On homogenization for non-self-adjoint locally periodic elliptic operators (Q1676344)

The author describes a homogenization result for non self-adjoint operators from \(H^{1}(\mathbb{R}^{d})^{n}\) to \(H^{-1}(\mathbb{R}^{d})^{n}\) written as \( \mathcal{A}^{\varepsilon }=D^{\ast }A^{\varepsilon }D=\sum_{k,l=1}^{d}D_{k}A_{kl}^{\varepsilon }D_{l}\) where \(D=-i\nabla \) and \( A^{\varepsilon }(x)=A(x,x/\varepsilon )\) with \(A_{kl}\in C^{0,s}(\mathbb{R} ^{d};L_{\infty }(Q))^{n\times n}\), the space of complex-valued functions in \( L_{\infty }(\mathbb{R}^{d}\times \mathbb{R}^{d})\) which are Hölder continuous with exponent \(s\in [ 0,1]\) in the first variable and periodic with respect to the lattice \(\mathbb{Z}^{d}\) in the second variable (here \(Q\) is the unit cell \([-1/2,1/2]^{d}\)). The author assumes that \( \mathcal{A}^{\varepsilon }\) is uniformly coercive in \(\varepsilon \), hence that it is \(m\)-sectorial, which allows to define \((\mathcal{A}^{\varepsilon }-\mu )^{-1}\) whenever \(\mu \) is correctly chosen. For the main result of his paper, the author introduces the solution \(N_{\xi }(x,\cdot )\) of the cell problem \(D^{\ast }A(x,\cdot )(DN_{\xi }(x,\cdot )+\xi )=0\), \( \int_{Q}N_{\xi }(x,y)dy=0\) and then the effective operator \(\mathcal{A} ^{0}=D^{\ast }A^{0}D\) with \(A^{0}(x)=\int_{Q}A(x,y)(I+D_{2}N(x,y))dy\). The main result of the paper proves that if \(\mu \notin spec(\mathcal{A}^{0})\) then, if \(s=0\) \((\mathcal{A}^{\varepsilon }-\mu )^{-1}\) converges in the operator norm on \(L_{2}(\mathbb{R}^{d})^{n}\) to \((\mathcal{A}^{0}-\mu )^{-1}\) and if \(s\in (0,1]\), the following convergence rate \(\left\| (\mathcal{A} ^{\varepsilon }-\mu )^{-1}f-(\mathcal{A}^{0}-\mu )^{-1}f\right\| _{L_{2}( \mathbb{R}^{d})}\leq C_{\mu }\varepsilon ^{s}\left\| f\right\| _{L_{2}( \mathbb{R}^{d})}\) holds true for every \(f\in L_{2}(\mathbb{R}^{d})\) and \( \varepsilon \) close to 0. For the proofs, the author establishes continuity properties of the operator \(\mathcal{K}_{\mu }^{\varepsilon }\) from \(L_{2}( \mathbb{R}^{d})^{n}\) to \(H^{s}(\mathbb{R}^{d})^{n}\) defined as \(\mathcal{K} _{\mu }^{\varepsilon }f(x)=\int_{Q}N(x+\varepsilon z,\varepsilon ^{-1}x)D( \mathcal{A}^{0}-\mu )^{-1}f(x+\varepsilon z)dz\). He then defines a corrector operator and he establishes properties of this operator. When \(s=0\), the author essentially uses tools from his previous paper [Funct. Anal. Appl. 50, No. 1, 71--75 (2016; Zbl 1343.35022); translation from Funkts. Anal. Prilozh. 50, No. 1, 85-89 (2016)]. When \(s\in (0,1]\) the author uses tools from [\textit{T. A. Suslina}, St. Petersbg. Math. J. 26, No. 4, 643--693 (2015; Zbl 1316.35024); translation from Algebra Anal. 26, No. 4, 195--263 (2014); \textit{M. Sh. Birman} and \textit{T. A. Suslina}, St. Petersbg. Math. J. 17, No. 6, 897--973 (2006; Zbl 1175.35007); translation from Algebra Anal. 17, No. 6, 1--104 (2005)].