Untwisting information from Heegaard Floer homology (Q2398919): Difference between revisions

The unknotting number of knots is a well studied invariant in knot theory or 4-dimensional topology. The definition of this invariant is quite simple but it is very hard to compute. In the paper under review, the author considers a generalization of the unknotting number of knots, which is called the ``untwisting number''. Let \(K\) be an oriented knot in the 3-sphere and \(U\) be the unknot. Suppose that \(lk(U, K)=0\). Then, a generalized crossing change on \(K\) is defined by performing \(\pm 1\)-surgery along \(U\). A usual crossing change is a generalized crossing change because it can be regarded as adding a full-twist near a crossing. The untwisting number \(tu(K)\) of a knot \(K\) is defined as the minimum number of generalized crossing changes needed to obtain the unknot from \(K\). It is known that if the classical unknotting number of \(K\) is one, then by the Montesinos trick, the double branched cover \(\Sigma(K)\) along \(K\) is represented by the Dehn surgery along some other knot whose framing is half the determinant of \(K\). The author proves that this fact cannot be generalized to untwisting number one knots (Theorems 1.1 and 3.2). Moreover, for an oriented knot \(K\), the author gives a \(4\)-manifold bounded by \(\Sigma(K)\) by considering the numbers of ``positive'' and ``negative'' generalized crossing changes needed to obtain the unknot from \(K\) (Theorems 1.2 and 4.4). By using this \(4\)-manifold and the Heegaard Floer correction terms of \(\Sigma(K)\), the author gives an obstruction to untwisting number one. As an application, the untwisting numbers of several alternating knots with \(10\) or \(11\) crossings are computed.