On commuting and noncommuting complexes (Q5929457)

In this article methods of combinatorial topology are applied to the study of simplicial complexes associated with a finite group \(G\). The complexes of interest are defined as follows: \(\text{NC}(G)\) is the complex of chains in the set of subsets of \(G\) containing no pair of commuting elements; \(\text{BNC}(G)\) is the full subcomplex of \(\text{NC}(G)\) on the sets of non-central elements of \(G\). For a prime \(p\), \(\text{NC}_p(G)\) is the full subcomplex on the sets of elements of order \(p\). Similarly, \(C(G)\) (resp. \(C_p(G)\)) is the order complex of the poset of subsets of pairwise commuting elements (of order \(p\)). The primary thrust of the paper is directed toward the subcomplexes of \(\text{NC}(G)\). To begin, the authors prove that \(\text{BNC}(G)\) is simply-connected. A covering space argument shows that \(\text{BNC}(G)\) is not contractible if \(G\) has non-trivial center or \(|G|\) is odd (and \(G\) is not Abelian); in either case there is a non-trivial free action on \(\text{BNC}(G)\). Similar reasoning shows that \(\text{BNC}_p(G)\) is not contractible for \(p\) odd, or for \(p=2\) with \(|G|\) odd or \(|Z(G)|\) even. The authors conjecture that the Euler characteristic of \(\text{BNC}(G)\) is odd for any nonabelian simple group \(G\); if true this assertion would imply the Feit-Thompson theorem that odd-order groups are solvable. The main results arise from application of an unpublished theorem of A.~Björner, M.~Wachs, and V.~Welker on ``blowup'' or ``inflated'' simplicial complexes. A blowup complex \(S\) is obtained from a simpler ``core'' complex \(\overline S\) by replacing each vertex \(v\) with a discrete set \(S_v\). Simplices of \(S\) are obtained by choosing arbitrary elements from the sets \(S_v\) corresponding to vertices of simplices of \(\overline S\). For \(\overline S\) connected, Björner, Wachs, and Welker give a wedge decomposition of the homotopy type of \(S\) in terms of \(\overline S\) and the simplices of \(\overline S\). Here the authors realize \(\text{BNC}(G)\) as a blowup complex, and use the general result to obtain lower bounds on the homology of \(\text{BNC}(G)\). As a corollary, it is shown that \(H_{s-1}(\text{BNC}(G))\not=0\) if \(G\) contains a maximal non-commuting set of size \(s\), again for odd-order groups and groups with nontrivial center. In the special case where commutativity defines a transitive relation on \(G-Z(G)\), \(\text{NC}(G)\) is homotopy equivalent to a bouquet of spheres of dimension \(\text{nc}(G)-1\), where \(\text{nc}(G)\) is the size of a maximal noncommuting set in \(G\). A duality theory is developed which yields some connections between the commuting and non-commuting complexes. The complexes \(C_p(G)\) are shown to be homotopy equivalent to the more familiar complexes \(A_p(G)\) of chains of elementary Abelian \(p\)-subgroups. Then results of Quillen on \(A_p(G)\) yield an estimate on the homotopy connectivity of \(\text{NC}_p(G)\). The homological results are obtained in a more general formulation, involving the notion of ``commuting structure'', being a symmetric, reflexive relation on a set. The reviewer remarks that such a commuting structure is essentially just a graph without multiple edges. The commuting and non-commuting complexes of a commuting structure are the clique complexes of the graph and its complement. This is the duality referred to above. The wedge decomposition obtained by the authors is in fact a decomposition of the clique complex of a graph, provided the ``core'' is connected. For instance the result cited above concerning groups with transitive commutativity relation is in fact an assertion about the clique complex of a complete \(n\)-partite graph, the complement of the graph of an equivalence relation.