(45a) Discovery of Light-Driven Hydrogen Evolution Catalysts Using High Throughput Simulation, Experiments and Machine Learning

The renewable production of hydrogen is an important technology. Current approaches to catalyst discovery for thermo/electro and photochemical methods for hydrogen production have been difficult to optimize. This difficulty arises in part because of the separation of the synthesis and characterization of catalysts from their performance under operating conditions. It is challenging to build models that relate catalyst performance to synthesis conditions. We have taken a different approach to this problem. We use a photo-driven process that couples a photoabsorber with the synthesis of multi-component metal particles and hydrogen production. This approach is readily parallelized in a multi-well plate, enabling us to explore a wide range of synthesis conditions that are directly connected to catalyst performance measurements. We will show how we used this approach to screen thousands of possible catalyst synthesis conditions to identify promising combinations. The high-throughput experiments along, however, do not tell us what metals to test. For this we augment the approach with a parallel high-throughput computation approach where hundreds of thousands of DFT calculations are performed on multi-component alloy surfaces to screen for promising candidates with desired hydrogen and CO adsorption energies. Finally, we combine these efforts to drive focused studies of the most promising candidate synthesis conditions to develop a more conventional understanding of how the synthesis conditions led to the observed activity. We will show how this approach has led to confirmation of some known results from the electrochemical production of hydrogen from water, as well as some new results we may not have discovered with any single approach.