Noncommutative knot theory (Q1880636)

The aim of this paper is to introduce new invariants for knots, which are higher order analogs of the classical Alexander module and Alexander polynomial. Given a knot \(K\) in \(S^3\), let \(G\) be its knot group, and for \(n\geq 0\) let \(G^{(n+1)}\) be the \((n+1)\)-st term in the derived series of \(G\). The \(n\)-th integral Alexander module \(\mathcal A_n^{\mathbb Z}(K)\) is defined as \(G^{(n+1)}/G^{(n+2)}\), viewed as a right module over the ring \(\mathbb Z[G/G^{(n+1)}]\). For \(n>0\), this ring is non-commutative, non-Noetherian and not an UFD in general, which a priori makes handling these modules difficult. In spite of these problems, the author shows that these modules lead to effective invariants with applications to estimating knot genus, detecting fibered, prime, and alternating knots, and producing new obstructions to the existence of symplectic structures on certain four manifolds. To deal with modules over \(\mathbb Z[G/G^{(n+1)}]\), the author uses the theory of poly-(torsion-free abelian) groups. Via a localized version \(\mathcal A_n(K)\) of \(\mathcal A_n^{\mathbb Z}(K)\), he introduces the \(n\)-th order Alexander polynomial. While the indeterminacy in this polynomial makes it difficult to handle, its degree is a well defined numerical knot invariant \(\delta_n(K)\). This generalizes the classical Alexander module and Alexander polynomial, which is obtained for \(n=0\). The fundamental result for most of the applications discussed in the paper is then that for a non-trivial knot \(K\) and each \(n>0\), one has \(\delta_0(K)\leq\delta_1(K)+1\leq\dots\leq\delta_n(K)+1\leq 2\cdot\text{genus}(K)\). If \(K\) is fibered or alternating or if the \(0\)-surgery \(M_K\) at \(K\) has the property that \(S^1\times M_K\) admits a symplectic structure, then all these inequalities have to be equalities. An important part of the paper is devoted to a construction for modifications of a knot, called genetic infection, whose influence on the higher order Alexander module can be well controlled. For a knot \(K\) with non-trivial classical Alexander polynomial, this leads to families of modifications, whose effect on \(\mathcal A_n^{\mathbb Z}\) and on \(\mathcal A_n\) can be controlled very well. In particular, one obtains modifications for which some of the above inequalities become proper. Finally, the author also studies higher order bordism invariants generalizing the Arf invariant, higher order von Neumann signatures of knots, and higher order generalizations of the Blanchfield linking forms.