On gradient structures for Markov chains and the passage to Wasserstein gradient flows (Q258483)

The authors consider spatially discrete approximations of the Wasserstein gradient structure of the Fokker-Planck equation and pass to the limit in the mesh size. The proof is based on the variational characterization of gradient structures on metric spaces going back to de Giorgi. Hereby, a functional on curves is defined which has solutions to the gradient flow equation as its only minimizers. The authors use on one hand the Wasserstein gradient structure discovered by \textit{R. Jordan} et al. [SIAM J. Math. Anal. 29, No. 1, 1--17 (1998; Zbl 0915.35120)] for the Fokker-Planck equation and the more recently discovered one for reversible Markov chains [\textit{S.-N. Chow} et al., Arch. Ration. Mech. Anal. 203, No. 3, 969--1008 (2012; Zbl 1256.35173); \textit{J. Maas}, J. Funct. Anal. 261, No. 8, 2250--2292 (2011; Zbl 1237.60058); \textit{A. Mielke}, Calc. Var. Partial Differ. Equ. 48, No. 1--2, 1--31 (2013; Zbl 1282.60072)]. The discrete equation is obtained as a spatial finite-volume discretization, which has a natural interpretation as a time-continuous Markov chain. In this way, both the discretization and the limit have a gradient structure. The main result shows that the functional characterizing the discrete gradient structure satisfies a liminf estimate with respect to the one associated with the Fokker-Planck equation. They complete the picture by constructing a recovery sequence. Hence, the content of the paper is an evolutionary \(\Gamma\)-convergence result for a discrete gradient structure converging towards the Wasserstein gradient structure of the Fokker-Planck equation.