Cobordisms of sutured manifolds and the functoriality of link Floer homology (Q2629120)

This extensive paper resolves several important questions about Heegaard Floer theory. It proves the functoriality of Heegaard Floer theory under cobordisms and confirms its status as a topological quantum field theory, completing a series of papers by the author, together with others including D. Thurston, aimed at establishing these properties. Link Floer homology is a powerful link invariant, which categorifies the Alexander polynomial and can for instance detect the genus of a knot, and whether it is fibred. \textit{P. Ozsváth} and \textit{Z. Szabó} [Adv. Math. 202, No. 2, 326--400 (2006; Zbl 1099.53058)] showed that 4-dimensional cobordisms of closed 3-manifolds induce maps on Heegaard Floer homology. It is therefore a natural question whether link cobordisms induce maps on link Floer homology. The present paper answers this question -- indeed a more general question -- in the affirmative. It deals with sutured Floer homology (\(SFH\)), which was introduced by the author in [Algebr. Geom. Topol. 6, 1429--1457 (2006; Zbl 1129.57039)], generalising both link Floer homology and the ``hat'' Heegaard Floer homology of a closed 3-manifold. The paper is largely devoted to a careful demonstration that \(SFH\) yields a topological quantum field theory (TQFT), in the sense of \textit{M. Atiyah} [Publ. Math., Inst. Hautes Étud. Sci. 68, 175--186 (1988; Zbl 0692.53053)]. The author considers sutured manifolds \((M, \gamma)\), and cobordisms between them of the following type. A cobordism from \((M_0, \gamma_0)\) to \((M_1, \gamma_1)\) is a triple \[ \mathcal{W} = ( W, Z, [ \xi ] ) \] where \(W\) is a compact oriented 4-manifold with boundary, \(Z\) is a 3-dimensional submanifold of \(\partial W\) with \(\partial W \backslash \text{Int}(Z) = - M_0 \sqcup M_1\), and \(\xi\) is a positive contact structure on \(Z\) with dividing set \(\gamma_i\) on the convex boundary \(\partial M_i\). With appropriate restrictions on Euler class and up to an appropriate notion of equivalence, these form the objects and morphisms of a cobordism category \(\mathbf{BSut}\). The main theorem of the paper (11.11) is that \(SFH\) yields a functor from \(\mathbf{BSut}\) to the category of \(\mathbb{Z}_2\)-vector spaces, forming a (3+1)-dimensional TQFT. To a balanced sutured 3-manifold \((M,\gamma)\) is associated the \(\mathbb{Z}_2\) vector space \(SFH(M, \gamma)\), and to a cobordism \(\mathcal{W}\) from \((M_0, \gamma_0)\) to \((M_1, \gamma_1)\) is associated a linear map \(F_\mathcal{W} : SFH(M_0, \gamma_0) \rightarrow SFH(M_1, \gamma_1)\). When specialised to the case of links, this results in notions of \textit{decorated links} and \textit{decorated link cobordisms}, forming a category \(\mathbf{DLink}\). Associated constructions of sutured manifolds and cobordisms then give a functor \(\mathbf{DLink} \rightarrow \mathbf{BSut}\), and hence a natural map on link Floer homology is associated to decorated link cobordisms. Demonstrating that SFH forms a TQFT has presented several substantial problems. A first problem is that of assigning a well-defined group to a sutured manifold, not just an isomorphism class. These and related naturality issues were dealt with by the author with \textit{D. Thurston} in [Naturality and mapping class groups in Heegaard Floer homology, Preprint, \url{arXiv:1210.4996}]. More general issues regarding the definition of TQFTs by surgery descriptions were dealt with by the author in [Defining and classifying TQFTs via surgery, Preprint, \url{arXiv:1408.0668}]. The present paper resolves the remaining issues, and the amount of effort required is substantial. The structure of the argument can be summarised as follows. The author first introduces the category \(\mathbf{BSut}\) (Section 2), and discusses Spin\(^c\) structures on the 3- and 4-manifolds involved (Section 3), as well as link cobordisms and their decorations (Section 4). A key notion in the argument is that of a \textit{special} cobordism, which is a balanced cobordism \(\mathcal{W} = (W, Z, [\xi])\) for which \(Z\) is a product \(Z = \partial M_0 \times I\) and \(\xi\) is an \(I\)-invariant contact structure. Maps on \(SFH\) are demonstrated for special cobordisms, following ideas of \textit{J. E. Grigsby} and \textit{S. M. Wehrli} [Adv. Math. 223, No. 6, 2114--2165 (2010; Zbl 1205.57015)] generalising Ozsváth-Szabó triangle maps. The author uses his previous work describing TQFTs via surgery. Maps are associated with a framed link (Section 6), or equivalently, with 2-handle addition. Given a framed link \(\mathbb{L}\) in a manifold \((M, \gamma)\), one can attach 2-handles along \(\mathbb{L}\) to obtain a sutured manifold \((M(\mathbb{L}), \gamma)\), and a special cobordism from \((M, \gamma)\) to \((M(\mathbb{L}), \gamma)\). These have standard Heegaard triple diagrams and the author shows that triangle maps yield well-defined natural maps \(SFH(M, \gamma) \rightarrow SFH(M(\mathbb{L}), \gamma)\). The simpler cases of 1- and 3-handles are then treated (Section 7), allowing the author to define (Section 8) a map on \(SFH\) for special cobordisms. It is shown, carefully, that these maps split along Spin\(^c\) structures in an appropriate sense. The author then turns to contact invariants (Section 9). The \(SFH\) gluing maps of \textit{K. Honda}, \textit{W. H. Kazez} and \textit{G. Matić} [Contact structures, sutured Floer homology and TQFT, Preprint, \url{arXiv:0807.2431}] are recalled and refined by their Spin\(^c\) classes. The map \(F_\mathcal{W}\) associated to a cobordism \(\mathcal{W}\) from \((M_0, \gamma_0)\) to \((M_1, \gamma_1)\) can then be defined (Section 10). Roughly speaking, the cobordism \(\mathcal{W} = (W, Z, [\xi])\) is factored into two parts, the first essentially consisting of an inclusion of sutured manifolds \((-M_0, -\gamma_0) \subset (-M_0 \cup Z, -\gamma_1)\), with the contact structure \(-\xi\) between them, and the second being a special cobordism. The map \(F_\mathcal{W}\) is thus defined as a composition of an SFH gluing map, and the map associated to a special cobordism. This map is shown to split along Spin\(^c\) structures in a natural way, to be independent of any choices involved, and to be natural and functorial (Section 11), giving the desired TQFT. It is also shown that for the cobordism \(\overline{\mathcal{W}}\) obtained by ``turning \(\mathcal{W}\) upside down'', the maps \(F_{\mathcal{W}}\) and \(F_{\overline{\mathcal{W}}}\) are dual, in an appropriate sense (theorem 11.8). It is also shown that the maps associated to Weinstein cobordisms preserve contact invariants.