Highly efficient parallel grand canonical simulations of interstitial-driven diffusion-deformation processes

Abstract

Diffusion of interstitial alloying elements like H, O, C, and N in metals and their continuous relocation and interactions with their microstructures have crucial influences on metals properties. However, besides limitations in experimental tools in capturing these mechanisms, the inefficiency of numerical tools also inhibits modeling efforts. Here, we present an efficient framework to perform hybrid grand canonical Monte Carlo and molecular dynamics simulations that allow for parallel insertion/deletion of Monte Carlo moves. A new methodology for calculation of the energy difference at trial moves that can be applied to many-body potentials as well as pair ones is a primary feature of our implementation. We study H diffusion in Fe (ferrite phase) and Ni polycrystalline samples to demonstrate the efficiency and scalability of the algorithm and its application. The computational cost of using our framework for half a million atoms is a factor of 250 less than the cost of using existing libraries.