Adaptive Iterative Numerical Homogenization for Quasilinear Nonmonotone Elliptic PDE

Abstract

We propose and analyze an adaptive iterative numerical homogenization method to approximate the solution of a class of quasilinear nonmonotone elliptic problems that is of multiscale nature. The method is based on the technique of the Localized Orthogonal Decomposition (LOD) applied to the linear problems in each step of a Kačanov iteration. In this approach, the multiscale basis is recomputed adaptively in each iteration and a linear problem is solved with this updated multiscale space, where in both steps the nonlinearity is evaluated using the approximation of the previous iteration. As a key component of the proposed approach, we present a locally computable error indicator, which at each iteration identifies the basis functions requiring updates, while previously computed basis functions are retained whenever possible. We provide a priori error estimates and show convergence of the method, requiring only higher integrability of the right-hand side, but no higher differentiability of the solution itself. Furthermore, we discuss how to adapt the proposed adaptive iterative LOD in the context of Newton's method as iteration scheme. Numerical experiments illustrate the theory and validate the applicability of the proposed method.