdot\vol (M)^{-3n}$, for some positive constant $c'$. 

3325: 

3326: The proposition follows by applying a good covering argument with ordinary 

3327: balls of radius $\gep =c'\cdot\vol (M)^{-3n}$. The needed number of balls in such a

3328: cover is $\leq c''\frac{\vol(M)}{\gep^n}\leq\ga\cdot\vol (M)^{3n^2+1}$.

3329: \end{proof}     

3330: 

3331: In fact, it is easy to obtain stronger results by means of elementary 

3332: computations of volumes of neighborhoods of short closed geodesics. 

3333: We shall demonstrate this in the real hyperbolic case. We remark that in all

3334: other rank-$1$ cases, one can obtain similar estimates by using the same

3335: means, and applying Rauch's comparison theorems (see \cite{CH} 1.10). 

3336: However, also the following estimate is probably not tight, and

3337: it should be possible (and not necessarily very hard) to obtain better 

3338: estimates. 

3339: 

3340: \begin{prop}

3341: For $n\geq 4$, there are constants $\ga =\ga (n), d=d(n)$, such that any 

3342: compact hyperbolic $n$-manifold $M$ is homotopically 

3343: equivalent to some 

3344: $\big(d,\ga\cdot \vol (M)^{(1+n( [\frac{n-1}{2}]+1))/(n-2-[\frac{n-1}{2}])}

3345: \big)$-simplicial complex. 

3346: \end{prop}

3347: 

3348: \begin{proof}[Sketched proof for $n=4$]

3349: In this case any connected

3350: component of the thin part of the ordinary thick-thin decomposition (with 

3351: respect to some fixed $\gep$) is a neighborhood of a short closed geodesic 

3352: which is topologically a ball bundle over a circle.

3353: In order to understand the geometry of the thin components it is most 

3354: convenient to look at the upper half-space model for the hyperbolic space

3355: $\BH^4$ (see \cite{BePe} for details). Lift the component so that the short 

3356: closed geodesic is lifted to the line connecting $0$ to $\infty$. 

3357: Then our lifted 

3358: component is a cone, centered by this line. The intersection of this cone with

3359: a horosphere perpendicular to this line is a union of coaxial ellipsoids (with 

3360: respect to the induced $(n-1)$-Euclidean structure on the horosphere). 

3361: 

3362: Using the fact that any abelian subgroup of $\textrm{SO}_3(\BR )$ 

3363: ($\cong$ to the fixator group of this line) is contained 

3364: in a $1$-dimensional torus, one can show, by a simple pigeonhole argument on the powers of the corresponding 

3365: isometry, that if the length of our short closed

3366: geodesic is $a<<\gep$ then, for some constant $c$, the $3$ dimensional

3367: Euclidean ball of radius $c\frac{1}{a^{1/2}}$ is contained in the union of 

3368: the above ellipsoids. This implies that the volume of the component is at least

3369: $c'(1/a^{1/2})^3\cdot a$. Thus $a\geq c''\frac{1}{\vol (M)^2}$. 

3370: 

3371: Thus the injectivity radius at any point of $M$ is at least 

3372: $\rho =c''\frac{1}{\vol (M)^2}$, 

3373: and one can construct, as above, a simplicial complex with at most

3374: $c'''\frac{\vol (M)}{\rho^4}=\ga\cdot\vol (M)^9$ vertices, all of them of 

3375: degree bounded by some constant $d$.

3376: 

3377: We remark that the proof for general dimension $n$ uses the fact that any 

3378: abelian subgroup of $\textrm{SO}_{n-1}(\BR )$ is contained in some 

3379: $[\frac{n-1}{2}]$-dimensional torus.

3380: \end{proof}

3381: 

3382: \begin{rem}

3383: In order to obtain analogous estimates for non-compact rank-$1$ manifolds, one 

3384: should use the thick-thin decomposition with $\gep =\gep \big(\vol (M)\big)$

3385: as above, so that all the components of $M_{\leq\gep}$ would be cusps, and 

3386: then estimate explicitly the function $\gb (\gep)$ which is defined in the

3387: proof of Theorem \ref{thm1},

3388: and detect its influence on the determination of the constant $b=b(\gep )$

3389: in Proposition \ref{simplicial-complex} in order to finally calculate the resulting 

3390: simplicial complex.

3391: \end{rem}

3392: 

3393: 

3394: 

3395: %subsection

3396: 

3397: 

3398: 

3399: \subsection{Wang's theorem for products of $\SL_2$'s}\label{SL_2}

3400: 

3401: This paragraph is not precisely a part of the main theme of this paper, but only a part of the 

3402: same subject of mathematics. 

3403: Moreover, we are not presenting any new result here, but only clarify things which are

3404: evidently known to some peoples. The author decided to write this paragraph because it might serve

3405: as a complement to Wang's paper \cite{Wa}.

3406: 

3407: Wang's theorem states that if $G$ is a connected semisimple

3408: Lie group without compact factors, and $G$ is not locally isomorphic to 

3409: $\textrm{PSL}_2(\BR )$ or $\textrm{PSL}_2(\BC )$,

3410: then for any $v>0$, there are only 

3411: finitely many conjugacy classes of irreducible lattices in $G$ of covolume 

3412: $\leq v$.

3413: 

3414: In \cite{Wa} Wang didn't consider the case where $G$ is of higher rank and

3415: has factors locally 

3416: isomorphic to $\textrm{PSL}_2(\BR )$ or $\textrm{PSL}_2(\BC )$.

3417: Margulis' arithmeticity theorem implies that if $G$ has both a 

3418: factor which is locally isomorphic to $\textrm{PSL}_2(\BR )$ or 

3419: $\textrm{PSL}_2(\BC )$ and a factor which is not locally isomorphic

3420: $\textrm{PSL}_2(\BR )$ or $\textrm{PSL}_2(\BC )$ then $G$ contains no 

3421: irreducible lattices (see \cite{Mar1} corollary 4.5, page 315). 

3422: It was remarked by Borel (see \cite{Bor} 8.3 and 8.1)

3423: that Wang's argument implies also the finiteness

3424: of the number of conjugacy classes of irreducible lattices for groups locally

3425: isometric to $G_{a,b}=\SL_2(\BR )^a\times\SL_2(\BC )^b$ when 

3426: $(a,b)\neq (1,0),(0,1)$. 

3427: Let as now explain this remark of Borel.

3428: 

3429: It follows from Margulis' super rigidity theorem that irreducible higher rank

3430: lattices are locally rigid. Thus, the missing 

3431: ingredient in Wang's argument (\cite{Wa}, 8.1) when applied to groups 

3432: locally isometric to $G_{a,b}$ is the following statement (which was also 

3433: noted without proof in \cite{Bor}).

3434: 

3435: \begin{prop}\label{irr-lim}

3436: Let $G$ be a semi-simple Lie group with no compact factors, and let 

3437: $\Gamma_n\leq G$ be a sequence of irreducible lattices. Assume that 

3438: $(\Gamma_n)$ converges to a lattice $\gD$ in the topology of closed 

3439: subgroups (Hausdorff convergence on compact sets). Then $\gD$ is also 

3440: irreducible. 

3441: \end{prop}

3442: 

3443: \begin{proof}

3444: Assume that $\gD$ is reducible. Then we can write $G$ as an almost direct 

3445: product $G=G_1\cdot G_2$ in such a way that $\gD$ is commensurable with

3446: $\gD_1\cdot\gD_2$ where $\gD_i=\gD\cap G_i$. Fix a finite generating set for

3447: $\gD$, and for large $n$, denote by $f_n:\gD\to\gC_n$ the homomorphism induced

3448: by sending each generator to the closest element in $\gC_n$. As explained in 

3449: \cite{Wa}, since $\gD$ is finitely presented, $f_n$ is a well defined 

3450: homomorphism whenever $\gC_n$ is close enough to $\gD$.

3451: 

3452: For any $\gd\in\gD$, $f_n(\gd )\to \gd$.  

3453: We will show that $f_n(\gd )$ is central for each non-central $\gd\in\gD_1$, and for any large enough $n$. Since the center of $G$ is 

3454: discrete, this will imply the desired contradiction. 

3455:      

3456: Fix $\gd\in\gD_1$ non-central. As $\gC_n$ is irreducible, 

3457: we will show that $f_n(\gd )$ is central by showing that its projection

3458: to the second factor $\pi_2(f_n(\gd ))$ is the unit element in $G_2$. 

3459: 

3460: $\gD_2$ is a lattice in $G_2$.

3461: Let $\{\gd_{2,1},\gd_{2,2},...,\gd_{2,k}\}\in\gD_2$ be a finite set of 

3462: generators for $\gD_2$. By Borel's density theorem, 

3463: $\{\text{Ad}(\gd_{2,1}),...,\text{Ad}(\gd_{2,k})\}$ generates the algebra 

3464: $$

3465:  \langle \text{Ad} (G_2)\rangle \leq

3466:  \textrm{End}(\mathfrak{g}_2)

3467: $$ 

3468: (here $\mathfrak{g}_2$ denotes the Lie algebra 

3469: of $G_2$). Since this algebra is finite dimensional, it is generated by

3470: $\{ \text{Ad}\big(\pi_2 ( f_n (\gd_{2,i} ))\big)\}_{i=1}^k$ 

3471: whenever $n$ is 

3472: large enough. Since $G_2$ is semi-simple, the adjoint representation 

3473: $\text{Ad}:G_2\to\textrm{GL}(\mathfrak{g}_2)$ has no invariant vectors.

3474: 

3475: Let $\gep_n=\pi_2 (f_n (\gd ))$. Since $f_n (\gd )$ is close to $\gd$,

3476: $\gep_n=\pi_2 (f_n (\gd ))$ is close to the identity of $G_2$. We can therefore

3477: assume that $\gep_n$ is contained in an identity neighborhood of $G_2$ where 

3478: $\log =\exp^{-1}:G_2\to\mathfrak{g}_2$ is a well defined diffeomorphism.

3479: As $\gd$ commutes with each $\gd_{2,i}$, $\gep_n$ commutes with each 

3480: $\pi_2 (f_n (\gd_{2,i} ))$, and it follows that

3481: $$

3482:  \text{Ad}\big(\pi_2 ( f_n (\gd_{2,i} ))\big)(\log \gep_n )=

3483:  \log \gep_n,

3484: $$

3485: which in turn implies $\log \gep_n=0$, i.e. $\gep_n=1$.

3486: \end{proof}

3487: 

3488: Together with Proposition \ref{irr-lim}, the original argument from \cite{Wa} 8.1 gives:

3489: 

3490: \begin{thm}[Wang's theorem]\label{WT}

3491: Let G be a connected semi-simple Lie group without compact factors, which is 

3492: not locally isomorphic to $\SL_2(\BR )$ or $\SL_2(\BC )$. Then for any $v>0$

3493: there are only finitely many conjugacy classes of irreducible lattices in $G$

3494: with covolume $\leq v$. 

3495: \end{thm}  

3496:        

3497: \begin{rem}

3498: In \cite{BP}, Borel and Prasad established a very strong and general finiteness result, but they

3499: omitted the cases of $G=G_{a,b}$ by requiring absolute rank $\geq 2$. This requirement was used in 

3500: their proof of the stronger finiteness statement, in which the ambient group $G$ can be varied. 

3501: However, Prasad remarked to the author that, when $G$ is fixed, this requirement is unnecessary

3502: in their argument, and hence, the finiteness of the number of 

3503: conjugacy classes of arithmetic lattices of covolume $\leq v$ in $G$

3504: could be proved also by using their methods.

3505: \end{rem}

3506: 

3507: \begin{rem}

3508: More generally, for any $G$, the finiteness statement holds for lattices (not necessarily 

3509: arithmetic) which are irreducible with respect to the $\SL_2$ factors of $G$.

3510: I.e. for the set of conjugacy classes of lattices in $G$ which project densely

3511: to any factor of $G$ which is locally isomorphic to $\SL_2(\BR )$ or

3512: $\SL_2(\BC )$.

3513: \end{rem}

3514: 

3515: 

3516: 

3517: 

3518: 

3519: 

3520: 

3521: 

3522: 

3523: 

3524: 

3525: 

3526: \begin{Ack}

3527: I would like to thank 

3528: Shahar Mozes (my Ph.D. adviser) for his guidance throughout 

3529: the last few years and for insightful suggestions for this work,

3530: to Pierre Pansu for hours of helpful conversations and for 

3531: suggestions which were essential to this research,

3532: to Alex Lubotzky for many discussions, suggestions and ideas, 

3533: and to Uri Bader, Yair Glasner, and Yehuda 

3534: Shalom for many discussions and clever suggestions. Proofs of some of the 

3535: statements presented here were established during these discussions. 

3536: I would also like to thank Emmanuel Breuillard, Assaf Naor and Gopal Prasad 

3537: for some remarks concerning early versions of this paper.

3538: Finally, I would like to express my sincere gratitude to the anonymous referees 

3539: for their careful reading of the manuscript and for many 

3540: suggestions and corrections which tremendously improved the exposition of this paper.

3541: 

3542: \end{Ack}

3543: 

3544: 

3545: 

3546: 

3547: 

3548: 

3549: \begin{thebibliography}{99}

3550: 

3551: \bibitem{BaNe} U. Bader, A. Nevo, Conformal actions of simple Lie groups on compact pseudo-Riemannian manifolds, 

3552: J. Differential Geom. {\bf 60} (2002), no. 3, 355--387.

3553: 

3554: \bibitem{BGS} W. Ballmann, M. Gromov, V. Schroeder, 

3555: {\it Manifolds of Nonpositive Curvature}, Birkhauser, 1985.

3556: 

3557: \bibitem{BePe} R. Benedetti, C. Petronio, {\it Lectures on Hyperbolic 

3558: Geometry}, Springer-Verlag, 1992.

3559: 

3560: \bibitem{Bor} A. Borel, Commensurability classes and volumes of hyperbolic 

3561: 3-manifolds, Ann. Scuola Norm. Sup. Pisa, Ser. IV, {\bf 8} (1981) 1-33.

3562: 

3563: \bibitem{BoHC} A. Borel, Harish-Chandra, Arithmetic subgroups of algebraic groups. Ann. Math.

3564: (2) {\bf 75} (1962) 485-535.

3565: 

3566: \bibitem{BP} A. Borel, G. Prasad, 

3567: Finiteness theorems for discrete subgroups of bounded covolume in 

3568: semi-simple groups, Publ. Math. I.H.E.S. {\bf 69} (1989), 119-171.

3569: 

3570: \bibitem{BT} R. Bott, L.W. Tu, {\it Differential Forms in Algebraic 

3571: Topology}, Springer-Verlag, 1982.

3572: 

3573: \bibitem{BH} R. Bridson, A. Haefliger, {\it Metric Spaces of Non-Positive 

3574: Curvature}, Springer, 1999.

3575: 

3576: \bibitem{BGLM} M. Burger, T. Gelander, A. Lubotzky, S. Mozes, 

3577: Counting hyperbolic manifolds, Geom. Funct. Anal. 12 (2002), no. 6, 1161--1173. 

3578: 

3579: \bibitem{BS} M. Burger, V. Schroeder, Volume, diameter and the first 

3580: eigenvalue of locally symmetric spaces of rank one, 

3581: J. Differential Geometry {\bf 26} (1987), 273-284.

3582: 

3583: \bibitem{CH} J. Cheeger, D.G. Ebin, {\it Comparison theorems in Riemannian 

3584: geometry}, North-Holland

3585: 

3586: \bibitem{chinburg} T. Chinburg, Volume of hyperbolic manifolds, 

3587: J. Differential Geometry {\bf 18} (1983), 783-789.

3588: 

3589: \bibitem{cooper} D. Cooper, The volume of a closed hyperbolic 3-manifold

3590: is bounded by pi times the length of any presentation of its fundamental group,

3591: Proc. Amer. Math. Soc. {\bf 127} (1999), 941-942.

3592: 

3593: \bibitem{cor} k. Corlette, Archimedean superrigidity and hyperbolic geometry. 

3594: Ann. of Math. {\bf 135} (1992), no. 1, 165--182. 

3595: 

3596: \bibitem{Ga-Ra} H. Garland, M.S. Raghunathan, Fundamental domains for 

3597: lattices in (R-)rank 1 Lie groups, Ann. of Math. {\bf 92} (1970), 279-326.

3598: 

3599: \bibitem{Gr} M. Gromov, {\it Metric Structures for Riemannian and 

3600: Non-Riemannian Spaces}, Birkhauser, 1998.

3601: 

3602: \bibitem{Gr1} M. Gromov, Hyperbolic manifolds according to Thurston and Jorgensen, Semin. 

3603: Bourbaki, 32e annee, vol. 1979/80, exp. 546, Lect. Notes Math. 842, (1981), 40-53.

3604: 

3605: \bibitem{GS} M. Gromov, R. Schoen, Harmonic maps into singular spaces and $p$-adic superrigidity 

3606: for lattices in groups of rank one. Inst. Hautes Études Sci. Publ. Math. {\bf 76} (1992), 165--246.

3607:  

3608: \bibitem{Kazhdan} D.A. Kazhdan, Connection of the dual space of a group with the

3609: structure of its closed subgroups, Functional Analysis and Application {\bf 1}

3610: (1967), 63-65. 

3611: 

3612: \bibitem{Lub} A. Lubotzky, Subgroup growth and congruence subgroups, Invent.

3613: Math. Springer-Verlag {\bf 119} (1995), 267-295.

3614: 

3615: \bibitem{Mar1} G.A. Margulis, {\it Discrete Subgroups of Semisimple Lie 

3616: Groups}, Springer-Verlag, 1990.

3617:  

3618: \bibitem{Mar2} G.A. Margulis, Arithmeticity of the irreducible lattices in the semi-simple groups

3619: of rank greater then $1$, (Russian), Invent. Math. {\bf 76} (1984) 93-120.

3620: 

3621: \bibitem{Mar3} G.A. Margulis, Non-uniform lattices in semisimple algebraic groups. 

3622: Lie groups and their representations (Proc. Summer School on Group Representations of the Bolyai János Math. 

3623: Soc., Budapest, 1971), Halsted, New York, (1975) 371--553. 

3624: 

3625: \bibitem{MR} G.A. Margulis, J. Rohlfs, On the proportionality of covolumes of

3626: discrete subgroups. Math. Ann {\bf 275} (1986) 197-205 

3627: 

3628: \bibitem{PR} V. Platonov, A. Rapinchuk, {\it Algebraic Groups and Number 

3629: Theory}, Academic Press, 1994.

3630: 

3631: \bibitem{P} G. Prasad, Volume of $S$-arithmetic quotients of semi-simple 

3632: groups, Publ. Math. I.H.E.S. {\bf 69} (1989), 91-117.

3633: 

3634: \bibitem{Rag} M.S. Raghunathan, {\it Discrete Subgroups of Lie Groups}, 

3635: Springer, New York, 1972.

3636: 

3637: \bibitem{Sp} E. Springer, {\it Algebraic Topology}, 

3638: Springer-Verlag, 1966.

3639: 

3640: \bibitem{sury} B. Sury, Arithmetic groups and Salem numbers, Manuscripta Math

3641: {\bf 75} (1992), 97-102.

3642: 

3643: \bibitem{Th} W.P. Thurston, {\it Three-Dimensional Geometry and Topology},

3644: Volume 1, Princeton univ. press, 1997.

3645: 

3646: \bibitem{Wa} H.C. Wang, Topics on totally discontinuous groups,

3647: {\it Symmetric Spaces}, edited by W. Boothby and G. Weiss (1972), 

3648: M. Dekker, 460-487. 

3649: 

3650: \bibitem{S-P.W} S.P. Wang, The dual space of semisimple Lie groups, Amer. J. 

3651: Math {\bf 91} (1969), 921-937. 

3652: 

3653:          

3654: \end{thebibliography}

3655: 

3656: 

3657: 

3658: 

3659: 

3660: 

3661: \end{document}

3662: