Hierarchical fractional-step approximations and parallel kinetic Monte Carlo algorithms (Q695157)

It is well known that kinetic Monte Carlo (KMC) algorithms are a very important tool for the simulation of out-of-equilibrium, spatially distributed processes. Such models have many applications in material science, catalysis and complex biological processes. Typically, they are used to simulate models in transport micro-mechanisms for atoms and molecules, adsorption, desorption processes, and diffusion on surfaces. Moreover, they have applications in agent-based, evolutionary games problems in epidemiology, ecology, and even traffic networks. In this paper, we have a presentation of a mathematical framework for constructing and analyzing parallel algorithms for lattice kinetic Monte Carlo (KMC) simulations. The obtained algorithms have the capacity to simulate a wide range of spatio-temporal scales in spatially distributed, non-equilibrium physiochemical processes with complex chemistry and transport micro-mechanisms. There is a possibility to use these algorithms for specific hierarchical parallel architectures such as multi-core processors or clusters of Graphical Processing Units (GPUs). This paper is divided into eight sections. After the preliminaries that show the general mathematical framework, we have a presentation of fractional step kinetic Monte Carlo algorithms. In Section 3, there is a description of key features of the fractional step methods in the Processor Communication Schedule (PCS). After this the authors present a formal argument for the error analysis of the fractional step approximations for KMC, which suggests the order of convergence of the schemes. Section 5 shows a possible application for a Graphical Processing Unit (GPU). Section 6 is devoted to the introduction of a probabilistic strategy to re-balance the workload dynamically during the simulation based on the following idea from Mass Transport methods. Next we have benchmark tests for the proposed methods.