Extrapolative-Machine-Learning-Guided Discovery of Multielemental Heterogeneous Catalysts for Low-Temperature NO Reduction by H₂

Selective catalytic reduction of NO_xwith hydrogen (H₂–SCR) in the presence of oxygen is an environmentally friendly technology that has attracted considerable attention. However, even the most promising currently available catalysts are not sufficiently active to effectively promote this reaction, particularly at low temperatures (<150 °C). Therefore, there is an urgent need for the development of highly active H₂–SCR catalysts. Although data-science approaches, including machine learning (ML), have been suggested to accelerate the development of catalysts for such important processes, the discovery of efficient catalysts using ML remains limited. This limitation stems from a common criticism of ML, namely, its perceived inability to extrapolate and identify extraordinary materials. Herein, we present an extrapolative ML approach for the development of efficient multielemental H₂–SCR catalysts. Starting with 45 catalysts as the initial dataset, we employed a closed-loop discovery system that combined ML predictions and experimental validation over 24 iterative cycles. The iterative workflow facilitated the experimental evaluation of 425 catalysts, leading to the discovery of several compositions surpassing previously reported systems in average N₂yield within 50–150 °C. The top-performing catalyst was identified as Pt(1.3)–Ir(0.2)/Ba(1.5)–Co(1)/H-ZSM-5 (Si/Al ratio = 11). Notably, the optimal catalyst contained Co, an element absent from the initial dataset; the optimal catalyst composition could hardly be predicted even by human experts.