On surgery curves for genus-one slice knots (Q387724)

This paper investigates a 40 year old question concerning the existence of an essential smoothly slice curve on a genus \(1\) Seifert surface bounded by a smoothly slice knot. Such a curve need not exist in the topological category, but seems to exist invariably in the the smooth category. The question is: does it always exist? For example \textit{R. Kirby} asked in [Proc. Symp. Pure Math., Vol. 32, Part 2, 273--312 (1978; Zbl 0394.57002)]: Is it true that the untwisted Whitehead double of a knot \(J\) is smoothly slice if and only if \(J\) is? If the genus \(1\) knot \(K\) is only assumed algebraically slice then there are two candidate curves on the Seifert surface \(F\) to be smoothly slice. The authors call such curves surgery curves for \(F\). The conjecture is that if \(K\) is smoothly slice, then one of these curves is smoothly slice. The authors remark that it has been noted that the Casson-Gordon invariants can be described in terms of signature invariants of curves on Seifert surfaces and interpreted as obstructions to slicing \(K\) by slicing a surgery curve on \(F\). These obstructions lead to a variety of constraints on putative surgery curves for a smoothly slice knot. Letting \(\sigma_{K}(t)\) denote the Levine-Tristram signature function of \(K\) [\textit{J. Levine}, Invent. Math. 8, 98--110 (1969; Zbl 0179.52401)], adjusted to be the average of the one-sided limits at the jumps, the authors say a knot \(J\) satisfies the \((m,p)\) signature condition for integers \(m>0\) and \(p\) relatively prime to \(m\) and \(m+1\) if \[ \sum_{i=0}^{r-1} \sigma_{J}(ca^{i}/p) = 0 \] for all \(c\in\mathbb Z_{p}^{*} , a=(m+1)/m\) mod \(p\) and \(r\) the order of \(a\) mod \(p\). The authors prove that for a genus \(1\) smoothly slice knot \(K\) with Alexander polynomial \(\Delta_{K}(t) = m(m+1)t^2 - (m^2 + (m+1)^2)t + m(m+1)\) one of the surgery curves satisfies the \((m,p)\) conditions for infinitely many primes \(p\). (\(m\) for such a knot is denoted by the function \(m(K)\)). A corollary to this is that if \(m>0\) the integral of the signature function of one of the surgery curves is \(0\). D. Cooper, who first observed this in his 1982 Univ. of Warwick Ph.D. thesis, asked if the vanishing of the sum \(\sum_{i=1}^{p-1} \sigma_{J}(i/p)\) for an appropriate infinite set of primes implies the vanishing of the signature function. The authors provide a negative answer to this question using any knot \(K\) with non-zero signature function and considering the knot \(J=K_{(m(K),1)} \# -K_{(m(K)+1,1)}\) where \(m(K)>0\) and \(K_{(r,s)}\) denotes the \((r,s)\)-cable of \(K\). The authors go on to use conditions still stronger than the \((m,p)\) signature conditions by means of a Witt group. They show that topologically slice knots have surgery curves that satisfy a version of these stronger so called \((m,p)\) Witt conditions. They further show that any of the knots \(J\) cited above, \(J=K_{(m(K),1)} \# -K_{(m(K)+1,1)}\) satisfy these stronger Witt conditions. Another illuminating example given is the Whitehead double with \(2\) twists of the knot \(J = (T_{2,3})_{(2,-3)}\), the \((2,-3)\) cable of the \(2,3\) torus knot. This knot, denoted \( K = Wh(J,2)\) is shown to satisfy 5 properties: (1) \(K\) is genus \(1\), algebraically slice, with both surgery curves having the knot type of \(J\). (2) \(J\) satisfies the \((1,p)\) Witt conditions. (3) The signature function of \(J\) is non-zero. (4) \(\Delta_{J}(t)\) is not the Alexander polynomial of a smoothly slice knot. (5) \(Arf (J) \neq 0\). The authors do not know if \(K\) is either topologically locally flat slice or smoothly slice, so \(K\) remains a possible counterexample to the conjecture that the Seifert surface of a genus 1 smoothly slice knot contains a smoothly slice surgery curve. Another example, this time of an algebraically slice and not smoothly slice genus \(1\) knot with surgery curves satisfying the same 5 conditions listed above is given by the authors. The paper ends with a brief discussion of the knot signature function and more generally symmetric (step) jump functions on the unit interval.