Quantitative nonlinear homogenization: control of oscillations (Q6165026)

It is classical in homogenization theory the construction of correctors which, contrarily to the solutions of the macroscale problem, provide an approximation of the solutions in a strong topology. The goal of the quantitative homogenization is to estimate the error between the corrector and the solution. This paper is devoted to solve this problem in the case of nonlinear stochastic homogenization. Namely, the authors consider a nonlinear elliptic problem corresponding to a monotone operator with \(p\)-growth. The model problem corresponds to \(-\nabla\cdot A(\frac{x}{\varepsilon})(1+|\nabla u_\varepsilon|^{p-2})\nabla u_\varepsilon=f\) where \(2\leq p<\infty\) in dimension \(d\leq 3 \) and \(2\leq p<2(d-1)/(d-3)\) for \(d\geq 4\). The matrix function \(A\) satisfies the usual uniform ellipticity and bounding conditions, it is smooth and chosen randomly. The authors estimate the difference between the solutions of this equation and the corrector in the \(H^1\) topology. For this purpose, they provide annealed Meyers' estimates for the linearized operator.