Sutured Floer homology and hypergraphs (Q390219)

Given a bipartite planar graph \(G\), there is a standard way of creating a link \(L_G\) in the 3-sphere \(S^3\) and a Seifert surface \(F_G\) for \(L_G\). To create \(F_G\), thicken \(G\) in the plane to make a planar surface and then, for each edge of the graph \(G\), insert a full twist in that portion of the Seifert surface (for example by doing \(\pm 1\) Dehn surgery along a meridian for each edge). These twists can be inserted so that the diagram of \(L_G\) (the boundary of \(F_G\)) is alternating. Links created in such a way are called \textit{special alternating} links. The main theorem creates an isomorphism between a certain set \(S(F_G)\) coming from the sutured Floer homology of the exterior of \(L_G\) and another set \(Q_{(E,R)}\) coming from the combinatorics of the graph. One corollary of this work is that if \(F\) is a standard Seifert surface for a homogeneous link \(L\), then \(S(F)\) is convex. Sutured Floer homology (recently discovered by \textit{A. Juhász}, [Algebr. Geom. Topol. 6, 1429--1457 (2006; Zbl 1129.57039)]) is an effective tool for studying Seifert surfaces. Given a Seifert surface \(F \subset S^3\) for a link \(L\) and a Spin\(^c\)-structure \(\mathfrak{s} \in \text{Spin}^c(S^3 - F, L)\), the sutured Floer homology \(SFH(S^3 - F, L, \mathfrak{s})\) is a group, which in the context of this paper, is either trivial or isomorphic to \(\mathbb{Z}\). The set \(S(F)\) is the set of Spin\(^c\) structures \(\mathfrak{s}\) such that \(SFH(S^3 - F, L , \mathfrak{s})\) is non-trivial. The set \(S(F)\) can be thought of as a certain subset of the lattice \(H_1(S^3 - F; \mathbb{Z})\), as \(\text{Spin}^c(S^3 - F, L)\) can be thought of as an affine copy of \(H_1(S^3 - F; \mathbb{Z})\). On the other hand, bipartite planar graphs have been useful combinatorial objects, at least since [\textit{W. T. Tutte}, Proc. Camb. Philos. Soc. 44, 463--482 (1948; Zbl 0030.40903)]. A bipartite planar graph \(G\) divides the plane into regions and its vertices are partitioned into two color classes. If \(E\) is one of the color classes and if \(R\) is the set of regions, we can define a hypergraph \((E,R)\) with vertex set \(E\) and with hyperedge set consisting all subsets \(\rho \subset E\) such that there is a region in \(R\) so that \(\rho\) is the intersection of \(E\) with the set of vertices in \(G\) incident to the region. A \textit{hypertree} for \((E,R)\) is a function \(f \to R \to \mathbb N\) such that there is a spanning tree for \(G\) such for each \(\rho \in E\), \(f(\rho)\) is one less than the number of elements of \(\rho\). The set of hypertrees was introduced by \textit{A. Postnikov} [Int. Math. Res. Not. 2009, No. 6, 1026--1106 (2009; Zbl 1162.52007)], and \textit{T. Kálmán} [Adv. Math. 244, 823--873 (2013; Zbl 1283.05136)] showed that \(Q_{(E,R)}\) is equal, in a certain precise sense, to the determinant of a certain type of adjacency matrix. In the context of this paper, [\textit{S. Friedl} et al., J. Topol. 4, No. 2, 431--478 (2011; Zbl 1270.57041)] showed that \(S(F_G)\) can be calculated from the Turaev torsion of the exterior of \(F_G\). The proof proceeds by rather intricate matrix calculations which are elucidated by a very helpful example, worked out in detail. The paper also includes a number of interesting questions for further research.