The present study attempts to find an alternate computational tool to model the complex physical interactions within the molten salt FLiNaK in a way that is both efficient and accurate. Additionally, this study seeks to describe the effects of several different types of impurities on the FLiNaK salt system. This study selects two different polarizable force fields, the AMOEBA polarizable approach and the polarizable ion model, to determine the density and the structure of the impure FLiNaK salt mixtures at typical operating temperatures in molten salt reactors (between 500-900 °C). This study conducts ab initio molecular dynamics (AIMD) simulations and classical molecular dynamics (CMD) for these salt mixtures to determine the correct parameter set for these two force fields. This study also uses an optimizer to minimize the difference between the forces calculated with AIMD and CMD simulation data. The AMOEBA polarizable approach is able to predict density for FLiNaK; however, it is unable to reliably predict other thermophysical properties due to the instability of its CMD simulations. The polarizable ion model is able to reliably determine density and salt structure for pure and impure FLiNaK mixtures. This model can be further used to determine other thermophysical properties. The polarizable ion model predicted densities for four impure salt mixtures: FLiNaK-MoF3, FLiNaK-UF3, FLiNaK-CsF, and FLiNaK-ZrF4. The predicted densities at 700 °C given in kg/m3 are 1929.94, 2454.15, 1650.67, and 1961.87, respectively with an error compared to the additive density model of -2.51%, -5.79%, -17.15%, and -1.67%, respectively. This study presents the radial distribution function and density correlation functions for each salt mixture. This study also presents a discussion of the shortcomings of the AMOEBA polarizable approach, as well as further work that may be done with these tools.

This work 1) draws insights on molten salt structure and properties directly from ab initio molecular dynamics (AIMD) simulations, 2) demonstrates the advantageous symbiosis of computational and experimental collaborations on molten salt research, and 3) simultaneously generates ab initio data sets for fitting an interatomic potential model for classical molecular dynamics (MD) simulations. This work discusses the motivations for AIMD simulations of molten salts, thermophysical properties and structural characteristics of interest, advanced methodologies for AIMD simulations, and several completed AIMD studies on molten salts. Of import are the methodological contributions of this work to AIMD simulations, primarily the radical increase in generalized gradient planewave energy cutoff used to more accurately model the electron distribution across a highly-polarizable molten salt. Cutoffs of up to 2500 Rydbergs are used in this work, but 2000 Rydbergs is found to be sufficient for most AIMD NpT modelling of molten fluorides. The equilibrium liquid density of eutectic FLiNaK as a function of temperature is found to agree with the experimental density reported by Chrenkova et al. to within 0.2%, and the equilibrium liquid density of eutectic FMgNaK is found to agree with experimental measurements reported herein to within 4%. Self-diffusion coefficients in FMgNaK are also considered, with applicability to other halide salts. Molybdenum, Cesium, iodide, nickel, hydrogen, oxide, and uranium complexation are examined. It is found that solvation strength can be qualitatively determine via AIMD simulations, and that poorly solvated solutes will minimize the surface area of interaction with the salt solution. Cesium in particular is shown to be volatile or retainable in FLiBe at 500, 650, and 800 ËšC based on complexation and validated experimentally. It is shown that the chemical potential of an anion varies between melts as influenced by the different cations present in each melt. Hence, attempts to use a common electrochemical reference reaction for different salt mixtures are at best an approximation.

No description available.