Surgery obstructions from Khovanov homology (Q438619): Difference between revisions

In this paper, the author shows that if the twofold branched cover of the complement of a strongly invertible knot admits a filling with finite fundamental group, then there is a rational closure of its branched set which has thin reduced Khovanov homology. Khovanov homology is a link invariant which categorifies the Jones polynomial. As the latter, its relationship with geometric properties of links remains rather confused. This paper brings some new light in this direction by pointing out some computable obstructions to lens and finite surgery arising from the width, that is the difference between maximal and minimal degrees of non vanishing homology, of some Khovanov homologies associated to strongly invertible knots (under some technical assumptions). The first section is introductory. The second section recalls the definition of reduced Khovanov homology over \(\mathbb{F}_2\) and the mapping cone presentation of it. It defines also a signature-related normalization of Khovanov grading which suits well the mapping cone presentation when the triplet is ``quasi-alternating''--like. In the third section, the author establishes the relationship between rational surgery on the complement of a strongly invertible knot (and more generally on a strongly invertible knot manifold) and the twofold branched cover over the different rational closure of an associated tangle. So one can consider the Khovanov homology -- and its width -- of the closed tangle associated to a surgery on a strongly invertible knot. In section four, integer surgery is shown to induce a simple mapping cone process at the level of associated Khovanov homology. It is then used to prove that the Khovanov width associated to the different integer surgeries on a strongly invertible knot differ at most by one. Moreover Khovanov width associated to a rational surgery is bounded above by the maximal width associated to integer surgeries. Along the way, it is also proven that the unknot is the only strongly invertible knot with Khovanov width 1 associated to every large integer surgery, and that it is bounded by 2 for every surgery on a Berge knot. On the other side, section five proves that, under some technical assumption, the Khovanov width associated to a rational surgery is also bounded below by the minimal width associated to integer surgeries. Section six is devoted to bounding by 1 the width associated to a lens surgery and by 2 the width associated to a finite surgery. The obstruction results follow then. This can be compared with obstructions arising from Heegaard--Floer homology and their counterparts on the Alexander polynomial [\textit{P. Ozsváth} and \textit{Z. Szabó}, Topology 44, No. 6, 1281--1300 (2005; Zbl 1077.57012)]. In the last section and as applications, the author recovers the fact that the figure eight knot, the \((-2,p,p)\)--pretzel knots for \(5\leq p\leq 31\) and the knot \(14^n_{11893}\) have no finite filling. Note that the latter couldn't be ruled out just using the Alexander polynomial.