Hard Simulation Problems in the Modeling of Magnetic Materials: Parallelization and Langevin Micromagnetics

Abstract

We present recent results on two attempts at understanding and utilizing large-scale simulations of magnetic materials. In the first study we consider massively parallel implementations on a Cray T3E of the n-fold way algorithm for magnetization switching in kinetic Ising models. We find an intricate relationship between the average time increment and the size of the spin blocks on each processor. This narrows the regime of efficient implementation. The second study concerns incorporating noise into micromagnetic calculations using Langevin methods. This allows measurement of quantities such as the probability that the system has not switched within a given time. Preliminary results are reported for arrays of single-domain nanoscale pillars.