Local topology in deformation spaces of hyperbolic 3-manifolds (Q633897)

Let \(\mathrm{AH}(M)\) denote the deformation space of marked hyperbolic \(3\)-manifolds homotopy equivalent to a compact \(3\)-manifold \(M\). An element of \(\mathrm{AH}(M)\) can be viewed as a faithful and discrete representation \(\rho\) of \(\pi_1(M)\) into \(\mathrm{PSL}_2({\mathbb C})\), the group of motions on the hyperbolic \(3\)-space \({\mathbb H}^3\). If \(M=S\times I\) is the product of the orientable closed surface \(S\) of genus \(>1\) and the unit interval \(I\), then \(\mathrm{int}(\mathrm{AH}(M))={\mathcal QF}(S)\) is the quasi-Fuchsian locus. Let \({\mathcal T}(S)\) denote the Teichmüller space of \(S\). There exists a homeomorphism \(Q: {\mathcal T}(S)\times {\mathcal T}(S)\to {\mathcal QF}(S)\) such that for \(Y\) fixed, \(X\mapsto Q(X,Y)\) is the Bers embedding [\textit{L. Bers}, Bull. Am. Math. Soc. 66, 94--97 (1960; Zbl 0090.05101)]. The set \(B_Y=Q({\mathcal T}(S)\times\{Y\})\) is called a Bers slice. As phenomena of ``bumping'' and ``self-bumping'' were discovered, the picture of the boundary of \(\mathrm{AH}(M)\) and that of \(B_Y\) were expected to be very wild. The fact that a deformation space of once-punctured torus groups is not locally connected also supports this expectation [\textit{K. W. Bromberg}, Duke Math. J. 156, No. 3, 387--427 (2011; Zbl 1213.30078)]. Two distinct components of \(\mathrm{int}(\mathrm{AH}(M))\) are said to bump at \(\rho\in \partial \mathrm{AH}(M)\) if their closures intersect at \(\rho\), and a component \(B\) is said to self-bump at \(\rho\) if \(\rho\in \partial B\) and every sufficiently small neighborhood of \(\rho\) intersects \(B\) in a disconnected set. The purpose of this paper is to show that there is no bumping or self-bumping at boundary points which are characterized in certain ways. Theorem 1.1. If the boundary of \(M\) is incompressible and \(\rho\in\partial \mathrm{AH}(M)\) is such that every parabolic element in \(\rho(\pi_1(M))\) lies in a rank-two free abelian group, then there is no bumping or self-bumping at \(\rho\). This theorem implies that \(\mathrm{AH}(M)\) is locally connected at generic boundary points. Theorems 1.2. There is no bumping at every quasiconformally rigid point of \(\partial \mathrm{AH}(M)\). Theorem 1.3. If \(M\) is acylindrical or homeomorphic to \(S\times I\), then there is no self-bumping at every quasiconformally rigid point in \(\partial \mathrm{AH}(M)\). Theorem 1.5. At every quasiconformally rigid boundary point of a Bers slice \(B\subset {\mathcal QF}(S)\) there is no self-bumping and, in particular, the closure \(\overline{B}\) is locally connected at \(\rho\). The proofs of these theorems are based on the recent advance in the understanding of geometry and topology of \(3\)-manifolds, to which the authors of the present paper have made a great contribution. In particular, the ending lamination theorem [\textit{Y. N. Minsky}, Ann. Math. (2) 171, No. 1, 1--107 (2010; Zbl 1193.30063); \textit{J. F. Brock}, \textit{R. D. Canary} and \textit{Y. N. Minsky}, ``The classification of Kleinian surface groups. II: The ending lamination conjecture'', Preprint, \url{arXiv:math/0412006}] and the geometry of curve complexes [\textit{H. A. Masur} and \textit{Y. N. Minsky}, Invent. Math. 138, No. 1, 103--149 (1999; Zbl 0941.32012)] present key tools in this work. Let us take Theorem 1.5 for the case \(M=S\times I\) to illustrate the roles played by these key tools in its proof. For a subsurface \(W\subset S\), let \({\mathcal C}(W)\) denote the curve complex of \(W\) and \({\mathcal C}(S,W)\) the set of curves in \(S\) which intersect \(W\) essentially. Then there is a naturally defined map \(\pi_W: {\mathcal C}(S,W)\to {\mathcal C}(W)\) called the subsurface projection map. For \(X\in {\mathcal T}(S)\), let \(\alpha\) be one of the shortest geodesics which meet \(W\) essentially. Then let \(\pi_W(X)\) mean \(\pi_W(\alpha)\). (When \(W\) is an annulus we need a modification for the definition of \(\pi_W\).) The subsurface projection is important, because, for a simple closed curve \(\gamma\), that \(\ell_{\gamma}(Q(X,Y))\) is very small implies that the quantity \[ m_{\gamma}(X,Y)=\max\left(\frac{1}{\ell_{\gamma}(X)},\frac{1}{\ell_{\gamma}(Y)},\sup_{\gamma\subset\partial W}d_W(\pi_W(X),\pi_W(Y))\right) \] is very large, where \(W\) ranges over all essential subsurfaces of \(S\) whose boundary contains a curve parallel to \(\gamma\). Here \(d_W\) is the metric on the curve complex \({\mathcal C}(W)\), and for \(N=Q(X,Y)\), \(X\) or \(Y\), \(\ell_{\gamma}(N)\) denotes the length of the geodesic representative of \(\gamma\) in \(N\) [Brock, Canary and Minsky, loc. cit.] Let \(\rho\) be a quasiconformally rigid point in \(\partial \mathrm{AH}(S\times I)\). Then the proof of Theorem 1.5 is done by showing that for every sufficiently small neighborhood \(U\) of \(\rho\) there exists a smaller neighborhood \(U'\) such that any two points in \(U'\) can be connected by a path in \(U\). Let \({\pmb\alpha}=\{\alpha_1,dots ,\alpha_p\}\) be a curve system on \(S\) associated to upward-pointed cusps and \({\pmb\beta}=\{\beta_1,\dots ,\beta_q\}\) the one associated to downward-pointed cusps. Let \(\{S_i\}\) and \(\{T_k\}\) be the collections of components of \(S\setminus {\pmb\alpha}\) and \(S\setminus {\pmb\beta}\), respectively, which are not thrice-punctured spheres. By the ending lamination theorem, there exist ending laminations \(\lambda_i\) on \(S_i\) and \(\mu_k\) on \(T_k\) such that \(\rho\) is completely determined by \({\pmb\alpha}\), \({\pmb\beta}\), \(\{\lambda_i\}\) and \(\{\mu_k\}\). If \(W =S_i\) or \(W=T_k\), then an ending lamination on \(W\) is identified with a point in the Gromov boundary \(\partial_{\infty}{\mathcal C}(W)\). Choose collections \({\mathbb U}=\{U_i\}\) and \({\mathbb V}=\{V_k\}\) of neighborhoods \(U_i\) of \(\lambda_i\) and \(V_k\) of \(\mu_k\) in the curve complexes \({\mathcal C}(S_i)\) and \({\mathcal C}(T_k)\). For \({\mathbb U}\), \({\mathbb V}\) and \(\delta>0\) define the set \({\mathcal U}(\delta,{\mathbb U},{\mathbb V})\) of those \(Q(X,Y)\) satisfying the following conditions : (1) \(\pi_{S_i}(X)\in U_i\) for all \(i\), (2) \(\ell_{\alpha_j}(Q(X,Y))<\delta\) for all \(\alpha_j\in {\pmb\alpha}\), (3) \(\pi_{T_k}(Y)\in V_k\) for all \(k\) and (4) \(\ell_{\beta_j}(Q(X,Y))<\delta\) for all \(\beta_j\in {\pmb\beta}\), (5) the geodesic representative \(\alpha_i^{\ast}\) of \(\alpha_i\) can be homotoped toward the upward-pointing end without being hindered by \(\alpha_j^{\ast}\) (\(j>i\)) and all \(\beta_k^{\ast}\), and \(\beta_k^{\ast}\) can be homotoped toward the downward-pointing end without being hindered by \(\beta_l^{\ast}\) (\(l>k\)) and all \(\alpha_j^{\ast}\). Roughly speaking, \(\alpha_1,\dots ,\alpha_p\), \(\beta_q,\dots ,\beta_1\) are arranged according to their ``closedness'' to the upward-pointing end (or equivalently ``remoteness'' from the downward-pointing end). Then Lemma 9.1 asserts that the sets \({\mathcal U}(\delta,{\mathbb U},{\mathbb V})\) are the intersections with \({\mathcal QF}(S)\) of a (not necessarily \textit{open}) neighborhood systems for \(\rho\). For a sufficiently small \(\epsilon>0\) and some small neighborhoods \({\mathbb U}'\subset {\mathbb U}\) and \({\mathbb V}'\subset {\mathbb V}\), any two points in the portion \({\mathcal W}(\epsilon, {\mathbb U}',{\mathbb V}')\) of \({\mathcal U}(\epsilon,{\mathbb U}',{\mathbb V}')\) consisting of those \(Q(X,Y)\) with \(\ell_{\alpha_j}(X)<\epsilon\) for all \(\alpha_j\in {\pmb\alpha}\) and \(\ell_{\beta_k}(Y)<\epsilon\) for all \(\beta_k\in {\pmb\beta}\) are connected by a path in \({\mathcal U}(\delta,{\mathbb U},{\mathbb V})\). One may guess this fact by anticipating that the portion looks like \({\mathcal T}_0({\pmb\alpha})\times {\mathcal T}_0({\pmb\beta})\) for \(\epsilon\) very small. Here \({\mathcal T}_0({\pmb\gamma})\) is the stratum of Riemann surfaces with nodes corresponding to curves in \({\pmb \gamma}\) in the augmented Teichmüller space \(\hat{\mathcal T}(S)\), and, being a product of (lower dimensional) Teichmüller spaces, it is path-connected. However, to make the argument precise, we need a result about a relation between Teichmüller geodesics and geodesics in the curve complex in the paper by Masur and Minsky cited above. If smaller \(\epsilon'<\epsilon\) and neighborhoods \({\mathbb U}''\subset {\mathbb U}'\) and \({\mathbb V}''\subset {\mathbb V}'\) are suitably chosen, any point \(Q(X,Y)\) in \({\mathcal U}(\epsilon',{\mathbb U}'',{\mathbb V}'')\) is connected to a point in \({\mathcal W}(\epsilon, {\mathbb U}',{\mathbb V}')\) by a path in \({\mathcal U}(\delta,{\mathbb U},{\mathbb V})\). The desired path is obtained by concatenation of a finite number of paths: along the first path the length \(\ell_{\alpha_1}(X)\) of \(\alpha_1\) in \({\pmb\alpha}\) closest to \(X\) shrinks to \(\epsilon'\) while other data \(\{\ell_{\alpha_j}(X)\}_{j>1}\), \(\{\ell_{\beta_l}(Y)\}\), \(\{\pi_{S_i}(X)\}\) and \(\{\pi_{T_k}(Y)\}\) change by a bounded amount, and then along the second path the length \(\ell_{\alpha_2}(X)\) shrinks to \(\epsilon'\) without disrupting other data, and so on. Then top conformal boundary \(X\) is deformed in \(B_Y\) to a surface \(X_T\) on which \(\ell_{\alpha_i}(X_T)<\epsilon\) for all \(\alpha_j\in {\pmb\alpha}\). After these paths follow paths along which \(\ell_{\beta_k}(Y)\) shrinks to \(\epsilon'\) successively. Those paths are found in the images of the Fenchel-Nielsen coordinates \(\Phi_{{\pmb\gamma}}: {\mathbb R}^m\times{\mathbb R}^m_+\to {\mathcal T}(S)\) associated to some (not necessarily maximal) curve systems \({\pmb\gamma}=(\gamma_1,...,\gamma_m)\). In Section 5, \(\Phi_{{\pmb\gamma}}\) is carefully constructed so that its image contains an arbitrary prescribed point of \({\mathcal T}(S)\) and so that its subsurface projections along annuli with cores \(\gamma_j\) and subsurfaces in the complement of \(S\setminus{\pmb\gamma}\) are well controlled. The authors conjecture that \(\mathrm{AH}(M)\) is not locally connected if \(M\) has a boundary component of genus at least \(2\).