The length of unknotting tunnels (Q969657)

A compact orientable \(3\)-manifold \(M\) with torus boundary component has tunnel number one if it contains a properly embedded arc \(\tau\) whose exterior is a handlebody. Such an arc is called an unknotting tunnel of \(M\). When \(M\) is hyperbolic, there exists a unique geodesic arc homotopic to \(\tau\). If \(\tau\) runs between distinct boundary components, then \textit{C. Adams} [Math. Ann. 302, No.~1, 177--195 (1995; Zbl 0830.57009)] showed that its geodesic representative has bounded length. Then he asked if an unknotting tunnel in a hyperbolic \(3\)-manifold always has bounded length. First, \textit{C. C. Adams} and \textit{A. W. Reid} [Comment. Math. Helv. 71, No.~4, 617--627 (1996; Zbl 0873.57006)] got an affirmative answer for two-bridge knot complements. \textit{H. Akiyoshi, Y. Nakagawa} and \textit{M. Sakuma} [KNOTS'96. Proceedings of the international conference and workshop on knot theory, Tokyo, Japan, July 22--26, 1996. Singapore: World Scientific. 433--448 (1997; Zbl 0960.57013)] confirmed it for punctured torus bundles. The main result of the paper under review is to give a negative answer to that question. That is, there exist finite volume one-cusped hyperbolic tunnel number one manifolds for which the geodesic representative of the unknotting tunnel is arbitrarily long, as measured between the maximal horoball neighborhood of the cusp. The construction starts with maximally cusped geometrically finite uniformizations of a compression body whose boundary consists of a torus and a genus two surface and a genus two handlebody, and glues them via an isometry, then performs Dehn surgeries. Finally, this construction is modified to give hyperbolic knots in homology \(3\)-spheres whose unknotting tunnel is arbitrarily long. The existence of such a knot in the standard \(3\)-sphere is open.