Integration of a Structural Coherence Model with a Semi-Universal Machine Learning Interatomic Potential to Predict Molten Salt Thermal Conductivity

Abstract

The accurate design and performance assessment of energy generation systems that use molten salts as working fluids require accurate characterization of thermal conductivity, with a sufficiently low uncertainty. However, experimental data on several molten salt systems, of interest for various energy generation applications, are generally lacking. Physics-informed models, as well as molecular dynamic (MD) simulations, have the potential to populate gaps in current thermophysical property databases. Existing predictive models using MD simulation (Green–Kubo, non-equilibrium MD) take significantly longer to predict thermal conductivity than classical MD models that otherwise predict salt density with reasonable accuracy. We present an approach that uses the SuperSalt machine learning interatomic potentials (MLIPs) for chloride salts, as well as a structural coherence model (SCM) to predict thermal conductivity of molten salts using volumetric heat capacity, adiabatic speed of sound, and the partial pair distribution function. The SCM uses this function to predict the mean free path of thermal transport and applies generalized kinetic theory to predict thermal conductivity near the melting temperature, achieving results one to two orders of magnitude faster than existing MD methods. By also predicting the isothermal bulk modulus, the temperature dependence of thermal conductivity above the melting temperature can be determined. This integrated SCM-MLIP (ISM) methodology serves as a proof of concept for using MLIPs to predict thermal properties within current experimental uncertainties at a fraction of the computational cost. This enables investigation into thermal conductivity as well as other thermophysical properties across a variety of molten salt compositions. When coupled with targeted experimental studies, the ISM methodology can help contribute to existing databases necessary to safely implement molten salts into power generation systems.