Generating High-Precision Force Fields for Molecular Dynamics Simulations to Study Chemical Reaction Mechanisms Using...

Area of Science:

• Computational Chemistry
• Organic Chemistry
• Artificial Intelligence in Chemistry

Background:

• Traditional theoretical studies of chemical reaction mechanisms rely on manual construction of transition states, which is experience-dependent.
• Enhanced sampling in molecular dynamics (MD) simulations can directly model reaction pathways but is computationally limited by high-precision potential energy functions.

Purpose of the Study:

• To develop a computational scheme for training high-precision force fields for molecular modeling.
• To enable accurate and efficient simulations of chemical reaction mechanisms.

Main Methods:

• Utilized a graph-neural-network-based molecular model, the molecular configuration transformer, to train a potential energy function.
• Integrated this trained potential energy function into molecular dynamics simulations with enhanced sampling.
• Applied the methodology to study a Claisen rearrangement and a manganese-catalyzed carbonyl insertion reaction.

Main Results:

• Achieved highly accurate molecular modeling simulations at a low computational cost.
• Enabled more precise calculations of chemical reaction mechanisms compared to previous methods.
• Successfully applied the approach to complex organic reactions.

Conclusions:

• The developed AI-driven potential energy function significantly improves the accuracy and efficiency of simulating chemical reaction mechanisms.
• This approach overcomes the computational limitations of traditional methods, paving the way for advanced theoretical studies in organic chemistry.