High order correctors and two-scale expansions in stochastic homogenization (Q1682502)

This paper deals with high order correctors in stochastic homogenization. Let \(\mathbf{B}\) be the set of nearest neighbor edges in \(\mathbb Z^d\), and \(\{e_i, i=1,\dots,d\}\) be the canonical basis of \(\mathbb Z^d\). On a probability space \((\Omega,\mathcal F, P)\) it is defined a sequence of i.i.d. non-degenerate random conductances, denoted by \(\{\omega_l\}_{l\in \mathbf{B}}\) with \(\omega_l\in (\delta,1)\) for some \(\delta>0\). Let us define \(a(x)=\mathrm{diag}(\omega_{(x,x+e_1)},\dots,\omega_{(x,x+e_d)})\), \(a:\mathbb Z^d\to \mathbb R^{d\times d}\). The author considers the problem of homogenization for random partial differential equation \(-\nabla\cdot a(x/\varepsilon)\nabla u_\varepsilon(x)=f(x)\). It is assumed that \(\{\omega_l\}_{l\in \mathbf{B}}\) satisfies the log-Sobolev inequality with constant \(\rho>0\). One of the main results is following. When \(n\geq 2\), \(d\geq 2n-1\), for any \(p\geq 1\), there exists \(C=C(d,\delta,\rho,p)>0\) such that \(\langle|\nabla\psi_n^\lambda|^p\rangle\leq C\) uniformly in \(\lambda>0\). When \(d\geq 2n+1\), we have \(\langle|\psi_n^\lambda|^p\rangle\leq C\) uniformly in \(\lambda>0\). In particular, for the \(n\)-th order corrector \(\psi_n^\lambda\), it has a stationary gradient when \(d\geq 2n-1\), and it is stationary when \(d\geq 2n+1\). As an application of high order correctors, the author obtains the first and high order fluctuations in stochastic homogenization in a strong sense.