Metric entropy of nonautonomous dynamical systems (Q5920338)

For sequences of spaces \(X_n\) and mappings \(f_n:X_n\to X_{n+1}\), \(n\geq 1\), this interesting paper introduces the concept of metric entropy for cocycles given by \(f_k^0:=I_{X_k}\), \[ f_k^n:=f_{k+n-1}\circ\cdots\circ f_k \] with integers \(k\geq 1\) and \(n\geq 0\). The author speaks of a nonautonomous dynamical system in this context, which can be ``metric'' for probability spaces \((X_n,\mu_n)\) and measurable \(f_n\), or ``topological'' when the \(X_n\) are compact metric spaces and the \(f_n\) are continuous mappings. Section~\(2\) deals with the concept of topological entropy and generalizes existing results to sequences of metric spaces \(X_n\) rather than having a single compact state space. Three characterization of entropy are possible (using open covers, spanning sets or separated sets). In the open cover definition, the sequences of open covers for \(X_n\) must have Lebesgue numbers being bounded away from \(0\). In order to obtain a power rule, one needs a uniform equicontinuity assumption on the mappings \(f_n\). Section~\(3\) introduces and discusses metric entropy for measure preserving systems, i.e., \(f_n\mu_n=\mu_{n+1}\). Thereto the notion of an ``admissible class'' of finite measurable partitions for \(X_{1,\infty}:=\{X_n\}_{n\geq 1}\) is defined. In the autonomous special case, it leads to the classical notion of metric entropy. In the general time-variant situation it is shown that several properties such as a Rokhlin inequality, invariance under ambient isomorphisms and a power rule hold. Finally, Section~\(4\) contains an inequality between metric and topological entropy for equicontinuous systems yielding one part of the variational principle. The paper closes with concluding remarks and open problems.