A full-twist inequality for the \(\nu^{+}\)-invariant (Q724261)

Let \(K\) be a knot and \(D\subseteq S^3\) be a disk with boundary disjoint from \(K\). Let \(n\) be the algebraic intersection of \(K\) and \(D\). Performing \((-1)\)-framed surgery along the boundary of \(D\) affects \(K\) by putting a positive full twist in the strands of \(K\) passing through \(D\). If \(J\) is the resulting knot then \(K\) is said to deform to \(J\) by a positive full twist with \(n\)-linking. The aim of this paper is a bound controlling the difference \(J\#(-K)\). The resulting theorem is stated in terms of an invariant \(\nu^+\) due to \textit{J. Hom} and \textit{Z. Wu} [J. Symplectic Geom. 14, No. 1, 305--323 (2016; Zbl 1348.57023)]. This concordance invariant of knots in \(S^3\) takes values in the set of non-negative integers. The main result of the paper (Theorem 1.1) says that if \(K\) deforms to \(J\) by a positive full twist with \(n\)-linking, then \(\dfrac{(n-1)(n-2)}{2}\leq \nu^+(J\#(-K))\leq \dfrac{(n)(n-1)}{2}\). When \(n=0\), \(\nu^+(J\#(-K)) = 0\). The vanishing of \(\nu^+\) implies that many other Heegaard-Floer invariants vanish. In [loc. cit.], Hom and Wu proved that \(\nu^+\) gives a lower bound on genus, which can be made arbitrarily better than the bound coming from the \(\tau\)-invariant of \textit{P. Ozsváth} and \textit{Z. Szabó} [Geom. Topol. 7, 615--639 (2003; Zbl 1037.57027)]. As an application of Theorem 1.1, the author derives a bound on the genus of cables and shows that the set of cables whose 4-genus is detected by \(\nu^+\) is closed under the operation of cabling. Because \(\nu^+\) sends slice knots to \(0\) and is sub-additive, one can introduce an equivalence relation on knot concordance by saying that \(K\sim J\) if \(\nu^+(J\#(-K)) = \nu^+(K\#(-J))=0\). There is also a partial ordering on the resulting quotient by saying \(J\leq K\) if \(\nu^+(J\#(-K))=0\). With respect to this partial order, \(\tau\), \(\nu\), \(\gamma\), \(\epsilon\), \(\Upsilon\) and \(d\)-invariants of surgery are all monotonic. As a consequence of Theorem 1.1 this ordering is geometric, in that if \(K\) deforms to \(J\) by a positive full twist with \(n\)-linking where \(n=0\) or \(1\), then \(J\leq K\). If \(n\geq 3\), then \(J > K\). In order to better appreciate the ideas in this paper, we will summarize the properties of \(\nu^+\) used in the proof of Theorem 1.1. In [J. Reine Angew. Math. 706, 1--17 (2015; Zbl 1328.57010)], \textit{Y. Ni} and \textit{Z. Wu} defined a sequence of knot invariants \(V_0(K)\geq V_1(K)\geq V_2(K)\geq \dots\) taking values in the set of non-negative integers. \(\nu^+(K)\) is defined to be the first index \(k\) with \(V_k(K)=0\). \(V_k(K)\) can be computed in terms of the difference between \(d\)-invariants of a rational surgery on \(K\) and on the unknot [loc. cit.]. In the case that \(K\) deforms to \(J\) by a positive full twist with \(n\)-linking then \(J\#(-K)\) bounds a disk in a punctured \(\overline{\mathbb{CP}^2}\). Performing surgery on this disk, the authors find a rational homology ball bounded by surgery on \(J\#(-K)\) and use it to deduce that certain \(d\)-invariants vanish. This allows them to compute \(V_k(J\#(-K))\) for some choices of \(k\), and so allows them to control \(\nu^+\).