Exploration of Ethanol-to-Butadiene Catalysts by High-Throughput Experimentation and Machine Learning

Meeting the high demand for synthetic rubbers and elastomers requires a cost-effective source of 1,3-butadiene, making on-purpose synthesis from ethanol a crucial process. To tackle this challenge, a comprehensive multi-elemental approach was employed in conjunction with a genetic algorithm-driven high-throughput experimentation technique. The study explored a vast catalyst space, comprising up to 14 elements - including Mg, Al, Cr, Ni, Cu, Zn, Ga, Y, Zr, Nb, Mo, Ag, La, and Hf - co-supported on mesoporous silica, with the aims to discover effective combinations and understand the roles of each element in the overall reaction mechanism. The discovered efficient catalysts were composed of primarily Mg, Zn, Y, and Hf, and secondary Zr, Nb, and La. Such highly multi-elemental design was suggested to achieve a balance for the complex reactions of ETB, where efficient conversion of acetaldehyde to butadiene while minimizing the production of ethylene was critical. The highest yield obtained was 73% for butadiene. Through the application of machine learning techniques on the collected dataset, we successfully derived important insights related to catalyst design and catalysis. In particular, we proposed a visualization method to facilitate a deeper understanding of the role of each element in the overall catalysis.