Evaluation of Rate Coefficients in the Gas Phase Using Machine-Learned Potentials.

Area of Science:

• Computational chemistry
• Chemical kinetics
• Astrochemistry

Background:

• Accurate computation of reaction rate coefficients is crucial for understanding chemical processes.
• The C5H5 potential energy surface is relevant to molecular weight growth in combustion and interstellar media.
• Machine-learned potentials offer a promising avenue for accelerating chemical kinetics simulations.

Purpose of the Study:

• To assess the capability of machine-learned potentials for computing rate coefficients.
• To develop and apply a neural network (NN) model for the C5H5 potential energy surface.
• To integrate the NN model with an automated kinetics workflow for comprehensive rate coefficient calculations.

Main Methods:

• Training a neural network (NN) model to represent the C5H5 potential energy surface.
• Coupling the NN model with the KinBot automated kinetics workflow code.
• Benchmarking the NN model's performance across various stages of kinetics calculations, including energy, barrier heights, and entropic contributions.

Main Results:

• The trained NN model effectively describes the C5H5 chemical landscape.
• The integrated workflow successfully computed rate coefficients using the NN potential.
• Exhaustive benchmarking confirmed the NN model's reliability from electronic energy to transition state theory calculations.

Conclusions:

• Machine-learned potentials, specifically NN models, are capable of accurately computing rate coefficients.
• The developed approach provides a robust method for studying complex chemical systems relevant to combustion and astrochemistry.
• This work highlights the potential of AI in advancing chemical kinetics and simulations.