Horseshoes, entropy and periods for graph maps (Q1312083)

The authors study topological entropy and cycles for continuous maps defined on graphs which are finite, connected and not directed. By applying some of the techniques developed to study interval maps they establish: (i) If a continuous map \(f\) of a graph into itself has an \(s\)-horseshoe (an interval \(I\) and \(s\) subintervals with pairwise disjoint interiors, each of them mapped by \(f\) onto the whole \(I\)), then \(h(f) \geq \log s\) (\(h(f)\) is the topological entropy of \(f\)). (ii) If a continuous map \(f\) of a graph into itself has positive topological entropy, then there exist sequences \((k_ n)\) and \((s_ n)\) of positive integers such that for each \(n\) the map \(f^{k_ n}\) has an \(s_ n\)-horseshoe and \(\lim\sup_{n\to \infty}{1\over k}\log s_ n = h(f)\). (iii) Topological entropy defines a lower semicontinuous \([0,\infty]\)- valued map of the space of all continuous maps from a graph into itself with the topology of uniform convergence. Moreover, the paper contains results that relate the topological entropy \(h(f)\) and the set \(\text{Per}(f)\) of periods of cycles of \(f\).