On the Conley decomposition of Mather sets (Q971959)

This paper studies invariant sets in the context of Mather's theory of Lagrangian systems. The author considers two ways of describing the invariant sets. The first is to use a purely topological point of view in the style of Conley, and look at the transitive components of compact metric spaces with flows. Alternatively, one can study invariant sets via minimizing the action and creating decompositions into so-called static classes. The author aims to make explicit the connections between these two approaches and their respective decompositions. The author concludes that the topological decomposition is finer than the variational decomposition, and that the decompositions coincide for most -- but not all -- systems. As an application of his analysis, the author proves the semicontinuity of the Aubry set as a function of the Lagrangian -- under certain non-degeneracy conditions. Mather has provided a counterexample to show this does not hold in general.