SlotFFT techniques for fast computation of large and periodic electromagnetics problems

Large periodic structures in computational electromagnetics need the management of large number of unknowns, implying high memory use and computation time. Apart from fast techniques such as the multilevel fast multipole algorithm, this kind of problems might be accelerated taking advantage of periodicity properties. We will present some techniques based on the application of the FFT to the method of moments for periodic structures (techniques that we called slotFFT) and some compression utilities based on macrobasis calculation, in order to further increase the sizes of the structures analyzed. The costs of these techniques are roughly O(N) in memory usage and O(N log N) or less in CPU time.