Acceleration near field interaction in half-space MLFMA by means of genetic algorithm

The discrete complex images methods(DCIMs) are commonly used in half space multilevel fast multipole algorithm(MLFMA) to approximate sommerfeld integral(SI) in spatial domain Green´s functions. Although DCIMs provide a much faster solution to evaluate the SIs comparing to numerical integral methods, the near field interaction calculation is still a dominant load in MLFMA due to the large number of complex images from DCIMs. To reduce number of complex images, a genetic algorithm(GA) is used only for near field only for calculating near field of Green´s functions. Thus the efficiency of near field matrix filling of half space MLFMA is improved as demonstrated by numerical examples.