Multilevel inverse-based factorization preconditioner for solving sparse linear systems in electromagnetics

We introduce an algebraic recursive multilevel approximate inverse-based preconditioner, based on a distributed Schur complement formulation. The proposed preconditioner combines recursive combinatorial algorithms and multilevel mechanisms to maximize sparsity during the facorization. Experiments on selected sparse linear systems from electromagnetics applications demonstrate the potential of the proposed method, also against other state-of-the-art solvers.