Edward Valeev

Research Interests

Electronic structure theory is routinely used to interpret experimental results and guide designs of new experiments in chemistry and materials science. The workhorse of modern electronic structure, the Density Functional Theory (DFT), is fast and reasonably accurate for well-behaved systems, but it is too empirical to be predictive. Truly predictive Wave Function or Green's Function methods which can replace some or most experiments are too expensive to be applicable to experimentally-relevant systems in condensed phase chemistry or materials science. The goal of our research is to develop electronic structure methods that can predict properties of molecules and materials. This objective can only be achieved by advancing molecular structure theory in accord with modern applied mathematics and computational science. Some of the ongoing efforts include:

- Development of reduced-scaling wave function methods that can be routinely applied to systems with 1000 or more atoms at the cost comparable to DFT. Such reduced-scaling approaches utilize novel tensor compression (tensor network) techniques that we develop.

- Development of real-space representations for wave functions that can be used to describe electronic structure more accurately than the traditional Gaussian atomic orbitals, especially in relativistic (heavy element) chemistry, high-density materials, and in laser-driven processes.

- Development of self-interaction-free approaches to the density functional theory that seek to cure one of its fundamental ills.

- Development of effective hamiltonian approaches suitable for chemistry simulation on quantum computers.

- Development of efficient methods for periodic solids.

- Development of advanced software for computation of electronic structure, including chemistry software for deployment on the DOE exascale computers. These developments are enabled by our development of programming models for composing scientific applications on extreme-scale distributed-memory heterogeneous machines.

Our group deploys our ideas in open-source, free, and commercial software to be usable by other computational researchers and experimentalists alike. Our main research platform MPQC (mpqc.org) is designed for large scale distributed parallel memory machines; we also develop many components for electronic structure (Libint for Gaussian integrals, TiledArray for tensor computation) that are used by many research and production codes around the world. Our contributions are found in end-user packages for chemistry Orca, CP2K, Psi, and others.

Our work is currently supported by the US National Science Foundation and US Department of Energy. Past research has also been supported by Sloan Foundation, Camille and Henry Dreyfus Foundation, and ACS Petroleum Research Fund.