1: \documentclass{amsart}

2: \usepackage{amssymb,epsfig,amsmath,latexsym,amsthm}

3: \theoremstyle{plain} \newtheorem{theorem}{Theorem}[section]

4: \newtheorem{lemma}[theorem]{Lemma} \newtheorem*{claim}{Claim}

5: \newtheorem{proposition}[theorem]{Proposition}

6: \newtheorem{corollary}[theorem]{Corollary}

7: \newtheorem{question}[theorem]{Question}

8: \newtheorem{conjecture}[theorem]{Conjecture}

9: \newtheorem{problem}[theorem]{Problem}

10: \newtheorem{ComplexityConjecture}[theorem]{Hyperbolic Complexity

11: Conjecture} \theoremstyle{definition}

12: \newtheorem{SmallestManifoldConjecture}[theorem]{Smallest Hyperbolic

13: Manifold Conjecture} \theoremstyle{definition}

14: \newtheorem{definition}[theorem]{Definition}

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

16: \newtheorem{acknowledgements}[theorem]{Acknowledgements}

17: \newtheorem{addendum}[theorem]{Addendum}

18: \newtheorem{remarks}[theorem]{Remarks}

19: \newtheorem{correction}[theorem]{Correction}

20: \newtheorem{warning}[theorem]{Warning}

21: \newtheorem{notation}[theorem]{Notation}

22: \newtheorem{example}[theorem]{Example}

23: \DeclareMathOperator{\Diff}{Diff} \DeclareMathOperator{\Emb}{Emb}

24: \DeclareMathOperator{\Hyp}{Hyp} \DeclareMathOperator{\id}{id}

25: \DeclareMathOperator{\RM}{RM} \DeclareMathOperator{\Bd}{Bd}

26: \DeclareMathOperator{\length}{length}

27: \DeclareMathOperator{\handle}{handle}

28: \DeclareMathOperator{\base}{base}

29: \DeclareMathOperator{\handles}{handles}

30: \DeclareMathOperator{\valence}{valence}

31: \DeclareMathOperator{\area}{area} \DeclareMathOperator{\Isom}{Isom}

32: \DeclareMathOperator{\Diam}{Diam} \DeclareMathOperator{\Lim}{Lim}

33: \DeclareMathOperator{\Sum}{Sum} \DeclareMathOperator{\rank}{rank}

34: \DeclareMathOperator{\genus}{genus} \DeclareMathOperator{\kernal}{ker}

35: \DeclareMathOperator{\rel}{rel} \DeclareMathOperator{\textin}{in}

36: \DeclareMathOperator{\textint}{int}

37: \DeclareMathOperator{\bridges}{bridges}

38: \DeclareMathOperator{\plates}{plates}

39: \DeclareMathOperator{\beams}{beams}

40: \newcommand{\cirA}{\overset\circ{A}}

41: \newcommand{\cirB}{\overset\circ{B}}

42: \newcommand{\cirC}{\overset\circ{C}}

43: \newcommand{\cirD}{\overset\circ{D}}

44: \newcommand{\cirE}{\overset\circ{E}}

45: \newcommand{\cirF}{\overset\circ{F}}

46: \newcommand{\cirH}{\overset\circ{H}}

47: \newcommand{\cirI}{\overset\circ{I}}

48: \newcommand{\cirJ}{\overset\circ{J}}

49: \newcommand{\cirL}{\overset\circ{L}}

50: \newcommand{\cirM}{\overset\circ{M}}

51: \newcommand{\cirN}{\overset\circ{N}}

52: \newcommand{\cirP}{\overset\circ{P}}

53: \newcommand{\cirR}{\overset\circ{R}}

54: \newcommand{\cirS}{\overset\circ{S}}

55: \newcommand{\cirT}{\overset\circ{T}}

56: \newcommand{\cirU}{\overset\circ{U}}

57: \newcommand{\cirV}{\overset\circ{V}}

58: \newcommand{\cirW}{\overset\circ{W}}

59: \newcommand{\cirX}{\overset\circ{X}}

60: \newcommand{\cirY}{\overset\circ{Y}}

61: \newcommand{\cirZ}{\overset\circ{Z}}

62: \newcommand{\cirsigma}{\overset\circ{\sigma}}

63: \newcommand{\cirgamma}{\overset\circ{\gamma}}

64: \newcommand{\cirtau}{\overset\circ{\tau}}

65: \newcommand{\cirlambda}{\overset\circ{\lambda}}

66: \newcommand{\cirkappa}{\overset\circ{\kappa}}

67: \newcommand{\cirmu}{\overset\circ{\mu}}

68: \newcommand{\cire}{\overset\circ{e}} \newcommand{\eqdef}{\overset

69: {def}{=}} \newcommand{\Sinfty}{S^2_{\infty}}

70: \newcommand{\lambdaij}{\lambda_{ij}}

71: \newcommand{\sigmaij}{\sigma_{ij}} \newcommand{\solidtorus}{D^2\times

72: S^1} \newcommand{\finv}{f^{-1}} \newcommand{\kinv}{k^{-1}}

73: \newcommand{\hinv}{h^{-1}} \newcommand{\ginv}{g^{-1}}

74: \newcommand{\gprimeinv}{g^{\prime -1}} \newcommand{\pinv}{p^{-1}}

75: \newcommand{\piinv}{\pi^{-1}} \newcommand{\qinv}{q^{-1}}

76: \newcommand{\BH}{\mathbb H} \newcommand{\BR}{\mathbb R}

77: \newcommand{\BB}{\mathbb B} \newcommand{\BN}{\mathbb N}

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

79: \newcommand{\osolidtorus}{\overset\circ{D^2}\times S^1}

80: \newcommand{\Vprime}{V^\prime} \newcommand{\piM}{\pi_1(M)}

81: \newcommand{\Mg}{M_\delta} \newcommand{\Wone}{W_{\Delta_1}}

82: \newcommand{\Wtwo}{W_{\Delta_2}} \newcommand{\Wi}{W_{\Delta_i}}

83: \newcommand{\mD}{\mathcal{D}} \newcommand{\mE}{\mathcal{E}}

84: \newcommand{\mF}{\mathcal{F}} \newcommand{\mG}{\mathcal{G}}

85: \newcommand{\mP}{\mathcal{P}} \newcommand{\mQ}{\mathcal{Q}}

86: \newcommand{\mR}{\mathcal{R}} \newcommand{\mW}{\mathcal{W}}

87: \newcommand{\Vol}{\mathrm{Vol}}

88: \newcommand{\Nthick}{N_{\ge\epsilon}}

89: \newcommand{\Nthin}{N_{\le\epsilon}} \def\CAT{{\text{CAT}}}

90: \def\inte{{\text{int}}} \def\E{\mathcal{E}} \def\C{\mathcal{C}}

91: \def\H{\mathbb{H}} \def\R{\mathbb{R}} \def\Z{\mathbb{Z}}

92: 

93: 

94: \begin{document}

95: 

96: \title{Mom Technology and volumes of hyperbolic 3-manifolds}

97: 

98: \author{David Gabai}\footnote{Partially supported by NSF grants

99: DMS-0071852, DMS-0346270, and DMS-0504110.}\address{Department

100:   of Mathematics\\Princeton University\\Princeton, NJ 08544}

101: 

102: \author{Robert Meyerhoff}\address{Department of Mathematics, Boston

103:   College, Chestnut Hill, MA 02467}

104: 

105: \author{Peter Milley}\address{Department of Mathematics, University of

106: Calfornia Riverside, Riverside, CA 92521}

107: 

108: 

109: 

110: \thanks{Version 0.80, July 27, 2006}

111: 

112: \maketitle

113: 

114: \setcounter{section}{-1}

115: 

116: \section{Introduction}\label{S0}

117: 

118: This paper is the first in a series whose goal is to understand the

119: structure of low-volume complete orientable hyperbolic 3-manifolds.

120: Here we introduce \emph{Mom technology} and enumerate the hyperbolic

121: Mom-$n$ manifolds for $n\le 4$. Our long-term goal is to show that all

122: low-volume closed and cusped hyperbolic 3-manifolds are obtained by

123: filling a hyperbolic Mom-$n$ manifold, $n\le 4$ and to enumerate the

124: low-volume manifolds obtained by filling such a Mom-$n$.

125: 

126: William Thurston has long promoted the idea that volume is a good

127: measure of the complexity of a hyperbolic 3-manifold

128: (see, for example, \cite{Th1} page 6.48).

129: Among known low-volume manifolds, Jeff

130: Weeks (\cite{We}) and independently Sergei Matveev and

131: Anatoly Fomenko (\cite{MF}) have observed that there is a close connection

132: between the volume of closed hyperbolic 3-manifolds

133: and combinatorial complexity.  One goal of this project is

134: to explain this phenomenon, which is summarized by the following:

135: 

136: \begin{ComplexityConjecture} (Thurston,

137: Weeks, Matveev-Fomenko) The complete low-volume hyperbolic 3-manifolds

138: can be obtained by filling cusped hyperbolic 3-manifolds of small

139: topological complexity.\end{ComplexityConjecture}

140: 

141: \begin{remark}  Part of the challenge of this conjecture is to 

142: clarify the undefined adjectives \emph{low} and \emph{small}.  In the

143: late 1970's, Troels Jorgensen proved that for any positive constant

144: $C$ there is a finite collection of cusped hyperbolic 3-manifolds from

145: which all complete hyperbolic 3-manifolds of volume less than or equal

146: to $C$ can be obtained by Dehn filling.  Our long-term goal stated

147: above would constitute a concrete and satisfying realization of

148: Jorgensen's Theorem for ``low'' values of $C$.

149: \end{remark}

150: 

151: A special case of the Hyperbolic Complexity Conjecture is the

152: long-standing

153: \begin{SmallestManifoldConjecture}The Weeks Manifold $M_W$, obtained

154:   by $(5,1)$, $(5,2)$ surgery on the two components of the Whitehead

155:   Link, is the unique oriented hyperbolic $3$-manifold of minimum

156:   volume.

157: \end{SmallestManifoldConjecture}

158: Note that the volume of $M_W$ is $0.942\ldots$.

159: 

160: All manifolds in this paper will be orientable and all hyperbolic

161: structures are complete.  We call a compact manifold \emph{hyperbolic}

162: if its interior supports a complete hyperbolic structure of

163: finite volume.

164: 

165: \begin{definition}  A \emph{Mom-n structure} $(M,T,\Delta)$ consists of a

166: compact 3-manifold $M$ whose boundary is a union of tori, a preferred

167: boundary component $T$, and a handle decomposition $\Delta$ of the

168: following type.  Starting from $T\times I$, $n$ 1-handles and $n$

169: 2-handles are attached to $T\times 1$ such that each 2-handle goes

170: over exactly three 1-handles, counted with multiplicity.  Furthermore,

171: each 1-handle encounters at least two 2-handles, counted with

172: multiplicity. We say that $M$ is a \emph{Mom-n} if it possesses a

173: Mom-$n$ structure \((M,T,\Delta)\).

174: \end{definition}

175: 

176: \begin{remarks}  On a Mom-$n$, the handle decomposition $\Delta$ deformation

177: retracts to an almost simple 2-complex which has $2n$ true vertices,

178: in the sense of Matveev \cite{Mv2}. Therefore Mom-$n$ manifolds are a

179: subset of those with Matveev complexity at most $2n$.\end{remarks}

180: 

181: Here is the fundamental idea at the foundation of our project.  Given

182: a complete finite-volume hyperbolic 3-manifold $N$, start with either

183: a slightly shrunken maximal horotorus neighborhood $V$ of a cusp or

184: slightly shrunken maximal tube V about a geodesic.  After expanding

185: $V$ in the normal direction, it eventually encounters itself, thereby

186: creating a 1-handle.  Subsequent expansions give rise to the creation

187: of 1, 2, and 3-handles.  In the presence of low volume we expect that

188: $V$ will rapidly encounter 1 and 2-handles and $\partial V$ together

189: with a subset of these handles (perhaps somewhat perturbed to allow

190: for the ``valence-3 2-handle condition'') will create a Mom-$n$

191: manifold $M$, for some $n\le 4$. Furthermore, the complement of $M$

192: will consist of cusp neighborhoods and tubular neighborhoods of

193: geodesics. In practice, the handle structure may arise in a somewhat

194: different manner; e.g., as a sub-complex of the dual triangulation of

195: the Ford Domain (see \cite{GMM3}).

196: 

197: The papers \cite{GM} and \cite{GMM} can be viewed as steps in this

198: direction when $V$ is a tubular neighborhood about a geodesic

199: $\gamma$.  Indeed, \cite{GM} gives a lower bound on $\Vol(N)$ in

200: terms of the tube radius of $\gamma$ and \cite{GMM} gives a lower

201: bound in terms of the first two ortholengths, or equivalently the

202: radii of the expanding $V$ as it encounters its first and second

203: 1-handles.

204: 

205: \begin{definition}  If $i:M\to N$ is an embedding, then we say that the

206: embedding is \emph{elementary} if $i_*\pi_1(M)$ is abelian, and

207: \emph{non-elementary} otherwise.

208: \end{definition}

209: 

210: In \S 1 we give the basic definitions regarding Mom-$n$ manifolds

211: embedded in hyperbolic 3-manifolds and state for later use some

212: standard results about hyperbolic 3-manifolds and embedded tori in

213: such.  The end result of \S 2 - \S 4 is that if $n\le 4$, then given a

214: non-elementary Mom-$n$ in a hyperbolic manifold one can find a

215: non-elementary hyperbolic Mom-$k$, where $k\le n$. Weaker results are

216: given for general values of $n$.  In \S5 we enumerate the hyperbolic

217: Mom-$n$ manifolds for $n\le 4$; Theorem

218: \ref{list of mom-2's and mom-3's} and Conjecture \ref{number of mom-4's}

219: together imply the following:

220: 

221: \begin{theorem}There are $3$ hyperbolic Mom-$2$ manifolds, 21 hyperbolic

222: Mom-$3$ manifolds (including the 3 hyperbolic Mom-$2$'s, which are also

223: Mom-$3$'s).

224: \end{theorem}

225: 

226: \begin{conjecture}There are 138 hyperbolic Mom-$4$ manifolds (including

227: the hyperbolic Mom-$2$'s and Mom-$3$'s, which are also Mom-$4$'s).

228: \end{conjecture}

229: 

230: In \S 6 we show that any non-elementary embedding of a hyperbolic

231: Mom-$n$ manifold $M$, $n\le 4$, into a compact hyperbolic manifold $N$

232: gives rise to an \emph{internal Mom-$n$ structure} on $N$, i.e.\ every

233: component of $\partial M$ either splits off a cusp of $N$ or bounds a

234: solid torus in $N$.

235: 

236: In \S 7 we give examples of internal Mom-2

237: structures on cusped hyperbolic 3-manifolds, including in particular

238: a detailed

239: exposition of one of our key motivating examples.

240: 

241: In future papers we will use the Mom Technology to directly address

242: the Hyperbolic Complexity Conjecture, and the Smallest Hyperbolic

243: Manifold Conjecture. Indeed, in \cite{GMM3} we will identify all

244: $1$-cusped hyperbolic $3$-manifolds with volume less than $2.7$ by showing

245: that all such manifolds possess internal Mom-$n$ structures with

246: $n\le 3$. This result, in combination with work of Agol-Dunfield (see

247: \cite{AST}), gives a lower bound of $0.86$ for the volume of an orientable

248: hyperbolic 3-manifold (but see the note below).  The Agol-Dunfield

249: result is an improvement of an earlier result of Agol which provides a

250: tool for controlling the volume of a hyperbolic $3$-manifold in terms of

251: the volume of an appropriate cusped hyperbolic $3$-manifold from which

252: it is obtained via Dehn filling; the improved version utilizes recent

253: work of Perelman.

254: 

255: This leads to three very promising directions towards the Smallest

256: Hyperbolic Manifold Conjecture.  Either, improve the technology of

257: \cite{GMM3} to identify the 1-cusped manifolds of volume less than $2.852$,

258: and then apply Agol-Dunfield.  Or, extend the tube radius results of

259: \cite{GMT} from $\log(3)/2$ to $0.566$ (the Agol-Dunfield volume bound

260: involves the radius of a solid tube around a short geodesic in the

261: original closed hyperbolic 3-manifold).

262: Or, extend the method of \cite{GMM3} to the closed case, thereby providing

263: an essentially self-contained proof of the Smallest Hyperbolic

264: Manifold Conjecture.

265: 

266: Note that all three approaches require an analysis of volumes of

267: hyperbolic 3-manifolds obtained by Dehn filling a Mom-2 or Mom-3

268: manifold.  These Dehn filling spaces have been extensively studied by

269: J. Weeks and others, and it is highly likely that all low-volume

270: manifolds in these Dehn filling spaces have been identified.  However,

271: some work will need to be done to bring these studies up to a suitable

272: level of rigor.

273: 

274: The authors wish to thank Morwen Thistlethwaite for his assistance in

275: identifying the hyperbolic manifolds in \S 5.

276: 

277: \begin{section} {Basic Definitions and Lemmas}\end{section}

278: 

279: \vskip 12pt

280: 

281: \begin{definition}  Let $M$ be a compact connected

282: 3-manifold $M$ with $B\subset \partial M$ a compact surface which may

283: be either disconnected or empty.  A \emph{handle structure} $\Delta $

284: on $(M,B)$ is the structure obtained by starting with $B\times I$

285: adding a finite union of 0-handles, then attaching finitely many 1 and

286: 2-handles to $B\times 1$ and the 0-handles. We call $B\times I$

287: (resp. $B\times I \cup$ 0-handles) the \emph{base}

288: (resp. \emph{extended base}) and say that the handle structure is

289: based on $B$.  The \emph{valence} of a 1-handle is the number of

290: times, counted with multiplicity, the various 2-handles run over it

291: and the \emph{valence} of a 2-handle is the number of 1-handles,

292: counted with multiplicity, it runs over.

293: 

294: Following the terminology of Schubert \cite{Sch} and Matveev

295: \cite{Mv1} we call the 0-handles, 1-handles and 2-handles

296: \emph{balls}, \emph{beams} and \emph{plates} respectively.  We call

297: \emph{islands} (resp.  \emph{bridges}) the intersection of the

298: extended base with beams (resp. plates) and the closed complement of

299: the islands and bridges in $B\times 1 \cup \partial$(0-handles) are

300: the \emph{lakes}.  We say that $\Delta$ is \emph{full} if each lake is

301: a disc.  If $B=\emptyset$, then we say that $\Delta$ is a

302: \emph{classical handle structure}.

303: 

304: Let $M$ be a compact 3-manifold with $\partial M$ a union of tori and

305: let $T$ be a component of $\partial M$.  We say that $(M,T,\Delta)$ is

306: a \emph{weak Mom-n} if $\Delta$ is a handle structure based on $T$

307: without 0-handles or 3-handles and has an equal number of 1 and 2-handles, such

308: that each 1-handle is of valence $\ge 2$ and each 2-handle is of

309: valence 2 or 3.  Furthermore, there are exactly $n$ 2-handles of valence 3.

310: A weak Mom-$n$ with no valence-2 2-handles is a \emph{Mom-n}.  A weak

311: Mom-$n$ is \emph{strictly weak} if there exists a valence-2 2-handle.

312: \end{definition}

313: 

314: The following is a well-known existence result stated in our language.

315: 

316: \begin{proposition}  A compact 3-manifold $M$ has a weak Mom-structure if

317: and only if $\partial M$ is a union of at least two tori.

318: \end{proposition}

319: 

320: \begin{proof}

321: If $M$ has a weak Mom-$n$ structure, then by definition all of its

322: boundary components are tori and there is at least one such boundary

323: component.  Further, because there are no 3-handles in $\Delta$, there

324: must be another (torus) boundary component.

325: 

326: The converse is not much more difficult.  In fact, if $M$ has at least

327: two boundary components, then it is standard to create it by starting

328: with a thickened boundary component and adding 1 and 2-handles where

329: the 1 and 2-handles are of valence $\ge 2$.  By subdividing the

330: 2-handles with 1-handles we satisfy the condition that the 2-handles

331: have valence $\le 3$.  Since $\chi(M)=0$, there are an equal number of

332: 1 and 2-handles.\end{proof}

333: 

334: \begin{definition}  Call a torus that bounds a solid torus a

335: \emph{tube} and call a torus bounding a tube with knotted hole a \emph

336: {convolutube}.  Recall that a \emph{tube with knotted hole} is a

337: $B^3-\cirN(\gamma)$, where $\gamma$ is a knotted proper arc.

338: \end{definition}

339: 

340: The following standard result follows from the loop theorem (see for

341: example \cite{Jac}).

342: 

343: \begin{lemma}\label{torus}  If $S$ is a torus in an irreducible

344: 3-manifold $N$, then either $S$ is incompressible or $S$ is a tube or

345: a convolutube.  If $S\subset \cirN$, $\partial N$ is incompressible

346: and $S$ is compressible, and there exists an embedded essential

347: annulus connecting $S$ to a component of $\partial N$, then $S$ is a

348: tube.  \qed\end{lemma}

349: 

350: \begin{proposition}  \label{convoluted embedding}If $M$ is a

351: non-elementary compact, connected 3-manifold embedded in the compact

352: hyperbolic 3-manifold $N$ and $\partial M$ is a union of tori, then,

353: up to isotopy, $N$ is obtained from $M$ by first filling a subset of

354: the components of $\partial M$ by solid tori to obtain the manifold

355: $M_1$, then compressing a subset of the components of $\partial M_1$

356: to obtain the manifold $M_2$, then attaching 3-balls to the 2-sphere

357: components of $\partial M_2$ to obtain $M_3$.  Furthermore all of

358: these operations can be performed within $N$.\end{proposition}

359: 

360: \begin{proof}  The components

361: of $\partial M$ that bound solid tori in $N$ are exactly those

362: boundary components which compress to the non-$M$ side.  Fill in all

363: such tori to obtain the manifold $M_1$.  If $P$ is a component of

364: $\partial M_1$ which is not boundary parallel in $N$, then $P$ is

365: compressible in $N$ and hence is a convolutube.  These convolutubes

366: can be isotoped to lie in pairwise disjoint 3-balls in $N$.  Therefore

367: we can compress all the compressible components of $\partial M_1$

368: (to obtain $M_2$) and

369: cap the resulting 2-spheres with 3-cells to obtain $M_3$ which is

370: isotopic to $N$.

371: 

372: Since $M_3$ must have all boundary components boundary parallel

373: in $N$ and $M_3$ is non-elementary, the result follows.\end{proof}

374: 

375: 

376: \begin{corollary}  Let $M\subset \cirN$ be a connected compact

377: non-elementary submanifold in the compact hyperbolic 3-manifold $N$.

378: If $\partial M$ is a union of tori, then each component of $\partial

379: N$ is parallel to a component of $\partial M$ via a parallelism

380: disjoint from $M$.\qed\end{corollary}

381: 

382: 

383: The following result is due to Kerckhoff (see \cite{Koj}).

384: \begin{lemma}\label{geodesic complement}  If $\gamma$ is a simple

385: closed geodesic in the complete, finite-volume hyperbolic 3-manifold

386: $N$, then $N-\gamma$ has a complete finite-volume hyperbolic

387: structure.\qed\end{lemma}

388: 

389: \begin{lemma}  \label{pi image}  Let $M$ be a compact submanifold of 

390: the compact hyperbolic 3-manifold $N$.

391: 

392: i) If $M_1=M-\cirX$, where $X\subset N$ is either a solid torus or 3-ball

393: with $\partial X\subset M$, or

394: $M_1$ is obtained by deleting a 2-handle or more generally

395: deleting an open regular neighborhood of a properly embedded arc from

396: M, then the inclusion $M\to N$ is a non-elementary embedding if and only

397: if the inclusion $M_1\to N$ is a non-elementary embedding.

398: 

399: ii) Suppose $M$ is non-elementary, $\partial M$ is a union of tori and $A$

400: is an essential annulus in $M$. Split $M$ along $A$; then some

401: component $M_1$ of the resulting manifold is non-elementary.

402: 

403: iii) Suppose $F\subset M$ is an embedded torus essential in $M$, and

404: $M\subset N$ is non-elementary. Split $M$ along $F$; then exactly one

405: component of the resulting manifold is non-elementary.

406: \end{lemma}

407: 

408: \begin{proof}  The conclusion is immediate in case (i) 

409: because both $M_1$ and $M$ have the same $\pi_1$-image.

410: 

411: Under the hypotheses of (ii) the boundary of each component of $M$

412: split along $A$ is also a union of tori.  We consider the case where

413: the split manifold connected, for the general case is similar.  Since all tori

414: in $N$ separate, $M$ is obtained from $M_1$ by attaching a thickened

415: annulus $A$ to a boundary parallel torus, a tube or a convolutube.  In

416: the first case $M $ and $M_1$ have the same $\pi_1$-image since $N$ is

417: anannular.  If $A$ is attached to the outside of a convolutube, then

418: $\pi_1(M)\subset \BZ$.  If $A$ is attached to the outside of a tube,

419: then $M$ can be enlarged to a Seifert fibered space in $N$ and hence

420: is elementary.

421: 

422: To prove (iii) note that $F$ is either boundary parallel or

423: is a tube or convolutube.  In each case the result follows

424: immediately.

425: \end{proof}

426: 

427: \begin{definition}  Let $N$ be a compact hyperbolic 3-manifold.

428: An \emph{internal Mom-n structure on N} consists of a non-elementary

429: embedding $f:M\to N$ where $(M,T,\Delta)$ is a Mom-$n$ and each

430: component of $\partial M$ is either boundary parallel in $N$ or bounds

431: a solid torus in $N$.  We will sometimes suppress mention of the

432: embedding and simply say that $(M,T,\Delta)$ is an internal Mom-$n$

433: structure on $N$.  In the natural way we define the notion of

434: \emph{weak internal Mom-$n$ structure on $N$}.

435: \end{definition}

436: 

437: \begin{lemma}  A non-elementary embedding of the Mom-$n$ manifold $M$ 

438: into the compact hyperbolic 3-manifold $N$ will fail to give an

439: internal Mom-$n$ structure on $N$ if and only if some component of

440: $\partial M$ maps to a convolutube.  In that case, a reimbedding of

441: $M$, supported in a neighborhood of the convolutubes gives rise to an

442: internal Mom-$n$ structure on $N$.\qed\end{lemma}

443: 

444: \begin{definition}  A \emph{general based Mom-n} $(M,B,\Delta)$ 

445: consists of a compact manifold $M$ with $\partial M$ a union of tori,

446: $B\subset \partial M$ a compact codimension-0 submanifold of

447: $\partial M$ that is

448: $\pi_1$-injective in $\partial M$, and $\Delta$ a handle structure

449: based on $B$ without 0-handles such that every 1-handle is of

450: valence-$\ge 2$, every 2-handle is of valence-3 and there are exactly

451: $n$ of each of them.  A \emph{weak general based Mom-n} is as above with

452: $\Delta$ perhaps having $k\ge 0$ extra valence-2 2-handles in addition

453: to the $n$ valence-3 2-handles and hence has $k+n$ 1-handles.

454: 

455: A \emph{general based internal Mom-n structure on N} consists of a

456: non-elementary

457: embedding $f:M\to N$, where $N$ is a compact hyperbolic 3-manifold and

458: $(M,B,\Delta)$ is a general based Mom-$n$ structure.  Along similar

459: lines we have the notion of \emph{weak general based internal Mom-n

460: structure on N}.

461: \end{definition}

462: 

463: \begin{notation}  If $X$ is a space, then let $|X|$ denote the number of

464: components of $X$, $\cirX$ denote the interior of $X$ and $N(X)$ denote

465: a regular neighborhood of $X$.  If $\sigma$ is a 2-handle, then let

466: $\delta \sigma$ denote the lateral boundary, i.e. the closure of that

467: part of $\partial \sigma$ which does not lie in lower index handles.

468: If $b$ is a bridge which lies in the 2-handle $\sigma$, then define

469: $\delta b=b\cap\delta\sigma$.  \end{notation}

470: 

471: \begin{section} {Handle structures and Normal surfaces}\end{section}

472: 

473: We slightly modify Haken's \cite{Ha} theory of surfaces in

474: handlebodies to our setting.  We closely parallel the excellent

475: exposition given by Matveev in \cite{Mv1}.

476: 

477: \smallskip

478: \begin{definition}  Let $\Delta$ be a handle structure on $M$ based 

479: on $B\subset \partial M$.  A compact surface $F\subset M$ is called

480: \emph{normal} if

481: \begin{enumerate}

482: \item $F$ intersects each plate $D^2\times I$ in parallel copies of

483: the form $D^2\times$ pt.

484: \item Each component of the intersection of $F$ with a beam $D^2\times

485: I$ is of the form $\alpha\times I$, where $\alpha$ is a proper arc

486: whose endpoints are disjoint from $\delta$(bridges).  Furthermore,

487: each component of $D^2\times 0-\alpha$ intersects $\delta$(bridges) in

488: at least two points.

489: \item Each component $U$ of $F\cap B\times [0,1]\cup \text{0-handles}$

490: is $\pi_1$-injective in $B\times [0,1]\cup \text{0-handles}$.  If

491: $U\cap B\times 0\neq\emptyset$, then $U$ is a product disc or annulus,

492: i.e. The inclusion $(U,U\cap(B\times 0),U\cap(B\times

493: 1),U\cap(\partial B\times I)) \to (B\times I, B\times 0, B\times 1,

494: \partial B\times I)$, can be relatively isotoped to a vertical

495: embedding.

496: \item If $U$ is a component of $F\cap B\times I$ with $F\cap B\times

497: 0\neq\emptyset$, then $U$ is an essential vertical disc or annulus,

498: i.e. $U=\alpha\times I$ where $\alpha\times 0$ is either an essential

499: simple closed curve or essential proper arc in $B\times 0$.

500: 

501: \end{enumerate}\end{definition}

502: 

503: 

504: \begin{remark}  i)  For $F$ closed, the second condition can be 

505: restated by requiring that $\alpha$ intersect distinct components of

506: $D^2\times 0\cap (\bridges)$.  When $\partial F\neq\emptyset$, the

507: second condition implies that $F$ is locally efficient in that it

508: neither can be locally boundary compressed nor can its weight be

509: reduced via an isotopy supported in the union of a 2-handle and its

510: neighborhing 1-handles.

511: 

512: ii) Note that $\partial F$ lies in the union of the beams, lakes and

513: $B\times 0$.\end{remark}

514: 

515: \begin{lemma}(Haken) If $F$ is a compact, incompressible,

516: boundary-incompressible surface in a compact irreducible 3-manifold,

517: then $F$ is isotopic to a normal surface.\qed\end{lemma}

518: 

519: 

520: \begin{definition}\label{def:valence}

521: Let $\Delta$ be a handle structure on $M$ based 

522: on $B\subset \partial M$.

523: The \emph{valence} $v(b)$ of a beam (resp. 

524: plate) is the number of plates (resp. beams) that attach to it,

525: counted with multiplicity.  Define the \emph{complexity} $C(\Delta)$

526: to be $(\rho_1(\Delta),|\Delta^1|)$ lexicographically ordered, where

527: $\rho_1(\Delta)=\Sigma_{\beams b} \max(v(b)-2,0)$ and $|\Delta^1|$ is

528: the number of 1-handles.  In particular we have\\

529: 

530: \noindent$\rho_1$-\textbf{formula}:

531: $\rho_1(\Delta)=\sum_{\text{2-handles}\ \sigma}\valence(\sigma) -

532: 2|\Delta^1|+ |\text{valence-1}\ \text{1-handles}| +

533: 2|\text{valence-0}\ \text{1-handles}|$

534: \end{definition}

535: 

536: 

537: %RRRR

538: 

539: \begin{lemma}(\text{Matveev}) \label{reduction} Let $\Delta$ be a 

540: handle structure on $M$ based on $B$, $F\subset M$ a closed normal

541: surface and let $M'$ be $M$ split along $F$.  If each component of

542: $M'\cap B\times [0,1]$ disjoint from $B\times 0$ is a 3-ball, then

543: $M'$ has a handle structure $\Delta'$ based on $B$ with

544: $\rho_1(\Delta')= \rho_1(\Delta)$.

545: \end{lemma}

546: 

547: \begin{proof}  This follows almost exactly as in \S 3 and

548: \S 4 of \cite{Mv1}: $M'$ naturally inherits a

549: handle structure $\Delta_1$ from $\Delta$ as follows.  The surface $F

550: $ splits $B\times I$ into various submanifolds one of which is

551: homeomorphic to $B\times [0,1]$ with $B\times 0=B$.  All of the other

552: submanifolds which lie in $M'$ are 3-balls.  This new $B\times [0,1]$

553: becomes the base and the 3-balls become 0-handles.  The various 1 and

554: 2-handles are split by F into 1 and 2-handles and as in \cite{Mv1},

555: $\rho_1(\Delta_1)=\rho_1(\Delta)$.\end{proof}

556: 

557: \begin{lemma} \label{no trivial 1-handles}

558: Given the handle structure $\Delta$ on $(M,B)$, if some 1-handle is

559: valence-1, then there exists another structure $\Delta_1$ on $(M,B)$

560: with $C(\Delta_1)<C(\Delta)$.\qed\end{lemma}

561: 

562: 

563: 

564: \begin{lemma}\label{Mom-n complexity}  If $(M,T,\Delta)$ is a Mom-$n$,

565: then $C(\Delta)=(n,n)$.\qed\end{lemma}

566: 

567: 

568: 

569: 

570: 

571: 

572: \begin{lemma}  \label{classical reduction}

573: Let $\Delta$ be a handle structure on $(M,B)$, $F\subset M$ a

574: connected separating normal surface and $M_1$ be the component of

575: $M-\cirN(F)$ which does not contain $B$.  If each component of $F\cap

576: B\times [0,1]$ is a disc, then $M_1$ has a classical handle structure

577: $\Delta_1$ with $\rho_1(\Delta_1)\le \rho_1(\Delta)$.  \end{lemma}

578: 

579: \begin{proof}  This follows as in the proof of

580: Lemma \ref{reduction} after noting that each component of $M_1\cap

581: B\times [0,1]$ is a 3-ball and these 3-balls correspond to the

582: 0-handles of the induced handle structure on $M_1$.\end{proof}

583: 

584: \begin{lemma}  If $\partial M$ is a union of tori,

585: and $\Delta$ is a handle structure on $(M,T)$ with $T$ a component of

586: $\partial M$, then there exists a weak Mom-$n$ $(M,T,\Delta_1)$ with

587: $n\le \rho_1(\Delta_1)$.  \end{lemma}

588: 

589: \begin{proof}  First apply Lemma \ref{no trivial 1-handles}, then add

590: 1-handles to subdivide the valence-$k$, $k\ge 4$, 2-handles into

591: valence-3 2-handles.  \end{proof}

592: 

593: \begin{definition}  If $B\neq\emptyset$ is a compact submanifold of

594: $\partial M$, then define $\mathrm{rank}_{\rho_1}(M,B)$ to be the least

595: $n$ such that there exists a handle decomposition $\Delta$ on $(M,B)$

596: with $\rho_1(\Delta)=n$.\end{definition}

597: 

598: \begin{problem}\label{problem} Is there an example of a compact 

599: hyperbolic 3-manifold $N$ with $T$ a component of $\partial N$ and $A$

600: an essential annulus in $T$ such that $\mathrm{rank}_{\rho_1}(N,A)<$

601: $\mathrm{rank}_{\rho_1}(N,T)$?

602: \end{problem}

603: 

604: \begin{section} {Estimates for the reduction of $\rho_1$ under splitting}

605: 

606: The main result of these next two sections is Theorem \ref{hyperbolic

607: Mom-n} which shows that if a hyperbolic 3-manifold $N$ has an internal

608: Mom-$n$ structure $(M,T,\Delta)$ with $\Delta$ full and $n\le 4$, then

609: it has an internal Mom-$k$ structure $(M_1,T_1,\Delta_1)$ where

610: $k\le n$, $\Delta_1$ is full, and $M_1$ is hyperbolic. 

611: If $n>4$, we obtain the similar

612: conclusion except that ``full'' is replaced by ``general based'' and hence

613: $T_1$ can be a union of tori and annuli.

614: 

615: As far as we know, transforming a structure based on an annulus lying

616: in $T_1$ to one based on the whole torus $T_1$ may require an increase in

617: $\rho_1$.  This issue is responsible for many of the technicalities of

618: this section and the next.  See Problem \ref{problem}.

619: 

620: 

621: 

622: \begin{lemma}\label{free edge}  Let $f:M\to N$ be a non-elementary

623: embedding of a compact connected manifold into a compact irreducible

624: 3-manifold.  Suppose $\partial M$ is a union of tori, $T$ is a

625: component of $\partial M$ and that $\Delta$ is a handlebody structure

626: on $(M,T)$ without 0-handles such that each 2-handle is of valence

627: $\ge 3$.  Then there exists a non-elementary embedding $g:M'\to N$

628: with $\partial M'$ a union of tori and a handle structure $\Delta'$ on

629: $(M',T')$ with $T'=T$ such that

630: $$\rho_1(\Delta')+2|\text{valence-1 1-handles of

631: $\Delta$}|+3|\text{valence-0 1-handles of $\Delta$}|\le

632: \rho_1(\Delta).$$

633: 

634: If instead each 2-handle of $\Delta$ is of valence $\ge 2$ then we

635: have

636: $$\rho_1(\Delta')+|\text{valence-1 1-handles of

637: $\Delta$}|+2|\text{valence-0 1-handles of $\Delta$}|\le

638: \rho_1(\Delta).$$

639: \end{lemma}

640: 

641: \begin{proof}  Both assertions follow similarly by

642: induction on the number of 1-handles of $\Delta$.  If $\eta$ is a

643: valence-1 1-handle, then cancelling $\eta$ with its corresponding

644: 2-handle creates a handle structure $\Delta_1$. The Lemma follows by

645: applying the $\rho_1$-formula to $\Delta_1$ and induction.  If $\eta$

646: is a valence-0 1-handle, then the manifold $M_1$ obtained by deleting

647: $\eta$ is connected and $\chi(M_1)=1$, hence has a 2-sphere boundary

648: component $S$ which bounds a 3-ball disjoint from $\cirM_1$.  Let

649: $(M_2,T,\Delta_2)$ be obtained from $(M,T,\Delta)$ by deleting $\eta$

650: as well as a 2-handle which faces $S$. $M_2$ is a non-elementary

651: embedding by Lemma \ref{pi image}.  Now apply the $\rho_1$-formula and

652: induction.\end{proof}

653: 

654: 

655: 

656: \begin{lemma}  \label{boundary reduction} Let $f:M\to N$ be a 

657: non-elementary embedding of a manifold into a compact hyperbolic

658: 3-manifold $N$, where $\partial M$ is a union of tori.  Suppose that

659: $M$ has a full handle structure $\Delta$ without 0-handles based on

660: the component $T$ of $\partial M$ such that every 2-handle is of

661: valence $\ge 3$.  If either of the following are true then there

662: exists a non-elementary embedding $f:M' \to N$ with handle structure

663: $\Delta'$ on $(M',T')$ such that $\rho_1(\Delta')+2\le

664: \rho_1(\Delta)$, $T'=T$ and $\partial M'$ is a union of tori:

665: 

666: i) There exists a valence-1 2-handle $\sigma\subset N-\cirM$ that can

667: be added to $\Delta$.

668: 

669: ii) There exists a disc $D\subset\partial M$ such that $\partial D$ is

670: the union of two arcs $\alpha\cup\beta$, where $\beta$ lies in a lake

671: and $\alpha$ lies in a 2-handle $\lambda$.  Furthermore, within

672: $\lambda\cap \partial M$, $\alpha$ separates components of

673: $\lambda\cap (\mathrm{1-handles})$.\end{lemma}

674: 

675: \begin{proof}  By Lemma

676: \ref{free edge} we can assume that every 1-handle of $\Delta$ is of

677: valence $\ge 2$.  To prove (i) let $\Delta_1$ be the handle structure

678: on the manifold $M_1$ obtained by attaching $\sigma$ to $\Delta$ along

679: $\partial M$.  Let $\eta$ denote the 1-handle which $\sigma$ meets.

680: Let $\Delta_2$ and $M_2$ be obtained by deleting a 2-handle $\lambda

681: \neq \sigma$, which faces the resulting 2-sphere boundary component.

682: Let $\Delta_3$ be obtained by cancelling $\sigma$ and $\eta$.

683: Finally, in the usual way, reduce to a non-elementary $M_4$ with

684: structure $\Delta_4$ on $(M_4,T)$ whose 1 and 2-handles are of valence

685: $\ge 2$.  Applying the $\rho_1$-formula shows that

686: $\rho_1(\Delta_4)+2\le \rho_1(\Delta)$ unless, measured in $\Delta$,

687: $\valence(\eta)=\valence(\lambda)=3$ and $\lambda$ attaches to $\eta$

688: at least twice.  If $\lambda$ attaches to $\eta$ twice, then in the

689: passage from $\Delta_1$ to $\Delta_2$ delete a 2-handle $\lambda_1\neq

690: \lambda$ which faces the 2-sphere.  If $\lambda$ attaches to $\eta$

691: thrice, then either $\Delta$ is not full or $\eta$ is the unique

692: 1-handle of $\Delta$.  In the latter case $M_2=T\times I$ which is

693: elementary, a contradiction.

694: 

695: ii) Under these hypotheses we can attach a 2-handle $\sigma\subset

696: N-\cirM$ to $\Delta$ such that either valence($\sigma$)=1 or

697: $\valence(\sigma)\le \valence(\lambda)+2$ and $\lambda$ faces the

698: resulting 2-sphere boundary component.  If valence$(\sigma)=1$, then

699: apply (i).  Otherwise let $\Delta_1$ be obtained by deleting $\lambda$

700: and apply the $\rho_1$-formula and if necessary Lemma \ref{free edge}.

701: \end{proof}

702: 

703: \begin{lemma}\label {0-handle reduction}  Let $M$ be a non-elementary

704: embedding of a compact 3-manifold into the compact hyperbolic

705: 3-manifold $N$ with $\partial M$ a union of tori, and let $\Delta$ be a

706: handle structure of $M$ based on $R\subset \partial M$.  If there

707: exists a valence $\ge 3$ 1-handle $\eta$ of $\Delta$ which attaches to

708: a 0-handle $\zeta$, then there exists a non-elementary embedding

709: $M'\to N$, and a handle structure $\Delta'$ based on $R'\subset

710: \partial M'$ such that $\rho_1(\Delta')<\rho_1(\Delta)$.  Here either

711: $(M',R')=(M,R)$ or $M'=M-\cirV$ and $R'=R\cup\partial V$ where $V$ is

712: an embedded solid torus in $\cirM$.

713: \end{lemma}

714: 

715: \begin{proof}  If $\eta$ also attaches to either the base or a 0-handle

716: distinct from $\zeta$, then cancelling $\eta$ with $\zeta$ gives rise

717: to $\Delta'$ on $(M,R)$ with $\rho_1(\Delta')<\rho_1(\Delta)$.  If

718: $\eta$ attaches only to $\zeta$, then let $M'$ be obtained by deleting

719: $\cirV$, the open solid torus gotten by hollowing out $\zeta$ and

720: $\eta$.  Let $R'=R\cup \partial V$ and let $\Delta'$ be the induced

721: structure on $(M',R')$.\end{proof}

722: 

723: \begin{lemma} \label{reducible reduction} Let $N$ be a compact hyperbolic

724: 3-manifold.  If $(M,T,\Delta)$ is a full internal weak Mom-n structure

725: on $N$ and $M$ is reducible then there exists $(M_p,T_p,\Delta_p)$, a

726: weak internal Mom-k structure on N such that

727: $k+2=\rho_1(\Delta_p)+2\le\rho_1(\Delta)=n$. If $T$ does not lie in a

728: 3-cell, then $T_p=T$.\end{lemma}

729: 

730: \begin{proof}  We first consider the case that no reducing sphere in 

731: $M$ bounds a ball in $N$ containing $T$.  Let $F$ be a least-weight

732: normal reducing 2-sphere and note that $F\cap T\times I$ is a union of

733: discs.  Let $M_0$ and $M_0'$ be the components of $M$ split along

734: $F$. By Lemma \ref{pi image} and the irreducibility of $N$ exactly one

735: of $M_0$, $M_0'$ is non-elementary.  We let $M_0$ (resp. $M_0'$) denote

736: the non-elementary (resp. elementary) component with $\Delta_0$

737: (resp. $\Delta_0'$) its induced structure.  By hypothesis $T\subset

738: M_0$.

739: 

740: We show that $\rho_1(\Delta_0^\prime)>0$.  Let $X$ denote the union of

741: the islands and bridges of $\Delta$ and $Y'=X\cap M_0^\prime$.  If

742: $\rho_1(\Delta_0^\prime)=0$, then each component $A$ of $Y'$ is an

743: annulus.  If some component $A$ of $Y'$ is disjoint from the lakes of

744: $\Delta$, then $Y'$ would be disjoint from the lakes and so $F$ would

745: 2-fold cover a projective plane, which contradicts the fact that $N$

746: is irreducible.  Therefore each component of $Y'$ has one boundary

747: component in a lake and one component in $\cirX$.  Since $\Delta$ is

748: full, this implies that $F$ is a boundary parallel 2-sphere and

749: contradicts the fact that $\partial M$ is a union of tori.

750: 

751: If $Y=X\cap M_0$ then a similar argument shows that some component of

752: $Y$ is not an annulus and furthermore some 2-handle $\sigma$ of

753: $\Delta_0$ faces $F$ and attaches to a valence-$\ge 3$ 1-handle.

754: Delete $\sigma$ from $\Delta_0$ to obtain $(M_1,T,\Delta_1)$ with

755: $\rho_1(\Delta_1)+1\le\rho_1(\Delta_0)$.  The standard simplifying

756: moves as in Lemma \ref{free edge} transform $(M_1,T,\Delta_1)$ to

757: $(M_2,T,\Delta_2)$, a weak Mom-k with

758: $k=\rho_1(\Delta_2)\le\rho_1(\Delta_1)$. Since

759: $\rho_1(\Delta_0)+\rho_1(\Delta_0^\prime)=\rho_1(\Delta)=n$ we have

760: $k+2=\rho_1(\Delta_p)+2\le\rho_1(\Delta)=n$.  If $M_2$ is reducible,

761: then split along an essential least-weight sphere $F_2$ which is normal with

762: respect to $\Delta_2$.  Retain the component which contains $T$ and do

763: the usual operations to obtain the weak Mom-$k_2$ $(M_3,T,\Delta_3)$

764: with $k_2\le k$ and $M_3$ non-elementary.  By Haken finiteness, this

765: procedure terminates in a finite number of steps, completing the proof

766: in this case.

767: 

768: \smallskip We now consider the case that $T$ is compressible in $M$.

769: Let $F$ be a least-weight compressing disc for $T$.  Note that $F\cap

770: T\times I$ consists of discs and a single annulus.  If $M_1$ is $M$

771: split along $F$ with induced handle structure $\Delta_1$ and $A$ is

772: $T$ split along $F\cap T$, then $\Delta_1$ is based on $A$.  Since $M$

773: is obtained by attaching a 1-handle to a spherical component of

774: $\partial M_1$, it follows that $M_1$ is non-elementarily embedded in

775: $N$.  Let $M_2$ be obtained by attaching 2-handles to $\partial M_1$

776: along the components of $\partial A$, then capping off the resulting

777: 2-sphere which faces $A$ with a 3-cell.  The resulting manifold $M_2$

778: has two 2-sphere boundary components and the induced handle structure

779: $\Delta_2$ is classical.  Since $\Delta$ is full some 2-handle of

780: $\Delta_2$ facing a 2-sphere of $\partial M_2$ attaches to a 1-handle

781: of valence $\ge 3$.  Delete this 2-handle to obtain the non-elementary

782: $M_3$ with handle structure $\Delta_3$ which satisfies

783: $\rho_1(\Delta_3)+1\le \rho_1(\Delta)$.  Delete another 2-handle to

784: create $M_4$ and $\Delta_4$ such that $\partial M_4$ is a union of

785: tori.  Now apply Lemma \ref{0-handle reduction} to $\Delta_4$ to

786: create $(M_5,T_5,\Delta_5)$ so that $\rho_1(\Delta_5)+1\le

787: \rho_1(\Delta_4)$.  Finally cancel the extraneous 0-handles and

788: valence-1 1-handles to create the desired weak internal Mom-k

789: structure $(M_6,T_6,\Delta_6)$.

790: 

791: \smallskip

792: Let $F$ be a least-weight essential normal 2-sphere for $\Delta$.

793: Since we can assume that $T$ is incompressible in $M$, $F\cap T\times

794: I$ is a union of discs.  If $T$ does not lie in the ball bounded by

795: $F\subset N$, then proceed as in the first part of the proof.

796: Otherwise let $M_0$ and $M_0'$ be the components of $M$ split along

797: $F$, with $M_0'$ the component containing $T$. Since $\Delta$ is full

798: and some component of $T\times 1\cap M_0'$ is nonplanar, it follows

799: that $\rho_1(\Delta_0')\ge 1$ and hence

800: $\rho_1(\Delta_0)+1\le\rho_1(\Delta)$.  To complete the proof apply

801: Lemma \ref{0-handle reduction} to $\Delta_0$, delete a 2-handle to

802: create a manifold with torus boundary components and cancel low

803: valence handles as in the previous paragraph.

804: \end{proof}

805: 

806: 

807: 

808: \begin{remark} If $(M, T, \Delta)$ is a weak internal Mom-n structure 

809: on the compact hyperbolic manifold $N$, and $\partial M$ is

810: compressible in $M$, then $M$ is reducible.  Therefore, if $M$ is

811: irreducible and $\Delta$ has an annular lake $A$ that is homotopically

812: inessential in $T\times 1$, then the core of $A$ bounds a disc in

813: $T\times I$ which separates off a 3-cell in $M$.  Absorbing this

814: 3-cell into $T\times I$ simplifies $\Delta$ and transforms $A$ into a

815: disc lake.

816: 

817: From now on we will assume that if a homotopically inessential lake

818: appears it is immediately removed via the above operation.

819: \end{remark}

820: 

821: 

822: \begin{definition}  If $\Delta$ is a handle structure on $M$, then 

823: the \emph{sheets} of $\Delta$ are the connected components of the

824: space $\mathcal{S}$ which is the union of the 2-handles and the

825: valence-2 1-handles.  So sheets are naturally thickened surfaces which

826: are attached to a 3-manifold along their thickened boundaries.  The

827: valence of a sheet is the number of times the boundary runs over

828: 1-handles counted with multiplicity.

829: \end{definition}

830: 

831: 

832: \begin{lemma}  \label{sheet reduction} Let $N$ be a compact hyperbolic

833: 3-manifold and $f:M\to N$ a non-elementary embedding where $\partial M$

834: is a union of tori.  Let $\Delta$ be a handle structure on $M$ with no

835: 0-handles based on a component $T$ of $\partial M$ such that the

836: valence of each 2-handle is at least 3 and $T$ does not lie in a

837: 3-cell of $N$.  If some sheet $S$ of $\Delta$ is not a thickened disc

838: then there exists a non-elementary $M'\subset N$ with handle structure

839: $\Delta'$ based on a component $T'$ of $\partial M'$ such that

840: $\rho_1(\Delta')+1\le \rho_1(\Delta)$.  If equality must hold then

841: $S$ is a thickened annulus or Mobius band and if $\Delta$ is full

842: then $S$ is a thickened Mobius band and $\Delta'$ is full.

843: \end{lemma}

844: 

845: 

846: \begin{proof}  By Lemma \ref{free edge} and the proof of Lemma 

847: \ref{reducible reduction} we can assume that $M $ is irreducible and

848:   each 1-handle of $\Delta$ has valence $\ge 2$. If $M_1$ denotes the

849:   manifold obtained by deleting the sheet $S$, then

850:   $\chi(M)=\chi(M_1)+\chi(S)$.  Since $\chi(M)=0$, if $\chi(S)<0$,

851:   then $\partial M_1$ contains a 2-sphere.  This implies that $M$ is

852:   reducible.  Note that if $\chi(S)=0$, then $M_1$ is non-elementary.

853: 

854: 

855: Now assume that $\chi(S)=0$.  In this case $S$ is either an

856: annulus$\times I$ or a non-trivial $I$-bundle over a Mobius band.

857: Since $\chi(S)=0$, if $S$ contains more than one 2-handle, then

858: $\valence(S)\ge 2$.  If $\valence(S)>1$, then $\rho_1(\Delta_1)+2\le

859: \rho_1(\Delta)$ holds where $\Delta_1$ is the induced structure on

860: $M_1$.  If $S$ is a valence-1 annulus$\times I$, then $\Delta$ has an

861: annular lake and $\rho_1(\Delta_1)+1=\rho_1(\Delta)$.

862: 

863: If $S$ is a thickened Mobius band of valence-1, then

864: $\rho_1(\Delta_1)+1= \rho_1(\Delta)$.  If $S$ attached to the

865: component $R$ of $\partial M_1$ and $R$ had an annular lake, then $R$

866: must be a tube with a compressing disc $D$ whose boundary goes over a

867: 1-handle $\eta$ of $\Delta_1$ exactly once. Let $M_2$ be obtained by

868: attaching the 2-handle $\sigma$ with core $D$ to $M_1$ and $\Delta_2$

869: the induced handle structure.  Proceed as in the proof of Lemma \ref

870: {boundary reduction} to show that exists a non-elementary embedding

871: $(M_3,T)$ with handle structure $\Delta_3$ with

872: $\rho_1(\Delta_3)+2\le\rho_1(\Delta_1)+1\le \rho_1(\Delta)$.

873: \end{proof}

874: 

875: 

876: \begin{lemma}  \label{annulus reduction}   Let $(M,T,\Delta)$ be an 

877: internal Mom-n structure on the compact hyperbolic manifold $N$.

878: Assume that every sheet of $\Delta$ is a disc and $\Delta$ is full.

879: If there exists an embedded annulus $A$ connecting the component $S$

880: of $\partial M-T$ to $T$, then there exists a non-elementary embedding

881: of a manifold $M'$ into $N$ such that $\partial M'$ is a union of tori

882: and $\rho_1(\Delta')+2\le \rho_1(\Delta)$.  Here $\Delta'$ is a handle

883: structure on $M'$ based on $T'$.  Either $T'=T$ or $T'$ is the union

884: of an essential annulus in $T$ and possibly a component of $\partial

885: M'$.

886: \end{lemma}

887: 

888: \begin{proof}  By Lemma \ref{reducible reduction} we  can

889: assume that $M$ is irreducible.  Since all sheets are discs and

890: letting $M_2=M$, we obtain from $\Delta$ a full handle structure $\Delta_2$

891: with no 0-handles based on the component $T_2=T$ of $\partial M_2$

892: where each 1 and 2-handle of $\Delta_2$ is of valence $\ge 3$.  Finally

893: $\rho_1(\Delta_2)\le \rho_1(\Delta)$.  We can assume that $A\subset

894: M_2$ is a least-weight normal annulus connecting $T_2$ to the boundary

895: component $S\neq T_2$.  Since $A$ is least weight, $A\cap T_2\times I$

896: is a union of discs and a single annulus.

897: 

898: Let $\partial A_0$ (resp.  $\partial A_1$) denote $A\cap T_2$

899: (resp. $A\cap S$).  Our $A$ has an induced handle structure $\Phi$

900: based on $\partial A_0$ as follows.  The base consists of the annular

901: component of $A\cap T_2\times [0,1]$, the 0-handles consist of the

902: disc components of $A\cap T_2\times[0,1]$, the 1-handles

903: (resp. 2-handles) consist of the intersections of $A$ with the

904: 1-handles (resp. 2-handles).

905: 

906: Let $M_3$ denote the manifold obtained by splitting $M_2$ along $A$

907: and let $\Delta_3$ denote the induced handle structure.  As in \S 2,

908: the ball components of $T_2\times I$ split along $A$ to become 0-handles

909: of $\Delta_3$.  The remaining component is $B' \times I$, where $B'$

910: is $T_2$ split along $\partial A_0$ and hence $\Delta_3$ is based on

911: $B'\subset T_2$.  By Lemma \ref{pi image}, $M_3$ is non-elementarily

912: embedded in $N$.

913: 

914: 

915: If $\eta$ is a 1-handle of $\Delta_2$ and if $\{\eta_i\}$ denotes the

916: 1-handles of $\Delta_3$ which descended from $\eta$, then $\sum_i

917: (\valence(\eta_i)-2) \le \valence(\eta)-2$ with equality if and only

918: if $A_1$ does not run over $\eta$.  In fact, counting with

919: multiplicity, if $A_1$ runs over the 1-handles of $\Delta_2$ more than

920: once then by operations as in Lemma \ref{free edge} we can pass to

921: $M_4$ and $\Delta_4$ with $\rho_1(\Delta_4)+2\le\rho_1(\Delta)$.  We

922: will now assume that $A_1$ ran over a unique 1-handle $\eta$ and it

923: did so with multiplicity one.  This implies that

924: $\rho_1(\Delta_3)+1=\rho_1(\Delta_2)$.  Also $A_1$ is the union of two

925: arcs $\alpha$ and $\beta$, where $\alpha$ lies in a 1-handle of

926: $\Delta_2$ and $\beta$ lies in a lake.  $\beta$ lies either in the

927: base of $\Phi$ or in a 0-handle $v^*$ of $\Phi$.

928: 

929: 

930: 

931: Give $A$ a transverse orientation. Call a 0-handle $v$ of $\Phi$

932: \emph{plus} (resp.  \emph{minus}) if the transverse orientation of the

933: disc $v\subset T_2\times I$ points away from (resp. towards) $T_2$.

934: Each such disc $v$ separates off, in $T_2\times I$, a 3-ball $v_B$.

935: Let $v_D$ denote $v_B\cap T_2\times 1$.

936: 

937: 

938: If $\beta$ lies in a 0-handle $v^*$, then Lemma \ref{0-handle

939: reduction} applies to $\Delta_3$ and the Lemma is proved.  Indeed,

940: since $A$ is least weight, the disc $v^*_D$ contains a bridge $b$ in

941: its interior and this bridge is not parallel to $\beta$.  In other

942: words, there

943: is no embedded disc in a lake whose boundary is a concatenation of four

944: arcs $\beta_1, \beta_2, \beta_3$ and $\beta_4$ such that $\beta_1=\beta,

945: \beta_3$ lies in $\partial b$, and $\beta_2$ and $\beta_4$ are arcs

946: lying in islands.  Otherwise, since $A_1$ runs over a unique 1-handle,

947: this implies that $\Delta_2$ had a valence-1 2-handle.  If $v^*_D$ is

948: disjoint from the other 0-handles of $\Phi$, then Lemma \ref{0-handle

949: reduction} applies to the 0-handle $v^*_B$ of $\Delta_3$.  Otherwise,

950: $v^*_B$ is split into balls by the various 0-handles of $\Phi$ and

951: Lemma \ref{0-handle reduction} applies to one of these balls.

952: 

953:  From now on we assume that $\beta$ lies in the base of $\Phi$.  Let

954: $X\subset T_2\times 1$ denote the union of the islands and bridges of

955: $\Delta_2$.  A similar but easier argument to the one given in the

956: previous paragraph shows that for each 0-handle $v$ of $\Phi$ either

957: $\partial v$ is boundary parallel in $X$ or Lemma \ref{0-handle

958: reduction} applies.  This implies that if $v\neq w$ are 0-handles of

959: $\Phi$, and $v_B\subset w_B$, then $\partial v$ and $\partial w$ are

960: normally parallel in $X$.  Furthermore, no 1-handle of $\Phi$ connects

961: a plus 0-handle to a minus 0-handle of $\Phi$.  Also, if $v^0$ and

962: $v^1$ are two 0-handles of $\Phi$ that are connected by a 1-handle,

963: then $v^0_B\cap v^1_B=\emptyset$.  Finally, there do not exist

964: 0-handles $w^0, w^1, \cdots w^n$ of $\Phi$ such that for $i=1, 2,

965: \cdots, n-1,\ w^i$ is connected to $w^{i+1}$ by a 1-handle and

966: $w^n_B\subset w^0_B$.

967: 

968: 

969: 

970: It follows that there exists a disc $E\subset A$ whose boundary is the

971: union of two arcs $\phi$ and $\psi$ where $\phi$ is a proper arc in a

972: 2-handle of $\Phi$ and $\psi$ lies in the $T\times 1\cap (\base(A))$.

973: Furthermore $E\cap (0,1\text{-handles}\ \Phi)\neq \emptyset$ and is

974: connected.  By the previous paragraph if $v,w$ are 0-handles of $\Phi$

975: lying in $E$, then they are of the same parity and $v_B\cap

976: w_B=\emptyset$.

977: 

978: 

979: 

980: 

981: Therefore, $E$ can be normally isotoped, with respect to $\Delta_2$ to

982: a disc $G\subset \partial M$ such that $\partial G$ is a union of two

983: arcs, one lying in a lake and the other in a 2-handle.  Now apply (ii)

984: of Lemma \ref{boundary reduction} to $\Delta_2$.\end{proof}

985: 

986: 

987: 

988: 

989: \begin{lemma} \label{annulus reduction II}  Let $(M,T,\Delta)$ be a 

990: full internal Mom-$n$ structure on the compact hyperbolic 3-manifold $N$

991: such that every sheet is a disc.  Suppose that there exists an

992: essential embedded annulus $A$ with $\partial A\cap T=\emptyset$.

993: Then either there exists a full internal Mom-$k$ structure

994: $(M',T',\Delta')$ on $N$ with $k<n$ or there exists a non-elementary embedding

995: $M'\to N$ with handle structure $(M',T',\Delta')$ where $\partial M'$

996: is a union of tori, $T'$ is a component of $\partial M'$, and

997: $\rho_1(\Delta')+2\le \rho_1(\Delta)$.\end{lemma}

998: 

999: \begin{proof}  Assume that $M$ is irreducible.  By Lemma \ref{annulus 

1000: reduction} we can assume that no essential annulus connects $T$ to a

1001: component of $\partial M-T$.  Let $A$ be a least-weight essential

1002: normal annulus with $A\cap T=\emptyset$.  By the second sentence,

1003: $A\cap T\times I$ is a union of discs.  Let $M_1$ be the non-elementary

1004: component of M split along $A$ and let $(M_1,T_1,\Delta_1)$ be the induced

1005: handle structure.  As in the previous proof, $\rho_1(\Delta_1)+1\le

1006: \rho_1(\Delta)$.  Indeed, since $\partial A$ traverses at least two

1007: 1-handles of $\Delta$, counted with multiplicity, the inequality will

1008: be strict unless $A$ normally double covers a Mobius band.  If $M_1$

1009: is disjoint from $T$, then the result follows from Lemma \ref{0-handle

1010: reduction}.  Assume now that $T\subset\partial M_1$ and $\Delta_1$ is

1011: full.

1012: 

1013: Cancel the 0-handles with 1-handles to obtain $(M_2,T_2,\Delta_2)$.

1014: If $\Delta_2$ is not full, then some 1-handle of valence $\ge 3$ was

1015: cancelled and hence $\rho_1$ is reduced.  Next cancel valence-1

1016: 2-handles to obtain $(M_3,T_3,\Delta_3)$.  Again fullness is preserved

1017: or $\rho_1$ is reduced.  If $\Delta_3$ has a valence-2 2-handle

1018: $\sigma$ connecting distinct 1-handles, then cancel one of these

1019: 1-handles with $\sigma$.  If the 2-handle $\sigma$ attaches to the

1020: same 1-handle $\alpha$, then the core of $\sigma\cap\alpha$ can be

1021: viewed as an embedded annulus or Mobius band $C$ in $M_3$.  Splitting

1022: $M_3$ along $C$ we obtain $(M_4,T_4,\Delta_4)$ where $T_4$ is the

1023: newly created boundary component of $M_4$ and $\Delta_4$ is the

1024: induced structure.  Note that $\rho_1(\Delta_4)<\rho_1(\Delta_3)$

1025: since $\Delta_3$ is full.  After finitely many such operations the

1026: lemma is proved.  \end{proof}

1027: 

1028: \begin{lemma}\label{annulus reduction III}  Let $(M,T,\Delta)$ be a 

1029: full internal Mom-$n$ structure on the compact hyperbolic 3-manifold $N$

1030: such that every sheet is a disc.  Suppose that there exists an

1031: essential embedded annulus $A$ with $\partial A\subset T$.  Then

1032: either there exists a full internal Mom-$k$ structure

1033: $(M',T',\Delta')$ on $N$ with $k<n$ or there exists a non-elementary embedding

1034: $M'\to N$ with handle structure $(M',T',\Delta')$ where $\partial M'$

1035: is a union of tori, $T'$ is a component of $\partial M'$, and

1036: $\rho_1(\Delta')+2\le \rho_1(\Delta)$.\end{lemma}

1037: 

1038: \begin{proof}  If $M$ is reducible or there exists an essential 

1039: annulus with some boundary component disjoint from $T$, then apply

1040: Lemma \ref{reducible reduction}, Lemma \ref{annulus reduction} or

1041: Lemma \ref{annulus reduction II}.  Let $(M,\partial M-T, \Sigma)$ be

1042: the dual handle structure.  I.e. the 1-handles (resp. 2-handles) of

1043: $\Sigma$ are in 1-1 correspondence with the 2-handles

1044: (resp. 1-handles) of $\Delta$ and for each handle $\sigma$, the core of

1045: $\sigma$ is the cocore of the dual handle and vice versa.  Note that

1046: $\rho_1(\Sigma)=\rho_1(\Delta)$ and each 1-handle of $\Sigma$ is of

1047: valence $3$, but $\Sigma$ may have valence-2 2-handles.

1048: 

1049: Let $A$ be an essential annulus, least weight with respect to

1050: $\Sigma$.  Since each essential annulus has its entire boundary in

1051: $T$, $A\cap (\partial M-T)\times I$ is a union of discs.  Let $M_1$ be

1052: the non-elementary component of $M$ split along $A$ with

1053: $(M_1,R_1,\Sigma_1)$ the induced structure.  As before

1054: $\rho_1(\Sigma_1)<\rho_1(\Sigma)$.  Eliminate the 0-handles and low

1055: valence-1 and 2-handles of $\Sigma_1$ to obtain $(M_2,R_2,\Sigma_2)$

1056: and note that either $\Sigma_2$ is full or

1057: $\rho_1(\Sigma_2)+2\le\rho_1(\Sigma)=\rho_1(\Delta)$.  If $\Sigma_2$

1058: is full, then $(M_3,T_3,\Delta_3)$ the handle structure dual to

1059: $(M_2,R_2,\Sigma_2)$ is full.  Let $(M_4,T_4,\Delta_4)$ be obtained by

1060: eliminating the valence-2 2-handles as in the previous proof.  As

1061: before $\rho_1(\Delta_4)+1\le \rho_1(\Delta)$ with equality holding

1062: only if $\Delta_4$ is full.

1063: \end{proof}

1064: 

1065: 

1066: 

1067: \begin{lemma} \label{torus reduction}  Let $(M,T,\Delta)$ be a full 

1068: internal Mom-$n$ structure on the compact hyperbolic 3-manifold $N$.  If

1069: $M$ is not hyperbolic, then either there exists a full internal Mom-$k$

1070: structure $(M',T',\Delta')$ on $N$ with $k<n$ and $M'$ hyperbolic or

1071: there exists a general based internal Mom-l, $l+2\le n$, structure

1072: $(M',B',\Delta')$ on $N$.\end{lemma}

1073: 

1074: \begin{proof}  By Lemma

1075: \ref{reducible reduction}, \ref{sheet reduction}, \ref{annulus

1076: reduction}, \ref{annulus reduction II} or \ref{annulus reduction III}

1077: it suffices to consider the case that $M$ is irreducible and

1078: anannular.  By Thurston \cite{Th2}, if $M$ is not hyperbolic, then it

1079: contains an essential torus.  Let $F$ be a least-weight essential

1080: torus.  Since there are no essential annuli, $F\cap T\times I$ is a

1081: union of discs.

1082: 

1083: 

1084: 

1085:   Let $M_1$ denote the component of $M$ split along $F$ which contains

1086: $T$ and let $M_1'$ denote the other component. Let $\Delta_1$ and

1087: $\Delta_1'$ denote the induced handle structures.  Note that

1088: $\Delta_1$ is based on $T$, $\Delta_1'$ is a classical structure and

1089: $\rho_1(\Delta_1)+\rho_1(\Delta_1')=\rho_1(\Delta)$.  By Lemma \ref{pi

1090: image} one of $M_1$ or $M_1'$ is non-elementary .  Let $X\subset

1091: T\times 1$ be the union of the islands and bridges of $\Delta$,

1092: $Y=M_1'\cap X$ and $Z=M_1\cap X$.

1093: 

1094: If $M_1$ is elementary, $\rho_1(\Delta_1)\ge 1$ since $Z$ is

1095: nonplanar.  In the usual way obtain a handle structure $\Sigma$ on

1096: $M_1'$ such that $\rho_1(\Sigma)\le \rho_1(\Delta')$, $\Sigma$ has

1097: exactly one 0-handle and each 1-handle is of valence $\ge 2$.  Since

1098: $M_1'$ is non-elementary, some 1-handle is of valence $\ge 3$.  To

1099: complete the proof apply Lemma \ref{0-handle reduction}.

1100: 

1101: 

1102:  From now on assume that $M_1$ is non-elementary.  We first show that

1103: $\rho_1(\Delta_1')\ge 1$.  If $\rho_1(\Delta_1')=0$, then $Y$ is a

1104: union of annuli, each component of which intersects $\partial X$ in

1105: $\le 1$ circle, since $\Delta$ is full.  If $Y\cap \partial

1106: X\neq\emptyset$, then each component must have this property and hence

1107: $F$ is boundary parallel in $M$.  If $Y\cap \partial X=\emptyset$,

1108: then $M_1'$ is an I-bundle over a Klein bottle, which is impossible in

1109: an orientable hyperbolic 3-manifold.

1110: 

1111: In the usual way pass from $(M_1,T_1,\Delta_1)$ to a Mom structure

1112: $(M_2,T_2,\Delta_2)$ with $\rho_1(\Delta_2)\le

1113: \rho_1(\Delta_1)$. Since $M$ is anannular, $\Delta_2$ has no annular

1114: lakes.  If some component of $\partial M_2$ is a convolutube, then

1115: reimbed $M_2$ in $N$ to get a full internal Mom-k structure where

1116: $k<n$.\end{proof}

1117: 

1118: 

1119: \begin{section} {From  Mom-$n$ to  Hyperbolic

1120: Mom-$n$}\end{section}

1121: 

1122: 

1123: The following is the main result of this section:

1124: 

1125: \smallskip

1126: \begin{theorem} \label{hyperbolic Mom-n} If $(M,T,\Delta)$ is a full 

1127: internal Mom-$n$ structure on the compact hyperbolic 3-manifold $N$ and

1128: $n\le 4$, then there exists a full internal Mom-$k$ structure

1129: $(M',T',\Delta')$ on $N$ where $M'$ is hyperbolic and $k\le n$.

1130: 

1131: For general $n$, either there exists an internal Mom-$k$ structure

1132: $(M',T', \Delta')$ in $N$ with $k\le n$ and $M$ hyperbolic or there

1133: exists a general based internal Mom-$k$ structure $(M',B',\Delta')$

1134: on $N$ such

1135: that $M'$ is hyperbolic and each component of $B'$ is either a component

1136: of $\partial M'$ or an annulus which is essential in a component of $\partial

1137: M$.  If $M\neq M'$, then $k+2\le n$.\end{theorem}

1138: 

1139: We first prove some preliminary lemmas about complexity (see Definition \ref{def:valence}).

1140: 

1141: \begin{lemma}(Clean-Up Lemma)\label{clean up}  Let $M$ be a compact 

1142: 3-manifold with $\partial M$ a union of tori and $N$ a compact

1143: hyperbolic 3-manifold.  If $M\to N$ is a non-elementary embedding and

1144: $\Delta$ is a handle structure on $(M,B)$, then there exists

1145: $(M',B',\Delta')$ a weak general based internal Mom-$n$ structure on $N$

1146: with $C(\Delta')\le C(\Delta)$.

1147: 

1148: If $M$ is hyperbolic, then either $\rho_1(\Delta')<\rho_1(\Delta)$ or

1149: $M'=M$.

1150: \end{lemma}

1151: 

1152: \begin{proof} Proof by induction on $C(\Delta)$.  By the usual

1153: cancellation operations pass to a handle structure $\Delta_2$ on

1154: $(M,B)$ without 0-handles or 1 and 2-handles of valence 1.  If $\Delta$

1155: has a valence-0 1-handle, then delete it and proceed as in Lemma

1156: \ref{free edge} to obtain $\Delta_3$ on $(M_3,B_3)$ with $M_3$

1157: non-elementary and $\partial M_3$ a union of tori.  All of these

1158: operations reduce $C(\Delta)$.

1159: 

1160: If $\Delta_2$ has a valence-0 2-handle $\sigma$, then compress $M_3$

1161: along a disc $D$ which passes once through $\sigma$ and intersects

1162: $B\times I$ in an annulus.  Let $\Delta_4$ be the resulting handle

1163: structure on $(M_4,B_4)$ where $M_4$ is the non-elementary component of

1164: $M_3$ split along $D$ and $B_4$ is $B_3\cap M_4$.  This splits

1165: $\sigma$ into two 2-handle components, at least one of which lies in

1166: $\Delta_4$.  Absorb such 2-handles into $B_4\times I$.  This creates

1167: $(M_5,B_5,\Delta_5)$ where $M_5=M_4$ and $B_5$ is obtained by

1168: attaching one or two discs to $\partial B_4$.  If some

1169: component of $B_5$ is a 2-sphere $S$, then create $(M_6,B_6,\Delta_6)$

1170: by filling in $S$ with a 3-cell $E$ and identifying $S\times I\cup E$

1171: as a 0-handle of $\Delta_6$.  Now cancel with a 1-handle to obtain

1172: $(M_7,B_7,\Delta_7)$ with $C(\Delta_7)< C(\Delta_3)$.  If one or two

1173: components of $B_5$ are discs and there exists a non-simply connected

1174: component of $B_5$, then the ball components of $B_5\times I$ are now

1175: re-identified as 0-handles and cancelled with 1-handles to obtain

1176: $(M_7,B_7,\Delta_7)$.  If all the components of $B_5$ are discs, then

1177: transforming the components of $B_5\times I$ to 0-handles produces a

1178: classical handle structure $\Delta_{5.5}$ on $M_5$.  Now apply the

1179: proof of Lemma \ref{0-handle reduction} to create $(M_6,B_6,\Delta_6)$

1180: with $C(\Delta_6)$ reduced.  If some component of $\partial M_6$ is a

1181: 2-sphere, then delete a 2-handle facing it to obtain

1182: $(M_7,B_7,\Delta_7)$.

1183: 

1184: By repeatedly applying the above operations we can assume that our

1185: $\Delta_7$ has no 0-handles, and each 1-handle and 2-handle is of

1186: valence $\ge 2$.  Furthermore, no component of $B_7$ is a disc.  If

1187: some component $G$ of $B_7$ is not $\pi_1$-injective in $\partial

1188: M_7$ then attach a 2-handle to a component of $\partial G\times I$

1189: whose restriction to $\partial M_7$ bounds a disc, to create

1190: $(M_8,B_8,\Delta_8)$.  Delete a 2-handle which faces the resulting

1191: 2-sphere boundary component and simplify as in the previous paragraphs

1192: to create $(M_9,B_9,\Delta_9)$ with $C(\Delta_9)<C(\Delta_7)$.  We can

1193: now assume that each component of $B_9$ is essential in $\partial

1194: M_9$, $\Delta_9$ is 0-handle free and all 1 and 2-handles are valence

1195: $\ge 2$.

1196: 

1197:   If $\partial M_9$ contains convolutubes, then reimbed $M_9$ in $N$

1198: to eliminate them.  If $M$ is not hyperbolic, then the resulting

1199: $(M_{10},B_{10},\Delta_{10})$ satisfies the conclusions of our clean

1200: up lemma.

1201: 

1202: If $M$ is hyperbolic, then the proof also follows as above.  However,

1203: note that $\partial M$ is incompressible so in the above process

1204: valence-0 1-handles or lakes compressible in $B\times I$ can be

1205: eliminated without changing $M$ and with reducing complexity.  The

1206: topology of $M$ may change if we delete a 2-handle or apply Lemma

1207: \ref{0-handle reduction}; however, if such operations must be done,

1208: they can be done in such a manner that reduces $\rho_1$.

1209: \end{proof}

1210: 

1211: 

1212: 

1213: \begin{lemma}  \label{weak} If

1214: $(M,B,\Delta)$ is a strictly weak general based Mom-$n$ structure with

1215: $M$ hyperbolic, then there exists a general based Mom-$k$

1216: structure $(M,B',\Delta')$ with $C(\Delta')<C(\Delta)$.

1217: \end{lemma}

1218: 

1219: \begin{proof}  Proof

1220: by induction on $C(\Delta)$.  If $\sigma$ is a valence-2 2-handle

1221: which goes over distinct 1-handles, then cancelling a 1-handle with

1222: $\sigma$ creates $(M,B, \Delta_1)$ with $C(\Delta_1)<C(\Delta)$.

1223: 

1224: We now assume that no valence-2 2-handle $\sigma$ goes over distinct

1225: 1-handles.  Suppose that $\sigma$ goes over the same 1-handle $\alpha$

1226: twice.  Then $\sigma\cup\alpha$ can be viewed as an embedded annulus

1227: or Mobius band $A$ with boundary on $B$.  Since $M$ is hyperbolic and

1228: orientable, $A$ is either a boundary parallel annulus or a

1229: compressible annulus.

1230: 

1231: If $A$ is boundary parallel, then it together with an annulus on $T$

1232: cobound a solid torus $V\subset B$ such that $A$ wraps longitudinally

1233: around $V$ exactly once.  Melting $V$ into $B\times [0,1]$ eliminates

1234: both $\alpha$ and $\sigma$, together with all the 2-handles and

1235: 1-handles inside of $V$.  The resulting manifold $(M,B_1,\Delta_1)$ is

1236: a weak provisional Mom-$k$ with $k\le n$ whose induced handle structure

1237: $\Delta_1$ satisfies $C(\Delta_1)<C(\Delta)$.

1238: 

1239: If $A$ is not boundary parallel, then $M$ is reducible, a

1240: contradiction.

1241: \end{proof}

1242: 

1243: 

1244: 

1245: 

1246: \begin{lemma}  \label{lower bound} If $(M,B,\Delta)$ is a general 

1247: based Mom-$n$ and $M$ is hyperbolic, then $n\ge 2$.

1248: \end{lemma}

1249: 

1250: \begin{proof}  This  follows by direct calculation.\end{proof}

1251: 

1252: 

1253: \begin{lemma}\label{general base 2}  If the compact hyperbolic 

1254: 3-manifold $N$ has a general based internal Mom-2 structure

1255: $(M,B,\Delta)$, then it has a full internal Mom-2

1256: structure.\end{lemma}

1257: 

1258: \begin{proof}  By splitting along annuli in $B\times I$ we can assume 

1259: that every non-peripheral lake of $B$ is a disc.  If $B$ is the union

1260: of two tori $T_1$ and $T_2$, then each 1-handle must connect $T_1$ to

1261: $T_2$.  This implies that each 2-handle is of even valence which is a

1262: contradiction.  If $B$ consists of a single torus, then $(M,B,\Delta)$

1263: is a full Mom-2 structure.  We finally assume that $B$ contains an

1264: annulus.  We only discuss the case that $B$ is connected; the case

1265: where $B$ is either the union of an annulus and torus or two annuli is

1266: similar and easier.

1267: 

1268: First suppose that $\Delta$ has a single sheet of valence four.  Let

1269: $\eta$ denote the valence-4 1-handle of $\Delta$. There exists an

1270: essential compressing disc $D$ for $B\times I$ which cuts across the

1271: bridges in at most three components.  View $N(D)$ as a 1-handle and

1272: $B\times I-\cirN(D)$ as a 0-handle to obtain a classical handle

1273: structure with two 1-handles respectively of valence 4 and $\le 3$.

1274: Now as in Lemma \ref{0-handle reduction} hollow out the 0-handle and

1275: $\eta$ to get a non-elementary $M_1$ with handle structure $\Delta_1$

1276: based on a torus with a single 1-handle of valence $\le 3$.  Therefore

1277: we obtain a Mom-1 structure on a hyperbolic 3-manifold, which is a

1278: contradiction.

1279: 

1280: Now consider the case that $\Delta$ has two valence-3 1-handles.  If

1281: some essential disc $D$ in $B\times I$ cuts the bridges in $\le 2$

1282: components, then as above, we obtain a handle structure with one

1283: 0-handle and three 1-handles.  Hollowing out a valence-3 1-handle and

1284: the 0-handle produces a handle structure $\Delta_1$ on a

1285: non-elementary manifold $M_1$ with $\rho_1(\Delta_1)=1$, which is a

1286: contradiction.  If no essential disc $D$ exists as above, then we can

1287: find one which cuts the bridges in exactly three components.  One

1288: readily enumerates the possible handle structures that satisfy our

1289: assumptions of valence-3 1-handles and non-peripheral disc lakes.

1290: After applying the hollowing out procedure, a Mom-2 handle structure

1291: is created.  Of the two choices of which 1-handle is to be hollowed

1292: out, one will produce a full Mom-2 structure.

1293: 

1294: \end{proof}

1295: 

1296: \begin{remark} As an example of the type of non-full Mom-2 discussed

1297:   in the previous lemma, note that the figure-8 knot complement

1298:   contains a non-full Mom-2 structure with two 1-handles of valence 3,

1299:   in addition to having a full Mom-2 structure.

1300: \end{remark}

1301: 

1302: \noindent\emph{Proof of Theorem \ref{hyperbolic Mom-n}} The proof is

1303: by induction on $\rho_1(\Delta)$.  If $(M,T,\Delta)$ is not hyperbolic,

1304: then apply Lemma \ref{torus reduction} to obtain $(M_1,T_1,\Delta_1)$

1305: where $M_1\to N$ is a non-elementary embedding, $\partial M_1$ is a

1306: union of tori and either $(M_1,T_1,\Delta_1)$ is a full internal Mom-$k$

1307: structure, $k<n$, or $\rho_1(\Delta_1)+2\le \rho_1(\Delta)$. In the former

1308: case the proof follows by induction. In the latter case as seen in the next

1309: paragraph we will produce a general based internal Mom-$k$ structure

1310: $(M_4,B_4,\Delta_4)$ with $k\le n-2$ and $M_4$ hyperbolic.  Lemma

1311: \ref{lower bound} then implies that $k\ge 2$ and hence $n\ge 4$.  If

1312: $k=2$, then Lemma \ref{general base 2} implies that $N$ has a full

1313: internal Mom-2 structure $(M_5,B_5,\Delta_5)$.  $M_5$ must be

1314: hyperbolic, or else the above arguement for $n=2$ will give a

1315: contradiction to Lemma \ref{lower bound}, completing the proof.

1316: 

1317: So assume that $(M_1,T_1,\Delta_1)$ is not a full internal Mom

1318: structure.  If $M_1$ is reducible, then split along a normal reducing

1319: 2-sphere, retain the non-elementarily embedded component, cap off the

1320: resulting 2-sphere boundary component with a 3-handle and cancel that

1321: 3-handle with a 2-handle.  After a sequence of such operations we

1322: obtain $(M_2,T_2,\Delta_2)$ with $\rho_1(\Delta_2)\le\rho_1(\Delta_1)$

1323: and $M_2$ irreducible.  If $M_2$ contains an embedded essential torus

1324: $R$, then split along $R$ and retain the non-elementarily embedded

1325: component.  After a sequence of such operations we obtain

1326: $(M_3,T_3,\Delta_3)$ with $\rho_1(\Delta_3)\le \rho_1(\Delta_2)$ and

1327: $M_3$ is irreducible and geometrically atoroidal.  Since $M_3$ is

1328: non-elementarily embedded in $N$, it is not a Seifert fibered space and

1329: hence is anannular and so by Thurston \cite{Th2} it is hyperbolic.

1330: After repeatedly applying Lemmas \ref{clean up} and \ref{weak} we

1331: obtain a general based internal Mom-k structure $(M_4,T_4,\Delta_4)$

1332: on $N$ where $M_4$ is hyperbolic and $\rho_1(\Delta_4)\le

1333: \rho_1(\Delta_3)$. \qed

1334: 

1335: \end{section}

1336: 

1337: %% BEGINNING OF NEW CONTENT

1338: \begin{section}{Enumeration of hyperbolic Mom-$n$'s

1339: for $2\le n \le 4$}\label{enumeration}

1340: \end{section}

1341: 

1342: % Constructing the ideal triangulation

1343: Let \((M,T,\Delta)\) be a full hyperbolic Mom-$n$, with \(2\le n\le

1344: 4\). The handle

1345: structure $\Delta$ collapses to a cellular complex $K$ in the

1346: following fashion. Each 1-handle

1347: collapses to the arc at its core, and each 2-handle collapses to the

1348: disc at its core (expanded as necessary so that it is still attached

1349: to the cores of the appropriate 1-handles). Also, $T\times I$

1350: collapses to $T\times 1$, subdivided into 0-cells, 1-cells, and

1351: 2-cells corresponding to the islands, bridges, and lakes of

1352: $(M,T,\Delta)$. (Note that if $(M,T,\Delta)$ were not full, we might

1353: have a non-simply connected lake and $K$ would not be a proper

1354: cellular complex.)

1355: 

1356: The resulting complex $K$ is a \emph{spine} for $M$ in the sense of

1357: \cite{MF}. If all of the 1-handles of $\Delta$ are of valence 3, then

1358: it is also a \emph{special spine} in the sense of \cite{MF}; however

1359: $K$ is not a special spine in general. In particular, in a special

1360: spine the link of each point is either a circle or a circle with two

1361: or three radii, but if $\Delta$ has a 1-handle of valence $n$ then the

1362: endpoints of the corresponding arc in $K$ will have links which are a

1363: circle with $n$ radii. This, however, is the only way in which $K$

1364: fails to be a special spine.

1365: 

1366: In section 2 of \cite{MF}

1367: Matveev and Fomenko describe how a manifold with a special spine can

1368: be reconstructed by gluing together truncated or ideal simplices dual

1369: to the vertices of the spine. This construction is easily generalized

1370: to our situation, and shows that $M$ can be reconstructed from $K$ by

1371: gluing together ideal polyhedra dual to the vertices of $K$. The

1372: result is an ideal cellulation of $M$ which is dual to the cellular

1373: complex $K$.

1374: 

1375: % No ``digonal pyramids''

1376: 

1377: The $3$-cells of this cellulation will be dual to the elements of

1378: \(K^0\), which consist of the endpoints of the cores of the

1379: $1$-handles of $\Delta$.  In addition, since we've assumed each

1380: $1$-handle of $\Delta$ meets at least two $2$-handles, each point

1381: \(v\in K^0\) will be the endpoint of at least two curves in \(T\times

1382: 1 \cap K^1\). Hence if \(n_v\) is the valence of $v$ in the 1-skeleton

1383: of $K$ then

1384: \(n_v\ge 3\).  If \(n_v \ge 4\) then $v$ is dual to an

1385: \((n_v-1)\)-sided pyramid: the base of the pyramid is dual to the core

1386: of a $1$-handle while the sides are dual to curves in \(T\times 1 \cap

1387: K^1\). If \(n_v = 3\) then $v$ is dual to a ``digonal pyramid'', which

1388: we eliminate from the cellulation by collapsing it to a face in the

1389: obvious fashion.

1390: % Pairing up into dipyramids

1391: Thus $K$ is dual to a cellulation of $M$ by ideal pyramids. Since the

1392: bases of these pyramids correspond to the ends of the $1$-handles of

1393: $\Delta$, we can pair them up into a collection of ideal dipyramids.

1394: 

1395: % Nature of the resulting triangulations

1396: 

1397: We can say more concerning the possible types and combinations of

1398: dipyramids. On one hand, each vertex $v$ is adjacent to \(n_v-1\)

1399: edges in \(T\times 1\cap K^1\), and each such edge has two endpoints;

1400: on the other hand, the core of each $2$-cell of $\Delta$ contributes

1401: three edges to \(T\times 1 \cap K^1\), and there are $n$ such cores in

1402: a Mom-$n$.  Therefore \(\Sigma_v (n_v-1)=6n\) in a

1403: Mom-$n$. Furthermore, \(n_v-1\) must be at least $2$ and (if it's

1404: greater than $2$) equals the number of sides of the pyramid dual to

1405: $v$. Finally the vertices $v$ occur in pairs since each one

1406: corresponds to an end of a $1$-handle, and the vertices in each pair

1407: have the same valence. Therefore for a Mom-2, there are only two

1408: possibilities: four three-sided pyramids, which glue together to form

1409: two three-sided dipyramids, or two four-sided pyramids and two

1410: ``digonal pyramids'', which (after eliminating the ``digonal

1411: pyramids'') glue together to form a single ideal

1412: octahedron. Similarly, there are only three possibilities for a Mom-3:

1413: three three-sided dipyramids, a three-sided dipyramid together with an

1414: octahedron, or a five-sided dipyramid by itself. The five

1415: possibilities for a Mom-4 are: four three-sided dipyramids, two

1416: three-sided dipyramids and an octahedron, one three-sided dipyramid

1417: and one five-sided dipyramid, two octahedra, or one six-sided

1418: dipyramid.

1419: 

1420: Thus, if $(M,T,\Delta)$ is a hyperbolic Mom-2, Mom-3, or Mom-4 then

1421: $M$ can be obtained by gluing together the faces of one of these ten

1422: sets of ideal polyhedra. Enumerating the possibilities for $M$ then

1423: becomes a matter of enumerating the ways in which the faces of these

1424: polyhedra can be glued together to form a hyperbolic $3$-manifold.

1425: 

1426: This task is simplified somewhat by the following observation: the

1427: faces of each dipyramid always have exactly one vertex which is dual

1428: to the cusp neighborhood \(T\times [0,1)\).  When gluing the polyhedra

1429: together to form $N$, all such vertices must be identified with one

1430: another and with no other vertices. Thus given any two faces, there is

1431: only one possible orientation-preserving way that those two faces

1432: could be glued together.

1433: 

1434: %  Eliminating symmetric conjugates

1435: 

1436: Hence it is sufficient to enumerate the number of ways in which the

1437: faces of one of the ten sets of polyhedra can be identified in

1438: pairs. Although it is almost trivial to program a computer to do this,

1439: care must be taken as the number of possibilities is a factorial

1440: function of the number of faces, and a naive approach can rapidly

1441: exhaust a computer's memory. To reduce the demands on the computer, a

1442: refinement to the naive approach was employed. First, for each

1443: possible set of polyhedra a symmetry group was computed. Each

1444: dipyramid has dihedral symmetry, while if a given set of polyhedra

1445: contains two dipyramids with the same number of sides then they can be

1446: exchanged to provide an additional symmetry. Secondly, an ordering was

1447: chosen for the set of all possible pairings of faces, namely the

1448: lexicographic ordering of the pairings when represented as

1449: permutations. Our computer program considered the set of pairings in

1450: order, and any pairing was immediately rejected if it was conjugate to

1451: a previous pairing via an element of the symmetry group. This

1452: considerably reduced the running time of the program.

1453: 

1454: %  Computing \chi(L), L=link of a vertex

1455: %  Only one ``polar vertex''

1456: 

1457: The next step in the process is to eliminate pairings which result in

1458: obviously

1459: non-hyperbolic manifolds.  While the program SnapPea can in principle

1460: handle this, for reasons of speed our program checked one necessary

1461: criterion itself: whether the link of every ideal vertex was

1462: Euclidean. Computing the Euler characteristic of the link of each

1463: ideal vertex in the cellular complex resulting from a pairing was easy

1464: to do and eliminated many cases from consideration. Our program also

1465: eliminated any pairing in which the vertices supposedly dual to the

1466: original cusp neighborhood or solid torus in fact glued together to

1467: form two or more ideal vertices.

1468: 

1469: % Intermediate results

1470: The above considerations resulted in a list of gluing descriptions of

1471: 4,231 manifolds which might be hyperbolic Mom-2's or Mom-3's.

1472: At this point, SnapPea was employed to try and compute

1473: hyperbolic metrics for each of these manifolds, and to find further

1474: hyperbolic symmetries among the manifolds which admitted such

1475: metrics. SnapPea claimed to find hyperbolic metrics in 164 cases. In

1476: three of those cases SnapPea had experienced an obvious floating-point

1477: error and ``found'' a hyperbolic metric with an absurdly low

1478: volume. In those three cases the programs Regina and GAP (which,

1479: unlike SnapPea, do not rely on floating-point arithmetic) were used to

1480: calculate the fundamental groups of the corresponding manifolds. In

1481: all three cases the fundamental group was isomorphic to the group

1482: \(\langle a,b|[a,b^n]\rangle\) where \(n=3\) or $5$, which has a

1483: non-trivial center. Therefore, these three manifolds could not

1484: possibly be hyperbolic and were rejected. That left 161

1485: cases, which SnapPea identified as belonging to a total of 21 isometry

1486: classes of hyperbolic manifolds.

1487: 

1488: Some comments about rigor are in order here. Since SnapPea relies on

1489: floating-point arithmetic, some of its results are unavoidably

1490: inexact. In particular, there is no guarantee that SnapPea will find

1491: a hyperbolic metric on a manifold even if one exists, or that SnapPea

1492: will correctly discern the absence of a hyperbolic metric in cases

1493: where it doesn't

1494: exist. In practice it is our experience that if one is careful to

1495: allow SnapPea to simplify a triangulation before attempting to find a

1496: metric, then if a metric exists SnapPea will either find it or fail to

1497: make a determination, while if a metric doesn't exist SnapPea will

1498: either correctly say so or on rare occasions ``find'' a metric with

1499: absurdly low volume due to floating-point error. Still, from a

1500: standpoint of rigor this is problematic. Fortunately there is at

1501: least one task which SnapPea does perform exactly, and that is

1502: finding isometries between two different cusped manifolds: SnapPea will only

1503: report that an isometry exists if it finds identical triangulations of

1504: the two manifolds. (See \cite{We}; in particular see the comments in

1505: the source code file \emph{isometry.c}.) This is a combinatorial operation,

1506: not a floating-point one, and hence we are confident that SnapPea

1507: performs this operation rigorously.

1508: 

1509: Those familiar with SnapPea's source code may object that SnapPea

1510: re-triangu\-lates each manifold before determining if an isometry

1511: exists, and that SnapPea uses floating-point information to choose the

1512: re-triangu\-lation. To this objection we would reply that while

1513: floating-point information is used to \emph{choose} the

1514: re-triangulation, the actual re-triangu\-lating is still a combinatorial

1515: operation, i.e. it uses integer arithmetic. The new triangulation is

1516: guaranteed to have the same topological type (see the comments in

1517: \emph{canonize\_part\_1.c} from \cite{We}), and hence the

1518: possibility of floating-point error does not invalidate the result

1519: when SnapPea reports that it has found an isometry.

1520: 

1521: Thus while we are trusting SnapPea when it says that the 161 manifolds

1522: mentioned above are all in the isometry class of one of 21 manifolds

1523: from the SnapPea census, we are confident that we are not sacrificing

1524: rigor in so doing. Furthermore, the census manifolds were

1525: recently confirmed to be hyperbolic by Harriet Moser in \cite{Mos},

1526: establishing that we have found 21 different hyperbolic Mom-2's and

1527: Mom-3's.

1528: 

1529: Unfortunately, we still can't trust SnapPea when it fails to

1530: find an hyperbolic metric for a given manifold,

1531: as that result is not guaranteed to be

1532: rigorous. This means that there are 4,067 manifolds from the above

1533: list of 4,231 which may still be hyperbolic despite SnapPea evidence

1534: to the contrary. These manifolds were analyzed separately in the same

1535: way as the three manifolds for which SnapPea claimed to have found an

1536: absurdly low hyperbolic volume. Namely,

1537: we used Regina and GAP as before to compute the fundamental

1538: groups of the manifolds in question, and then examined the list of groups

1539: to see if any of them might be the fundamental group of a hyperbolic

1540: manifold. The vast majority of the groups on the list either had a

1541: non-trivial center, or else had two rank-2 Abelian subgroups

1542: which intersected in a rank-1 Abelian subgroup (also impossible in the

1543: fundamental group of a hyperbolic 3-manifold). Some of the groups required

1544: further analysis but were still eventually rejected; for example, many

1545: groups had an index-two subgroup with one of the above properties even

1546: when it was not clear that the whole group had such properties.

1547: 

1548: In the end the hand analysis did not reveal any new hyperbolic 3-manifolds

1549: in the list of gluing descriptions. This completes the proof of the

1550: following:

1551: 

1552: \smallskip

1553: \begin{theorem} \label{list of mom-2's and mom-3's}

1554: There are 3 hyperbolic manifolds $M$ such that \((M,T,\Delta)\) is a

1555: Mom-2 for some $T$ and $\Delta$: the manifolds known in SnapPea's

1556: notation as m125, m129, and m203. There are 18 additional hyperbolic manifolds

1557: $M$ such that \((M,T,\Delta)\) is a Mom-3 structure for some $T$ and

1558: $\Delta$: the manifolds known in SnapPea's notation as m202, m292,

1559: m295, m328, m329, m359, m366, m367, m391, m412, s596, s647, s774,

1560: s776, s780, s785, s898, and s959.

1561: \end{theorem}

1562: % Some details about this list: how many (4)-types, (33)-types, etc?

1563: Some comments about this list are in order. The manifold m129, better

1564: known as the complement of the Whitehead link, is the only manifold on

1565: this list which is obtained by gluing together the faces of an ideal

1566: octahedron. Also, all but one of these manifolds have two cusps. The

1567: exception is the three-cusped s776, which is the complement in \(S^3\)

1568: of a three-element chain of circles (the link \(6_1^3\) in Rolfsen's

1569: notation).

1570: 

1571: Enumerating hyperbolic Mom-4's was more difficult: merely

1572: enumerating the possible gluing descriptions resulted in a list of 1,033,610

1573: possibilities (compared to 4,231 possibilities in the previous case).

1574: From this list, SnapPea identified 138 different

1575: hyperbolic manifolds. In another 493 cases, SnapPea was either unable

1576: to make a determination or else experienced an obvious floating-point

1577: error. In each such case, the fundamental group of the corresponding

1578: manifold was again computed by Regina and GAP, and in each case the

1579: fundamental group was isomorphic to \(\langle a,b |[a^n,b]\rangle\)

1580: for some \(n\), or else had two rank-2 Abelian subgroups which

1581: intersected in a rank-1 Abelian subgroup.  Therefore these exceptional

1582: cases do not correspond to hyperbolic manifolds. Note that all of the

1583: Mom-2's and Mom-3's appear in the Mom-4 list; the same manifold can

1584: admit multiple handle structures.

1585: 

1586: Based on the above result, we propose the following:

1587: 

1588: \begin{conjecture}\label{number of mom-4's}

1589: There are 138 hyperbolic manifolds $M$ such that \((M,T,\Delta)\) is a

1590: Mom-2, Mom-3, or Mom-4 for some $T$ and $\Delta$. Of these, 117 are

1591: \emph{strict} Mom-4's, i.e. Mom-4's which are not Mom-2's or Mom-3's.

1592: \end{conjecture}

1593: 

1594: Of the 117 strict Mom-4's, SnapPea was successfully used to identify

1595: 83 of them as manifolds from the SnapPea census. Those manifolds appear

1596: in

1597: \begin{figure}

1598: \begin{center}

1599: \begin{tabular}{|c|c|c|c|c|c|c|}

1600: \hline\hline

1601: m357 & s579 & s883 & v2124 & v2943 & v3292 & v3450 \\

1602: m388 & s602 & s887 & v2208 & v2945 & v3294 & v3456 \\

1603: s441 & s621 & s895 & v2531 & v3039 & v3376 & v3468 \\

1604: s443 & s622 & s906 & v2533 & v3108 & v3379 & v3497 \\

1605: s503 & s638 & s910 & v2644 & v3127 & v3380 & v3501 \\

1606: s506 & s661 & s913 & v2648 & v3140 & v3383 & v3506 \\

1607: s549 & s782 & s914 & v2652 & v3211 & v3384 & v3507 \\

1608: s568 & s831 & s930 & v2731 & v3222 & v3385 & v3518 \\

1609: s569 & s843 & s937 & v2732 & v3223 & v3393 & v3527 \\

1610: s576 & s859 & s940 & v2788 & v3224 & v3396 & v3544 \\

1611: s577 & s864 & s941 & v2892 & v3225 & v3426 & v3546 \\

1612: s578 & s880 & s948 & v2942 & v3227 & v3429 &  \\

1613: \hline\hline

1614: \end{tabular}

1615: \end{center}

1616: \caption{Conjectured list of SnapPea manifolds which are strict Mom-4's.}

1617: \end{figure}

1618: Figure 1. SnapPea was not able to identifiy the remaining 34

1619: manifolds, and in fact 33 of those manifolds have volumes which do not

1620: appear anywhere in the SnapPea census, presumably because the Matveev

1621: complexity of the corresponding manifolds is greater than 7 (see

1622: \cite{MF}). The remaining manifold has the same volume and homology as

1623: the census manifold v3527; it is conceivable that SnapPea was simply

1624: unable to find a corresponding isometry.

1625: 

1626: The unidentified manifolds are listed in

1627: \begin{figure}

1628: \begin{center}

1629: \begin{tabular}{|c|}

1630: \hline\hline

1631: $(3,3,4\ ;\ 3,6,8,0,13,19,1,15,2,17,14,18,16,4,10,7,12,9,11,5)$ \\

1632: $(3,3,4\ ;\ 3,6,11,0,10,9,1,15,14,5,4,2,16,18,8,7,12,19,13,17)$ \\

1633: $(3,3,4\ ;\ 3,6,11,0,9,18,1,19,13,4,15,2,16,8,17,10,12,14,5,7)$ \\

1634: $(3,3,3,3\ ;\ 3,6,9,0,13,19,1,22,14,2,17,12,11,4,8,23,18,10,16,5,21,20,7,15)$ \\

1635: $(3,3,3,3\ ;\ 3,4,6,0,1,19,2,9,13,7,14,15,23,8,10,11,20,21,22,5,16,17,18,12)$ \\

1636: $(3,3,3,3\ ;\ 3,6,12,0,17,9,1,11,18,5,23,7,2,20,15,14,21,4,8,22,13,16,19,10)$ \\

1637: $(3,3,4\ ;\ 3,4,6,0,1,18,2,11,16,10,9,7,15,19,17,12,8,14,5,13)$ \\

1638: $(3,3,4\ ;\ 3,4,6,0,1,18,2,14,12,13,19,17,8,9,7,16,15,11,5,10)$ \\

1639: $(3,3,4\ ;\ 3,4,6,0,1,18,2,14,16,13,19,17,15,9,7,12,8,11,5,10)$ \\

1640: $(3,3,4\ ;\ 3,6,10,0,13,9,1,16,19,5,2,17,14,4,12,18,7,11,15,8)$ \\

1641: $(3,3,4\ ;\ 3,6,10,0,8,13,1,16,4,18,2,17,15,5,19,12,7,11,9,14)$ \\

1642: $(3,3,4\ ;\ 3,6,10,0,8,13,1,18,4,16,2,19,15,5,17,12,9,14,7,11)$ \\

1643: $(4,4\ ;\ 15,10,13,8,11,14,9,12,3,6,1,4,7,2,5,0)$ \\

1644: $(4,4\ ;\ 15,14,5,6,9,2,3,10,13,4,7,12,11,8,1,0)$ \\

1645: $(4,4\ ;\ 15,14,9,8,11,10,13,12,3,2,5,4,7,6,1,0)$ \\

1646: $(4,4\ ;\ 15,4,13,6,1,8,3,10,5,14,7,12,11,2,9,0)$ \\

1647: $(4,4\ ;\ 15,14,4,5,2,3,11,10,12,13,7,6,8,9,1,0)$ \\

1648: $(4,4\ ;\ 15,14,6,7,11,10,2,3,12,13,5,4,8,9,1,0)$ \\

1649: $(4,4\ ;\ 15,7,13,10,9,14,11,1,12,4,3,6,8,2,5,0)$ \\

1650: $(4,4\ ;\ 15,5,13,7,9,1,11,3,14,4,12,6,10,2,8,0)$ \\

1651: $(3,3,4\ ;\ 3,6,10,0,15,17,1,18,14,16,2,19,13,12,8,4,9,5,7,11)$ \\

1652: $(3,3,4\ ;\ 3,6,11,0,8,19,1,15,4,17,14,2,16,18,10,7,12,9,13,5)$ \\

1653: $(3,3,4\ ;\ 6,7,10,8,13,17,0,1,3,15,2,19,16,4,18,9,12,5,14,11)$ \\

1654: $(3,3,3,3\ ;\ 3,4,6,0,1,9,2,15,17,5,13,18,19,10,23,7,22,8,11,12,21,20,16,14)$ \\

1655: $(3,3,4\ ;\ 3,6,10,0,8,14,1,16,4,18,2,17,13,12,5,19,7,11,9,15)$ \\

1656: $(3,3,4\ ;\ 3,6,10,0,8,14,1,18,4,16,2,19,13,12,5,17,9,15,7,11)$ \\

1657: $(3,3,4\ ;\ 3,6,7,0,16,19,1,2,10,12,8,14,9,18,11,17,4,15,13,5)$ \\

1658: $(3,3,4\ ;\ 6,10,19,8,13,17,0,12,3,15,1,16,7,4,18,9,11,5,14,2)$ \\

1659: $(3,3,4\ ;\ 6,7,10,8,9,13,0,1,3,4,2,19,16,5,17,18,12,14,15,11)$ \\

1660: $(3,3,4\ ;\ 3,6,10,0,8,18,1,14,4,16,2,19,13,12,7,17,9,15,5,11)$ \\

1661: $(3,3,3,3\ ;\ 3,6,12,0,9,16,1,18,23,4,20,22,2,19,15,14,5,21,7,13,10,17,11,8)$ \\

1662: $(3,3,3,3\ ;\ 3,4,6,0,1,9,2,15,17,5,14,13,19,11,10,7,23,8,22,12,21,20,18,16)$ \\

1663: $(3,3,3,3\ ;\ 3,6,12,0,9,16,1,10,18,4,7,22,2,20,15,14,5,21,8,23,13,17,11,19)$ \\

1664: $(3,3,3,3\ ;\ 3,6,12,0,9,16,1,18,11,4,23,8,2,19,15,14,5,21,7,13,22,17,20,10)$ \\

1665: \hline\hline

1666: \end{tabular}

1667: \end{center}

1668: \caption{Conjectured Mom-4's not identified by SnapPea.}

1669: \end{figure}

1670: Figure 2. The notation used can be interpreted as follows: the

1671: numbers before the semi-colon describe the type of ideal polyhedra

1672: used to construct the manifold. For example, the first entry in the

1673: figure has the numbers ``$3,3,4$'' to the left of the semi-colon; each

1674: ``$3$'' indicates an ideal triangular dipyramid, while each ``$4$''

1675: indicates an ideal square dipyramid (i.e. an ideal

1676: octahedron). Each ideal dipyramid has two ``polar'' vertices and

1677: either three or four ``equatorial vertices''. Number the faces of all

1678: the polyhedra sequentially in such a way that the faces ``north'' of

1679: each equator are numbered before the faces ``south'' of each

1680: equator. For example, in the first entry the first triangular

1681: dipyramid has faces 0, 1, and 2 next to one polar vertex, and faces

1682: 10, 11, and 12 next to the other polar vertex. The next triangular

1683: dipyramid has faces 3, 4, and 5 as well as faces 13, 14, 15, and the

1684: square dipyramid has faces 6 through 9 and 16 through 19. (This

1685: somewhat unintuitive numbering scheme was chosen for convenience when

1686: writing the computer software for this part of the paper.) Then the

1687: numbers to the right of the semi-colon form a permutation which

1688: describes how to glue together the faces of the ideal polyhedra.

1689: For example, in the first entry the string of numbers which begins

1690: with ``3, 6, 8, 0, \ldots'' imply that face 0 is glued to face 3, face

1691: 1 is glued to face 6, and so on. Since

1692: we are requiring ``polar'' vertices to be identified solely with other

1693: ``polar'' vertices, no other information is needed to reconstruct the

1694: polyhedral gluing.

1695: 

1696: One additional point of information:

1697: all but eight of the manifolds in the list satisfy

1698: \(|\partial M|=2\); seven satisfy \(|\partial M|=3\) and one satisfies

1699: \(|\partial M|=4\). Thanks to the timely assistance of Morwen

1700: Thistlethwaite, the authors were able to positively identify all eight

1701: of these manifolds:

1702: 

1703: \begin{figure}

1704: \begin{center}

1705: \includegraphics{link_pics.ps}

1706: \end{center}

1707: \caption{Eight links whose complements are Mom-4's with 3 or more

1708: cusps.}

1709: \end{figure}

1710: 

1711: \begin{conjecture}

1712: There are 8 hyperbolic manifolds $M$ such that \(M,T,\Delta\) is a

1713: Mom-$n$ for some $2\le n\le 4$ and \(|\partial M|>2\). All eight

1714: manifolds are complements of links in $S^3$: the links $6^3_1$,

1715: $6^3_2$ (the Borromean rings), $8^3_1$, $8^3_9$, $8^4_2$, and the

1716: links with Gauss codes jcccddEGHiJBFCa, jcbecceaHbIJDGF, and

1717: mccdfiEhAjKLcmdbFG.

1718: \end{conjecture}

1719: 

1720: At the time of writing we are still searching for an efficient way to

1721: verify SnapPea's computations in the Mom-$4$ case; clearly, examining

1722: over a million fundamental groups by hand is not a practical solution. Until

1723: a better way is found, our enumeration results in the Mom-$4$ case should

1724: properly be considered speculative.

1725: 

1726: \begin{section} {Hyperbolic Mom-$n$'s in  hyperbolic 3-manifolds are internal

1727: Mom-$n$ structures for $n\le 4$}\end{section}

1728: 

1729: Let $R$ be a convolotube in the interior of a compact hyperbolic 

1730: 3-manifold $N$ and let $V$ be the cube

1731: with knotted hole bounded by $R$.  By drilling out solid tori from 

1732: $N-\cirV$, we can create a

1733: manifold

1734: $M $ which is non-elementarily embedded in $N$ and whose boundary 

1735: contains a convolutube. We call such an embedding \emph{knotted}.

1736: The goal

1737: of this chapter is to show that if $n\le 4$, any  embedding

1738: of a Mom-n manifold $(M,T)$ into a compact hyperbolic manifold 

1739: $(N,T)$ is unknotted.

1740: 

1741: \smallskip

1742: \begin{definition} Let $M$ be a compact 3-manifold and T a possibly 

1743: empty union of components of

1744: $\partial M$.  We say that

1745: $(M,T)$ is

1746: \emph{hereditarily unknotted}, if \emph{every} non-elementary

1747: embedding into a  compact hyperbolic 3-manifold

1748: $N$, taking $T$ to components of $\partial N$,  has the property that 

1749: each component of  $\partial M$

1750:   is either boundary parallel or bounds a solid torus.

1751: \end{definition}

1752: 

1753: \begin{remark}  If $(M,T)$ is herditarily unknotted and

1754: $M_1$ is obtained by filling a component of

1755: $\partial M - T$, then $(M_1,T)$ is

1756: hereditarily unknotted.\end{remark}

1757: 

1758: \begin{lemma}  If $(M,T)$ is a hereditarily unknotted Mom-$n$ manifold

1759: non-elementarily embedded in the hyperbolic 3-manifold $N$ such that $T$ 

1760: bounds a tubular neighborhood

1761: of a geodesic, then $(M,T)$ is an internal Mom-$n$ structure.\end{lemma}

1762: 

1763: \begin{proof}  Let $V$ be the solid torus bounded by $T$.

1764: By Lemma \ref{geodesic complement}, if $N_1=N-\cirV$

1765: with cusp neighborhoods deleted, then $N_1$ is compact

1766: hyperbolic.  Therefore

1767: $(M,T)\subset (N_1,T)$ is a non-elementary embedding and

1768: hence any  component of $\partial M-T$ either bounds a solid

1769: torus or is boundary parallel in $N_1$.  Therefore similar

1770: properties hold in $N$ and hence $(M,T)$ is an internal Mom-$n$

1771: structure on

1772: $N$.\end{proof}

1773: 

1774: \begin{remark} The condition that $T$ bounds a neighborhood of a 

1775: geodesic is essential. \end{remark}

1776: 

1777: \begin{lemma}  \label{two}Let $M$ be a compact hyperbolic 3-manifold 

1778: with $T$ a union of components of

1779: $\partial M$.  If $\partial M-T$ is connected,  $(M,T)$ is hereditarily

1780: unknotted.\end{lemma}

1781: 

1782: \begin{proof}  If under a non-elementary embedding $(M,T)\to (N,T)$,\ 

1783: \ $\partial M-T$ was a

1784: convolutube, then $M$ would be reducible.\end{proof}

1785: 

1786: 

1787: 

1788: The following result establishes criteria for showing that $(M,T)$ is 

1789: hereditarily unknotted.

1790: 

1791: \begin{lemma}  \label{H criteria} Let $M$ be a compact  hyperbolic

1792: 3-manifold with $V_1$, \ldots, $V_n$ components of $\partial M$ and $T$ 

1793: a nonempty union of some other

1794: components.  If any of the following hold, there exists no 

1795: non-elementary embedding $(M,T)\to

1796: (N,T)$ such that $N$ is compact hyperbolic and $\{V_1,\ldots, V_n\}$ is exactly

1797: the set of convolutubes of $\partial M\subset N$.

1798: 

1799: i) The manifold obtained by some filling of $M$ along $V_1$, \ldots,

1800: $V_n$ is a 3-manifold without any hyperbolic part.  (That is, after

1801: applying sphere and torus decompositions there are no hyperbolic

1802: components.)

1803: 

1804: ii) After some filling of $M$ along $V_1$, \ldots, $V_n$, the surface $T$

1805: is compressible.

1806: 

1807: iii) For every filling on a non-empty set of components of $\partial

1808: M-T\cup V_1\cup \cdots\cup V_n$, either $ V_1\cup\cdots\cup V_n$ is 

1809: incompressible or the

1810: filled manifold has no hyperbolic part.

1811: \end{lemma}

1812: 

1813: \begin{proof}   Suppose that $(M,T)$ embeds in $(N,T)$, where

1814: among the components of $\partial M$, 

1815: $V_1,\cdots, V_n$ are the set of convolutubes and $W_1,\cdots, W_m$

1816: are the tubes.  Let $W_i^*$ denote the solid torus bounded by $W_i$ 

1817: and $V_i^*$ denote the cube with

1818: knotted hole bounded by $V_i$.  Let $B_1,\cdots,B_n$ be pairwise disjoint 3-balls 

1819: in $N$ such that for each

1820: $i$, $V_i\subset B_i$.

1821: 

1822: i) Let $\hat M$ be a manifold obtained by filling the $V_i$'s.  Let 

1823: $\hat N$ be obtained by deleting the $V_i^*$'s and doing the 

1824: corresponding fillings along the $V_i$'s.  Therefore $\hat N$ is 

1825: obtained from $\hat M$ by Dehn filling and $\hat N$ is a connected 

1826: sum of $N$ with $S^2\times S^1$'s and/or lens spaces and/or $S^3$'s. 

1827: This implies that $\hat M$ has a hyperbolic part.

1828: 

1829: ii) If $T$ is compressible  in $\hat M$ it is compressible in $\hat 

1830: N$ and hence in $N$, which is a contradiction.

1831: 

1832: iii) First observe that $V_i$ compresses in the manifold $M$' 

1833: obtained by filling $M$  where each

1834: $W_i$ is filled with $W_i^*$.  Topologically, $M^\prime$ is 

1835: homeomorphic to $N$ with $n$ open

1836: unknotted and  unlinked solid tori removed and so has a hyperbolic part.

1837: \end{proof}

1838: 

1839: 

1840: 

1841: \begin{theorem}

1842: If the Mom-$n$ manifolds for $n\le 4$ with three or more boundary

1843: components are exactly those listed in Figure 3 (i.e. if Conjecture

1844: \ref{list of mom-2's and mom-3's} is true), then any hyperbolic Mom-$n$

1845: manifold $(M,T)$ with $n\le 4$ is hereditarily unknotted.

1846: \end{theorem}

1847: 

1848: \begin{proof}  By Lemma \ref{two} it suffices to consider the case 

1849: where $M$ is one of the eight Mom-4 manifolds with at least three

1850: boundary components listed in Figure 3. If $M$ is any of the first six

1851: manifolds and $T$ is any component of $\partial M$, then $(M,T)$ is

1852: hereditarily unknotted by criterion (i) of Lemma \ref{H criteria}.

1853: For manifolds 7 and 8, depending on which boundary component is used

1854: for $T$, applications of (i) and (iii) imply that they are

1855: hereditarily unknotted.

1856: \end{proof}

1857: 

1858: \begin{section} {Examples of Mom-$n$ structures}\end{section}

1859: 

1860: % Examples

1861: 

1862: In this section we give some representative examples of hyperbolic

1863: manifolds $N$ which contain an internal Mom-2 or Mom-3 structure

1864: \((M,T,\Delta)\).  Our goal in this section is to give the

1865: reader an intuitive feel for how these particular cell complexes arise

1866: inside hyperbolic manifolds. All of the manifolds in this section

1867: involve manifolds $N$ with torus boundary, with the base torus of the

1868: Mom-structure \((M,T,\Delta)\) being \(\partial N\).  To obtain Mom-$n$

1869: structures on closed manifolds, note that if \(T=\partial N\) then a

1870: Mom-$n$ structure \((M,T,\Delta)\) on $N$ passes to a Mom-$n$ structure on

1871: any manifold obtained by filling \(\partial N\).

1872: 

1873: % The figure-8 knot complement

1874: 

1875: \smallskip

1876: \begin{example}

1877: The first example is the figure-$8$ knot complement. We construct a

1878: Mom-2 \((M,T,\Delta)\) inside this manifold as follows. The torus $T$

1879: is just the boundary of the manifold. The $1$-handles \(\lambda_1\)

1880: and \(\lambda_2\) span the two tangles which make up the standard

1881: diagram of this knot, as seen in

1882: \begin{figure}[tb]

1883: \begin{center}

1884: \includegraphics{figure8.ps}

1885: \end{center}

1886: \caption{The figure-8 knot complement equipped with a Mom-2.}

1887: \end{figure}

1888: figure 4. Finally the $2$-handles \(\sigma_1\) and \(\sigma_2\) are

1889: symmetrically placed as shown in the diagram. Note that, as required,

1890: each $2$-handle meets three $1$ handles counting

1891: multiplicity. Specifically, each $2$-handle meets \(\lambda_1\) twice

1892: and \(\lambda_2\) once. Also, one can see from the diagram that the

1893: complement of $T\cup \{\lambda_i\}\cup \{\sigma_j\}$ consists of a

1894: solid torus, and that the solid torus retracts onto a

1895: homotopically non-trivial simple closed curve (which is a geodesic in

1896: $N$).  Thus this is a valid hyperbolic Mom-2 structure on $N$.

1897: 

1898: Moreover, we can quickly determine the nature of the ideal

1899: triangulation of $M$ described in the Section \ref{enumeration}. The

1900: ends of \(\lambda_1\) are each dual to a four-sided pyramid in this

1901: triangulation, and the two endpoints of \(\lambda_2\) are each dual to

1902: a ``digonal pyramid'', so that each get eliminated. Thus the figure-8

1903: knot complement possesses a Mom-2 structure \((M,T,\Delta)\) where $M$

1904: is a two-cusped hyperbolic manifold which is in turn obtained by

1905: gluing together the faces of an ideal octahedron. By the comments

1906: after Theorem \ref{list of mom-2's and mom-3's}, $M$ must be the

1907: complement of the Whitehead link.  And indeed, it is easy to verify

1908: that if one drills out the core of the solid torus in the complement

1909: of $M$ one obtains a manifold homeomorphic to the complement of the

1910: Whitehead link.\end{example}

1911: 

1912: % A triangulation example: m016

1913: 

1914: \begin{example}

1915: Next we will let $N$ be the manifold known as m003 in the SnapPea

1916: census. This manifold has first homology group \(\BZ + \BZ/5\), and

1917: hence is not a knot complement; instead, we will present this manifold

1918: as the union of two regular ideal hyperbolic tetrahedra; see

1919: \begin{figure}[tb]

1920: \begin{center}

1921: \includegraphics{m003.ps}

1922: \end{center}

1923: \caption{The two ideal tetrahedra making up the manifold m003.}

1924: \end{figure}

1925: figure 5. Note that in the diagram each face is glued to the

1926: corresponding face on the other tetrahedron, in such a way that the

1927: edges match up into two equivalence classes as shown. To make $N$ a

1928: compact manifold with torus boundary, assume the ideal tetrahedra are

1929: truncated.  Now suppose that we construct $(M,T,\Delta)$ in this case

1930: as follows.  For the $1$-handles, we use neighborhoods of the two

1931: edges shown in the diagram, truncated by the torus \(T=\partial N\).

1932: And for the $2$-handles, we use neighborhoods of the two truncated

1933: triangles which are formed by gluing together the faces on the front

1934: of each tetrahedron in the diagram. It is a simple exercise to confirm

1935: that the complement of the resulting embedded manifold $M$ consists of

1936: a solid torus, and that the solid torus retracts onto a simple closed

1937: geodesic curve, and that therefore this manifold possesses a valid

1938: hyperbolic Mom-2 structure. Each of the $1$-handles in this Mom-2

1939: meets three of the $2$-handles, counting multiplicity; therefore we

1940: can conclude that m003 contains a Mom-2 \((M,T,\Delta)\) where $M$ is

1941: obtained by gluing together two ideal three-sided dipyramids. From

1942: Theorem \ref{list of mom-2's and mom-3's} and the comments

1943: following it we know this must be either m125 or m203. Further

1944: investigation with SnapPea shows that it must in fact be m125.

1945: 

1946: It is instructive to get another view of this Mom-2 by constructing a

1947: cusp diagram for this manifold. Specifically, consider the

1948: triangulation induced on $T$ by the given ideal triangulation of

1949: m003. The two ideal tetrahedra in m003 will appear as eight triangles,

1950: the four ideal triangles will appear as twelve edges, and the two

1951: edges will appear as four vertices. The resulting cusp diagram is

1952: shown in

1953: \begin{figure}[tb]

1954: \begin{center}

1955: \includegraphics{m003_cusp.ps}

1956: \end{center}

1957: \caption{The cusp diagram for m003, with the components of the Mom-2

1958: highlighted.}

1959: \end{figure}

1960: figure 6; keep in mind this is a diagram of a torus, so the edges of

1961: the parallelogram are identified with one another. (The labels inside

1962: each triangle indicate which of the ideal simplices contributes that

1963: triangle to the cusp diagram.)  The highlighted edges in the cusp

1964: diagram are those that correspond to the $2$-handles of the handle

1965: structure $\Delta$; in other words, they along with the four vertices

1966: of the diagram comprise \(\Delta^1 \cap T\).\end{example}

1967: 

1968: \begin{example}

1969: As another example in this vein, consider the manifold $N$=m017. This

1970: manifold has first homology group \(\BZ+\BZ/7\), so again it is not a

1971: knot complement in \(S^3\), but for brevity's sake we only present a

1972: cusp diagram here. In

1973: \begin{figure}[tb]

1974: \begin{center}

1975: \includegraphics{m017_cusp.ps}

1976: \end{center}

1977: \caption{The cusp diagram for m017, with the components of a Mom-2

1978: highlighted.}

1979: \end{figure}

1980: figure 7, the corners of the three ideal hyperbolic tetrahedra which

1981: make up m017 can be seen. And again, the highlighted edges in the

1982: cusp diagram correspond to two faces of those tetrahedra which provide

1983: the $2$-handles for an internal Mom-2 in this manifold. Note that we can

1984: determine from the cusp diagram alone that the $1$-handles of this

1985: Mom-2 meet four and two $2$-handles respectively, counting

1986: multiplicity, and that therefore in the resulting Mom-2 structure

1987: \((M,T,\Delta)\) the manifold $M$ is obtained by gluing together an

1988: ideal octahedron. As before, this implies that $M$ must be

1989: homeomorphic to the complement of the Whitehead link. Some further

1990: work with SnapPea confirms this: m017 is obtained by (-7,2) Dehn

1991: surgery on either component of the link.\end{example}

1992: 

1993: \begin{example}Finally, we include the motivating example for this paper.

1994: \begin{figure}[tb]

1995: \begin{center}

1996: \includegraphics{m011_cusp.ps}

1997: \caption{SnapPea's cusp diagram for m011, with the components of a Mom-2

1998: highlighted.}

1999: \end{center}

2000: \end{figure}

2001: Figure 8 shows the maximal cusp diagram of the 1-cusped manifold m011

2002: as provided by Weeks' SnapPea program. Unlike the previous cusp diagrams

2003: in this section, it also shows all the horoballs at

2004: hyperbolic distance at most 0.51 from the maximal horoball at

2005: infinity. The paralellogram shows a fundamental domain for the

2006: $\BZ\oplus\BZ$ action.  Note that the ideal triangulation presented in

2007: this diagram is dual to the Ford decomposition of the manifold.

2008: In particular the 1-simplices of the triangulation are geodesics

2009: orthogonal to pairs of horoballs; these 1-simplices appear either as

2010: edges in the figure joining the endpoints of the simplex in $\Sinfty$, or

2011: as the vertical geodesics passing from the center of each horoball to the

2012: horoball at infinity.

2013: 

2014: Let $H_\infty$ denote the horoball at infinity. There are six

2015: horoballs in the diagram up to the $\BZ\oplus\BZ$-action, labelled

2016: $1-$, $1+$, $2-$, $2+$, $3-$, and $3+$. This notation

2017: means that if $\gamma\in \pi_1(m011)$ takes horoball $n\pm$ to

2018: $H_\infty$, the horoball at infinity, then $H\infty$ is transformed to

2019: one labelled $n\mp$.  The geodesic from $n-$ to $H_\infty$ is oriented

2020: to point into $H_\infty$ and hence the geodesic from $n+$ to

2021: $H_\infty$ is oriented out of $H_\infty$.  These orientations induce,

2022: via the $\pi_1(m011)$-action, the indicated orientations on the edges

2023: of the diagram.  We explain, by example, the meaning of the edge

2024: labels.  The edge $3$ from $2+$ to $2-$ corresponds to a geodesic

2025: $\sigma$ with the property that when $2+$ is transformed to

2026: $H_\infty$, then $2-$ is transformed to $3+$ and $\sigma$ is

2027: transformed to the vertical geodesic oriented from $H_\infty$ to $3+$.

2028: (Had the edge been oriented oppositely, then $2-$ would have been

2029: transformed to $3-$.)  SnapPea did not provide the orientation

2030: information, however such information can be derived from the SnapPea

2031: data.

2032: 

2033: By staring at this diagram we can see how m011 contains a Mom-$2$.  Let

2034: $V_0$ be the maximal horotorus neighborhood of the cusp, slightly

2035: shrunken.  By expanding $V_0$, the expanded $V_0$ touches the

2036: (expanded) horoballs labeled $1$, thereby creating a 1-handle denoted

2037: $E_1$.  Let $V_1$ denote this expanded $V_0$.  Further expansion

2038: creates $V_2$ which is topologically $V_1$ together with another

2039: 1-handle $E_2$, this 1-handle occuring between horoball 2 and

2040: $H_\infty$.  The edge labelled 1 between horoballs $2-$ and $2+$

2041: corresponds to a valence three 2-handle which goes over $E_1$ once and

2042: $E_2$ twice.  Similary the edge labelled $2$ between horoballs $1-$

2043: and $1+$ gives rise to a valence three 2-handle going twice over $E_1$

2044: and once over $E_2$.  The parallelogram of Figure 8 can also be viewed

2045: as $\partial V_0$, with the centers of $1-,1+, 2-, 2+$ as the

2046: attaching sites of the 1-handles and the thick black lines

2047: corresponding to where the 2-handles cross over $\partial V_0$.

2048: \end{example}

2049: 

2050: \vfill \pagebreak

2051: 

2052: \begin{thebibliography}{[GMM3]}

2053: 

2054: \bibitem[AST]{AST} I. Agol \& P. Storm \& W. Thurston, Lower bounds on

2055:   volumes of hyperbolic Haken 3-manifold (with an appendix by Nathan

2056:   Dunfield), preprint, math.DG/0506338.

2057: 

2058: \bibitem[GM]{GM} F. W. Gehring and G. J. Martin, Precisely invariant

2059: collars and the volume of hyperbolic 3-folds, \emph{J. Diff. Geom.}

2060: \textbf{49} (1998), 411--435.

2061: 

2062: \bibitem[GMM]{GMM} D. Gabai \& R. Meyerhoff \& P. Milley,

2063: Volumes of tubes in hyperbolic 3-manifolds, \emph{J. Diff. Geom.},

2064: \textbf{57} (2001), 23--46.

2065: 

2066: \bibitem[GMM3]{GMM3} D. Gabai \& R. Meyerhoff \& P. Milley,

2067: Minimum volume cusped hyperbolic 3-manifolds, to appear.

2068: 

2069: \bibitem[GMT]{GMT} D. Gabai, R. Meyerhoff \& N. Thurston,

2070: Homotopy hyperbolic 3-manifolds are hyperbolic, \emph{Annals of

2071: Math.}, \textbf{157} (2003), 335--431.

2072: 

2073: \bibitem[Ha]{Ha} W. Haken, Theorie der Normalflachen, \emph{Acta

2074: Math.}, \textbf{105} (1961), 245--375.

2075: 

2076: \bibitem[Jac]{Jac} W. Jaco, \emph{Lectures on three-manifold

2077:   topology}, American Mathematical Society, Providence, R.I., 1980.

2078: 

2079: \bibitem[Koj]{Koj} S. Kojima, Deformations of hyperbolic

2080:   3-cone-manifolds, \emph{J. Diff. Geom.}, \textbf{49} (1998), 469--516.

2081: 

2082: \bibitem[MF]{MF} S. Matveev and A. Fomenko, Constant energy

2083: surfaces of Hamiltonian systems, enumeration of three-dimensional

2084: manifolds in increasing order of complexity, and computation of

2085: volumes of closed hyperbolic manifolds, \emph{Russian Math. Surveys},

2086: \textbf{43}:1 (1988), 3--24.

2087: 

2088: \bibitem[Mos]{Mos} H. Moser, Proving a manifold to be hyperbolic once

2089:   it has been approximated to be so, Ph.D. thesis, Department of

2090:   Mathematics, Columbia University, 2004.

2091: 

2092: \bibitem[Mv1]{Mv1} S. Matveev, An algorithm for the recognition

2093: of 3-spheres (according to Thompson), \emph{Sbornik Mathematics},

2094: \textbf{186}:5 (1995), 69--84.

2095: 

2096: \bibitem[Mv2]{Mv2} \bysame, Algorithmic topology and

2097: classification of 3-manifolds, \emph{Algorithms and Computation in

2098: Mathematics}, \textbf{9} (2003), Springer - Verlag, Berlin.

2099: 

2100: \bibitem[Sch]{Sch} H. Schubert, Bestimmung der Prifaktorzerlegung von

2101: verkettungen, \emph{Math. Zeit.}, \textbf{76} (1961), 116--148.

2102: 

2103: \bibitem[Th1]{Th1} W. P. Thurston, ``The geometry and topology of

2104: 3-manifolds'', original 1978 Princeton notes.

2105: 

2106: \bibitem[Th2]{Th2} W. P. Thurston, Three-dimensional manifolds,

2107: Kleinian groups and hyperbolic geometry, \emph{Bull. AMS (N.S.)},

2108: \textbf{6} (1982), 357--381.

2109: 

2110: \bibitem[Wa]{Wa} F. Waldhausen, On irreducible 3-manifolds which

2111: are sufficiently large, \emph{Annals of Math.}, (2) \textbf{103} (1968),

2112: 56--88.

2113: 

2114: \bibitem[We]{We} J. R. Weeks, SnapPea: A computer program for

2115: creating and studying hyperbolic 3-manifolds, available from

2116: www.northnet.org/weeks.

2117: 

2118: \end{thebibliography}

2119: \end{document}

2120: