A proof of the finite filling conjecture (Q1609840)

The Finite Filling Conjecture of C. McA. Gordon states that if \(M\) is a \(3\)-manifold with boundary a single torus and the interior of \(M\) is hyperbolizable, then there are at most \(5\) Dehn fillings of \(M\) whose fundamental groups are finite or infinite cyclic. Moreover, defining the distance \(\Delta(p/q,r/s)\) between two filling slopes to be \(|ps-rq|\), the distance between any two of these exceptional slopes is at most \(3\), and there is at most one pair at distance \(3\). In this important paper, the authors prove the Finite Filling Conjecture. The proof uses methodology pioneered and extensively developed by M. Culler, J. Morgan and P. Shalen, and further developed by D. Cooper, D. D. Long, and the authors, among others. The proof provides additional information about finite filling slopes (those that produce a manifold with finite fundamental group). For example, when \(M\) is the exterior of a knot \(K\) in \(S^3\), so that there is a well-defined meridian slope \(\mu_K\), there is at most one finite filling slope \(r\) satisfying \(\Delta(r,\mu_K)\geq 2\), and if there is one, then \(\Delta(r,\mu_K)= 2\). If moreover \(K\) is amphichiral, then the only possible finite filling slopes are \(\pm 1\), and the only finite fundamental group possible from such a filling is the binary icosahedral group. Much information of a more technical nature is obtained, as well, especially about the \(A\)-polynomial which we discuss below. The approach starts from the variety of characters of \(\text{SL}(2,\mathbb{C})\)-representations of \(\pi_1(M)\). Any algebraic component of it that contains the character of a discrete, faithful representation is a curve, and determines a Culler-Shalen norm on \(H_1(\partial M; \mathbb{R})\). Roughly speaking, the norm of a primitive element \(\alpha\in H_1(\partial M; \mathbb{R})\) corresponding to a slope \(r\) measures the number of characters of representations of \(\pi_1(M(r))\) to \(\text{SL}(2,\mathbb{C})\), where \(M(r)\) is the manifold obtained by Dehn filling using the slope \(r\). The unit ball \(B\) for this norm encodes a great deal of information about \(M\). For example, its vertices are rational multiples of loops that occur in the boundaries of certain essential surfaces in \(M\). Previous work of the authors proved the Finite Filling Conjecture in all cases except when the norm is one of a finite number of explictly given norms. In this paper, the authors use (a slight variant of) the \(A\)-polynomial, due to \textit{D. Cooper, M. Culler, H. Gillet, D. D. Long}, and \textit{P. B. Shalen} [Invent. Math. 118, 47-84 (1994; Zbl 0842.57013)] to resolve the remaining cases. The \(A\)-polynomial \(A_M\) of \(M\) is a defining polynomial for a curve obtained from the \(\text{SL}(2,\mathbb{C})\)-representations of \(\pi_1(M)\) that are upper triangular on \(\pi_1(\partial M)\). It is a polynomial in two variables \(u\) and \(v\), and the Newton polygon \(N_M\) of \(A_M\) is then defined to be the convex hull of the pairs \((m,n)\) such that \(u^mv^n\) has a nonzero coefficient in \(A_M\). Using \textit{P. Shanahan}'s [Topology Appl. 108, 7-36 (2000; Zbl 0963.57009)] concept of width, the authors establish that \(N_M\) is dual to \(B\) in the sense that the line through any pair of antipodal vertices of \(B\) is parallel to a side of \(N_M\), and the same holds with the roles of \(B\) and \(N_M\) interchanged. The authors first show that a finite filling slope leads to a specialization of the variables of \(A_M\) to a polynomial all of whose roots are \(\pm 1\) or certain roots of unity. Then, using known results about \(A\)-polynomials, and their own technical advances, they are able to deduce the Finite Filling Conjecture in all except one case. For that one remaining norm, they use a special argument to show that the \(A\)-polynomial does not actually occur as an \(A\)-polynomial of any \(M\). In a closing section, the authors apply their theory to small hyperbolic knots (those whose complements contain no closed incompressible surface), and deduce the result of \textit{C. Delman} [Topology Appl. 63, 201-221 (1995; Zbl 0873.57007)] and D. Tanguay that the only finite filling slope for a hyperbolic \(2\)-bridge knot is the slope of the meridian.