An atomic cluster expansion (ACE) potential for water under extreme conditions

Area of Science:

• Materials Science
• Computational Chemistry
• Physics

Background:

• Water exhibits complex multiphase behavior under extreme conditions.
• Accurate simulation of these phases, including superionic ice, is computationally challenging.
• Existing interatomic potentials often fail to capture the full range of water's behavior.

Purpose of the Study:

• To develop a machine learning interatomic potential for water.
• To accurately capture water's complex multiphase behavior, including molecular and superionic ice phases.
• To enable high-fidelity molecular dynamics simulations of water under extreme conditions.

Main Methods:

• Utilized the Atomic Cluster Expansion (ACE) formulation for the interatomic potential.
• Parameterized the potential using ab initio molecular dynamics (AI-MD) simulations.
• Generated diverse configurations covering insulating and superionic ice, liquid water, and plasma phases.

Main Results:

• The H2O ACE potential accurately reproduces experimental and DFT-predicted isotherms and Hugoniots.
• The potential successfully captures the intricate phase behavior of water, including transitions to solid and superionic ice phases.
• Demonstrated high-fidelity molecular dynamics simulations for water up to 100 GPa and 6000 K.

Conclusions:

• The developed ACE potential provides a robust tool for large-scale, accurate simulations of water under extreme thermodynamic conditions.
• This work advances the capability to study water's behavior in planetary interiors and other extreme environments.
• The potential facilitates a deeper understanding of water's fundamental properties across its phase diagram.