Parallelized multilevel fast multipole algorithm for scattering by objects with anisotropic impedance surfaces

A parallelized multilevel fast multipole algorithm (MLFMA) is presented for simulating electromagnetic scattering from complex targets with anisotropic impedance surfaces. By employing both surface electric and magnetic currents as unknowns and weakly enforcing the anisotropic impedance boundary condition, a combined integral equation is formulated to generate a set of well-conditioned linear systems to be solved by MLFMA. To further improve the iterative convergence of the linear systems, a parallel sparse approximate inverse preconditioner is constructed from the near-field interaction of the system matrix. The MLFMA is parallelized to enable computation on a large number of processors for large-scale problems. Several numerical examples are presented to validate the algorithm and demonstrate its accuracy, scalability, and capability in handling large complex objects with anisotropic impedance surfaces. Copyright (c) 2014 John Wiley & Sons, Ltd.