Large deviations, central limit theorems and L^pconvergence for Young measures and stochastic homogenizations (Q4230643)

The authors study stochastic homogenization of a random ` medium'\ or ` material'\ given by a random function \( a : \Delta \rightarrow Z, \) where \(\Delta\) and \(Z\) are topological spaces. Instead of \(a\), one considers the random Young measure \(\nu\) given by \( \nu(B): = \int_\Delta I_B(x, a(x)) dx \) where \(B\) is a Borel set of \(\Delta \times Z\). Two models for homogenization are investigated: In the first model, called model B, introduced in the one-dimensional case by \textit{P. Baldi} [Ann. Mat. Pura Appl., VI. Ser. 151, 161-177 (1988; Zbl 0654.60024)], the material has random values on a fixed sequence of partitions (for instance, the \(n\)th partition could consist of cubes of the grid \(n^{-1/d} \mathbb Z\)). In the second model, called model FRp [` Poissonian model of Facchinetti and Russo' -- this is a modification of a model introduced by \textit{G. Facchinetti} and \textit{L. Russo}, Boll. Unione Mat. Ital., VI. Ser., C, Anal. Funz. Appl. 2, 159-170 (1983; Zbl 0534.49011)], the material has only two values but the size of the cells of the \(n\)th partition are random. For both models, the authors prove the following results on the speed of convergence of the Young measures \(\nu_n\) associated with the random function \(a_n\) belonging to the \(n\)th partition: 1. The distribution of the Young measures \(\nu_n\) satisfies a large deviation principle. The rate function is given explicitly for both models: it is convex, has compact level sets and vanishes at the limit measure \(\nu_\infty\). 2. For any bounded Hölder continuous function \(f\), \(E[| \nu_n(f) -\nu_\infty(f)| ^p]\) decreases polynomially in \(n\) with an explicitly given rate. 3. For any bounded continuous function \(f\), one has a central limit theorem, i.e., \(n ^{1/2}(\nu_n(f) - \nu_\infty(f))\) converges in law to a centered Gaussian random variable with an explicitly given variance \(\sigma ^2 = \sigma ^2 (f)\).