On the 3-torsion part of the homology of the chessboard complex (Q659606)

The chessboard complex \(M_{m,n}\) is the simplicial complex of matchings in the complete bipartite graph \(K_{m,n}\). A \((p-1)\)-simplex corresponds to the placement of \(p\) nontaking rooks on an \(m\times n\) chessboard (each rook occupies an edge \((i,j)\), \(i \in \{1,\ldots,m\}\) and \(j \in \{1,\ldots,n\}\)). There is an important literature on the chessboard complex and, more generally, on matching complexes of graphs which is recalled in the paper under review. In particular, the determination of the integral homology groups remains a difficult question and, in [J. Comb. Theory, Ser. A 115, No. 8, 1504--1526 (2008; Zbl 1206.55010)], the author obtained results concerning the 3-torsion of the integral homology groups of the matching complex \(M_n\) (the simplicial complex of matchings in the complete graph \(K_{n}\)). In the present paper, he uses a similar approach for the study of the integer homology of chessboard complexes. By considering different long exact sequences involving various chessboard complexes, the author extends an earlier result of \textit{J. Shareshian} and \textit{M. L. Wachs} [Adv. Math. 212, No. 2, 525--570 (2007; Zbl 1117.05110)], by showing that \(\widetilde{H}_d(M_{m,m};\mathbb{Z})\) contains 3-torsion whenever \( (2m-4)/3 \leq d \leq m-4\) and \(\widetilde{H}_d(M_{m,n};\mathbb{Z})\) contains 3-torsion whenever \(m+1\leq n \leq 2m-5\) and \( (m+n-4)/3 \leq d \leq m-3\). He also characterizes non trivial homology groups : for \(1\leq m \leq n\), \(\widetilde{H}_d(M_{m,n};\mathbb{Z})\) is nonzero if, and only if, either \((m+n-4)/3 \leq d \leq m-2\), or \(d=m-1\) and \( n\geq m+1\). Another result provides an upper bound on the rank of the 3-torsion part by proving that, for \(m\leq n\leq 2m-5\) and \((m+n-4)/3\leq d\leq (2m+n-7)/4\), \[ \dim \widetilde{H}_d(M_{m,n};\mathbb{Z}_3)\leq f_{3d-m-n+4}(n-m,m-d-1) \] where \(f_k\) is a polynomial of degree \(3k\); this last result remains valid over any coefficient field. Finally, he obtains by a computer-free proof that \(\widetilde{H}_2(M_{5,5};\mathbb{Z})\cong \mathbb{Z}_3\).