Northeastern professor sheds light on photochemistry using machine learning
The chemistry that you learned in high school or college is not the same chemistry going on in Professor Steven Lopez’s lab in the Interdisciplinary Science and Engineering Complex at Northeastern. He’s a new face on campus. He arrived one year ago after completing a Ph.D. in Chemistry at UCLA (Houk lab) and a postdoc at Harvard University (Aspuru-Guzik lab) funded by the Department of Energy. He was just awarded a $500,000 grant from the Office of Naval Research as part of a Multi University Research Initiative (MURI) for ‘Synthesis Planning and Reaction Discovery for Photochemistry and Chemistry in Novel Environments.’ He is part of a team of experimental and theoretical research groups from Stanford University, the University of Southern California, and the University of Toronto. The MURI team is collaborating to find new reaction pathways and new methods to get from compound A to compound B using light.
In the chemical reactions most people imagine, there is usually a heat source that helps to drive these reactions. “We have hundreds of successful and failed reactions that shape our chemical library of knowledge. Chemists have a seemingly supernatural ability to look at a molecule–or pair of molecules– and know if and how they will react with one another,” according to Lopez.
The key difference between that chemistry and what you may remember from chemistry is the driving force. In Lopez’s lab, it’s not heat, it’s light. When light is involved, things go a bit differently. “Reactions occur faster and there can be many unexpected side products,” says Lopez.
Lopez and his team focus on ‘green’ reactions and emphasize sustainability. Imagine the cost reduction when instead of using expensive earth-rare metals, you simply shine light on an organic compound (or plastic) to catalyze a reaction. Understanding how organic molecules behave when irradiated by sunlight is difficult because molecules enter an extremely short-lived ‘excited state’ when they absorb light (one thousandth of a billionth of a second). This excited state is too short-lived to determine the structure of a molecule, which is often a key component to understanding their chemistries. Light absorption rearranges the electrons in a molecule and when that happens, “you end up strengthening some bonds and weakening others, leading to the ‘unexpected’ products that computations are starting to predict.”
Professor Lopez’s lab space, which overlooks Northeastern’s campus from the Interdisciplinary Science and Engineering Center, substitutes lab coats and Bunsen burners with computers. Quantum mechanical computations and machine learning models are used to predict the outcome of light-activated (photochemical) reactions. The Lopez Lab uses Northeastern’s ‘Discovery’ cluster and Department of Defense High-Performance Computing (HPC) resources to carry out their calculations. An excited Lopez described that “machine learning connects one structure to the next in a high throughput way, where we can test millions of reactions. We’re doing tens or hundreds of years of chemistry. A pharmaceutical company might spend millions of dollars working out an optimal synthesis for a particular drug, and it might take some pretty expensive reagents along the way. But if we can predict a one-step light promoted reaction, that would be great,” Lopez says. Much of Lopez’s other work has centered around photovoltaics, materials that form the basis of solar panels, and light-based chemistry could open doors in that area as well as many others.
There’s a lot to look forward to from Lopez and his team. Learn more about the Lopez Lab on their website, and contact him directly if you’re interested in getting involved. Lopez says, “these complex problems require a multidisciplinary approach, so chemistry, computer science, physics, and engineering students should inquire about research opportunities.”