Heegaard Floer invariants of contact structures on links of surface singularities (Q2247189)

To the germ of a normal surface singularity \((S,0)\), one associates a contact structure \(\xi_{(S,0)}\) on its link \(Y_{(S,0)}\), called the canonical contact structure. \textit{P. Ozsváth} and \textit{Z. Szabó} associated to a contact 3-manifold \((Y,\xi)\) an element \(c^+(\xi) \in \mathrm{HF}^+(-Y)\), living in the `plus' version of Heegaard Floer homology [Duke Math. J. 129, No. 1, 39--61 (2005; Zbl 1083.57042)], which is a graded \(\mathbb{F}_2[U]\)-module (here \(\mathbb{F}_2\) is the field of 2-elements and \(U\) is a formal variable). \textit{Ç. Karakurt} introduced an invariant \(\sigma(\xi)\) of a contact structure \(\xi\) which says for which \(k\) its associated contact invariant \(c^+(\xi)\) can be written as \(U^k\cdot x\) for some element \(x \in \mathrm{HF}^+(-Y)\) [Kyoto J. Math. 54, No. 2, 271--294 (2014; Zbl 1300.57026)]. The article under review shows that \(\sigma(\xi_{(S,0)}) = 0\) unless \((S,0)\) is a \emph{rational} singularity. It generalises a result of \textit{Ç. Karakurt} and \textit{F. Öztürk}, who had addressed the case of \emph{almost-rational} singularities [Int. J. Math. 29, No. 3, Article ID 1850019, 21 p. (2018; Zbl 1386.57029)]. The motivation behind this statement comes from deformation theory: the authors were seeking to use Karakurt's \(\sigma\) as an obstruction to the existence of deformations, exploiting the monotonicity of \(\sigma\) under Stein cobordisms and the fact that a deformation induces such a cobordism. Their main result shows that such a strategy is, in fact, doomed to fail, unless one of the two singularities is rational and the other is not. (It was already known to singularity theorists that such deformations could not exist.) The proof uses in an essential way \textit{A. Némethi}'s lattice homology \(\mathbb{H}_0^+(\Gamma)\) [Publ. Res. Inst. Math. Sci. 44, No. 2, 507--543 (2008; Zbl 1149.14029)] associated to the dual graph \(\Gamma_{(S,0)}\) of the resolution of \((S,0)\), and the existence of a map of \(\mathbb{F}_2[U]\)-modules \(T^+ \colon \mathrm{HF}^+(-Y_{(S,0)}) \to \mathbb{H}_0^+(\Gamma_{(S,0)})\). It then analyses what happens to the contact invariant under this map, and shows that \(T^+(c^+(\xi_{(S,0)}))\) cannot be written as \(U\cdot \ell\) for any \(\ell \in \mathbb{H}_0^+(\Gamma_{(S,0)})\) unless \((S,0)\) is a rational singularity. This suffices to prove that \(\sigma(\xi_{(S,0)}) = 0\). In the case when the minimal resolution of \((S,0)\) is good (that is to say, when there is a minimal non-singular surface dominating \(S\) and such that the preimage of \(0\) is a normal crossing divisor), the arguments build on and expand ideas of Karakurt [loc. cit.] and Karakurt and Öztürk [loc. cit.]. The case when the smallest good resolution is not minimal requires a more careful and laborious analysis of the behaviour of both Heegaard Floer and lattice homology under blow-ups. An interesting remark is that, contrarily to the aforementioned work of Karakurt and Öztürk, the proof presented in this paper does not rely on the isomorphism between Heegaard Floer homology and lattice homology in full generality, which was not known at the time of publication. (Zemke has recently posted on the arXiv a proof of the isomorphism [\textit{I. Zemke}, ``The equivalence of lattice and Heegaard Floer homology'', Preprint, \url{arXiv:2111.14962}], which at the time of writing has not yet been peer-reviewed nor has it appeared in print.)