Scientific Achievement
Researchers in the MINES program determined that crystal nucleation from undercooled LiF melt passes through intermediates that can be predicted from equilibrium crystal–melt interface structure, simplifying our understanding of complex transformations.
Significance and Impact
Quantum-accurate molecular dynamics (MD) reveal that predicting transformation pathways far-from equilibrium may not require elaborate simulations, and be possible using simple equilibrium configurations.
Research Details
- LiF machine learning potential for liquid and solid
- >106 atom MD simulations of nucleation
- >10,000 repeats from an initial trajectory for statistical robustness
- Developed order parameters to distinguish phases
Publication Details
Z. Fan, D. Rawat, P. Zarzycki, M.L. Whittaker, M. Asta. Proceedings of the National Academy of Sciences (2025).
Work was performed in the Lawrence Berkeley National Lab and used in NERSC.
