Exploring catalytic reaction networks with machine learning

Chemical reaction networks form the heart of microkinetic models, which are one of the key tools available for gaining detailed mechanistic insight into heterogeneous catalytic processes. The exploration of complex chemical reaction networks is therefore a central task in current catalysis research. Unfortunately, microscopic experimental information about which elementary reaction steps are relevant to a given process is almost always sparse, making the inference of networks from experiments alone almost impossible. While computational approaches provide important complementary insights to this end, their predictions also come with substantial uncertainties related to the underlying approximations and, crucially, the use of idealized structure models. In this Perspective, we aim to shine a light on recent applications of machine learning in the context of catalytic reaction networks, aiding both the inference of effective kinetic rate laws from experiment and the computational exploration of chemical reaction networks.

Automated summary

Chemical reaction networks are crucial for understanding heterogeneous catalytic processes. However, inferring networks from experiments alone is challenging due to the lack of microscopic information. Computational approaches provide insights but come with uncertainties. This Perspective highlights the applications of machine learning in catalytic reaction networks, aiding in both experimental inference and computational exploration.