The kinetics of the ice-water interface from ab initio machine learning simulations

Area of Science:

• Computational physics
• Materials science
• Chemical physics

Background:

• Molecular simulations are crucial for understanding ice growth.
• Ab initio accuracy in simulations requires advanced computational techniques for large systems and long timescales.

Purpose of the Study:

• To investigate the kinetics of the ice-water interface using a neural-network potential.
• To study both ice melting and growth processes, including the effects of surface structure and pressure.

Main Methods:

• Utilized a neural-network potential for water trained on the revised Perdew-Burke-Ernzerhof functional.
• Performed molecular simulations of ice melting and growth, exploring basal and prismatic facets.
• Investigated effects of negative and high pressures (-1000 bar to 2000 bar).

Main Results:

• Observed non-monotonic behavior for ice growth kinetics and monotonic behavior for ice melting.
• Determined a maximum ice growth rate of 6.5 Å/ns at 14 K supercooling.
• Found prismatic facets grow faster than the basal facet; pressure has a limited impact on interface speed relative to supercooling/overheating.

Conclusions:

• Neural network potentials enable accurate ab initio simulations of ice-water interface kinetics.
• Surface structure significantly influences ice growth rates, with prismatic facets being more dynamic.
• Thermodynamic driving force, rather than pressure, is the key factor governing interface speed.