Bayesian-Optimization-Based Improvement of Cu-CHA Catalysts for Direct Partial Oxidation of CH4

In the present study, the catalytic performance of a Cu zeolite for partial CH4 oxidation is improved by refinement of catalyst composition using the Bayesian optimization method. For application to this challenging reaction, the activity of the catalyst needs to be improved while maintaining its selectivity for partial oxidation products; thus, both catalytic activity and selectivity are set as objective variables in this study. Accordingly, a model describing how zeolite composition affects its catalytic activity and selectivity is prepared from an experimental data set by the Bayesian optimization method. Optimal catalyst compositions identified from the model are evaluated experimentally, and the model and expected improvement (EI) are updated by another cycle of Bayesian optimization using the newly obtained experimental data. This trial cycle is repeated until the proposed optimal composition is confirmed experimentally. In the present work, Cu-CHA zeolites are used because they show relatively high catalytic performances among the various Cu zeolites. Cu-CHA catalysts having various Cu-ion exchange rates (Cu IERs) and four different Si/Al2 ratios (5.2, 7.2, 10, and 19.5) are synthesized and evaluated in CH4-O2-H2O flow reactor experiments. The Bayesian optimization process achieved optimization of Cu-IER and Si/ Al2 in only two cycles of the optimization process using 22 data. Therefore, Bayesian optimization is demonstrated as a means to improve Cu-CHA catalysts for partial CH4 oxidation.

Automated summary

The catalytic performance of Cu-CHA catalysts for partial CH4 oxidation was improved through refinement of catalyst composition using the Bayesian optimization method. Experimental validation showed that optimization of Cu-ion exchange rates (Cu IERs) and Si/Al2 ratios was achieved in only two cycles of the Bayesian optimization process.