Randomized computations in domain decomposition methods

An efficient and robust FETI-2λ domain decomposition method (DDM) framework for electromagnetic (EM) modeling is outlined. The proposed framework uses randomized algorithms to approximate the low-rank discrete Dirichlet-to-Neumann (DtN) map interactions that arise in FETI-2λ. The resulting approach is also combined with effective and robust local and global DDM preconditioners. A realistic numerical example is given to verify the effectiveness, efficiency and robustness.