Kolmogorov-Arnold Chemical Reaction Neural Networks for learning pressure-dependent kinetic rate laws

The introduction of Kolmogorov-Arnold Chemical Reaction Neural Networks (KA-CRNNs) marks a pivotal advancement in the field of chemical kinetics. Traditional Chemical Reaction Neural Networks (CRNNs) have limitations in representing pressure-dependent behaviors, often relying on empirical formulations like Troe or PLOG. KA-CRNNs address this gap by modeling kinetic parameters as functions of system pressure, allowing for assumption-free inference from data. A proof-of-concept study on the CH3 recombination reaction demonstrated that KA-CRNNs accurately reproduce pressure-dependent kinetics across various temperatures and pressures, outperforming conventional interpolative models. This framework not only enhances the understanding of complex reacting systems but also establishes a foundation for data-driven discovery of extended kinetic behaviors, emphasizing the importance of interpretability and physical consistency in machine learning applications within chemistry.