A Quantitative Theory of Stochastic Homogenization
Speaker: Scott Armstrong, CNRS Université Paris-Dauphine
Location: Warren Weaver Hall 1302
Date: Monday, February 1, 2016, 3:45 p.m.
Stochastic homogenization involves the study of solutions of partial differential equations with random coefficients, which are assumed to satisfy a "mixing" condition, for instance, an independence assumption of some sort. One typically wants information about the behavior of the solutions on very large scales, so that the ("microscopic") length scale of the correlations of the random field is comparatively small. In the asymptotic limit, one expects to see that the solutions behave like those of a constant-coefficient, deterministic equation. In this talk, we consider uniformly elliptic equations in divergence form, which has applications to the study of diffusions in random environments and effective properties of composite materials. Our interest is in obtaining quantitative results (e.g., error estimates in homogenization) and to understand the solutions on every length scale down to the microscopic scales. In joint work with Tuomo Kuusi and Jean-Christophe Mourrat, we introduce a new method for analyzing this problem, based on a higher-order regularity theory for equations with random coefficients, which, by a bootstrap argument, accelerates the exponent representing the scaling of the error the all the way to the optimal exponent given by the scaling of the central limit theorem.