Akiko Masaki, Takafumi Suzuki, Kenji Harada, Synge Todo, Naoki Kawashima
Based on the worm algorithm, we propose a general quantum Monte Carlo algorithm suitable for parallelization with domain-decomposition. Of particular importance is its application to simulations of huge lattice systems of bosons and spins on a distributed-memory computer. A large number of worms are introduced and its population is controlled by a fictitious transverse field. For a benchmark, we demonstrate the size-dependence of the Bose-condensation order parameter by applying the algorithm to the hardcore Bose-Hubbard model with $L\times L\times \beta t = 10,240\times 10,240\times 16$ using 3,200 computing cores of Fujitsu's FX10, which shows good parallelization efficiency.
View original:
http://arxiv.org/abs/1307.0328
No comments:
Post a Comment