A Status Report on "Gold Standard" Machine-Learned Potentials for Water.

Area of Science:

• Computational chemistry
• Materials science
• Chemical physics

Background:

• Water's fundamental role in life and unique properties drive extensive research.
• Developing accurate water potentials is crucial for molecular simulations.
• Recent advancements include machine learning potentials near the CCSD(T) gold standard.

Purpose of the Study:

• To provide a status report on recent high-level machine learning potentials for water.
• To analyze the methodology and applications of these potentials.
• To evaluate their performance in simulating water systems across different phases.

Main Methods:

• Review of five recent machine learning potentials for water.
• Focus on methodologies including different machine learning approaches.
• Assessment of performance using gas phase water clusters and condensed phase properties.

Main Results:

• Five machine learning potentials achieve accuracy near the CCSD(T) level.
• These potentials facilitate efficient and accurate simulations of water.
• Performance evaluation covers gas phase cluster energies and condensed phase dynamics.

Conclusions:

• High-level machine learning potentials represent a significant advancement for water simulations.
• These potentials enable accurate classical and quantum dynamical simulations.
• The discussed potentials offer robust tools for studying water across its phases.