Investigating the effect of metallic solutes on the thermodynamic and transport properties of FLiBe molten salt via molecular dynamics and machine learning

Molten salts have attracted immense interest due to their potential applications in advanced energy systems, including high-field fusion, next-generation fission, and thermal energy storage. Impurities such as Cr and Ni that enter the salt via corrosion can significantly influence the thermophysical properties of molten salts. However, a detailed study of the effects of such impurities on molten salt properties is limited. In this work, we systematically investigate the effects of adding metallic solutes, namely, Cr, Ni, and Fe, on the thermodynamics and transport properties of the lithium-based molten salt (FLiBe - 66% LiF and 33% BeF₂, a prototypical salt) using ab-initio molecular dynamics (AIMD) and machine learning potential. A deep neural network potential is trained on a large and diverse dataset generated from AIMD simulations by varying parameters such as oxidation states, concentration, and temperature. The trained potential is then deployed to predict the transport properties of FLiBe at a longer time scale. Our study provides valuable insights into impurity effects in molten salts and sets a foundation for developing generalized machine learning potential for a wide spectrum of molten salts.