On the accuracy of FFT based magnetostatic field evaluation schemes in micromagnetic hysteresis modeling

Micromagnetic hysteresis models for large, bulk like samples are useful for the identification of relations between microscopic material properties and macroscopic magnetic behavior. To bridge the gap between the nanometer space scale of the micromagnetic theory and the large sample dimensions, time and memory efficient numerical schemes are needed. In micromagnetic computations, fast Fourier transforms (FFTs) have been widely adopted to speed up magnetostatic field computations. In this paper, two FFT schemes are compared. The first scheme evaluates the magnetostatic field directly starting from the magnetization and has a large accuracy, while in the second scheme the magnetostatic field is derived from the scalar magnetic potential resulting in a reduced accuracy but also in a CPU time reduction for a magnetostatic field evaluation to 65% and a reduction of memory requirements to 55%. The influence of the low accuracy evaluations on the simulated macroscopic hysteresis behavior is studied. Therefore, comparison is made with the influence of thermal effects in hysteresis simulations. It is found that the resulting changes in macroscopic hysteresis behavior are of the same order of magnitude as the ones obtained when thermal fluctuations are taken into account in the high accuracy computations.