IU’s supercomputing cluster Karst, in concert with development of new computational algorithms, has brought new speed to computational modeling of quantum mechanics in chemistry, decreasing researchers’ calculation time by months.
These models are changing scientists’ understanding of the role of nuclei in chemical reactions, and setting the standard for future research.
In traditional chemistry classes, the nuclei present in a molecular or condensed-phase system are pictured as marbles residing on a surface of potential energy. Nicole DeGregorio, a graduate student in the Iyengar group in IU’s Chemistry Department, pictures these nuclei as fluid confined inside a very tiny bucket. In DeGregorio’s research pertaining to building quantum mechanical models, the nuclei are not marbles anymore but instead are fluid elements inside a confined space. DeGregorio says that in order to get the most accurate model of chemical reactions, the fast dynamics of tiny hydrogen atoms must be “observed as a wave, rather than a particle.”
The nucleus, when described as a wave, explains why changes to the mass of a particle in a reaction, “drastically changes the actual kinetics of a reaction,” said DeGregorio. For example, when replacing hydrogen with Deuterium (an isotope of hydrogen with twice its mass) within an enzyme. “In theory the reaction should only decrease by a small amount, but in actual experimental studies the rate is 81 times slower than with hydrogen in the case of some biological enzymes,” she explained.
Working with Dr. Iyengar as her advisor in the Chemistry Department, DeGregorio’s research focuses on efficiently computing the potential energy surfaces that dictate reactions. In other words, if the quantum mechanical definition of nuclei is taken to be as a fluid inside of a bucket, or a confined region in space, then, DeGregorio explained,“I work on building the bucket.”
Treating nuclei as a wave of data points instead of one fixed point (as might be the case for a particle) makes the computational modeling of these systems dense and expensive to run. “The size of the problem increases exponentially with the size of our system as well as the dimensions we add,” said DeGregorio. DeGregorio and Dr. Iyengar have developed new computational methods that invoke ideas from information theory and from graphical networks to reduce this complexity by several orders of magnitude. But before using Karst, the exponential scaling of these calculations slowed her research.
Using our computer clusters, it would take 3-4 months to do one calculation. When I moved over to Karst, it took 6 hours.
Nicole DeGregorio, IU Chemistry
By working to improve the models of chemical reactions as these arise from quantum mechanics, DeGregorio’s goal is to reduce the computational costs faced by future researchers, while shedding more light on foundational modeling through quantum mechanics. “On IU’s supercomputing resources like Karst and Big Red, you can request multiple nodes and run these calculations in parallel. It increases the impact for science quite drastically,” she said.