Efficient and accurate parallel computation of electromagnetic scattering by extremely large targets

For the parallelization of multilevel fast multipole algorithm (MLFMA), it is very efficient to adopt different strategies to parallelize plane waves and translators in MLFMA at different levels. But when the size of the targets increasing to several hundreds wavelengths, the procedures to compute distances of boxes becomes the bottle-neck to further improve the capability of parallel MLFMA. Encouraged by these, a new parallel recursive approach to efficiently handle distances of boxes is proposed. The validity of the proposed parallel approach is proved by numerical experiments. The accuracy and efficiency of the proposed parallel algorithm are checked by calculating the RCS of a perfect conducting sphere with a diameter of 200 wavelengths and modeled by over 23 millions unknowns. It shows that the approach successfully eliminates the bottle-neck mentioned and greatly improves the capability of parallel MLFMA. The parallel approach is further proved to be general and capable for extra large target with complex structure by calculating the RCS of a plane model with a fuselage of 500 wavelengths long, which is the largest target having been solved by MLFMA.