An optimal adaptive FEM for eigenvalue clusters (Q2350737)

The paper addresses the finite element approximation of multiple eigenvalues for the Laplacian which leads to the problem of eigenvalue clusters. The author presents and analyses an adaptive finite element method (FEM) based on \(P_1\) conforming elements for the eigenvalue computation. He proves the optimality of the algorithm in terms of the number of degrees of freedom compared to nonlinear approximation classes. All the estimates are explicit in the initial FE-meshsize, the eigenvalues and the cluster width, which allows to formulate explicit sufficient conditions for the optimal convergence. Numerical tests for non-convex 2d-domains confirm the optimal convergence rates of the method.