Assessing the EDIP potential for atomic simulation of carbon diffusion, segregation and solubility in silicon melt

Abstract

Molecular simulation based on empirical interatomic potential is an important tool in the study of impurity thermophysical properties, e.g. carbon diffusion, segregation and solubility in liquid silicon, which are key factors determining the carbon concentration and distribution during silicon crystal growth. The Tersoff potentials, although being very successful in describing many silicon and carbon properties, overestimate the silicon crystal melting temperature severely which incurs many criticism and restrictions when comparing the simulation results with experiments. In this study, the EDIP potential, which predicts an experimental comparable silicon melting temperature, was assessed whether it can reasonably capture the carbon diffusivity, segregation and solubility properties in silicon melt while avoiding the well-known temperature issue. We found that the EDIP potential, without requiring scaling the temperature, predicts consistent carbon and silicon diffusion coefficients with other potentials. The segregation coefficient estimated by EDIP is slightly higher than the experimental results. It fails in capturing the carbon solubility property by overestimating nearly two orders of magnitude higher than the experimental values. Based on potential energy and entropy calculation, we suggest that the problem of the EDIP potential is caused by the high entropy of the silicon melt phase. Overall, the research will provide useful information for other thermophysical studies based on EDIP potential and also be benefit to empirical potential optimization.