A group of undergraduate students, graduate students and a post-doctoral scholar in a BYU chemistry lab combine forces and use machine learning to solve a complex chemistry problem.
As a sophomore, BYU engineering student Nick Rollins remembers his lab advisor Prof. Daniel Ess assigning him a special mission for their upcoming research project: Rollins was charged with becoming an expert in machine learning (guided artificial intelligence).
“That became three months of headaches and pain killers, but I learned so much from it,” Rollins recalled. “The only reason I have the knowledge and skills I do now was really because he decided to sit an undergrad down and entrust him with this significant part of the project.”
Armed with advanced data science principles and techniques and using chemical theories created by Henry Eyring (the father of the Church leader Henry B. Eyring), the Ess Research Group was able to model how specific complex chemical reactions take place and design new catalysts for a critical industrial chemistry problem. That research was recently published in the journal Chemical Science.
“It took us about two years to start really getting a grasp of the problem and we still kind of spun our wheels for another two years,” Ess said. “Using machine learning became the catalyst for designing a new molecular catalyst.”
A catalyst makes chemical reactions go faster by lowering reaction energy barriers. Using catalysts in the chemical industry to make plastics and plastic co-monomers, linear alpha olefins in this case, is critical to save time, energy and resources. In collaboration with chemists at Chevron Phillips Chemical based in Kingwood, Texas, the group uniquely combined Eyring’s 90-year-old transition state theory with modern machine learning techniques to design new molecular chromium catalysts for selective ethylene oligomerization, which is a major challenge in the petrochemical industry.
“This is the first example of anyone using machine learning to predict the selectivity of a catalytic chemical reaction,” said Dr. Steven Maley, a post-doctorate scholar at BYU and co-author on the paper. “There have been other papers that have used machine learning to predict how reactive catalysts are but not how selective they are, which is a much more difficult thing to predict.”
While more than 92% of BYU students are undergraduates, it’s worth noting that the authors of this study included an amalgamation of undergraduate and graduate students as well as Maley, a post-doctoral scholar working in collaboration with industrial scientists. Like many high-performing science research groups, postdocs are a critical part of the recipe to help train undergraduate and graduate students and as a team generate high-level research.
“Having postdocs in the lab that already have their PhD provides a culture of cutting-edge research and is critical to help mentor younger students,” Ess said.
For Rollins and other undergraduates, a lab full of postdocs and graduate students is an arrangement that gives students the chance to learn the frontiers of science directly from their peers.
“I worked with one grad student in particular, Doo-Hyun Kwon. He was a really great mentor to us,” Rollins said. Kwon was a co-author on the manuscript and recently landed a job as a computational chemist at a major pharmaceutical company. “He went beyond just making sure we knew enough to do our work but really making sure we had a solid structure and background of understanding that we could apply to other things.”
