Phase Diagram of a Deep Potential Water Model

Area of Science:

• Computational chemistry
• Materials science
• Condensed matter physics

Background:

• Accurate modeling of water's complex phase behavior is crucial for understanding various physical and chemical processes.
• Density functional theory (DFT) approximations, like SCAN, offer a way to model these systems but can be computationally expensive.

Purpose of the Study:

• To develop a computationally efficient model for water's potential energy surface using the Deep Potential methodology.
• To accurately predict the phase diagram of water, including various ice polymorphs and fluid phases, across a wide range of temperatures and pressures.

Main Methods:

• Utilized the Deep Potential methodology to construct a machine learning model for water's potential energy surface.
• Employed molecular dynamics simulations to predict the phase diagram based on the developed model.
• Compared simulation results with experimental data for validation.

Main Results:

• The model accurately reproduces the potential energy surface of the SCAN approximation for water up to 2400 K and 50 GPa.
• The predicted phase diagram shows satisfactory agreement with experimental results, correctly identifying stable molecular and ionic fluid phases and most ice polymorphs.
• The model identified ice III and XV as metastable, differing from experimental observations where they are stable.

Conclusions:

• The Deep Potential model provides an efficient and accurate approach for simulating water's phase behavior.
• The study elucidates the complex atomic dynamics during phase transitions, including molecular dissociation and covalent fluctuations, which explains experimental observations of partial ionization.
• This methodology can be extended to study other complex materials and their phase transitions.