Exploring Photochemistry Using Correlated First Principles Simulation

BIOSKETCH: Dr. Fales’ PhD research at Michigan State with Professor Levine in the Department of Chemistry focused on electronic structure method development, electron dynamics, and semiconductor photochemistry. Specifically, Dr. Fales worked on understanding mechanisms for nonradiative decay in semiconductor nanoparticles, and he developed molecular electronic structure theories capable of describing electron correlation in molecular systems as large as proteins. To do this, Dr. Fales developed an efficient determinantbased full configuration interaction (FCI) algorithm using Graphical Processing Units (GPUs) to allow routine calculation of configuration spaces on the order of 108 determinants. He also formulated lowscaling GPU accelerated spin-purification approaches suitable for systems where spin contamination occurs as a result of numerical instability. These tools enabled me to study the effect of orbital relaxation in the context of complete active space CI (CASCI) methods at minimum energy conical intersection (MECI) geometries.

During Dr. Fales’ postdoctoral work with Professor Martinez at Stanford, he has focused on applying rank-reduction methodologies to correlated electronic structure theories, leveraging both element and rank sparsity of molecular wave functions to extend application of high-accuracy theories to larger systems. His recent scientific work has included high-performance first-principles method development, formulation of theories leveraging rank-sparsity in correlated wave functions, and application of wave function theory to better understand electronic characteristics of particular transition metal catalysts. Here, Dr. Fales implemented GPU-accelerated multireference configuration interaction singles (MRSCI) to describe high-lying excited states with multireference character. He extended GPU-accelerated direct configuration interaction (CI) to multiple devices, enabling configuration spaces larger than 109 determinants, and I developed fast configuration state function to determinant basis transformation approaches. Dr. Fales is currently investigating strategies to enable the use of single and mixed floating point precision in determinantal direct CI.