Center for Computational Electromagnetics of Complex Structures

Abstract : On the frequency domain fast solvers front, we have: We have continued our focus on improving the Fast Illinois Solve Code (FISC) which has a user base of over 400 distributed by SAIC/DEMACO. To stress the code for large scale computing, we put FISC through a number of tour de force computations. Recently, we have put FISC through a broad bandwidth calculation where we can simulate the time-domain scattering from a fill-size aircraft with frequency content up to 1 GHz. We continued to make progress with the parallelization of MLFMA with the development of ScaleME. By the end of April of 2001, ScaleME was able to solve a 10 million unknown problem. The code is about 7-9 times faster that FISC (Fast Illinois Solver Code) that solved a similar size problem about over two year ago. The speedup compared to FISC essentially comes from the good scaling property of ScaleME for massively parallel computing. We used 128 processors on the NCSA 561 02K. In addition, we have looked at ways to solve Maxwell's equations with fast solvers all the way from static to microwave frequencies. This simulation capability will encompass circuit theory as a special case, allowing antenna impedances and circuit phenomena to he more accurately modeled. We have also developed a new fast solver when the scatterer is placed on top, or embedded in a layered earth. This capability will have significance in studying the radar returns from targets placed under tree canopy, or in studying scattering from buried objects. With this code, we can solve problems with over a million unknowns easily.