Novel, parallel Monte Carlo simulations of systems of interacting classical particles

This dissertation consists of two parts describing the use of Graphical Processing Unit (GPU) accelerated Monte Carlo simulations to attack problems in statistical physics. In Part (1) we use the parallel tempering algorithm on a GPU and perform large-scale Monte Carlo simulations oftwo different Ising models on square lattices subject to uniform magneticfields with competing interactions: (a) antiferromagnetic (repulsive) nearest-neighbor (NN) and next-nearest-neighbor (NNN)interactions of the same strength and (b) two-body and three-body interactions. In both cases, phase diagrams are obtained and critical behaviors are studied. In part (2) we study the temperature dependence of structural properties and thermodynamic behavior of water clusters using Wang-Landau sampling. Five different empirical water models are compared, base on which the suggestion for the choice of the water model in Monte Carlo simulations for water clusters and/or inclusions is provided. We also present an implementation for the massively parallel Wang-Landau sampling on multiple GPUs. For GPU applications in both parts, we obtain a 50 ~ 200 times speedup comparing to corresponding codes on the current generation Intel Central Processing Unit (CPU).