1: %% Copy to submit to ArXiv
2: %%
3: %% For proceedings of RIMS conference, October 1999
4:
5: \documentclass[11pt]{amsart}
6:
7: \input{amsmacs}
8: \input{psfig}
9:
10: % \newcommand{\segment}[1]{\medskip
11: % \begin{center} \bf #1 \end{center}
12: % \medskip}
13:
14: \newcommand{\T}{{\mathbf T}}
15: \newcommand{\inj}{\operatorname{inj}}
16: \newcommand{\pleat}{\operatorname{\mathbf{pleat}}}
17: \newcommand{\short}{\operatorname{\mathbf{short}}}
18: \newcommand{\vertices}{\operatorname{vert}}
19: \newcommand{\collar}{\operatorname{\mathbf{collar}}}
20: \newcommand{\good}{\operatorname{\mathbf{good}}}
21: \newcommand\carriedby{\prec}
22: \newcommand\subtrack{<}
23: \newcommand\strongly{\carriedby\carriedby}
24: \newcommand\UML{\operatorname{\UU\MM\LL}}
25: \newtheorem{question}[theoremSt]{Question}
26:
27:
28: \begin{document}
29:
30: \title{Short geodesics and end invariants}
31: \author{Yair N. Minsky}
32: \address{SUNY Stony Brook}
33: \date{March 27, 2000}
34: %\thanks{}
35:
36: \maketitle
37:
38: \renewcommand\marginpar[1]{} % Kill marginpars for the final version
39:
40: Even topologically simple hyperbolic 3-manifolds
41: can have very intricate geometry. Consider in particular a closed
42: surface $S$ of genus 2 or more, and the product $N=S\times\R$. This
43: 3-manifold admits a large family of complete, infinite-volume
44: hyperbolic metrics, corresponding to faithful representations
45: $\rho:\pi_1(S)\to \PSL 2(\C)$ with discrete image.
46:
47: The geometries of $N$ are very different from the product structure that
48: its topology would suggest. Typically, $N$ contains a complicated
49: pattern of ``thin'' and ``thick'' parts. The thin parts are collar
50: neighborhoods of very short geodesics, typically infinitely many. Each
51: one, called a ``Margulis tube'', has a well-understood shape, but the
52: way in which these are arranged in $N$, and in particular the
53: identities of the short geodesics as elements of the fundamental
54: group, are still something of a mystery.
55:
56: This issue is closely related to the basic classification conjecture
57: associated with these manifolds, Thurston's ``ending lamination
58: conjecture''. This conjecture states that certain asymptotic
59: invariants of the
60: geometry of $N$, called ending invariants, in fact determine $N$ completely.
61: (Actually the classification of hyperbolic structures for any manifold
62: with incompressible boundary reduces to this case, by restriction to
63: boundary subgroups.)
64:
65: In this expository paper we will focus on the following question: What
66: information do the ending invariants give about the presence of very
67: short geodesics in the manifold? We will summarize and discuss
68: the theorem below, part of whose proof appears in \cite{minsky:kgcc}
69: and part of which will be in \cite{minsky:bdgeom}, as well as a few
70: conjectures.
71:
72: \state{Bounded Geometry Theorem.}{Let $S$ be a closed surface, and
73: consider a Kleinian surface group
74: $\rho:\pi_1(S)\to \PSL 2(\C)$ with
75: no externally short curves, and ending invariants $\nu_+$ and $\nu_-$.
76: Then
77: $$
78: \inf_{\gamma\in\pi_1(S)} \ell_\rho(\gamma) > 0
79: \iff
80: \sup_{Y\subset S} d_Y(\nu_+,\nu_-) < \infty.
81: $$
82: }
83:
84: Here the supremum is over proper essential isotopy classes of
85: subsurfaces in $S$, and
86: the quantities $d_Y(\nu_+,\nu_-)$, called ``projection
87: coefficients'', are defined in Section \ref{define dY}. The quantity
88: $\ell_\rho(\gamma)$ is the translation distance of $\rho(\gamma)$ in
89: $\Hyp^3$, or
90: the length of the closed geodesic associated to $\gamma$ in the 3-manifold.
91: (The condition on externally short curves is not really necessary --
92: it is added to simplify the other definitions and discussions -- see
93: \S\ref{surface groups} below).
94:
95: Part of our goal is to advertise a combinatorial object known as the
96: {\em complex of curves on a surface}, as a tool for studying the
97: geometry of hyperbolic 3-manifolds. This object
98: is used for definining the coefficients $d_Y$, and
99: in general it encodes something about
100: the structure of the set of simple loops on a surface.
101: In particular, face transitions between simplices in this complex
102: correspond to elementary moves on pants decompositions of $S$, and
103: these in turn correspond to homotopies between elementary pleated
104: surfaces in a hyperbolic 3-manifold. The interaction between the
105: combinatorial and geometric aspects of these moves is our main object
106: of study, and seems to be worthy of further consideration.
107:
108:
109: \section{Definitions}
110: \label{defs}
111: \subsection{Surface groups and ending laminations}
112: \label{surface groups}
113:
114: Let $S$ be a closed surface of genus $g\ge 2$.
115: A {\em Kleinian surface group} will be a
116: representation $\rho:\pi_1(S)\to \PSL 2(\C)$, discrete and faithful.
117: The quotient $\Hyp^3/\rho(\pi_1(S))$ is denoted $N_\rho$, and comes
118: equipped with a homotopy class of homotopy equivalences $S\to N_\rho$,
119: determined by $\rho$. In fact $N_\rho$ is homeomorphic to $S\times\R$,
120: by Thurston's theory of tame ends \cite{wpt:notes} and Bonahon's
121: Tameness theorem \cite{bonahon}.
122:
123: We can associate to $\rho$ two {\em ending invariants} $\nu_-$ and $\nu_+$,
124: which we will describe in the special case that $\rho$ has {\em no
125: parabolics} (see also \cite{minsky:knoxville} and Ohshika
126: \cite{ohshika:ending-lams}).
127:
128: Let $C(N_\rho)$ be the {\em convex core} of $N_\rho$, the smallest convex
129: submanifold whose inclusion is a homotopy equivalence. Fixing an
130: orientation on $S$ and $N_\rho$, there is an orientation-preserving
131: homeomorphism of $N_\rho$ to $S\times\R$ taking
132: $C(N_\rho)$ onto exactly one of
133: $S\times\R$, $S\times [0,\infty)$, $S\times(-\infty,1]$ or
134: $S\times[0,1]$.
135:
136: The end of $N$ defined by neighborhoods $S\times(a,\infty)$ is called
137: $e_+$, and the one defined by $S\times(-\infty,a)$ is called
138: $e_-$. If an end's neighborhoods all meet the convex hull it
139: is called {\em geometrically infinite}, and otherwise it is {\em
140: geometrically finite}. Suppose $e_+$ is geometrically
141: finite. Then the component $\boundary_+(C(N_\rho))$ corresponding to
142: $S\times\{1\}$ is a convex surface, and its exterior
143: $S\times(1,\infty)$ develops out to a ``conformal structure at
144: infinity'' on $S$, which we call $\nu_+$. (This surface is obtained
145: from the action of $\rho(\pi_1(S)$ on the Riemann sphere). We define
146: $\nu_-$ in the same way when $e_-$ is geometrically finite.
147:
148: Thurston pointed out that boundary $\boundary_+(C(N_\rho))$ is itself
149: a hyperbolic surface; let us call its structure $\nu'_+$. A theorem of
150: Sullivan (proof in Epstein-Marden \cite{epstein-marden}) states that
151: $\nu'_+$ and $\nu_+$ differ by a uniformly bilipschitz distortion.
152: %%We will blur the distinction between them in the rest of the paper.
153:
154: To describe the invariant of a geometrically infinite end we need to
155: briefly recall the notion of a {\em geodesic lamination}. Fixing a
156: hyperbolic metric on $S$, a
157: geodesic lamination is a closed subset of $S$ foliated by geodesics.
158: Let $\GL(S)$ denote the set of all of these. A {\em measured
159: lamination} is a geodesic lamination equipped with a Borel measure on
160: transverse arcs, invariant under transverse isotopy. The space $\ML(S)$
161: of measured laminations admits a natural topology coming from weak-*
162: convergence of the measures. On the supporting geodesic laminations,
163: this is related to but not quite the same as the topology of Hausdorff
164: convergence. However the difference will not be important to us here.
165: Simple closed geodesics with positive weights are
166: dense in $\ML(S)$, and we will consider geodesic laminations obtained
167: as supports of limits in $\ML(S)$ of sequences of simple closed
168: curves. Finally we remark that the choice of metric on $S$ is
169: irrelevant, as any other choice yields naturally isomorphic spaces of
170: laminations. For more details on this topic see
171: Bonahon \cite{bonahon:curves,bonahon:laminations},
172: Canary-Epstein-Green \cite{ceg}, or Casson-Bleiler
173: \cite{casson-bleiler}.
174:
175: If $e_+$ is geometrically infinite then the convex hull contains an
176: infinite sequence of closed geodesics $\gamma_n$, all homotopic to
177: {\em simple} closed loops on $S$, and eventually contained in
178: $S\times(a,\infty)$ for any $a$. This is a theorem of Bonahon, and
179: Thurston (previously) showed that for such a sequence the curves on
180: $S$ must converge in the sense of the previous paragraph to a
181: unique geodesic lamination on $S$. We call this lamination
182: $\nu_+$, the ending lamination of $e_+$. The
183: corresponding discussion for $e_-$ gives $\nu_-$.
184:
185: Finally let us define the technical simplifying condition in the
186: statement of the Bounded Geometry Theorem.
187: Call a curve $\gamma$ in $S$ {\em externally short}, with respect to a
188: representation $\rho$, if it is either parabolic or has length less
189: than $\ep_1$ with respect to the structures $\nu_-$ and $\nu_+$ (if
190: these are not laminations),
191: where $\ep_1$
192: is some fixed constant small enough that there exist hyperbolic
193: structures on $S$ with no curves of length less than $\ep_1$.
194: Note in particular that if $\rho$ has two degenerate ends then it
195: automatically has no externally short curves.
196:
197: \subsection{Pleated surfaces}
198:
199: A {\em pleated surface} is a map $f:S\to N$ together with a hyperbolic
200: metric on $S$, written $\sigma_f$ and called the {\em induced
201: metric}, and a $\sigma_f$-geodesic lamination $\lambda$ on $S$,
202: called the pleating locus, so that the following holds:
203: $f$ is length-preserving on paths, maps leaves of $\lambda$ to geodesics, and
204: is totally geodesic on the complement of $\lambda$. These were
205: introduced by Thurston \cite{wpt:notes}, and we will see some
206: explicit examples in \S \ref{pleat on pants}.
207:
208: It is a consequence of the work of Thurston and Bonahon that a
209: geometrically infinite end of a surface group $\rho$ admits pleated
210: surfaces in the homotopy class of $\rho$ contained in any neighborhood
211: of the end. The pleating loci of these surfaces must converge to the
212: ending lamination, and their hyperbolic structures converge to this
213: lamination in Thurston's compactification of the Teichm\"uller space.
214:
215: \subsection{Complexes of arcs and curves:}
216: \label{complex defs}
217:
218: Let $Z$ be a compact finite genus surface, possibly with boundary.
219: If $Z$ is not an annulus,
220: define $\AAA_0(Z)$ to be
221: the set of essential homotopy classes of
222: simple closed curves or properly embedded arcs in $Z$. Here
223: ``homotopy class'' means free homotopy for closed curves, and
224: homotopy rel $\boundary Z$ for arcs.
225: ``Essential'' means the homotopy class does not contain the constant
226: map or a map into the boundary. If $Z$ is an annulus, we make the same
227: definition except that homotopy for arcs is rel endpoints.
228:
229: We can extend $\AAA_0$ to a simplicial complex
230: $\AAA(Z)$ by letting a $k$-simplex be any $(k+1)$-tuple
231: $[v_0,\ldots,v_k]$ with $v_i\in\AAA_0(Z)$ distinct and having
232: pairwise disjoint representatives.
233:
234: Let $\AAA_i(Z)$ denote the $i$-skeleton of $\AAA(Z)$, and let
235: $\CC(Z)$ denote the subcomplex spanned by vertices corresponding to
236: simple closed curves. This is the ``complex of curves of $Z$''.
237:
238: If we put a path metric on $\AAA(Z)$ making every simplex regular
239: Euclidean of sidelength 1, then it is clearly quasi-isometric to its
240: $1$-skeleton. It is also quasi-isometric to $\CC(S)$ except in a few
241: simple cases when $\CC(S)$ has no edges. When $\boundary Z=\emptyset$,
242: of course $\AAA=\CC$.
243:
244: It is a nice exercise to compute $\AAA(Z)$ exactly for $Z$ a one-holed
245: torus, and we leave this to the reader. The answer is closely related
246: to the Farey graph in the plane -- see \cite{minsky:taniguchi}.
247:
248: Fix our closed surface $S$ and let
249: $\GL(S)$ denote the set of geodesic laminations on $S$ (note that
250: $\AAA_0(S)=\CC_0(S)$ can identified with a subset of $\GL(S)$).
251: Let $Y\subset S$ be a proper essential closed subsurface (all boundary curves
252: homotopically nontrivial). We have a ``projection map''
253: $$
254: \pi_Y : \GL(S) \to \AAA(\hhat Y)\union\{\emptyset\}
255: $$
256: defined as follows: there is a unique
257: cover of $S$ corresponding to the inclusion
258: $\pi_1(Y)\subset\pi_1(S)$, which can be
259: naturally compactified using the circle at infinity
260: of the universal cover of $S$ to yield a surface $\hhat Y$ homeomorphic
261: to $Y$ (remove the limit set of $\pi_1(Y)$ and
262: take the quotient of the rest). Any lamination $\lambda\in\GL(S)$
263: lifts to this cover as a collection of closed curves or arcs that have
264: well-defined endpoints in $\boundary \hhat Y$.
265: Removing the trivial components, we have a simplex of $\AAA(\hhat Y)$
266: and we can take, say, its barycenter
267: (we can also get the empty set if there are no essential components).
268: A version of this projection also appears in Ivanov \cite{ivanov:subgroupsbook,ivanov:rank}.
269:
270: If $\beta,\gamma\in\GL(S)$ (in particular in $\CC(S)$ have non-trivial
271: intersection with $Y$, we denote their ``$Y$-distance'' by:
272:
273: $$
274: d_Y(\beta,\gamma) \equiv d_{\AAA(\hhat Y)}(\pi_Y(\beta),\pi_Y(\gamma)).
275: $$
276: Note that $\AAA(\hhat Y)$ can be naturally identified with $\AAA(Y)$,
277: except when $Y$ is an annulus, in which case the pointwise
278: correspondence of the boundaries matters.
279: In the annulus case $d_Y$ measures relative twisting of arcs
280: determined rel endpoints, and in all other cases we ignore twisting on
281: the boundary of $\hhat Y$. If $\alpha$ is the core curve of an annulus
282: $Y$ we will also write
283: $$
284: d_\alpha = d_Y.
285: $$
286: See \cite{farb-lubotzky-minsky} for an application of this
287: construction in the annulus case.
288:
289: The complex of curves $\CC(S)$ was first introduced by Harvey
290: \cite{harvey:boundary}.
291: It was applied by Harer \cite{harer:stability,harer:cohomdim} and
292: Ivanov \cite{ivanov:complexes1,ivanov:complexes2,ivanov:complexes3}
293: to study the
294: mapping class group of $S$. Similar complexes were introduced by
295: Hatcher-Thurston \cite{hatcher-thurston}. Masur-Minsky
296: \cite{masur-minsky:complex1} proved that $\CC(S)$ is
297: $\delta$-hyperbolic in the sense of Gromov, and then applied this in
298: \cite{masur-minsky:complex2} to prove the structural theorems on pants
299: decompositions that we will use in Section \ref{proof main}.
300:
301: \subsection{Projection coefficients}
302: \label{define dY}
303:
304: Let us now see how to define the coefficients
305: $$
306: d_Y(\nu_+,\nu_-)
307: $$
308: which appear in the main theorem, where $\nu_\pm$ are ending
309: invariants for a surface group.
310: Using $\pi_Y$ as above, we can already define this
311: whenever $\nu_\pm$ are laminations.
312: In the case of a geometrically finite end when $\nu_+$ or $\nu_-$ are
313: hyperbolic metrics, we can extend this definition as follows:
314:
315: If $\sigma $ is a hyperbolic metric on $S$, and $L_1$ a fixed
316: constant, define
317: $$
318: \short(\sigma)$$
319: to be the set of pants decompositions of $S$ with
320: total $\sigma$-length at most $L_1$. A theorem of Bers
321: (see \cite{bers:degenerating,bers:inequality} and
322: Buser \cite{buser:surfaces})
323: says that $L_1$ can be chosen,
324: depending only on genus of $S$, so that $\short(\sigma)$ is always non-empty.
325: Let us also choose $L_1$ sufficiently large that, if $\sigma$ has no
326: curves of length less than $\ep_1$ (the constant from the end of
327: \S\ref{surface groups}), then {\em every} curve in $S$
328: intersects some $P\in\short(\sigma)$.
329:
330: Thus e.g. if both $\nu_+$ and $\nu_-$ are hyperbolic structures, we
331: may consider distances
332: $$
333: d_Y(P_+,P_-)
334: $$
335: for any $P_\pm\in\short(\nu_\pm)$ that both intersect $Y$ essentially,
336: and notice that the numbers obtained cannot vary by more than a
337: uniformly bounded constant. We let $d_Y(\nu_+,\nu_-)$ be, say,
338: the minimum over all choices. The case when one of
339: $\nu_\pm$ is a lamination and the other is a hyperbolic metric is handled
340: similarly. Note that the condition that $\rho$ has no externally short
341: curves implies that $d_Y(\nu_+,\nu_-)$ is well-defined for all $Y$.
342:
343: \section{Margulis tubes}
344:
345: Let $\gamma$ be a loxodromic element of a Kleinian group $\Gamma$.
346: We denote its complex translation length by $\lambda(\gamma) =
347: \ell+i\theta$ (determined mod $2\pi i$). Let $\TT_\ep$ be the
348: $\gamma$-invariant set
349: $\{x\in\Hyp^3: \inf_n d(x,\gamma^n(x)) \le \ep\}$. If $\ell(\gamma)<\ep$
350: This is a tube of some radius $r$ around the axis of $\gamma$, and
351: The Margulis Lemma and Thick-Thin decomposition tell us
352: (see e.g. \cite{kazhdan-margulis,wpt:textbook,benedetti-petronio}) that
353: there is a universal constant $\ep_0$ such that if $\ell(\gamma)<\ep_0$
354: then $\TT_{\ep_0}/\langle\gamma\rangle$ embeds as a solid torus $\T_\gamma$
355: in $N=\Hyp^3/\Gamma$, called a {\em Margulis tube}, and furthermore
356: that all Margulis tubes in $N$ are disjoint.
357:
358: The radius $r$ of the tube goes to $\infty$ as the length of the core
359: goes to 0. See Brooks-Matelski \cite{brooks-matelski} and Meyerhoff
360: \cite{meyerhoff:volumes}
361: for more precise bounds.
362:
363: Thus in some sense the geometry around a very short curve in $N$ is
364: very well understood. It is more difficult to determine the
365: pattern in which these tubes are arranged in the manifold, and in
366: particular which curves $\gamma$ have length less than a given $\ep$.
367:
368: \subsection{Margulis tubes in surface groups}
369: \label{margulis tubes}
370:
371: When $\Gamma$ is the image $\rho(\pi_1(S))$ of a Kleinian surface
372: group, there is a little more we can say. An observation of Thurston
373: \cite{wpt:II}, together with Bonahon's tameness theorem
374: \cite{bonahon}, imply that {\em only simple curves can be short:} that is,
375: $\ep_0$ may be chosen so that,
376: if $\ell_\rho(\gamma)<\ep_0$
377: and $\gamma$ is a primitive element of $\pi_1(S)$ then $\gamma$ is
378: represented by a simple loop in $S$. This is because, by Bonahon's
379: theorem, every point in $N_\rho$ is uniformly near the image of a
380: pleated surface. Thurston pointed out using a simple area bound that
381: if $\ep_0$ is sufficiently short a
382: $\pi_1$-injective pleated surface can only meet $\T_\gamma$
383: in the image of its own 2-dimensional Margulis
384: tube. The core of this tube must therefore be $\gamma$.
385:
386: % Another consequence of the presence of pleated surfaces is that, if
387: % $\ell_\rho(\gamma)<\ep_0$ then there is an upper bound to the
388: % translation distance of $\rho(\gamma)$ on $\TT_{\ep_0}(\gamma)$ (in
389: % general we only know that some power of $\rho(\gamma)$ has bounded
390: % distance there). DO WE USE THIS ANYWHERE??? See \cite{minsky:torus}
391: % for further discussion.
392:
393: \subsection{Bounds}
394:
395: An {\em upper bound} on the length of a curve in a surface group can
396: be obtained in terms of the conformal boundary at infinity. Bers
397: showed \cite{bers:boundaries} for a Quasi-Fuchsian representation $\rho$, that
398: $$
399: \frac1{\ell_\rho(\gamma)} \ge \frac12\left(\frac1{\ell_+(\gamma)} +
400: \frac1{\ell_-(\gamma)}\right)
401: $$
402: where $\ell_\pm$ denote lengths in the hyperbolic
403: structures on $S$ coming from the two conformal structures $\nu_\pm$
404: at infinity. The argument uses a monotonicity property for conformal
405: moduli and the action of $\gamma$ on the Riemann sphere.
406: When $S$ is a once-punctured torus this upper bound can be
407: generalized to an estimate in both directions (see
408: \cite{minsky:torus}). In general we have no such result, but in
409: Section \ref{conjectures} we will state a conjectural estimate.
410:
411: %% Define $\omega = 2\pi i/\lambda$. (c.f. Cusps are Dense).
412:
413:
414: \section{Bounded geometry}
415:
416: We say that $\rho$ has {\em bounded geometry} if there is a positive lower
417: bound on the translation lengths of all group elements.
418: This condition incidentally disallows parabolic elements (in a more
419: general discussion we would allow them and revise the condition),
420: but the real point is
421: that there is a positive lower bound on the lengths of all closed
422: geodesics in the quotient manifold.
423:
424: In \cite{minsky:slowmaps,minsky:endinglam}, we showed that
425: bounded geometry implies a positive solution
426: to the ending lamination conjecture. That is, if $\rho_1$ and $\rho_2$
427: both have bounded geometry, and have the same ending invariants, then
428: they are conjugate in $\PSL 2(\C)$. This result was accompanied by a
429: fairly explicit bilipschitz model for the metric on $N$, derived from
430: the Teichm\"uller geodesic joining the two ending laminations.
431:
432: The Bounded Geometry theorem gives us a way to strengthen
433: this result, since it implies that bounded geometry is detected by
434: the ending invariants:
435:
436: \begin{corollary}{Bounded Geometry ELC}
437: Let $\rho_1,\rho_2$ be Kleinian surface groups with the same ending
438: invariants, and suppose that $\rho_1$ has bounded geometry. Then
439: $\rho_1$ and $\rho_2$ are conjugate in $\PSL 2(\C)$.
440: \end{corollary}
441:
442: It is worth noting that bounded geometry is a rare condition. In the
443: boundary of a Bers slice, for example, there is a topologically
444: generic (dense $G_\delta$) set of representations each of which has
445: arbitrarily short elements (see McMullen
446: \cite[Cor. 1.6]{mcmullen:cusps}, and Canary-Culler-Hersonsky-Shalen
447: \cite{canary-culler-hersonsky-shalen:cusps} for generalizations).
448:
449: % \begin{pf}
450: % Order the set of conjugacy classes of $\pi_1(S)$ in some way:
451: % $\{c_1, c_2, c_3,\ldots\}$. Define $\Lambda_n$ to be this set of
452: % representations of $\pi_1(S)$:
453: %
454: % $$
455: % \{\rho : \text{for some $i>n$,
456: % $\ell_\rho(c_i) < 1/n$} \}.
457: % $$
458: %
459: % The representations with unbounded geometry are exactly $\bigcap_n
460: % \Lambda_n$.
461: %
462: % Each $\Lambda_n$ is open in the representation variety, so
463: % $\Lambda_n \intersect D$ is open in $D$.
464: % It remains to show that $\Lambda_n \intersect D$ is dense in $D$.
465: %
466: % Use the notation $(X,P)$ to mean the maximal cusp in $D$ corresponding to a
467: % pants decomposition $P$; that is, a representation $\rho$ with ending
468: % invariants $\nu_-=X$ and $\nu_+ = P$.
469: % Given any $P$, fix another decomposition $Q$ transverse to
470: % $P$, and let $Q_m$ be the $m$-fold Dehn twist of $Q$ around all
471: % components of $P$.
472: % So $Q_m \to P$ as projective measured laminations.
473: %
474: % The sequence $\rho_m = (X,Q_m)$ converges to $(X,P)$ -- see for
475: % example Kerckhoff-Thurston \cite{kerckhoff-thurston:limits}.
476: % \marginpar{Give length function continuity argument?}
477: %
478: % We next claim $\rho_m$ is eventually in $\Lambda_n$ for any $n$: This
479: % is because
480: % the conjugacy classes of $Q_m$ go to infinity in the list $\{c_i\}$, and
481: % $\ell_{rho_m}(Q_m) = 0$.
482: %
483: % Thus, each $\Lambda_n\intersect D$ accumulates onto $(X,P)$.
484: % McMullen showed in \cite{mcmullen:cusps} that maximal cusps are dense
485: % in $D$.
486: % Thus, $\Lambda_n\intersect D$ is dense in $D$.
487: % \end{pf}
488:
489: \section{The proof of the bounded geometry theorem}
490: \label{proof main}
491:
492: % In the case of a higher-genus surface $S$, the role of the
493: % continued-fraction coefficients from the torus case can be played by
494: % the subsurface projections $\pi_Y$. That is, if $\nu_-$ and $\nu_+$
495: % are the ending invariants of a representation $\rho$, we can form
496: % $d_Y(\nu_+,\nu_-)$ for every essential subsurface $Y$ of $S$, as in
497: % Section \ref{defs}. When $Y$ is an annulus $d_Y(\nu_+,\nu_-)$ measures
498: % relative twisting of $\nu_-$ and $\nu_+$ around $Y$, and these for the
499: % punctured torus were exactly the continued-fraction coefficients.
500:
501:
502:
503: The proof of the direction $(\Rightarrow)$ of
504: the Bounded Geometry Theorem appears in
505: \cite{minsky:kgcc}. The essential tool used there is Thurston's
506: ``efficiency of pleated surfaces'' theorem from \cite{wpt:II}.
507: We will outline the proof of $(\Leftarrow)$, for
508: which the details will appear in \cite{minsky:bdgeom}.
509:
510: \medskip
511:
512: In roughest form, the argument is this: Let $\gamma\in\pi_1(S)$ be an
513: element with $\ell_\rho(\gamma) < \ep_0$, and let $\T_\gamma$ be its
514: Margulis tube. We will use the condition $\sup d_Y(\nu_+,\nu_-) < \infty$ to
515: construct a sequence of pleated surfaces $\{f_i\}_{i=0}^M$ with the
516: following properties:
517: \begin{enumerate}
518: \item The size $M$ of the sequence is bounded by
519: $$M\le K (\sup d_Y(\nu_+,\nu_-))^a$$
520: where $K,a$ depend only on the genus of $S$.
521: \item Any successive $f_i,f_{i+1}$ are connected by a homotopy
522: $H:S\times[i,i+1]\to N_\rho$ which is {\em uniformly bounded}
523: except in a special case, described below.
524: \item The total homotopy $H:S\times[0,M]\to N_\rho$ homologically
525: encloses $\T_\gamma$.
526: \end{enumerate}
527:
528: Part (3) means that the image of $H$ must cover all of $\T_\gamma$.
529: Thus, if the ``special case'' of (2) does not occur, then the bounds of (1) and
530: (2) give a uniform diameter bound on $\T_\gamma$, and hence a lower
531: bound on $\ell_\rho(\gamma)$.
532:
533: The ``special case'' of (2) corresponds to the curve $\gamma$ itself
534: appearing in the pleating locus of some subsequence of the $f_i$. In
535: this case a more delicate argument is needed, using the
536: annulus projection
537: distance $d_\gamma(\nu_+,\nu_-)$ to bound the size of $\T_\gamma$.
538:
539: Let us now introduce the ingredients needed for this construction. In
540: \S\ref{finishing argument} we will return to the main proof.
541:
542: \subsection{Adapted pleated surfaces}
543: \label{pleat on pants}
544:
545: If $Q$ is a collection of disjoint, homotopically distinct curves on
546: $S$ (henceforth a ``curve system''), and $\rho$ a fixed Kleinian surface group,
547: we let $\pleat(Q,\rho)$ denote the set of pleated surfaces $f:S\to
548: N_\rho$, in the homotopy class determined by $\rho$, which map
549: representatives of $Q$ to geodesics.
550: There is the usual equivalence relation on this set,
551: in which $f\sim f\circ h$ if $h$ is a
552: homeomorphism of $S$ homotopic to
553: the identity. Let $\sigma_f$ denote the
554: hyperbolic metric on $S$ induced by $f$.
555:
556: In particular, if $Q$ is a maximal curve system, or ``pants
557: decomposition'', $\pleat(Q,\rho)$ consists of finitely many
558: equivalence classes, all
559: constructed as follows: Extend $Q$ to a triangulation of $S$ with
560: one vertex on each component of $Q$,
561: and ``spin'' this triangulation around $Q$,
562: arriving at a lamination $\lambda$ whose closed leaves are $Q$ and
563: whose other leaves spiral onto $Q$, as in Figure \ref{spun
564: lamination}.
565:
566: %\blankfig{spun lamination}{2in}{The lamination obtained by spinning a
567: %triangulation around a curve system.}
568: \realfig{spun lamination}{spin.ps}{The lamination obtained by spinning a
569: triangulation around a curve system. The picture shows one pair of
570: pants in a decomposition.}
571:
572: A unique pleated surface (up to equivalence) exists carrying
573: $\lambda$ to geodesics, since no element of $Q$ is parabolic (by
574: hypothesis on $\rho$). This was originally observed by Thurston (see
575: \cite{wpt:notes} and Canary-Epstein-Green \cite[Thm 5.3.6]{ceg} for a
576: proof).
577: The choices of $\lambda$ coming from the finite number of possible
578: triangulations up to isotopy, and the
579: different directions of spiraling, account for all of $\pleat(Q,\rho)$.
580:
581: \subsection{Elementary moves}
582:
583: An elementary move on a maximal curve system $P$ is a replacement of a
584: component $\alpha$ of $P$ by $\alpha'$, disjoint from the rest of $P$,
585: so that $\alpha$ and $\alpha'$
586: are in one of the two configurations shown in Figure
587: \ref{elementary move fig}.
588:
589: %\blankfig{elementary move fig}{2in}{The two types of elementary moves.}
590: \realfig{elementary move fig}{elmoves.ps}{The two types of elementary moves.}
591:
592: We indicate this by $P\to P'$ where
593: $P'=P\setminus\{\alpha\}\union\{\alpha'\}$ is the new curve system.
594: Note that there are infinitely many choices for $\alpha'$, naturally
595: indexed by $\Z$.
596:
597: Pleated surfaces associated to an elementary move are homotopic in a
598: controlled way. Let us first recall (see Buser \cite{buser:surfaces})
599: that a simple geodesic $\gamma$ in a hyperbolic surface $(S,\sigma)$
600: always admits a ``standard collar'', which is an annulus of radius
601: depending only on
602: $\ell_\sigma(\gamma)$, such that disjoint geodesics have disjoint
603: collars, and when $\ell_\sigma(\gamma) < \ep_0$ the collar covers all
604: but a bounded part of the
605: $\ep_0$-Margulis tube. We write this collar as
606: $\collar(\gamma,\sigma)$, or $\collar(P,\sigma)$ for the union of
607: collars over a curve system $P$.
608:
609: \begin{lemma+}{Elementary Homotopy}
610: If $P_0\to P_1$ is an elementary move exchanging $\alpha_0$ and
611: $\alpha_1$, $\rho$ is a Kleinian surface
612: group, and $f_i \in \pleat(P_i,\rho)$ for $i=0,1$, then
613: there exists a homotopy $H:S\times[0,1]\to N_\rho$ with the
614: following properties:
615: \begin{enumerate}
616: \item $H_0\sim f_0$ and $H_1\sim f_1$ under the usual equivalence.
617: \item If $\sigma_i$ is the induced metric of $H_i$ (for $i=0,1$)
618: then $\collar(P_j,\sigma_i) = \collar(P_j,\sigma_{1-i})$, for $j=0,1$.
619:
620: \item The metrics $\sigma_0$ and $\sigma_1$ are $K$-bilipschitz
621: except possibly when $\ell_\rho(\alpha_i) < \ep_0$ for $i=0$ or $1$.
622: In that case the metrics are locally $K$-bilipschitz
623: outside $\collar(\alpha_0)\union\collar(\alpha_1)$ (or just one
624: collar if only one curve is short in $N_\rho$).
625:
626: \item The trajectories $H(p\times[0,1])$ are bounded in length by $K$
627: except
628: possibly when $p\in\collar(\alpha_i)$ and $\ell_\rho(\alpha_i)<\ep_0$,
629: in which case they are bounded outside of $\T_{\rho(\alpha_i)}$.
630:
631: \end{enumerate}
632: The constant $K$ depends only on the genus of $S$.
633:
634: \end{lemma+}
635: (Note that $\collar(\alpha_i)$ in (3) and (4) makes sense without
636: specifying the metric $\sigma_j$, since the two are equal by (2).)
637:
638: It is worth pointing out that this theorem applies without any
639: a-priori bounds on the lengths $\ell_\rho(P_i)$. The proof is an
640: application of Thurston's Uniform Injectivity theorem for pleated
641: surfaces, and the closely related
642: Efficiency of Pleated Surfaces \cite{wpt:II} (see also Canary
643: \cite{canary:schottky}).
644: These theorems control the amount kind of bending that can occur in a
645: pleated surface, and in particular can be used to compare two pleated
646: surfaces that share part of their pleating locus.
647:
648: We also remark that part (2) is just for convenience -- it is easy to
649: arrange by an appropriate isotopy.
650:
651: \subsection{Resolution sequences}
652:
653: In \cite{masur-minsky:complex2}, we show the existence of special
654: sequences of elementary moves that are controlled in terms of the
655: geometry of the complex of curves, and particularly the projections
656: $\pi_Y$. First some terminology: if $P_0\to P_1 \to \cdots \to P_n$ is
657: an elementary-move sequence and $\beta$ is any simple closed curve,
658: denote
659: $$J_\beta = \{ i\in[0,n]: \beta\in P_i\}.$$
660: (Here $\beta\in P$ means $\beta$ is a component of $P$.)
661: We also denote $J_{\beta_1,\ldots,\beta_k} = \union J_{\beta_i}$.
662:
663: Note that if $\beta$ is a curve and $J_\beta$ is an interval $[k,l]$,
664: then the elementary move $P_{k-1}\to P_k$ exchanges some $\alpha$ for
665: $\beta$, and $P_l\to P_{l+1}$ exchanges $\beta$ for some
666: $\alpha'$. Both $\alpha$ and $\alpha'$ intersect $\beta$, and we call
667: them the {\em predecessor} and {\em successor} of $\beta$, respectively.
668:
669: \begin{theorem+}{Controlled Resolution Sequences}
670: Let $P$ and $Q$ be maximal curve systems in $S$.
671: There exists a geodesic
672: $\beta_0,\ldots,\beta_m$ in $\CC_1(S)$ and an elementary move sequence
673: $P_0\to\ldots\to P_n$, with the following properties:
674: \begin{enumerate}
675: \item $\beta_0\in P_0 = P$ and $\beta_m \in P_n = Q$.
676: \item Each $P_i$ contains some $\beta_j$.
677: \item $J_\beta$, if nonempty, is always an interval, and if
678: $[i,j]\subset[0,m]$ then
679: $$
680: |J_{\beta_i,\ldots,\beta_j}| \le K (j-i) \sup_Y d_Y(P,Q)^a,
681: $$
682: where the supremum is over only those subsurfaces $Y$ whose boundary
683: curves are components of some $P_k$ with $k\in
684: J_{\beta_i,\ldots,\beta_j}$.
685: \item
686: If $\beta$ is a curve with non-empty $J_\beta$, then its predecessor
687: and successor curves $\alpha $ and $\alpha'$ satisfy
688: $$
689: | d_\beta(\alpha,\alpha') - d_\beta(P,Q) | \le \delta.
690: $$
691: \end{enumerate}
692: The constants $K,a,\delta$ depend only on the genus of $S$.
693: The expression $|J|$ for an interval $J$ denotes its diameter.
694: \end{theorem+}
695: The sequence $\{P_i\}$ in this theorem is called a resolution
696: sequence. Such sequences are constructed in
697: \cite{masur-minsky:complex2} by an inductive procedure: beginning with
698: a geodesic $\{\beta_i\}$ in $\CC_1(S)$ joining $P$ to $Q$
699: (we are describing a geodesic here as a sequence of vertices where
700: successive ones are joined by edges),
701: we note that the link of each $\beta_i$ is itself a
702: curve complex for a subsurface. In each such complex we add a new
703: geodesic, and repeat. The final structure can then be
704: ``resolved'' into a sequence of maximal curve systems. Control of the
705: size of the construction at each stage is achieved by applying the
706: hyperbolicity theorem of \cite{masur-minsky:complex1}.
707:
708:
709: \subsection{Contraction and quasi-convexity}
710:
711: Let $\CC(S,\rho,L)$ denote the subcomplex of $\CC$ spanned by the
712: vertices with $\rho$-length at most $L$.
713: We will define a map
714: $\Pi_\rho:\CC(S) \to \PP(\CC(S,\rho,L_1))$,
715: where $\PP(X)$ is the power set of $X$, as follows.
716: For $x\in\CC(S)$, let $P_x$ be the curve system associated to the
717: smallest simplex containing $x$. We define
718: $$
719: \Pi_\rho(x) = \bigcup_{f\in\pleat(P_x,\rho)} \short(\sigma_f).
720: $$
721: This map turns out to have coarsely the properties of a
722: closest-point projection to a convex subset of a hyperbolic space.
723:
724: \begin{lemma+}{Contraction Properties}
725: There are constants $b,c>0$, depending only on the genus of $S$, such
726: that for any $\rho$ the map
727: $\Pi_\rho$ has the following properties:
728: \begin{enumerate}
729: \item (Coarse Lipschitz) If $d_\CC(x,y) \le 1$ then
730: $$\diam_\CC (\Pi_\rho(x)\union\Pi_\rho(y)) \le b.$$
731: \item (Coarse idempotence) If $x\in\CC(S,\rho,L_1)$ then
732: $$
733: d_\CC(x,\Pi_\rho(x)) \le b.
734: $$
735: \item (Contraction) If $r=d_\CC(x,\Pi_\rho(x))$ then
736: $$
737: \diam_\CC \Pi_\rho(B(x,cr)) \le b.
738: $$
739: \end{enumerate}
740: \end{lemma+}
741: Here $d_\CC$ and $\diam_\CC$ refer to distance and diameter measured
742: in $\CC(S)$, and $B(x,s)$ is a ball of $d_\CC$-radius $s$ around $x$.
743: By $\Pi_\rho(X)$ for a set $X$ we mean $\union_{x\in X} \Pi_\rho(x)$.
744:
745: Compare this with the contraction property in
746: \cite{masur-minsky:complex1}, which was used to prove hyperbolicity of
747: $\CC(S)$, and the property in \cite{minsky:projections}, which was
748: used to prove stability properties for certain geodesics in
749: Teichm\"uller space.
750:
751: An easy consequence of this theorem is the
752: following quasiconvexity property for $\CC(S,\rho,L_1)$:
753: \begin{lemma+}{Quasiconvexity}
754: If $\beta_0,\ldots,\beta_m$ is a geodesic in $\CC_1(S)$ and
755: $\beta_0,\beta_m\in \CC(S,\rho,L_1)$, then
756: $$
757: d_\CC(\beta_i,\Pi_\rho(\beta_i)) \le C
758: $$
759: for all $i\in[0,m]$ and a constant $C$ depending only on the genus of $S$.
760: \end{lemma+}
761: In particular a geodesic with endpoints in $\CC(S,\rho,L_1)$ never strays too
762: far from $\CC(S,\rho,L_1)$. This can be compared to the ``Connectivity''
763: lemma in \cite{minsky:torus}.
764:
765: The argument for this lemma is very simple, and has its origins in the
766: stability of quasi-geodesics argument in the proof of Mostow's
767: Rigidity Theorem \cite{mostow:hyperbolic}: We compare the path
768: \marginpar{CHECK if Mostow used this...}
769: $\{\beta_i\}$ to its image ``quasi-path'' $\{\Pi_\rho(\beta_i)\}$. If
770: the distance between these grows too much then the images slow
771: down because of the Contraction property (3). Since $\{\beta_i\}$ is a
772: shortest path and the two paths have nearly the same endpoints
773: (Coarse idempotence (2)), there is a bound on how far apart they can get.
774:
775: \medskip
776:
777: The proof of Lemma \ref{Contraction Properties} is another
778: application of Thurston's
779: Uniform Injectivity theorem, as well as some of the tools developed in
780: \cite{masur-minsky:complex1}. For example, to prove part (1),
781: we note that if two vertices of $\CC(S)$ are at distance 1 then they
782: correspond to disjoint curves, and hence a pleated surface exists that
783: maps both geodesically. Thus the argument reduces to bounding
784: $\Pi_\rho(x)$ for one $x$. Suppose two pleated surfaces share a curve $x$.
785: If $x$ is short then their short curve sets
786: intersect, and we finish by noting that $\diam_\CC(\short(\sigma))$ is
787: uniformly
788: bounded for any $\sigma$. If the curve $x$ is long, then in one of the
789: pleated surfaces we can find a curve $x'$ of bounded length that runs
790: along $x$ and then makes a very small jump in its complement
791: (a long curve in a hyperbolic surface must run very close to
792: itself). The Uniform Injectivity theorem is then applied to show that
793: $x'$ can be realized with bounded length on the second pleated surface
794: as well.
795:
796: Part (3) is the main point of the lemma. Its proof depends
797: on the analysis in \cite{masur-minsky:complex1}, which
798: shows roughly that if $x\in\CC(S)$ is far in $\CC(S)$ from the short
799: curves of a given hyperbolic metric $\sigma$ on $S$, then sets of the
800: form $B(x,R)$ for large $R$
801: can be carried in a long nested chain of ``train tracks'' (see
802: \cite{penner-harer})
803: whose branches mostly run nearly parallel to $x$.
804: These train tracks are then used to
805: control $\Pi_\rho(B(x,R))$, via a Uniform Injectivity argument similar
806: to the previous paragraph.
807:
808: \subsection{Building a resolution sequence for $\rho$}
809: \label{finishing argument}
810:
811: We can now use Theorem \lref{Controlled Resolution Sequences} and
812: Lemma \lref{Quasiconvexity}
813: to produce a resolution sequence adapted to the geometry of our
814: representation $\rho$.
815:
816: As a starting point we need an initial and terminal curve system:
817: \begin{lemma}{Pants near ends}
818: Given $\rho$ with no externally short curves, and a Margulis tube
819: $\T_\gamma$ in $N_\rho$, there exist maximal curve systems $P_+$ and
820: $P_-$, and pleated surfaces $f_+, f_-$ (in the homotopy class of
821: $\rho$) with the following properties:
822: \begin{enumerate}
823: % \item $\ell_\rho(P_\pm) \le L_1$
824: % \item If $f_\pm\in \pleat(P_\pm,\rho)$ then $f_+$ and $f_-$
825: % homologically encase $\T_\gamma$.
826: \item $P_\pm\in \short(\sigma_{f_\pm})$,
827: \item $f_+$ and $f_-$ homologically encase $\T_\gamma$.
828: \end{enumerate}
829: \end{lemma}
830:
831: This is done roughly as follows. If $\nu_+$ is a lamination then there
832: exists a sequence $g_i$ of pleated surfaces exiting the end of
833: $N_\rho$ corresponding to $\nu_+$. The curves in
834: $\short(\sigma_{g_i})$
835: converge to $\nu_+$ in the space of laminations (modulo measure),
836: and for large enough $i$, $g_i$ can be deformed to $e_+$
837: without meeting $\T_\gamma$. We can pick $f_+ = g_i$ and let
838: $P_+\in\short(\sigma_{g_i})$.
839: The same goes for
840: $f_-,P_-$, so if both invariants are laminations we have the
841: conclusion that $f_+$ and $f_-$ must encase $\T_\gamma$.
842:
843: If the end $e_+$ is geometrically finite we can let $f_+$ be the
844: pleated map to the convex hull boundary itself, and similarly for
845: $e_-$. Again choose $P_\pm\in\short(\sigma_{f_\pm}) = \short(\nu'_\pm)$.
846:
847:
848: \medskip
849:
850: Note that, if the pleated surfaces $g_i$ are chosen far enough out the
851: end (in the geometrically infinite case) then the homotopy from $g_i$
852: to a map in $\pleat(P_+,\rho)$ does not pass through $\T_\gamma$, and
853: so we may assume $f_\pm\in\pleat(P_\pm,\rho)$ and still have the
854: encasing condition. When there are geometrically finite ends this is
855: trickier because $\T_\gamma$ may be close to the convex hull
856: boundary. Slightly more care is needed in the rest of the construction
857: in that case. Let us from now on assume that
858: $f_\pm\in\pleat(P_\pm,\rho)$, and the encasing condition holds.
859:
860: \medskip
861:
862: Join $P_+$ to $P_-$ with a resolution sequence
863: $P_-=P_0\to\cdots\to P_n=P_+$, as in
864: Theorem \ref{Controlled Resolution Sequences}. Let
865: $\{\beta_i\}_{i=0}^m$ be the associated geodesic. This sequence may
866: be much longer than we need, so we will use Lemma \ref{Quasiconvexity}
867: to find a suitable subsequence. Recall that we would like our sequence
868: to have the property of homologically encasing $\T_\gamma$, so let us
869: try to throw away those surfaces that we are sure cannot meet
870: $\T_\gamma$. In particular, let $f\in\pleat(P_i,\rho)$ for some
871: $i\in[0,n]$, and let $P_i$ contain a curve $\beta_j$.
872: If $f(S)\intersect \T_\gamma \ne \emptyset$, then
873: $\gamma$ itself is short in $\sigma_f$ (as in \S\ref{margulis tubes})
874: and so $\gamma$ is distance 1 from $\Pi_\rho(\beta_j)$. It follows
875: from Lemma \ref{Quasiconvexity} that
876: $$
877: d_\CC(\beta_j,\gamma) \le C \eqno{(*)}
878: $$
879: where $C$ is a new constant depending only on the genus of $S$.
880: Thus we conclude that
881: there is a subinterval $I_\gamma$ of $[0,m]$ of diameter
882: at most $2C$, such that $f$ can only meet $\T_\gamma$ when $\beta_j$
883: satisfies $j\in I_\gamma$. Let us therefore restrict our elementary
884: move sequence to
885: $$P_{s-1}\to\cdots\to P_{t+1}$$
886: where $[s,t] = \union_{j\in I_\gamma} J_{\beta_j}$, and
887: %%$I_\gamma = [s,t]$, and
888: renumber it as $P_0\to\cdots\to P_M$.
889: This subsequence must also encase
890: $\T_\gamma$, since none of the pieces we have thrown away can meet
891: $\T_\gamma$.
892: Part (3) of Theorem \ref{Controlled Resolution Sequences} tells us
893: that
894: $$
895: M \le K(2C)\sup_Y d_Y(P_+,P_-)^a,
896: $$
897: where the supremum is over subsurfaces $Y$ whose boundaries appear
898: among the $P_i$ in our subsequence. This means by ($*$) that the
899: $\CC(S)$-distance $d_\CC(\boundary Y,\gamma)$ is bounded by $C+1$ for
900: all such $Y$. The analysis of \cite{masur-minsky:complex2} shows that,
901: for a fixed such bound,
902: $$d_Y(P_+,P_-) \le d_Y(\nu_+,\nu_-) + \delta$$
903: with $\delta $ depending only on the genus of $S$,
904: provided, when $e_+$ or $e_-$ are geometrically infinite, that the
905: surface $f_\pm$ are taken sufficiently far out in the ends (for
906: geometrically finite ends this is an easier consequence of Sullivan's
907: theorem comparing $\nu_\pm$ with $\nu'_\pm$, though here we must take
908: a bit more care with the constants to make sure that $\boundary Y$
909: intersects $P_\pm$).
910: Since the right side is a priori bounded by hypothesis, we obtain
911: our desired uniform bound on $M$.
912:
913: Now let
914: $H : S\times[i,i+1]\to N_\rho$
915: be the homotopy provided by Lemma \lref{Elementary Homotopy}, where
916: $H_i \in \pleat (P_i,\rho)$. After possibly adjusting by
917: homeomorphisms of $S$ homotopic to the identity, we can piece these
918: together to a map
919: $H:S\times[0,M]\to N_\rho$.
920:
921: Assume first that $\gamma$ is not a component of any $P_i$.
922: Then according to Lemma \ref{Elementary Homotopy}, $H$ can make only
923: uniformly bounded progress through the Margulis tube $\T_\gamma$. Thus
924: $\diam \T_\gamma$ is bounded above, and $\ell_\rho(\gamma)$ is bounded
925: below, and we are done.
926:
927: Now suppose that $\gamma$ does appear in the $\{P_i\}$. Then
928: $J_\gamma$ is some subinterval of $[0,M]$ by
929: Theorem \ref{Controlled Resolution Sequences}, and we let $\alpha$ and
930: $\alpha'$ be the predecessor and successor curves to $\gamma$ in the
931: sequence. Both of them cross $\gamma$, and we have
932: by part (4) of Theorem \ref{Controlled Resolution Sequences} that
933: $d_\gamma(\alpha,\alpha')$ is uniformly approximated by
934: $d_\gamma(P_+,P_-)$ and hence uniformly bounded.
935:
936:
937: For simplicity, let us consider now the case that both $\ell_\rho(\alpha)$ and
938: $\ell_\rho(\alpha')$ are uniformly bounded above and below.
939: (There is in fact a uniform upper bound on their lengths;
940: if they become too short a small additional argument is needed).
941:
942: Let $\sigma_i \equiv \sigma_{H_i}$ and note that, by Lemma \ref{Elementary
943: Homotopy}, for all $i\in J_\gamma$
944: the annuli $\collar(\gamma,\sigma_i)$ coincide. Name this common
945: annulus $B$. Write $J_\gamma = [k,l]$, and
946: consider in $S\times[0,M]$ the solid torus
947: $$
948: U = B\times[k-1,l+1].
949: $$
950: By Lemma \ref{Elementary Homotopy},
951: this is the only part of $S\times[0,M]$ that $H$ can map more than a
952: bounded distance into $\T_\gamma$. The height of this torus, $|k-l+2|$, is
953: at most $M$ and this is uniformly bounded. The top and bottom annuli
954: $B\times\{k-1\}$ and $B\times\{l+1\}$ have uniformly bounded geometry
955: (in $\sigma_{k-1}$ and $\sigma_{l+1}$, respectively),
956: by the length bounds we've assumed on $\alpha$ and $\alpha'$.
957: We will control the size of the {\em meridian} of $U$, and this will
958: in turn bound the size of $\T_\gamma$.
959:
960:
961: Assume $\alpha$ is a geodesic in $\sigma_{k-1}$ (where we note its
962: length is bounded above),
963: and let $a=\alpha\intersect B$. Similarly
964: assume $\alpha'$ is a geodesic in $\sigma_{l+1}$
965: and let $a'=\alpha'\intersect B$. The arc $a$ may a priori be
966: long in $\sigma_{l+1}$,
967: but its length is estimated by the number
968: of times it twists around $a'$, or $d_{\AAA(B)}(a,a')$.
969:
970: A lemma in 2 dimensional hyperbolic geometry establishes
971: $$
972: |d_{\AAA(B)}(a,a') - d_\gamma(\alpha,\alpha')| \le C
973: $$
974: where this $C$ depends only on $M$, which we have already bounded
975: uniformly.
976: The idea of this is that, in each elementary move, the metric $\sigma_i$
977: changes in a bilipchitz way outside the collars
978: of the curves involved in the elementary move. From this it follows
979: that, starting with a geodesic passing through a collar, we obtain a
980: curve which does only a bounded amount of additional twisting outside
981: the collar. After $M$ such moves the relative twisting of $\alpha$ and
982: $\alpha'$ can still be estimated by their twisting inside the collar,
983: up to an additive bound proportional to $M$.
984:
985: With this estimate and the bound on $d_\gamma(\alpha,\alpha')$ in
986: terms of $d_\gamma(P_+,P_-)$,
987: we find that $a$
988: and $a'$ intersect a bounded number of times, so that the length of
989: $a$ is uniformly bounded in $S\times\{l+1\}$. It follows that the
990: meridian of $U$
991: $$
992: m = \boundary (a\times[k-1,l+1])
993: $$
994: has uniformly bounded length in the induced metric. Thus its image is
995: bounded in $N_\rho$. It therefore spans a disk of bounded diameter,
996: and in fact we can homotope $H$ on all of $U$ to a new map of bounded
997: diameter. This bounds the diameter of $\T_\gamma$ from above, and again we
998: are done.
999:
1000:
1001: \section{Conjectures}
1002: \label{conjectures}
1003:
1004: \subsection{Length estimates}
1005:
1006: The reader may have noticed that in fact the argument
1007: outlined in the previous section shows that the infimum
1008: $\ep=\inf_\gamma\ell_\rho(\gamma)$ and the supremum
1009: $D=\sup_Yd_Y(\nu_+,\nu_-)$ can be bounded one in terms of the other.
1010: That is, any positive lower bound for $\ep$ implies some upper bound for
1011: $D$ independent of $\rho$, and vice versa. Thus there is a version of
1012: the theorem which
1013: yields non-empty information for quasi-Fuchsian groups (where $\ep>0$
1014: and $D<\infty$ automatically) as well. However it would be nice to
1015: have bounds that are more specific and more explicit.
1016:
1017: ``More specific'' means that we would like to know an estimate on
1018: $\ell_\rho(\gamma)$ for {\em a particular $\gamma$}. In
1019: \cite{minsky:kgcc} we actually show that for any subsurface $Y$,
1020: a large lower bound on $d_Y(\nu_+,\nu_-)$ implies a small upper bound
1021: for $\ell_\rho(\boundary Y)$. In the other direction something more
1022: complicated would need to be stated, since any curve $\gamma$ can be a
1023: boundary curve for many different subsurfaces.
1024:
1025: ``More explicit'' means we would like to know the estimate itself more
1026: explicitly. Furthermore it would be nice to estimate the {\em complex
1027: translation length} $\lambda$ and not just its real part $\ell$.
1028: In \cite{minsky:torus} this was done for the punctured-torus
1029: case. Here is a possible generalization, stated again in the
1030: case of $\rho$ with no externally short curves.
1031:
1032: \begin{conjecture}{Complex translation length}
1033: Let $\rho$ be a Kleinian surface group with no externally short curves.
1034: There exist $K,\ep>0$ depending only on the genus of $S$
1035: such that
1036: $$
1037: \ell_\rho(\gamma)>\ep \implies \sup_{\gamma\subset Y} d_Y(\nu_+,\nu_-)
1038: < K.
1039: $$
1040: Conversely, if $\sup_Y d_Y(\nu_+,\nu_-) \ge K$ then
1041: $$
1042: \frac{2\pi i}{\lambda_\rho(\gamma)} \asymp
1043: d_\gamma(\nu_+,\nu_-)
1044: +
1045: i
1046: {\sum^\sim_{\substack{Y\subset S\\
1047: \gamma\subset\boundary Y\\
1048: Y\not\sim\gamma}}} d_Y(\nu_+,\nu_-)
1049: $$
1050: \end{conjecture}
1051: Let us explain the notation used here.
1052: The expression $\displaystyle\sum^\sim_{x\in X} f(x)$ denotes
1053: $$
1054: 1 + \sum_{\substack{x\in X\\ f(x) \ge K}} f(x)
1055: $$
1056: where $K$ is our a-priori ``threshold'' constant.
1057: Our sum then is over all subsurfaces whose boundary contains the
1058: isotopy class of $\gamma$, except for the annulus homotopic to
1059: $\gamma$, excluding those where $d_Y(\nu_+,\nu_-)$ is below $K$.
1060: Both sides of the ``$\asymp$'' symbol are points in the
1061: upper half plane of $\C$, and we take ``$\asymp$'' to mean that the
1062: hyperbolic distance between them is bounded by an a-priori constant
1063: $D_0$. Implicit in the statement is that it holds for some $D_0$
1064: which depends only on the genus of $S$.
1065:
1066: The significance of the hyperbolic distance estimate on
1067: $2\pi i/\lambda(\gamma)$ is that we can interpret $2\pi
1068: i/\lambda(\gamma)$ as a
1069: Teichm\"uller parameter for the Margulis tube $\T_\gamma$, as follows
1070: (cf. \cite{minsky:torus} and McMullen \cite{mcmullen:cusps}).
1071: Normalize $\rho(\gamma)$ so that it acts on $\hat\C$ by $z\mapsto e^\lambda
1072: z$. The quotient $(\C\setminus\{0\})/\rho(\gamma)$ is then a torus, and there
1073: is a preferred marking of this torus by the pair $(\hat\gamma,\mu)$,
1074: where $\mu$ is the meridian of the torus, or the image of the unit
1075: circle in $\C$, and $\hat\gamma$ is the image of the curve
1076: $\{e^{t\lambda}:t\in[0,1]\}$. Note, this curve depends on the choice
1077: of $\lambda$ mod $2\pi i$. In \cite{minsky:torus} we point out that
1078: if $\ell_\rho(\gamma)$ is sufficiently short then we can
1079: choose $\hat\gamma$ to be a minimal representative of $\gamma$ on
1080: the torus just by choosing $\theta = \Im\lambda \in [0,2\pi)$.
1081:
1082: The quantity $2\pi i/\lambda$ turns out to be the point in the upper
1083: half-plane representation of the Teichm\"uller space of the torus
1084: which represents the marked quotient torus. Estimating this quantity
1085: up to bounded hyperbolic distance is then equivalent to estimating the
1086: torus structure up to bounded Teichm\"uller distance, which
1087: corresponds to knowing the action of $\rho(\gamma)$ up to bounded
1088: quasi-conformal conjugacy. This finally is equivalent to bounded
1089: bilipchitz conjugacy of the action on $\Hyp^3$, and thus is the
1090: ``right'' kind of estimate if we are interested in knowing the
1091: quotient geometry up to bilipschitz equivalence.
1092:
1093: The imaginary part of the conjectural estimate is supposed to estimate
1094: the ``height'' of the margulis tube boundary for $\gamma$, and its
1095: real part is supposed to measure the ``twist'' of the meridian around
1096: $\hat \gamma$. In our discussion of the Bounded Geometry Theorem,
1097: we essentially showed that the height was bounded by the
1098: number of elementary moves it took to pass $\T_\gamma$, and the
1099: twisting was bounded by the relative twisting of the predecessor and
1100: successor curves $\alpha$ and $\alpha'$. In general we expect that
1101: large values of $d_Y(\nu_+,\nu_-)$ with $\gamma\subset\boundary Y$
1102: will contribute to parts of the elementary move sequence that make
1103: progress along the sides of $\T_\gamma$, and thus give a good estimate
1104: for its height.
1105:
1106: In \cite{minsky:torus} we obtained a similar estimate for the case
1107: where $S$ is a once-punctured torus. (In this case we are not
1108: requiring $S$ to be closed, and our representations must satisfy the
1109: added condition that the conjugacy class corresponding to loops around
1110: the puncture is mapped to parabolics.) Let us state this just in the
1111: case that $\nu_\pm$ are both laminations. For the torus, a
1112: lamination are determined by its
1113: slope in $H_1(S,\R) = \R^2$, which takes values in
1114: $\hhat\R = \R\union\{\infty\}$.
1115: Simple closed curves correspond to rational points.
1116: For any simple closed curve $\alpha$ we defined a quantity
1117: analogous to $d_\alpha(\nu_-,\nu_+)$ as follows: after an appropriate
1118: basis change for $S$ (or equivalently action by an element of
1119: $\SL 2(\Z)$, we may assume that $\alpha$ is represented by $\infty$,
1120: and let $\nu_-(\alpha),\nu_+(\alpha)$ be the irrational numbers
1121: representing the ending laminations. Then define
1122: $$w(\alpha) = \nu_+(\alpha)-\nu_-(\alpha).$$
1123: We showed that $\ell_\rho(\alpha)$ can only be short if $w(\alpha)$ is
1124: above a uniform threshold, and in this case we estimated
1125: $$
1126: \frac{2\pi i}{\lambda_\rho(\alpha)} \asymp w(\alpha) + i.
1127: $$
1128: In fact $w(\alpha)$ is just a measure of relative twisting of $\nu_-$
1129: and $\nu_+$ around $\alpha$, and it is not hard to see that $|w(\alpha)|$
1130: is estimated by our $d_\alpha(\nu_-,\nu_+)$, up to a uniform additive
1131: error. Thus, this is really the same estimate as in Conjecture
1132: \ref{Complex translation length}, since there are no essential subsurfaces
1133: in $S$ other than annuli.
1134:
1135: \subsection{General representations}
1136:
1137: All the methods that we have presented here depend heavily on the
1138: assumption that $\rho$ is both faithful and discrete. It can be
1139: argued, however, that a full understanding of the deformation space of
1140: hyperbolic structures on a manifold would require some better geometric
1141: description of the whole representation variety, including indiscrete or
1142: non-faithful points, and it is tempting to try to enlist the complex
1143: of curves for this purpose.
1144:
1145: The only results I know that offer any hope are in a paper of
1146: Bowditch \cite{bowditch:markoff}, in which he studies general
1147: representations for the once-punctured torus (again with the
1148: parabolicity condition for the puncture). Such a representation
1149: determines a trace (closely related to complex translation length) for
1150: every conjugacy class, and in particular for the simple closed
1151: curves, which in this case correspond to $\Q\union\{\infty\}$, viewed as
1152: the vertices of the Farey tesselation of the disk. To every triangle and
1153: adjacent pair of triangles is associated a relation among the traces
1154: of the vertices, coming from the standard trace identities in $\SL
1155: 2(\C)$. Bowditch uses these relations alone, without discreteness, to
1156: analyze the global properties of the trace function, in particular obtaining
1157: a connectedness property for sublevel sets closely analogous to the
1158: quasi-convexity
1159: property of Lemma \ref{Quasiconvexity}. Using this he is able to
1160: define an invariant of the representation that generalizes the ending
1161: lamination for discrete representations; but it is hard to know how to
1162: extract more information from this invariant.
1163:
1164: In the higher genus case, no such analysis has been done, and it would
1165: be very interesting to try it. Elementary moves between pants
1166: decompositions still give rise to trace identities among the curves
1167: involved, although they are a bit more complicated. One wonders at
1168: least whether a result like Lemma \ref{Quasiconvexity} can be
1169: generalized to all representations.
1170:
1171: Bowditch is led to the following question: Consider the quantity
1172: $$
1173: \frac{\ell_\rho(\gamma)}{\ell_{\rho_0}(\gamma)}
1174: $$
1175: where $\rho_0$ is some fixed Fuchsian representation, $\rho$ is a
1176: general representation, and $\gamma$ is a non-trivial element of $\pi_1(S)$.
1177: The infimum of this ratio is positive for quasi-Fuchsian
1178: representations. For a non-quasi-Fuchsian discrete, faithful
1179: representation, the infimum is 0, and can be achieved by considering
1180: only $\gamma$ with {\em simple} representatives.
1181: The limit points of minimizing sequences in the space of laminations
1182: give the ending laminations for $\rho$.
1183:
1184: If $\rho$ is indiscrete or non-faithful the infimum is
1185: again 0 (indeed $\inf\ell_\rho$ is 0 as well), but the question is,
1186: is the infimum also 0 for the simple elements. In
1187: other words:
1188:
1189: \begin{question}
1190: Let $S$ be a closed surface of genus at least 2, and let
1191: let $\rho:\pi_1(S) \to \PSL 2(\C)$ be a representation. If
1192: $$
1193: \inf
1194: \frac{\ell_\rho(\gamma)}{\ell_{\rho_0}(\gamma)}
1195: > 0
1196: $$
1197: where $\gamma$ varies over all simple loops in $S$, must
1198: $\rho$ be quasi-Fuchsian?
1199: \end{question}
1200:
1201: This question appears to be difficult, and a positive answer
1202: would be a good starting point in using the complex of curves to
1203: analyze general representations. To indicate its difficulty, note
1204: that it is closely related to the following:
1205:
1206: \begin{question}
1207: If $\rho:\pi_1(S) \to \PSL 2(\C)$ is any representation with
1208: non-trivial kernel, does the kernel contain elements represented by
1209: simple loops?
1210: \end{question}
1211: A positive answer to this question is at least as hard to prove
1212: as the simple loop conjecture for
1213: hyperbolic 3-manifolds; see Gabai \cite{gabai:simple} and Hass
1214: \cite{hass:simple}.
1215:
1216: \input{bib}
1217:
1218: % \bibliographystyle{amsplain}
1219: % \bibliography{math}
1220:
1221: \end{document}
1222: