A note on sign conventions in link Floer homology (Q554511)

This paper investigates signs issues for link Floer homology with \(\mathbb{Z}\)-coefficients. Knot Floer homology is a link invariant developed by \textit{P. Ozsváth} and \textit{Z. Szabó} in [``Holomorphic disks and knot invariants'', Adv. Math. 186, 58--116 (2004; Zbl 1062.57019)]. It is defined as the homology of a bigraded chain complex associated to any Heegaard diagram of a knot in a closed 3-manifold. The differential is defined by counting domains of the Heegaard diagram which satisfy some analytical properties. Since there is a natural way of seeing links in \(S^3\) as knots in a connected sum of \(S^1\times S^2\)'s, knot Floer homology can be readily extended to them. Initially defined with \(\mathbb{F}_2\)-coefficients, there are \(2^{\ell-1}\) ways to lift this chain complex to \(\mathbb Z\)-coefficients, where \(\ell\) is the number of components of the studied link. In [``Holomorphic disks, link invariants and the multi-variable Alexander polynomial'', Algebr. Geom. Topol. 8 (2), 615--692 (2008; Zbl 1144.57011)], the same authors refined their invariant into a \((\ell+1)\)-graded one, called link Floer homology and defined for any null-homologous link in any closed 3-manifold \(Y\). However, due to signs problems, their construction is given only for \(\mathbb{F}_2\)-coefficients. The present paper fills this gap by determining the essential \(2^{b_1(Y)+\ell-1}\) ways of introducing signs to get \(\mathbb{Z}\)-versions of link Floer homology. The first part of the paper is devoted to this construction. The author defines orientation systems which are consistent choices of signs for all domains of the Heegaard diagram. He uses it to refine the differential as a counting ``with signs'' of domains. He proves that the resulting homology depends only on the signs assigned to certain specific domains, namely empty periodic ones. They form an abelian group of rank \(b_1(Y)+\ell-1\), and signs can be chosen independently on any basis. Then the author shows that, for two Heegaard diagrams of the same link, there is a natural bijection between the associated orientation systems. Ozsváth and Szabó's proofs of invariance then apply. Finally the author proves that, for links in \(S^3\), one recovers the \(2^{\ell-1}\) \(\mathbb{Z}\)-versions of knot Floer homology by collapsing \(\ell\) gradings into one. A second part is devoted to a combinatorial approach. As a matter of fact, for some Heegaard diagram as toroidal grids, the link Floer homology construction becomes combinatorial and computationable. In [``On combinatorial link Floer homology'', Geom. Topol. 11, 2339--2412 (2007; Zbl 1155.57030)], \textit{C. Manolescu, P. Ozsváth, Z. Szabó} and \textit{D. Thurston} used this idea to give a so-called signs assignment which lifts the construction to \(\mathbb{Z}\)-coefficients. The author shows that it fits his orientation system approach and gives \(2^{\ell-1}-1\) other signs assignments which correspond to the other possibilities. The paper ends with computations for the 2-unlink and the Hopf link, showing that different orientation systems may or may not give the same link Floer homology. Note that Ozsváth and Szabó invariants exist in many variants. The present paper deals with the graded hat one. The author plans to consider the minus and infinity versions in a forthcoming paper which should shed light on a canonical choice among all the possibilities for lifting link Floer homology to \(\mathbb{Z}\)-coefficients.