(154c) Enabling Catalyst Discovery through High-Throughput Experimentation and Machine Learning

The development of high-throughput experimentation over the past two decades allows researchers to screen large compositional and operational spaces more efficiently and create large datasets under reproducible conditions ¹. Machine learning is an avenue to extract knowledge and unravel multidimensional relationships present in such large experimental datasets ^{2, 3}. A framework will be described that incorporates machine learning algorithms with experimental high-throughput catalytic data, spectroscopic data, and elemental properties to discover new materials. The framework allows to rank chemically descriptive features, to predict future catalyst performance, and to guide synthesis. Application of this framework predicted several novel catalyst compositions for low-temperature ammonia decomposition, which were experimentally validated against state-of-the-art ammonia decomposition catalysts and were found to have exceptional low-temperature performance at substantially lower weight loadings of Ru.

For many reactions, there is also a vast amount of scientific literature available over a large range of parameters. We extracted experimental data from the published literature for ammonia synthesis. However, gaps exist in the parameter space that have not been explored in the literature. These gaps can be rapidly filled using high-throughput measurements. Machine learning models were developed, trained, and compared to establish multi-dimensional correlations between catalyst formulations, elemental properties, support properties, synthesis parameters, reaction conditions, and ammonia synthesis rates. The activity of unknown catalyst formulations was predicted, and the best predicted catalyst candidates were synthesized and tested to validate the predictions. Consequently, this approach also led to the finding of new ammonia synthesis catalyst formulations with higher activity compared to some state-of-the-art catalysts in the literature.

1. Hattrick-Simpers, J.; Wen, C.; Lauterbach, J., The Materials Super Highway: Integrating High-Throughput Experimentation into Mapping the Catalysis Materials Genome. Catalysis Letters 2015, 145 (1), 290.
2. McCullough, K.; Williams, T.; Mingle, K.; Jamshidi, P.; Lauterbach, J., High-throughput experimentation meets artificial intelligence: a new pathway to catalyst discovery. Phys Chem Chem Phys 2020, 22 (20), 11174.
3. MacQueen, B.; Jayarathna, R.; Lauterbach, J., Knowledge extraction in catalysis utilizing design of experiments and machine learning. Current Opinion in Chemical Engineering 2022, 36, 100781.