Multipath Metropolis Simulation of Classical Heisenberg Model

Abstract

Processor cores are becoming less expensive and thus more accessible. To utilize increasing number of available computing elements, good parallel algorithms are necessary. In light of these changes in contemporary computing, multipath Metropolis simulation of classical Heisenberg model is explored. In contrast to the original single-path algorithm, multipath simulation approach is inherently parallel because different random-walk paths are mutually independent. This independence enables easy and efficient harnessing of numerous cores' computing power in embarrassingly parallel algorithms. Aside form being inherently parallel, multipath simulation approach results in independent and normally distributed simulation output. Normal distribution enables simple and straightforward statistical processing. Thus, multipath simulation results can be easily computed with arbitrary and statistically known precision.