Efficient discovery and optimization of complex high-throughput experiments

As the pace of experimentation in materials science and catalysis has increased, experimental tactics and strategies have had to adapt to meet the demands of goals of experimentalists, and the spaces they explore. This pace has increased from runs/year to runs/day and sometimes to runs/minute in high-throughput experimentation. Although much of this capacity is used to simply speed up conventional experimental designs, the leading-edge application is discovery of low-probability, high-value occurrences (hits) by searching extensive, complex experimental spaces. Conventional design of experiments (DoE) is not capable of dealing with these issues. Instead, more advanced experimental tactics and strategies must be implemented. After introducing the elements that make an experimental campaign complex, here we present a novel statistical model-based evolutionary experimental strategy and apply it to the optimization of a family of artificial complex systems. With our experiments, we show that such a strategy may significantly reduce the experimental effort required for finding the optima compared to other state-of-the-art evolutionary strategies.