We develop a new abstract strategy for proving ergodicity with explicit computable rate of convergence for diffusions associated with a degenerate Kolmogorov operator L. A crucial point is that the evolution operator L may have singular and nonsmooth coefficients. This allows the application of the method e.g. to the degenerate N -particle Langevin dynamics.
IntroductionStudying the decay to equilibrium of degenerate kinetic equations is still an active and demanding mathematical research area lying in between modern Stochastics and Functional Analysis. Especially in the last decade, many results concerning the relaxation to equilibrium of the degenerate Langevin dynamics have been obtained in case the underlying potential is sufficiently smooth and nonsingular; the required mathematical tools e.g. rely on hypoellipticity, hypocoercivity or stochastic Lyapunov type techniques. However, in Statistical Mechanics the underlying potential in the Langevin dynamics is usually of Lennard-Jones type. Thus it is singular and, as far as we know, analyzing the decay to equilibrium in this singuar situation is not possible with the abovementioned methods. Nevertheless, suitable tools to handle this situation are provided in article [1] in which ergodicity with rate of convergence for the N -particle Langevin dynamics including singular interaction potentials is established. The method used in [1] is itself based on the strategy from [3]. Hence it is of interest to generalize the method from [1] and [3] to an abstract setting which is done below. We present a new abstract method in a Hilbert space framework which is suitable for proving ergodicity with explicit computable rate of convergence for various diffusion processes associated with a degenerate non-coercive Kolmogorov evolution operator L. Again the generator may have singular and nonsmooth coefficients. We mention that our newly developed method is based on an interplay between Functional Analysis (using operator semigroups) and Stochastics and relies on martingale methods. It is further interesting to note that we require conditions in the abstract setting similar to the assumptions made in an abstract hypocoercivity setting in [2] (and see [4]).The underlying proceeding summarizes the abstract method obtained in our article [5] (or see [6, Ch. 3]). We do not present any proofs. For all details, the reader is referred to the mentioned reference. Therein, also applications of the abstract method to the Langevin dynamics and the generalized fiber lay-down process arising in industrial mathematics are presented.