About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Circular Metallurgy: Design, Technology, Application
|
| Presentation Title |
Predicting Melting Temperature of Molten Salt Electrolytes with Universal Machine Learning Interatomic Potentials |
| Author(s) |
Nathan Zou, Alexander Urban |
| On-Site Speaker (Planned) |
Nathan Zou |
| Abstract Scope |
Mitigating carbon emissions associated with conventional metal extraction requires cleaner metallurgical technologies, and molten salt electrolysis offers a promising alternative. Experimental characterization of molten salt electrolytes across diverse chemistries is challenging, motivating predictive modeling of electrolyte properties. Melting temperature is critical for defining the operating window of molten electrolytes and provides a benchmark for atomistic models to describe both crystalline and molten phases. Here, we predict melting temperatures of halide salts using first-principles-based universal machine-learning interatomic potentials (MLIPs) with a coexistence simulation approach, demonstrating efficient modeling of solid-liquid phase behavior with improved transferability over classical potentials. By incorporating long-range dispersion into the MLIP-driven simulation, we show that physics-informed corrections can compensate for missing interactions in the underlying DFT training data. This work establishes a predictive atomistic framework for molten halides and provides a foundation for predicting eutectic phase behavior, solvation, and ionic conductivity, key properties in molten salt electrolyte design. |
| Proceedings Inclusion? |
Undecided |
| Keywords |
Modeling and Simulation, Computational Materials Science & Engineering, Pyrometallurgy |