Remarks on optimal rates of convergence in periodic homogenization of linear elliptic equations in non-divergence form (Q2218204)

The paper under review deals with the study of the optimal convergence rate in periodic homogenization of linear elliptic equations in non-divergence form. More precisely, the authors consider the boundary value problem \[ \left\{ \begin{array}{ll} -a_{ij}(x/\varepsilon)u^\varepsilon_{x_ix_j}=f(x) & \mathrm{in}\ U, \\ u^\varepsilon=g & \mathrm{on}\ \partial U, \end{array} \right. \] where \(U\) is a smooth bounded domain in \(\mathbb{R}^n\), \(A(y)=(a_{ij})_{1\leq i,j \leq n}\in C^2(\mathbb{R}^n,\mathbb{R}^{n^2})\) is a symmetric, \(\mathbb{Z}^n\)-periodic and positive definite matrix function, \(f \in C^2 (\overline{U})\), \(g \in C^4(\partial U)\). As is well-known, as \(\varepsilon \to 0\), \(u^\varepsilon\) converges to the solution of the homogenized problem \[ \left\{ \begin{array}{ll} -\overline{a}_{ij} u_{x_ix_j}=f(x) & \mathrm{in}\ U, \\ u=g & \mathrm{on}\ \partial U, \end{array} \right. \] where \(\overline{A}=(\overline{a}_{ij})_{1\leq i,j \leq n}\) is the effective matrix. The authors study the optimal rate of convergence of \(u^\varepsilon\) to \(u\) and they prove that is either \(O(\varepsilon)\) or \(O(\varepsilon^2)\) depending on the diffusion matrix \(A\), the source term \(f\), and the boundary data \(g\).