Data of the figures and machine learning potentials for "A molecular dynamics study on local structural features and thermophysical properties of the NaCl-UCl₃ molten salt with machine learning potential"

NaCl-UCl₃molten salt is a potential fuel for the molten salt reactor due to its ability to dissolve higher actinides and excellent thermophysical properties. The composition of the NaCl-UCl₃salt changes due to the burn-up during the operation of the reactor, which makes it important to study the local structural features and determine the thermophysical properties for the entire composition space.Among the existing molecular dynamics (MD) methods, <i>ab initio</i> molecular dynamics (AIMD) simulations provide the highest accuracy. However, the AIMD calculations are computationally expensive and not feasible to calculate dynamic properties such as viscosity and thermal conductivity. In this work, using atomic cluster expansion (ACE), we developed the first transferable machine learning interatomic potential (MLIP) trained with density functional theory (DFT) data that covers the entire compositional space of the binary NaCl-UCl₃ melt, and calculates the thermophysical properties of the binary melt with an accuracy of <i>ab initio</i> level and at a low computational cost equivalent to empirical interatomic potentials.The calculated useful thermophysical properties for the nuclear applications, such as density, excess molar volume, heat capacity, viscosity, and thermal conductivity will enrich the database of salt properties. In addition, a comprehensive analysis of the local structural features correlating the changes in the thermophysical properties will enhance the understanding of salt chemistry.