Quantitative results on the corrector equation in stochastic homogenization (Q1676092)

The authors derive optimal estimates in stochastic homogenization of linear elliptic equations in divergence form in dimensions \(d\geq2\). The first main result of the paper is following. Let \(\langle\cdot\rangle\) be a stationary ensemble that satisfies spectral gap, and let \(\bar\phi,\bar\phi'\) and \(\bar\phi_{T},\bar\phi'_{T}\) denote the corrector and adjoint corrector, and modified corrector and modified adjoint corrector for respective directions \(\xi,\xi'\in R^{d}\), \(|\xi|=|\xi'|=1\), and \(T>0\). For all \(L>0\) consider the random matrix \(A_{T,L}\) defined by \(\xi'\cdot A_{T,L}\xi := \int_{R^{d}}(T^{-1}\bar\phi'_{T}(x)\bar\phi_{T}(x)+(\xi'+\nabla\bar\phi'_{T}(x))\cdot A(x)(\xi+\nabla\bar\phi_{T}(x)))\eta_{L}(x)dx\), where \(x\to\eta_{L}(x)\) is a smooth averaging function on \(B_{L}=\{z:\;| z|<L\}\) such that \(\int_{R^{d}}\eta_{L}(x)dx=1\) and \(\sup|\nabla\eta_{L}|\leq C L^{-d-1}\), where \(C\) is a constant which only depends on the dimension, the ellipticity constant, and the spectral gap constants. Then, for all \(T\gg1\), \(\text{var}[\xi'\cdot A_{T,L}\xi]\leq C L^{-2}\ln(2+\sqrt{T}/L)\) for \(d=2\), and \(\text{var}[\xi'\cdot A_{T,L}\xi]\leq C L^{-d}\) for \(d>2\). The second main result of the paper is following. Let \(\langle\cdot\rangle\) be a stationary ensemble that satisfies spectral gap, and let \(\phi_{T},\phi'_{T}\) denote the modified corrector and modified adjoint corrector for directions \(\xi,\xi'\in R^{d}\), respectively, \(|\xi|=|\xi'|=1\), and \(T>0\). The approximation \(A_{T}\) of the homogenized matrix \(A_{\mathrm{hom}}\) defined by \(\xi'\cdot A_{T}\xi:=\langle(\xi'+D\phi'_{T})\cdot A(0)(\xi+D\phi_{T})\rangle\), where \(D\) is differential operator, satisfies, for \(T\gg1\), \(| A_{\mathrm{hom}}-A_{T}|\leq CT^{-1}\) for \(d=2\), \(| A_{\mathrm{hom}}-A_{T}|\leq CT^{-3/2}\) for \(d=3\), \(| A_{\mathrm{hom}}-A_{T}|\leq CT^{-2}\ln T\) for \(d=4\), \(| A_{\mathrm{hom}}-A_{T}|\leq CT^{-2}\) for \(d>4\).