Design of N2O Decomposition Catalysts with SO2 Poisoning Resistance Through Data Assimilation Using Machine Learning Potentials

Nitrous oxide (N2O), present in exhaust gases and sewage, is a potent greenhouse gas, and its decomposition or suppression is crucial. Current research focuses on the catalytic thermal decomposition of N2O using heterogeneous catalysts. Transition metal oxides such as Ni (Nickel) and Co (Cobalt) have shown high conversion rates in this process. However, sulfur dioxide (SO2), which often accompanies N2O, can poison these catalysts, leading to their degradation during operation. Common practice involves using ammonia (NH3) as a reaction gas to extend catalyst life, but this can lead to the formation of nitrogen oxides (NOx) through side reactions and necessitates additional NH3 supply infrastructure. This not only reduces catalyst efficiency but also increases raw material costs and auxiliary equipment expenses. The development of catalyst designs that improve resistance to SO2 without using NH3 remains an unexplored area. In this report, we introduce a high-speed automated computational method using machine learning potentials to automatically generate possible surface states of heterogeneous catalysts and comprehensively calculate the adsorption patterns of N2O and SO2. This method integrates computational results to create equivalent alternative metrics that can be directly compared with experimental Diffuse Reflectance Infrared Fourier Transform (DRIFT) analysis spectra. These metrics enable the identification of heterogeneous catalyst surfaces with only a few data points, significantly aiding in the elucidation of poisoning mechanisms. We have successfully identified specific adsorption peaks of N2O and SO2 on oxide catalysts at room and reaction temperatures using DRIFT analysis, calculated the similarity between these surrogate values and actual analytical data, and pinpointed the catalyst surface. Furthermore, we discovered a unique adsorption pattern where SO2 specifically desorbs near reaction temperatures—a critical finding for addressing catalyst poisoning. We propose a catalyst composition optimized to maximize the occurrence of this adsorption pattern. Experimental validation confirmed the expected desorption of SO2, demonstrating a successful catalyst design that is resistant to poisoning