Accelerated discovery of multi-elemental reverse water-gas shift catalysts using extrapolative machine learning approach

Designing novel catalysts is key to solving many energy and environmental challenges. Despite the promise that data science approaches, including machine learning (ML), can accelerate the development of catalysts, truly novel catalysts have rarely been discovered by ML because of one of its most common limitations and criticisms—the assumed inability of the models to extrapolate and identify extraordinary materials beyond those present in the training data set. Herein, we demonstrate an extrapolative ML approach to develop new multi-elemental catalysts based on supported Pt as an active metal and TiO2 as a support for the low-temperature reverse water-gas shift (RWGS) reaction. Using 45 catalysts as the initial data points and performing 44 cycles of the closed loop discovery system (ML prediction + experiment), we experimentally tested a total of 300 catalysts and identified more than 100 catalysts with superior activity as compared to the previously reported high-performance catalysts. The composition of the optimal catalyst discovered by this approach was Pt(3)/Rb(1)-Ba(1)-Mo(0.6)-Nb(0.2)/TiO2. Notably, Nb was not included in the original dataset, and the catalyst composition identified was unpredictable even by human experts.