Multiscale techniques for solving eigenvalue problems

Axel Målqvist
Department of Mathematical Sciences
University of Gothenburg


We consider numerical approximation of linear and non-linear eigenvalue problems, using the Localized Orthogonal Decomposition (LOD) technique. In this approach a low dimensional generalized finite element space is constructed, by solving localized (in space) independent linear stationary problems. The eigenvalue problem is then solved in the computed low dimensional space at a greatly reduced computational cost.