Helly numbers of acyclic families (Q2445970)

The \textit{Helly number} \(\text{h}(\mathcal F)\) of a family of sets \(\mathcal F\) \textit{with empty intersection} (i.e. \(\bigcap_{X\in \mathcal F} X= \emptyset\)) is the largest cardinal of a sub-family \({\mathcal A}\subset \mathcal F\) of empty intersection such that any proper sub-family \(\mathcal B\) of \(\mathcal A\) has non-empty intersection. \textit{Ed. Helly}'s theorem [Jahresber. Dtsch. Math.-Ver. 32, 175--176 (1923; JFM 49.0534.02)] can be formulated by saying that any finite family of convex sets in \({\mathbb R}^d\) has Helly number at most \(d+1\). In this paper, the authors generalizes this type of results by allowing certain families with disconnected intersections. Let \(\mathcal F\) be a finite family of open subsets (with empty intersection) of a locally arc-wise connected topological space \(\Gamma\). The nerve \(\mathcal N(\mathcal F)\) of \(\mathcal F\) is the simplicial complex obtained by associating a \(k\)-dimensional simplex to every sub-family of cardinal \((k+1)\) and with non-empty intersection. When \(\mathcal F\) is a good cover, the nerve captures the topology of the union of the sets (i.e. \({\mathcal N}(\mathcal F)\) and \(\bigcup_{X\in \mathcal F} X\) have the same homotopy type, see \textit{A. Björner} [in: Handbook of combinatorics. Vol. 1--2. Amsterdam: Elsevier (North-Holland). 1819--1872 (1995; Zbl 0851.52016)]) but this is no longer true with families having disconnected intersections, and the authors introduce the \textit{multinerve} \(\mathcal M(\mathcal F)\) of \(\mathcal F\) which conserves more information than \(\mathcal N(\mathcal F)\): by definition, the multinerve associates to each sub-family of \(\mathcal F\) of cardinal \((k+1)\) with non-empty intersection a number of \(k\)-dimensional simplices equal to the number of connected components of the intersection of the sub-family. The multinerve has a structure of simplicial poset and there is a natural projection from the multinerve to the nerve. Generalizing the \textit{nerve theorem}, the authors get a \textit{homological multinerve theorem} which, for acyclic families (a family \(\mathcal F\) is \textit{acyclic} if the intersection of every non-empty sub-family has trivial \(\mathbb Q\)-homology in dimension larger than zero; in particular, the intersection needs not be connected), states that \(\bigcup_{X\in \mathcal F} X\) has the same reduced homology as \(\mathcal M(\mathcal F)\). Actually, the \textit{homological multinerve theorem} is far more general, stating that \(\widetilde{H}_i(\mathcal M(\mathcal F)) \cong \widetilde{H}_i(\bigcup_{X\in \mathcal F} X)\) for \(i=0\) and \(i\geq s\) if the family is \textit{acyclic with slack s} (\textit{acyclicity with slack} -- a notion introduced by \textit{S. Hell} in [``On a topological fractional Helly theorem'', \url{arXiv:math/0506399}] -- allows non trivial homology in low dimension: the family \(\mathcal F\) is acyclic with slack \(s\) if \(\widetilde{H}_i(\bigcup_{X\in \mathcal G} X,\mathbb Q)=0\) for all nonempty \(\mathcal G\) \(\subset \mathcal F\) and all \(i\geq \max(1,s-| \mathcal G|)\)). Next, by an adaptation of a technique developed by \textit{G. Kalai} and \textit{R. Meshulam} in [J. Topol. 1, No. 3, 551--556 (2008; Zbl 1148.55014)], the authors prove a \textit{projection theorem} which relates the Leray number of the simplicial poset \(\mathcal M(\mathcal F)\) with the Leray number \(L(\mathcal N(\mathcal F))\) of the poset \(\mathcal N(\mathcal F)\). As \(\text{h}(\mathcal F) \leq L(\mathcal N(\mathcal F))+1\) and by using spectral sequence techniques, these two results are the main ingredients for proving the principal result of the paper (Theorem 3): If the family \(\mathcal F\) is acyclic with slack \(s\) and if any sub-family of \(\mathcal F\) of cardinality at least \(t\) intersects in at most \(r\) connected components, then \(\text{h}(\mathcal F) \leq r(\text{max}(d_{\Gamma},s,t)+1)\) where \(d_{\Gamma}\) is the smallest integer such that \(H_i(U,\mathbb Q)=0\) for every open subset \(U\) of \(\Gamma\) and every \(i\geq d_{\Gamma}\). This bound is tight. As an application, the authors get explicit bounds on Helly numbers in geometrical transversal theory. More generally, these results improve and unify various related results which are discussed in the paper.