Machine Learning-Assisted High-Temperature Nonequilibrium Kinetics of Air Components: From Microscale to Macroscale

The growing demand for reproducing state-to-state (StS) level accuracy in modeling the nonequilibrium kinetics of air components at temperatures above 5000 K has spurred the use of microscopic state-specific data to revise macroscopic thermochemical models. However, the computational cost of obtaining comprehensive state-specific kinetic data remains prohibitively high. Machine learning has emerged as a transformative tool to bridge this gap, enabling the efficient generation of accurate state-specific data sets. This advancement allows traditional engineering models to be replaced or augmented by solutions grounded in microscopic data, yielding a significant improvement in simulation accuracy under extreme nonequilibrium conditions, such as those encountered in atmospheric entry and hypersonic flight, plasma discharges, and combustion and detonation processes. This Perspective reviews the development of machine learning applications across the critical workflow of high-temperature nonequilibrium kinetics, focusing on electronic structure calculations, potential energy surface fitting, and kinetic database construction. It further highlights persistent challenges faced in nonadiabatic dynamics, complex multimolecular systems, and interface scattering dynamics. Our goal is to promote the adoption of machine learning-powered high-fidelity simulations for the aforementioned important applications while also establishing the foundations for more efficient workflows in future dynamics and kinetics research.