Investigation on Chemical Behavior of Ni and Fe Impurities in LBE by Deep Potential Molecular Dynamics Simulation

Abstract

The compatibility of liquid lead-bismuth eutectic (LBE) with structural materials is one of the problems in its application in advanced nuclear reactors. At present, the understanding of impurity atoms diffusion in liquid LBE is still limited. It is particularly difficult to study the chemical behaviors such as diffusion properties of impurity atoms in LBE by experimental methods. However, traditional theoretical calculations such as density functional theory (DFT) and classical molecular dynamics (MD) methods have dilemmas in simulation scale and simulation accuracy. To address those issues, a scheme based on DFT calculation, deep neural networks, and machine learning was introduced. By training on high-quality data sets generated by DFT calculations, three deep potential (DP) models of LBE, LBE-Ni, and LBE-Fe were constructed to describe the interaction between atoms. The results of AIMD calculation of radial distribution function (RDF) can be repeated by DPMD. By performing MD simulations with DP models, the microstructures of Ni, Fe impurities in LBE and the thermal physical properties of LBE were investigated. Meanwhile, the estimated thermophysical properties were discussed, including density, specific heat capacity, shear viscosity, and self-diffusion coefficient of Ni and Fe atoms. The higher RDF peaks of Bi-Ni and Bi-Fe reveal that both impurity atoms interact more strongly with Bi atoms. The predicted thermophysical properties are in good agreement with the experimental results, and have better accuracy than the results of the classical MD that based on the embedded atom method (EAM). In conclusion, a thorough understanding of the microstructure of impurities in LBE is provided and the data of the thermophysical properties of LBE are enriched.

Keywords

• lbe
• thermophysical property
• diffusion coefficient
• first-principle
• molecular dynamics
• machine learning