Khovanov homology, sutured Floer homology and annular links (Q600807)

For every link \(\mathbb{L}\) in a solid torus, this paper gives a spectral sequence whose second page is its triply-graded Khovanov homology for links in a thickened annular \(A\times I\), and which converges to the sutured Floer homology of the double branched cover of \(A\times I\) over \(\mathbb{L}\). Khovanov homology is a bigraded module \(\mathrm{Kh}\) associated to any link in \(S^3\). It has been refined for links in thickened surfaces into a triply graded homology \(\mathrm{Kh}^*\) in [\textit{M. Asaeda}, \textit{J. H. Przytycki} and \textit{A. S. Sikora}, Algebr. Geom. Topol. 4, 1177--1210 (2004; Zbl 1070.57008)]. Heegaard-Floer homology is a many-sided invariant of closed \(3\)-manifolds (graded module \(\widehat{\mathrm{HF}}\)) and links in \(3\)-manifolds (bigraded module \(\widehat{\mathrm{HFK}}\)). In [ibid. 6, 1429--1457 (2006; Zbl 1129.57039)], \textit{A. Juhász} unified them into an invariant \(\mathrm{SHF}\) of sutured manifolds. Although coming from different backgrounds, Khovanov and Heegaard-Floer homologies seem to be closely related. In [Adv. Math. 194, No.~1, 1--33 (2005; Zbl 1076.57013)], \textit{P. Ozsváth} and \textit{Z. Szabó} exhibit, for every knot \(K\), a spectral sequence whose second page is \(\mathrm{Kh}(K)\) and which converges to \(\widehat{\mathrm{HF}}\big(\boldsymbol{\Sigma}(S^3,K)\big)\), where \(\boldsymbol{\Sigma}(S^3,K)\) is the double branched cover of \(S^3\) over \(K\). This spectral sequence is restated and improved in terms of sutured manifolds by the present authors in [Adv. Math. 223, No.~6, 2114--2165 (2010; Zbl 1205.57015)]. In [On knot Floer homology in double branched covers, \url{arXiv:math.GT/0706.0741}], \textit{L. P. Roberts} extends Ozsvàth and Szabò's result to knots in the complement of the unknot. The main goal of this paper is to interpret Roberts' spectral sequence from a sutured point of view. The authors also relate it to their former work on tangles in the thickened disk. The paper is organized as follows. The section 2 is devoted to the spectral sequence. It heavily relies on the technique developed by the authors in the section 5 of [loc.cit.]. They begin by a review of \(\mathrm{Kh}^*\) and \(\mathrm{SHF}\) with \(\mathbb{Z}/2\mathbb{Z}\)-coefficients, in a way which fits their purpose. In particular, for any link \(\mathbb{L}\) in the thickened annulus \(A\times I\), they give a description of \(\mathrm{SFH}\big(\boldsymbol{\Sigma}(A\times I,\mathbb{L})\big)\) in terms of bigraded \(\widehat{\mathrm{HFK}}\). Then they prove that, for resolved links (links which project into \(A\) without any crossing), \(\mathrm{SFH}\big(\boldsymbol{\Sigma}(A\times I,\;.\;)\big)\) and \(\mathrm{Kh}^*(\;.\;)\) coincide as bigraded modules. For general links \(\mathbb{L}\subset A\times I\), \(\mathrm{SFH}\big(\boldsymbol{\Sigma}(A\times I,\mathbb{L})\big)\) can be described as the homology of a cubical complex whose vertices correspond to the sutured Floer homologies of the double branched covers over the resolutions of \(\mathbb{L}\) and edges to some surgery maps. Since the graded parts (for the missing grading of \(\mathrm{Kh}^*\)) of these surgery maps are equal to the elementary cobordisms involved in the definition of \(\mathrm{Kh}^*\), we obtain a spectral sequence starting at \(\mathrm{Kh}^*(\mathbb{L})\) and converging to \(\mathrm{SFH}\big(\boldsymbol{\Sigma}(A\times I,\mathbb{L})\big)\). If \(\mathbb{L}\) is a link inside a solid torus, then cutting along a meridian disk will lead to a balanced tangle \(T\) in the thickened disk \(D\times I\). Section 3 uses some naturality results for sutured Floer homology under decomposition along certain surfaces, given by the authors in [Int. Math. Res. Not. 2010, No.~21, 4159--4210 (2010; Zbl 1230.57012)], to prove that the spectral sequence from \(\mathrm{Kh}^*(T)\) to \(\mathrm{SFH}\big(\boldsymbol{\Sigma}(D\times I,\mathbb{L})\big)\) defined in [op. cit.] arises as a direct summand of the spectral sequence defined in section 2. Moreover, the authors prove that this summand is nontrivial iff the cut minimizes the intersection with \(\mathbb{L}\).