First-principles molten salt phase diagrams through thermodynamic integration.

Area of Science:

• Computational Materials Science
• Physical Chemistry
• Condensed Matter Physics

Background:

• Predicting material phase diagrams via molecular dynamics requires balancing accuracy with computational cost.
• Accurate interatomic potentials and large-scale simulations are often computationally prohibitive.

Purpose of the Study:

• To develop a rapid, first-principles method for predicting solid-liquid phase boundaries.
• To assess the accuracy of different density-functional theory (DFT) functionals for phase boundary prediction.
• To enable accurate ab initio predictions of phase diagrams for various materials.

Main Methods:

• Employing thermodynamic integration from low-cost force fields to neural network potentials.
• Training neural network potentials using density-functional theory (DFT) data.
• Calculating the solid-liquid phase boundary for the model salt NaCl.

Main Results:

• Achieved rapid, first-principles prediction of the solid-liquid phase boundary for NaCl.
• Identified the critical role of dispersion interactions in DFT functionals for accurate phase boundary prediction.
• Demonstrated that the majority of computational cost is at the level of classical potentials.

Conclusions:

• The developed method enables accurate ab initio prediction of solid-liquid phase boundaries for any material.
• This approach significantly reduces the computational expense compared to traditional methods.
• Accurate prediction of phase diagrams is crucial for materials design and discovery.