How do you design a high-performing catalyst? A typical scientist might make a guess based on observations and then test the idea in the lab to see what happens. In this scenario, progress is sluggish and breakthroughs are erratic.
A team of CEBC chemists hope to speed up discovery by using machine learning. This branch of artificial intelligence is gaining fresh momentum because of recent advances in computing capacity. Machine learning uses advanced algorithms to organize enormous datasets, spot patterns and make predictions—with only minimal input from humans.
But in order for machine learning to accelerate innovation in catalyst design, someone has to create and test algorithms for the system. Professors Marco Caricato, Ward Thompson and Brian Laird from KU’s Department of Chemistry have earned a four-year, $1.6 million award from the U.S. Department of Energy (DOE) to carry out this investigation.
Working with collaborators at the University of Illinois Urbana-Champaign, the team seeks to develop algorithms and software that decipher trends in materials used to catalyze ethylene epoxidation and polymerization.
The work is in line with DOE’s Exascale Initiative, which is a major push to capture the next era of high-end computing and stimulate U.S. competitiveness through aggressive technology development.