Connected Floer homology of covering involutions (Q785349)

One of the central questions in knot theory is determining which knots \(K \subset S^3 = \partial D^4\) are \textit{smoothly slice}, i.e. bound a properly and smoothly embedded disc \(D \subset D^4\). The paper under review attacks the question by applying Heegaard Floer theory to the double cover \(\Sigma(K)\) of \(S^3\), branched over \(K\), together with its covering involution, \(\tau \colon \Sigma(K) \to \Sigma(K)\). The involution \(\tau\) induces a chain endomorphism \(\tau_\sharp\) on \(\mathrm{CF}^-(\Sigma(K)\), a chain complex computing the Heegaard Floer homology of \(\Sigma(K)\). This \(\tau_\sharp\) is, up to chain homotopy, an involution. Adapting the machinery developed in [\textit{K. Hendricks} and \textit{C. Manolescu}, Duke Math. J. 166, No. 7, 1211--1299 (2017; Zbl 1383.57036)] in the context of involutive Heegaard Floer homology, the authors extract more information out of \(\mathrm{HF}^-(\Sigma(K))\) than what had been previously done. First, in analogy with Hendricks and Manolescu [loc. cit.], they define the \textit{branched knot Floer homology} of \(K\), denoted with \(\mathrm{HFB}^-(K)\), a graded \(\mathbb{F}_2[U,Q]/(Q^2)\)-module, which is an invariant of \(K\) up to isotopy. From this group, similarly to correction terms in both ordinary and involutive Floer homology, they extract two numerical invariants, \(\overline\delta(K)\) and \(\underline{\delta}(K)\) that are concordance invariants. Building on ideas of [\textit{K. Hendricks} et al., J. Inst. Math. Jussieu 20, No. 1, 187--224 (2021; Zbl 1473.57036)], they define the \textit{(reduced) connected branched Floer homology} of \(K\), which is now a graded module (defined up to isomorphism) over \(\mathbb{F}_2[U]\), which is again a concordance invariant. This invariant turns out to vanish for two very natural classes of knots: quasi-alternating knots and torus knots. The main application of the paper is to linear independence questions in the concordance group, achieved using connected branched Floer homology. First, the authors prove that no member of a well-studied family of pretzel knots, \(P(-2,3,q)\) with \(q \ge 7\) odd, is concordant to a connected sum of quasi-alternating and torus knots. With more work, they also show that a certain family of pretzel knots, namely \(P(-2q-1, 4q+1, 4q+3)\) for \(q \ge 1\), is linearly independent in the concordance group modulo quasi-alternating and torus knots.