Density Functional Theory of Molecular Fragments: Strong Electron Correlation Beyond Density Functional Approximations

With support from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry, Professor Adam Wasserman of Purdue University is developing advanced theories and methods for computer simulation of the electronic structure of molecules and materials with new capabilities to enable highly accurate prediction of properties and interpretation of experiment for these systems. In addition, the new method of density functional theory of molecular fragments aims to provide computational efficiency for systems of ever-increasing complexity. Dr. Wasserman and his research group will focus on improving both the accuracy and the efficiency of these methods for the challenging systems of the “strongly-correlated type.” These systems in which electron correlation effects dominate, are typically out of reach for standard approximations based on Kohn-Sham Density Functional Theory, the most widely used electronic structure method. In addition to the broader impact resulting from the proposed research efforts, the PI's leadership within the Latin-American community of Purdue students will continue. Other educational activities include a bi-annual School and Workshop on Time-dependent DFT co-organized by the PI and a study-abroad course on “Science and social progress” that the PI will continue organizing. Under this award, the Wasserman research group will focus on two major thrusts: (1) Improving the accuracy of density-functional theory (DFT) calculations for systems that lie beyond the reach of state-of-the-art density-functional approximations; and (2) Improving the efficiency of such calculations. This separation is made possible by the development of recent density-to-potential inversion techniques that allow for the calculation of numerically exact non-additive noninteracting kinetic energies. To achieve the first goal, Wasserman and his group will employ an overlap functional of the fragment densities to construct and test physically motivated approximations for the inter-fragment exchange-correlation energy functional. To achieve the second, a two-pronged approach will be followed in which (a) a semi-local fragment-density functional will be developed based on exact constraints for the non-additive kinetic-energy functional; and (b) a trained neural network will be used to incorporate non-local effects into a meta-generalized gradient approximation (GGA) functional for the full noninteracting kinetic energy functional. All advances will be made available to the broader scientific community through open-source codes for density embedding calculations. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.