On the number of Mather measures of Lagrangian systems (Q717159): Difference between revisions

A standard application of variational methods is to show the existence of periodic orbits of Lagrangian systems. John Mather had the idea of applying variational methods to measures instead of curves. If \(M\) is a compact connected manifold, \(TM\) its tangent bundle, and \(L: TM\to\mathbb{R}\) a Lagrangian, we can consider the set \(I(L)\) of compactly supported Borel probability measures on \(TM\) that are invariant under the Euler-Lagrange flow. It turns out that \(\int_{TM} Ld\mu\) is a natural definition of the action of a probability measure \(\mu\). A Mather measure is a minimizer of the action on \(I(L)\). The current paper addresses the question of how many Mather measures exist for a Lagrangian system. This work is a refinement of previous work by \textit{G. Contreras} and the author [Ann. Math. (2) 167, No. 3, 1099--1108 (2008; Zbl 1175.37067)]. The first general results in this area are due to Mañé who showed that Mather measures are, more broadly, minimizers of a kind of universal infinite-dimensional linear programming problem. The author of this paper applies tools of classical convex analysis to the infinite-dimensional problem. An important concept here -- of interest more broadly -- is the idea of countably rectifiable sets of finite codimension in Banach and Fréchet spaces.