A Machine Learning Approach for Rate Constants. II. Clustering, Training, and Predictions for the O(³P) + HCl → OH +...

Area of Science:

• Computational Chemistry
• Chemical Kinetics
• Machine Learning

Background:

• Accurate prediction of thermal rate constants is crucial for understanding chemical reactions.
• Existing machine learning models face challenges with limited exact rate constant databases.
• The Eckart transmission coefficient is a key factor in rate constant calculations.

Purpose of the Study:

• To develop an improved machine learning strategy for predicting thermal rate constants.
• To apply and validate the new method on the O(3P) + HCl reaction.
• To create a publicly available database and software for rate constant prediction.

Main Methods:

• Utilizing clustering based on corrections to the Eckart transmission coefficient.
• Applying Gaussian process regression within each identified cluster.
• Employing a training strategy on the full dataset for each cluster and random value testing.

Main Results:

• The new method was successfully applied to predict rate constants for the O(3P) + HCl reaction.
• The O(3P) + HCl reaction served as a benchmark for the ring polymer molecular dynamics method.
• The reaction data was integrated into the database, refining the training set.

Conclusions:

• The proposed machine learning approach enhances the prediction of thermal rate constants.
• The developed Python software and database offer valuable resources for computational chemistry.
• This work advances the application of machine learning in predicting chemical reaction dynamics.