Multistage nucleation pathway in LiF molten salt mirrors the crystal-melt interface structure

Area of Science:

• Materials Science
• Chemical Physics
• Computational Chemistry

Background:

• Fundamental questions persist regarding crystal nucleation processes from melts and solutions.
• Classical nucleation theory may not fully capture complex kinetic pathways observed in various systems.
• Studies on ionically bonded systems, particularly molten salts, are lacking despite their importance.

Purpose of the Study:

• To develop a machine learning interatomic potential for the model ionic system LiF.
• To investigate the homogeneous crystal nucleation pathway in undercooled LiF melts.
• To connect the crystallization pathway with the equilibrium crystal-melt interface structure.

Main Methods:

• Development of a machine learning interatomic potential with quantum-level accuracy for LiF.
• Microsecond-scale molecular dynamics simulations of LiF nucleation over a range of undercoolings.
• Establishment and application of local order parameters to analyze simulation data.

Main Results:

• The developed potential accurately reproduces experimental properties of LiF across wide temperature and pressure ranges.
• Homogeneous nucleation initiates from liquid regions exhibiting slow dynamics and high bond orientational order.
• Precritical and postcritical nuclei show specific local structures (HCP, BCC, FCC) during the nucleation process.

Conclusions:

• The study provides unprecedented insights into the complex nucleation pathways of ionically bonded melts.
• Machine learning potentials enable efficient, large-scale simulations crucial for understanding nucleation dynamics.
• A link is established between the crystallization pathway and the equilibrium crystal-melt interface structure.