Rate coefficients for C and O2 reactive collisions relevant to interstellar clouds from QCT and machine learning

Area of Science:

• Astrochemistry and Computational Chemistry
• Quantum Mechanics and Molecular Dynamics

Background:

• Interstellar molecules undergo exothermic and barrierless reactions, enabling rapid reactions at low astronomical temperatures.
• Accurate state-selected rate coefficients for reactions like C + O2 are vital for understanding interstellar and atmospheric environments.
• Traditional computational methods for determining these parameters are often computationally intensive.

Purpose of the Study:

• To develop and validate a computational approach combining quasi-classical trajectory (QCT) calculations with machine learning (ML) for determining state-selected rate coefficients.
• To significantly reduce computational requirements for calculating reaction cross sections and rate coefficients for the C + O2 system.
• To provide a comprehensive dataset of rovibrational rate coefficients for the C-O2 collision system for astrophysical modeling.

Main Methods:

• Utilized quasi-classical trajectory (QCT) calculations.
• Implemented machine learning techniques, specifically Neural Network (NN) and Gaussian Process Regression (GPR), to model the reaction dynamics.
• Validated the ML models against explicit numerical calculations over a temperature range of 50-1500 K.

Main Results:

• Achieved significant reduction in computational cost while maintaining accuracy in calculating state-selected reaction cross sections and rate coefficients.
• Both NN-based and GPR-based models demonstrated high accuracy, with a coefficient of determination (R2) > 0.96 within the studied temperature range.
• Generated the most extensive dataset to date for rovibrational rate coefficients (v = 0-4, j = 0-70 → v' = 0-15) for the C-O2 system.

Conclusions:

• The combined QCT and ML approach is a computationally efficient and accurate method for determining state-selected rate coefficients for interstellar reactions.
• The generated comprehensive dataset of rovibrational rate coefficients will significantly aid astrophysical modeling of the C-O2 collision system.
• This work paves the way for applying similar ML-accelerated methods to other complex astrochemically relevant reactions.