Barriers in metric spaces (Q1033051)

Call a subset of a connected topological space a barrier if it is connected and its complement is disconnected. The authors investigate certain barriers in the tight span \(T(D)\) of a metric space. Let \(X\) be a set and let \(D\) be a metric for \(X\). Write \[ T(D)= \{f\in\mathbb{R}^X: f(x)= \sup(D(x,y)- f(y)), y\in X,\forall x\in X\}; \] \(T(D)\) is called the tight span of \(D\); the authors consider sets of the form \(T_{(f,\varepsilon)}(D)= T(D)- B_\varepsilon(f)\), where \(B_\varepsilon(f)\) is the closed unit ball in \(T(D)\), with \(\varepsilon> 0\). Consider: \(\Gamma=\) the graph with vertex-set \(\text{supp}_\varepsilon(f)\) and edge set \[ \left\{(a,b)\in \begin{pmatrix}\text{supp}_\varepsilon(f)\\ 2\end{pmatrix}: f(a)+ f(b)> ab+\varepsilon\right\}. \] \(\pi_0(T_{f,\varepsilon}(D))=\) the set of connected components of \(T_{(f,\varepsilon)}(D)\). \(\pi_0(\Gamma)\) the set of connected components of \(\Gamma\). Then the authors establish: Result 1. There exists a canonical surjective mapping \[ \pi_f: \pi_0(T_{(f,\varepsilon)}(D))\to \pi_0(\Gamma). \] Result 2. Given a bipartition of \(\text{supp}_\varepsilon(f)\) into two nonempty subsets \(A\) and \(B\) such that the corresponding open sets \(O_{(f,\varepsilon)}(A)\) and \(O_{(f,\varepsilon)}(B)\) of \(T_{(f,\varepsilon)}(D)\) form a bipartition of \(T_{(f,\varepsilon)}(D)\) with \(f(a)+ f(b)\leq ab+ 2\varepsilon\), \(\forall a\in A\) and \(\forall b\in B\), then \(B_\varepsilon(f)\) is a barrier in \(T(D)\).