The crossing number of satellite knots (Q740544)

A knot \(K\) is a satellite of another knot \(L\) if, up to isotopy, it lies in a neighbourhood of \(L\) in a non-trivial way. The crossing number \(c(K)\) of a knot \(K\) is the minimal number of crossings in a diagram of \(K\). It has been conjectured that if \(K\) is a satellite of \(L\) then \(c(L)\leq c(K)\). In this paper, it is proved that this relationship holds up to a constant factor (more precisely, that \(c(L)\leq 10^{13} c(K)\)). The author notes that the method of proof has been chosen for simplicity rather than to minimize the bound. However, it does not seem possible that this method would yield the full conjecture. The proof builds on earlier work of the author [J. Topol. 2, No. 4, 747--768 (2009; Zbl 1190.57003)] that addresses the same question in the case that \(K\) is a connected sum with \(L\) as one summand. In fact, this earlier result is also used in this proof. A key special case that has to be considered separately is when \(K\) is a \(2\)--cable of \(L\). The large-scale idea of the proof is to find a `short' arc within a handle structure on a particular \(3\)--manifold. In most cases the arc is found using a normal surface. Although such a surface will in general have high complexity relative to the handle structure, the idea of `generalised parallelity bundles' allows the author to detect regions to avoid, so that the chosen path lies within a sufficiently small subsurface. Another important case is finding a `short' core curve in a handle structure for a solid torus. This is addressed in another paper by the author [Trans. Am. Math. Soc. 366, No. 11, 6027--6050 (2014; Zbl 1321.57024)]. In addition to the main result, the author gives two alternative versions that have smaller bounds but are weaker in other ways.