Now, Paco Laveille and colleagues combine Bayesian optimization, automated synthesis and high-throughput catalytic experiments in a closed-loop approach to design cost-effective heterogeneous catalysts with high activity and selectivity for the hydrogenation of CO2 to methanol. In their study they considered six metals, namely cerium, cobalt, copper, iron, indium and zinc, potassium as a promoter, and four supports, Al2O3, SiO2, TiO2 and ZrO2. Each catalyst contains up to three metals with a total metal content between 2.5 and 5 wt% and can contain or not potassium promoter with a maximum loading of 1 wt%. The catalysts are prepared by a robot via incipient wetness impregnation and calcined, and catalytic tests are performed at 275 °C, 50 bar, H2/CO2 ratio of 3 and a space velocity of 10,000 mL g–1 h–1. The outcomes to optimize after each iteration are maximizing CO2 conversion and methanol selectivity while minimizing methane selectivity and cost of supported metal.
A total of 24 catalysts are prepared in each iteration and 5 iterations are performed. The best-performing catalyst in the first generation contains 1.22 wt% Zn, 1.16 wt% Cu, 0.5 wt% In on ZrO2 and achieves 6.2% CO2 conversion and 45% methanol selectivity. From the second generation onwards, the algorithm turns into mostly Cu-based catalysts supported onto ZrO2. The optimal catalyst is obtained in the fourth generation, consisting of 1.85 wt% Cu, 0.69 wt% Zn and 0.05 wt% Ce on ZrO2, which achieves 10.2% CO2 conversion and 41.8% methanol selectivity, with a much-reduced cost than those derived from the previous iterations. Interestingly, if the cost minimization is removed as an objective after the first generation, in the second generation the catalysts become mostly In-based and the methanol selectivity increases substantially.
This is a preview of subscription content, access via your institution