97 / 2026-03-29 17:48:40

Molecular Dynamics Simulation of Primary Radiation Damage in Tungsten using a Deep Learning Interatomic Potential

deep learning,Tungsten,irradiation,grain boundaries

极端条件下的基础物理 > 极端条件下的原子物理、核物理与等离子体物理

Tungsten as an excellent material for fusion reactor applications, is subjected to extremely harsh irradiation environments within fusion reactor settings. Molecular dynamics (MD) simulations offer a powerful approach to elucidate irradiation damage mechanisms at the atomic scale, which is critical for understanding the macroscopic property degradation of tungsten under such complex and severe conditions. However, the accuracy of MD simulations is significantly affected by the inherent limitations of existing interatomic potentials. In this study, we employed a deep learning potential function, dp-hyb-zbl, trained within a framework that integrates the three-body embedded descriptor with the deep potential (DP) model, to simulate the primary irradiation collision cascade in tungsten. The effects of primary knock-on atom (PKA) energy, temperature, and grain boundaries on defect evolution were systematically investigated. Our results demonstrate that the dp-hyb-zbl model exhibits superior predictive capability in forecasting dislocation loop formation compared to conventional potentials. Furthermore, the influence of irradiation-induced dislocation loops on the mechanical properties of tungsten was analyzed, providing insights into how these nano-scale defects modulate the macroscopic performance of irradiated materials.