Ian Dunn finds research so engrossing that he didn’t even take a break before graduate school. On the same day that he finished at Harvard University in 2014, he packed up and left the Boston area with his father. That evening he arrived at his New York City apartment to start his Ph.D. at Columbia University.
Dunn, a Department of Energy Computational Science Graduate Fellowship (DOE CSGF) recipient, always enjoyed math, but it was his high school physics teacher’s mad-scientist enthusiasm for simple, elegant equations that inspired him to stay late and work problem sets. As an undergraduate, he and his roommates threw mathematical “proof parties,” did swing-set physics experiments at the local playground and learned to program in the Python language. They even veered off the computing-math track to take organic chemistry together.
Over the following summer, he combined chemistry with computing in Walter Englander’s laboratory at the University of Pennsylvania. Dunn honed his Python skills, parsing mass spectrometry data to understand patterns in protein folding.
But Dunn soon realized that his core chemistry interests centered on physical theory, differential equations and mathematics. He remembers thinking, “I don’t just want to know how molecules react with each other. I want to know why they react with each other.” He decided to major in chemical physics and dig deeper into quantum mechanics, rules that govern the behavior of super-small particles.
Dunn continued to pursue research at Harvard. Working with computational chemist Alán Aspuru-Guzik (now at the University of Toronto), he analyzed the thermodynamics of metabolic reactions – those that maintain life – using density functional theory, a quantum mechanical modelling method. “It excited me so much that I could get out of the lab and predict something, even though not perfect, from equations, mathematics, computations and physical theory.”
Dunn carried that enthusiasm to Columbia and his Ph.D. research with David Reichman. Many-body quantum physics deals with complex interacting phenomena that can be challenging to model. Dunn incorporates more realistic – but complicating – factors such as modeling both nuclei and electrons using quantum mechanics, examining how these particles travel over time and analyzing materials where electrons’ behavior is strongly correlated. At times the complexity can feel nearly impossible, he says. “When you can predict something, that’s pretty magical.”
These simulations often require approximations or clever computational schemes to make calculations manageable. For example, Dunn has used a computational model to examine with quantum mechanical precision how energy packets, or excitons, move through solids. As such packets travel through closely packed molecules, they deflect off both nuclei and electrons jiggling around in these materials. He says that activity creates a deformation, as if a quantum mechanical bowling ball has rolled across a memory-foam mattress, forming a wake. Dunn’s work is examining fundamental physics, but some day such models could be used to design and optimize the properties of new materials such as solar cells.
Dunn also has observed that efficient solutions to this model based on the hierarchical equations of motion can break down. He’s since examined how and why that happens. With that knowledge, he can find ways to make such simulations more tractable without sacrificing reliable physical answers.
For his 2016 Lawrence Livermore National Laboratory practicum, Dunn took a completely different approach to modeling chemical systems. That year Livermore’s Jean-Luc Fattebert and Daniel Osei-Kuffuor were named finalists for the Gordon Bell prize honoring outstanding achievements in high-performance computing applications for developing a version of density functional theory that scales linearly rather than exponentially. The achievement enables chemical simulations of more than a million electrons at once. Dunn has tried to lengthen those simulations, from 100 picoseconds (billionths of a second) up to 300 picoseconds. He’s continuing to collaborate with the Livermore scientists on this project.
Though the practicum varied from his doctoral studies, the exposure to applied mathematicians and computational scientists who wrote high-performance programs rubbed off on Dunn. Thinking more about applied math and clean codes and taking a numerical methods course as part of his DOE CSGF program of study have let him bring a computational rigor that matches theoretical precision to his Ph.D. research. “It really added to my toolbox in a big way,” he adds.
Dunn plans to graduate in 2019 and is deciding whether he’d like to work in academia, in industry or at a national lab.
Image caption: Ian Dunn uses detailed quantum mechanical models to depict how an energy packet (yellow ball) hops between molecules (balls on springs) in perylene diimide (PDI) crystals. PDI and similar molecules have properties that could prove useful in photovoltaics. Credit: Ian Dunn.