Synthesis of metal alloy catalysts using high-throughput experiments and machine learning optimization

Abstract

The periodic table comprises over a hundred elements, offering numerous possibilities for the discovery of novel materials that have superior properties and could therefore be used to address current technological and societal challenges. However, exploring the extensive range of combinations are resource-intensive: slow and costly, particularly for materials significantly affected by the synthesis procedures. In this final year project, a workflow for the high throughput synthesis of multimetallic alloys is presented. The two-step workflow is comprised by a liquid mixing step and an annealing step. An acceleration factor of 2.4 relative to the traditional auto combustion sol gel synthesis method is achieved by synthesizing 24 samples in 620 minutes. To evaluate the effectiveness of this methodology and with the assistance of previous computational work carried out by collaborators at Meta AI, copper and three other copper alloys, namely binary Cu-Ag, Cu-Zn, and ternary Cu-Zn-Ag, are synthesized, due to their predicted promising use in CO2 reduction. The synthesized samples show homogeneously distributed elemental composition and high phase purity. The catalytic performance is evaluated by collaborators at the University of Toronto. The initial findings from measuring pure Cu, which serves as a baseline, demonstrate consistent performance when compared to commercially available Cu nanoparticles. Crucially, the Faradaic efficiencies show different results compared to Cu nanoparticles. Firstly, a substantial amount of H2 gas is produced, accompanied by low CO. This is due to the large amount of carbon in our powders, stemming from the annealing step, and large particle size of the pure Cu. To guide future experiments and optimize the Faradaic efficiencies, the experimental data collected in this project is used to deploy a Bayesian Optimization (BO) algorithm. Specifically, q-Noisy Expected Hypervolume Improvement based Bayesian Optimization (qNEHVI-BO) model is implemented, providing insight to guide the next experimental steps to achieve dry samples and minimize the absolute difference between the obtained composition and the target.