Probabilistic approach to homogenization of a non-divergence form semilinear PDE with non-periodic coefficients (Q2453522)

In their paper, the authors investigate homogenization for a semilinear partial differential equation (PDE) of non-divergence form. There has been a lot of work by different authors on stochastic approaches for homogenization of PDEs, but nearly exclusively in one of the both classical situations, that of deterministic periodic and that of random stationary coefficients. Unlike these classical cases, the authors do not suppose periodicity or ergodicity for the coefficients. Instead it is supposed that the coefficients admit Cesàro limits, and null-recurrence is assumed with respect to the fast variable. In the case of linear PDEs, \textit{R. Khasminskii} and \textit{N. Krylov} [Stochastic Processes Appl. 93, No. 2, 229--240 (2001; Zbl 1053.60060)] studied this problem. In the present paper, it is extended to semilinear PDEs. For this, the authors combine PDE methods with BSDE ones. Considering the stochastic interpretation of the PDE in form of a backward stochastic differential equation (BSDE), the authors make use Jakubowski's S-topology to study the averaged BSDE. Using compactness methods, they show that the associated limit PDE admits a continuous Sobolev solution, and the application of the Itō-Krylov formula allows them to give a stochastic interpretation of the solution of this limit PDE in form of the solution of the averaged BSDE.