Performing Path Integral Molecular Dynamics Using an Artificial Intelligence-Enhanced Molecular Simulation Framework.

Area of Science:

• Computational Chemistry
• Molecular Dynamics
• Artificial Intelligence

Background:

• Path Integral Molecular Dynamics (PIMD) is crucial for studying nuclear quantum effects but is computationally intensive.
• Conventional PIMD methods face limitations in computational complexity and resource requirements.
• Developing efficient simulation techniques is essential for understanding molecular behavior.

Purpose of the Study:

• To develop an artificial intelligence-enhanced molecular simulation framework for efficient PIMD.
• To mitigate the computational complexity and resource demands of traditional PIMD simulations.
• To enable the accurate capture of nuclear quantum effects in complex molecular systems.

Main Methods:

• An AI-enhanced molecular simulation framework was developed with a modular architecture and high-throughput capabilities.
• Machine Learning Force Fields (MLFFs) were integrated into the simulation framework.
• The framework's performance was validated using two systems: double-proton transfer in formic acid dimer and water-ice phase transformation.

Main Results:

• The AI-enhanced framework demonstrated accelerated PIMD simulations.
• Quantum mechanical accuracy was preserved during the accelerated simulations.
• The framework successfully captured nuclear quantum effects for both tested systems.

Conclusions:

• The proposed AI-enhanced framework significantly improves the efficiency of PIMD simulations.
• This approach effectively reduces the computational cost associated with studying nuclear quantum effects.
• The framework offers a viable method for investigating complex molecular systems with quantum mechanical accuracy.