Geodesic systems of tunnels in hyperbolic 3-manifolds (Q2441264)

A system of unknotting tunnels for a 3-manifold \(M\) with torus boundary consists of \(n\) embedded arcs, with endpoints on the boundary of \(M\), such that the complement of the arcs is homeomorphic to a handlebody (of genus \(n+1\), giving a Heegaard splitting of the 3-manifold); the minimal possible \(n\) is called the \textit{tunnel number} of the 3-manifold. For hyperbolic 3-manifolds of tunnel number one it is not known, in general, whether every unknotting tunnel is isotopic to a geodesic (this is known in special cases, for example for the complements of 2-bridge knots and for hyperbolic 3-manifolds with two boundary components). The present paper offers some evidence that the analogue for hyperbolic 3-manifolds of tunnel number \(n>1\) may not be true, by showing that there exist finite-volume hyperbolic 3-manifolds with a single cusp and a system of \(n\) unknotting tunnels, \(n-1\) of which are homotopic to geodesics that come arbitrarily close to self-intersecting. In order to show this result, the authors prove that, for \(n \geq 2\), there is a geometrically finite hyperbolic structure on a \((1;n+1)\)-compression body with a system of \(n\) core tunnels, \(n-1\) of which are homotopic to self-intersecting geodesics and hence cannot be isotopic to simple geodesics (Lackenby and Purcell conjectured that for the (1;2)-compression body, core tunnels are always isotopic to geodesics); this is proved by studying Ford domains of geometrically finite structures on such compression bodies.