Tunnel number degeneration under the connected sum of prime knots (Q387962): Difference between revisions

Given a knot \(K\) in \(S^3\), an unknotting tunnel system for \(K\) is a collection of arcs \(t_1,t_2,\dots ,t_n\) properly embedded in the exterior of \(K\) such that \(S^3 - N(K\cup t_1\cup t_2\cup\dots \cup t_n)\) is a handlebody. The minimal number of arcs in an unknotting tunnel system for \(K\), denoted \(t(K)\), is called the tunnel number of \(K\) and is a knot invariant. \textit{T. Kobayashi} and \textit{Y. Rieck} [Commun. Anal. Geom. 14, No. 5, 1037--1077 (2006; Zbl 1113.57006)] showed that tunnel number can degenerate arbitrarily under connected sum, that is, they showed that for any \(n\), there exist knots \(K_1\) and \(K_2\) such that \(t(K_1\# K_2)\leq t(K_1)+t(K_2)-n\). However, all of these examples were composite knots and it was conjectured by \textit{Y. Moriah} [Geom. Topol. Monogr. 12, 191--232 (2007; Zbl 1138.57008)] that such behavior does not occur for prime knots and in particular, if \(K_1\) and \(K_2\) are prime, then \(t(K_1\# K_2) \geq t(K_1)+t(K_2)-1\). This paper disproves this conjecture by showing that there are infinitely many pairs of prime knots \(K_1\) and \(K_2\) such that \(t(K_1\# K_2)=t(K_1)+t(K_2)-2\). This follows from an investigation of 2--string free tangle decompositions of knots with tunnel number two.