Derivatives of knots and second-order signatures (Q969663)

The authors study systematically obstructions to a knot in \(S^3\) being slice. Let \(K\) be such a knot and \(M_K\) the closed manifold obtained by zero-framed surgery along that knot. Cochran-Orr-Teicher introduced sliceness obstructions based on the \(L^2\)-rho invariants \(\rho(M_K,\phi)\) where \(\phi: \pi_1(M_K)\to \Gamma\) is a group homomorphism. By \textit{W. Lück} and \textit{T. Schick} [High-dimensional manifold topology. Proceedings of the school, ICTP, Trieste, Italy, May 21--June 8, 2001. River Edge, NJ: World Scientific, 362--399 (2003; Zbl 1044.58026)] this is the difference between the ordinary signature of a zero bordism \(W\) for \(M_K\) and the \(L^2\)-signature of the \(\Gamma\)-covering of \(W\) (provided an appropriate \(W\) exists). These invariants of the knot are called signature. If \(\Gamma\) is abelian, one talks about zeroth order signatures, if \(\Gamma\) is metabelian they are called first-order signatures, examples are Casson-Gordon invariants. Finally, the second order signatures of the paper are obtained if \(\Gamma\) is two-step solvable. The main problem with these invariants is that there are many possibilities for the choice of \(\Gamma\) and \(\phi\). The observation is that the slice condition implies that for a specific choice one obtains zero. To use the obstruction one therefore has to look at a large (often infinite set) of invariants and check that none of them is zero. The main new idea of the paper is a very systematic identification and study of the relevant situations: in particular, the authors cut down the set of invariants to be considered, and often arrive at a finite set. The main tool to do this is via so called partial derivatives of the given knot (which works best if the knot has genus one): choose a Seifert surface \(\Sigma\) of minimal genus and a summand \(m\subset H_1(\Sigma)\) on which the Seifert form vanishes (called a metabolizer). A partial derivative of \(K\) with respect to \(m\) is a link \(\partial K/\partial m\) embedded into \(\Sigma\) whose components represent a basis of \(m\). A key observation of the paper is that the relevant second order signatures of \(K\) can be expressed in terms of first order signatures of derivatives of \(K\). The main theorem of the paper is the following: Theorem: Let \(K\) be a slice knot of genus \(1\) with Seifert surface \(\Sigma\) of genus \(1\), \(A_0(K)= H_1(S^3-K;\mathbb{Q}[t,t^{-1}])\) the rational Alexander module. Then there is a Lagrangian \(P\subset A_0(K)\) (with respect to the Blanchfield pairing) such that for any metabolizer \(m\) whose image in \(A_0(K)\) generates \(P\) such that the set of first order signatures of \(\partial K/\partial m\) contains \(0\). Specifically, there is a homologically essential simple closed curve \(J\) of self-linking zero on \(\Sigma\) which has vanihing zero-th and first order signature. The authors prove generalizations to knots of higher genus. A remaining main tool in the proof is the concept of \textit{null-bordism} for \(M_K\), defined via conditions on the homology and the fundamental group. They show that the slice condition for \(K\) allows to construct a null-bordism for \(M_{\partial K/\partial m}\), and they show that null-bordant manifolds have trivial zero order and first order signatures.