College of Science & Engineering
Electronic structure calculations invariably feature a compromise between computational efficiency and accuracy. The computational scaling of the most accurate methods makes calculations beyond just tens of atoms infeasible. Methods that exploit the intrinsic locality of molecular interactions show significant promise in making tractable accurate the electronic structure calculation of large-scale systems. These multiscale methods treat different regions of the system at different levels of accuracy. This allows for a high chemical accuracy in a small region, such as an active site of a catalyst, and a less accurate, but more computationally efficient description of the remainder. An appealing strategy is thus to partition the system into a small subsystem that is treated at the wavefunction level of theory (CCSD(t), MRCI, etc.), while the large remaining subsystem treated at a more computationally tractable level of theory, such as density functional theory (DFT). These researchers are pioneering the development of new wavefunction-in-DFT methods along with machine learning methologies to allow for high accuracy calculations in large systems as significantly reduced computational cost. The new methods are then applied to large, condensed phases systems with complicated electronic structure, such as metal-oxide catalysts and electrocatalysts.