Dimension gaps between representability and collapsibility (Q1042449)

Let \({\mathcal S}=\{S_1, \ldots, S_n\}\) be a family of sets. The \textit{nerve} \(N({\mathcal S})\) of \({\mathcal S}\) is a simplicial complex whose vertex set is \([n]=\{1,\ldots , n\}\) and its simplices are the subsets \(\sigma\subseteq [n]\) with the property that \(\cap_{i\in \sigma}S_i\neq\emptyset\). If a simplicial complex is isomorphic to the nerve of a family of convex sets in \({\mathbb R}^d\), then it is called \(d\)-representable. \textit{G. Wegner} [Arch. Math. 26, 317--321 (1975; Zbl 0308.52005)] introduced the notion of \(d\)-collapsibility of simplicial complexes. A simplicial complex is called \(d\)-collapsible if it can be reduced to an empty simplicial complex by repeatedly removing a face of dimension at most \(d-1\) that is contained in a unique maximal face. Another important related concept is a \(d\)-Leray simplicial complex. A simplicial complex is \(d\)-Leray if its every induced subcomplex has zero homology over \({\mathbb Q}\) in dimensions \(d\) and larger. Wegner proved that a \(d\)-representable simplicial complex is \(d\)-collapsible and that every \(d\)-collapsible simplicial complex is \(d\)-Leray. Although the notions of \(d\)-representable, \(d\)-collapsible, and \(d\)-Leray are very similar in some aspects, no two of them coincide. Helly's theorem and other results in discrete geometry can be described in a very natural way using the language of \(d\)-representability of simplicial complexes. Moreover, in many of these statements, \(d\)-representability can be replaced by \(d\)-collapsibility or the \(d\)-Leray property in the assumptions. In this paper, the authors prove stronger differences between these notions by establishing ''dimension gaps''. More precisely, they construct for all \(d\geq 1\), a \(2d\)-Leray complex that is not \((3d-1)\)-collapsible, and they construct a \(d\)-collapsible complex that is not \((2d-2)\)-representable. In the course of the proofs, the authors obtain two results of independent interest as well. These are the following: (i) The nerve of every finite family of sets, each of size at most \(d\), is \(d\)-collapsible. (ii) If the nerve of a simplicial complex is \(d\)-representable, then the complex embeds in \({\mathbb R}^d\).