KKM -- a topological approach for trees (Q2567406)

Given a hypergraph \(\mathcal H\), \(\tau(\mathcal H)\) (the covering number of \(\mathcal H\)) is the minimal cardinality of a set intersecting all edges of \(\mathcal H\), and \(\nu(\mathcal H)\) is the maximal number of disjoint sets (i.e. a maximum matching) in \(\mathcal H\). While \(\nu(\mathcal H) \leq \tau(\mathcal H)\), to bound \(\tau(\mathcal H)\) by a function of \(\nu(\mathcal H)\) is much harder. If \(\mathcal H\) is a \(2\)-interval hypergraph [see \textit{G. Tardos}, Combinatorica 15, 123--134 (1995; Zbl 0823.05022)], then \(\tau({\mathcal H}) \leq 2\nu(\mathcal H)\). Tardos used topological methods (notions of contractibility); Berger's approach is also topological, but the main tool (KKM theorm) is different. First the author reproves an old theorem about a special hypergraph (namely if \(\mathcal H\) is a finite family of connected subsets of a tree \(T\), then \(\tau(\mathcal H) = \nu(\mathcal H)\)). Then the method is used to extend the results of \textit{T. Kaiser} [Discrete Comput. Geom. 18, 195--203 (1997; Zbl 0883.05124)]. It states that \(\tau({\mathcal H}) = (d^2-d)\nu({\mathcal H})\) for a \(d\)-interval hypergraph \(\mathcal H\), while its extension says the same for \(d\)-tree hypergraphs.