Calderón-Zygmund estimates for stochastic homogenization (Q892687): Difference between revisions

The authors consider qualilinear uniformly elliptic equation of the form \(-\nabla\cdot(a(\nabla u,x))=\nabla\cdot f\) in \(B_{R}\subseteq \mathbb{R}^{d}\), where \(\xi \to a(\xi,x)\) is a Lipschitz, uniformly monotone map, the maps \(x\to a(\xi,x)\) are stationary random fields satisfying a finite range of dependence assumption. For fixed \(2<m<p\), \(s\in (0,4/(m+2))\) and \(R\geq10\), the authors prove that there exist \(C\geq1\), \(k\geq1\) and nonnegative random variable \({\mathcal Y}_{R}\), \(E[\exp({\mathcal Y}_{R}^{s})]\leq C\) such that for every \(a\in\Omega\), \(f\in L^2(B_{R};R^{d})\) and solution \(u\in H^1(B_{R})\) of considered equation, we have the estimate \[ \||\nabla u|^2\|_{{-\!\!\!\! L}_1^{m/2}(B_{R/2})}\leq {\mathcal Y}_{R}^2 M^2(\log(2+M))^{k}, \] where \(M=\left(\||\nabla u|^2\|_{{-\!\!\!\! L}^1(B_{R})}+\|| f|^2\|_{{-\!\!\!\! L}_1^{p/2}(B_{R})}\right)^{1/2}.\) Here, for \(h>0, s\in[1,\infty)\), \(\phi\in L_{loc}^1(R^{d})\) and \(U\subseteq \mathbb{R}^{d}\) it is defined \[ \|\phi\|_{-\!\!\!\!L_{h}^{s}(U)}=\left(-\!\!\!\!\!\!\!\!\;\int_{U}\left(-\!\!\!\!\!\!\int_{B_{h}(x)}|\phi(y)| dy\right)^{s}dx\right)^{1/s}, \quad \|\phi\|_{-\!\!\!\!L^{s}(U)}=\left(-\!\!\!\!\!\!\!\!\;\int_{U}|\phi(y)|^{s} dx\right)^{1/s}. \]