Coupled Cluster Molecular Dynamics of Condensed Phase Systems Enabled by Machine Learning Potentials: Liquid Water...

Area of Science:

• Quantum chemistry
• Computational materials science
• Machine learning for physics

Background:

• Coupled cluster theory, particularly CCSD(T), provides high accuracy for electronic structure but is computationally expensive.
• The "gold standard" CCSD(T) method is limited to small molecular systems.
• Simulating condensed phase systems with high accuracy remains a significant challenge.

Purpose of the Study:

• To develop a method for transferring CCSD(T) accuracy from small clusters to large condensed phase systems.
• To enable accurate quantum chemistry simulations for extended materials.
• To perform coupled cluster molecular dynamics (CCMD) on complex systems.

Main Methods:

• A novel framework utilizing high-dimensional neural network potentials.
• Automated transfer of CCSD(T) accuracy from finite molecular clusters.
• Incorporation of nuclear quantum effects.

Main Results:

• Demonstrated high-quality coupled cluster molecular dynamics for liquid water.
• Successfully transferred CCSD(T) accuracy to a condensed phase system.
• Developed an efficient, generic, and systematically improvable machine learning strategy.

Conclusions:

• The developed machine learning approach overcomes the system size limitations of CCSD(T).
• This framework enables accurate CCMD simulations for complex systems, including liquid water with quantum effects.
• The method is efficient, versatile, and offers a pathway for future advancements in computational materials science.