(709f) Efficient Predictions of Methane Steam Reforming Pathway Energetics

Methane steam reforming (MSR) is a major process in the production of hydrogen and synthesis gas. Metal alloy catalysts have been used to facilitate this reaction. Designing new catalysts is difficult, due to the time and expense required to screen candidates that could potentially not work. Computational screening methods have been implemented to predict a small number of descriptors, but the descriptor-based approach can give inaccurate predictions in some cases. Predicting the energetics of the entire reaction pathway would give better predictions, but is computationally intensive using density functional theory (DFT). To remedy this, we created a machine learning model that can predict the adsorption energy and transition state energy of each step of MSR. We used the model to predict which states are kinetically relevant on more than 10,000 alloy surfaces. We then identified 50 metal alloy catalysts that showed the most promising catalytic behavior for MSR, and did further testing with DFT. We found several metal alloy catalysts with significantly higher predicted performance than currently used MSR catalysts, such as nickel and ruthenium. This method and model could help identify other potential metal alloy catalysts for various reactions.