Unknotting with a single twist (Q2070059)

The theme of the paper under review is the unknotting of a knot with a single twist. For a given knot \(K\), there may be several ways of twisting to untie it. The linking number between the knot and the twisting circle is assumed to be positive, and the direction of twist is recorded as positive or negative. For example, \(2^+\) means that there exists a positive unknotting twist with linking number two. These data define the set \(\mathcal{U}(K)\) of unknotting twist indices for a knot \(K\). The first result claims that for each knot, this set \(\mathcal{U}(K)\) is a subset of the form \(\{a^-,(a+1)^-,0^-,0^+,b^+,(b+1)^+\}\) for some \(a,b>0\). Moreover, the containment is proper unless \(a=b=1\). For example, it turns out that \(\mathcal{U}=\{2^-,3^-,0^+\}\) for the right-handed trefoil, and \(\mathcal{U}=\{2^-,0^-,0^+,2^+\}\) for the figure eight knot. The values of \(a\) and \(b\) are determined by the Heegaard Floer homology complex \(\mathrm{CFK}^\infty(K)\). Interestingly, the unknot has \(\{2^-,1^-,0^-,0^+,1^+,2^+\}\), and so does the knot \(8_9\). In general, finding knots for which the set \(\mathcal{U}\) contains several elements is hard. The authors give one constraint. If \(\{k^-,(k+1)^+\}\subset \mathcal{U}(K)\) for \(k\ge 1\), then \(g(K)\ge (2k^3+3k^2-11k+6)/6\). The main body of the paper describes various obstructions for unknotting such as \(4\)-dimensional aspects, homology of cyclic branched covers, linking forms, signatures, Casson-Gordon invariants, Arf invariants, and Heegaard Floer invariants.