Density Functional Theory and Linear Scaling Computation
Quantum Mechanical Studies of Biological Systems
Research Directions: Density Functional Theory and Linear Scaling Computation
Electronic structure calculations provide fundamental information of molecular structure and interaction, and the potential energy surfaces necessary for studying molecular dynamics. Description of electrons is necessarily based on quantum mechanics principles. Both conventional ab initio wave function methods and the Kohn-Sham density-functional theory are successful approaches. Because of the nature of quantum particles and the many-body interaction between the electrons, quantum mechanical calculations are most demanding in computational effort. Therefore the accuracy and efficiency of the methods are major concerns.
We believe that density functional theory is the method of choice for large systems. In terms of computational efficiency, DFT calculations, along with the Hartree-Fock (HF) method, are of the lowest cost of first-principle methods. In terms of accuracy, DFT in its current state is already better than HF method in describing chemistry and, because of its rigorous theoretical foundation, potentially it can provide much better accuracy of prediction through the development of energy functionals.
We plan to develop methods for efficient and accurate determination of electronic structure of large molecules based on DFT. Specifically, we plan: (1) to make DFT calculations very efficient by the development of optimal linear scaling approaches based on the newly developed absolute minimum principles with nonorthogonal localized orbitals for the diagonalization problem and the recursive bisection method for the classical Coulomb interaction of electrons, and (2) to enhance the accuracy of DFT by the construction of new and improved density functionals for electron exchange and correlation based on the adiabatic connection, and the formulation of functionals from wave function approach with localized orbitals.
Our development aims at linear scaling DFT computational methods, which are optimal and robust, and exchange-correlation functionals in terms of localized orbitals, which are accurate, open for systematic improvement, and feasible for calculations within the linear scaling computational formalism. We emphasize the use of localized molecular orbitals in our approaches, because we believe that localized orbitals provide not only computational efficiency for linear scaling algorithms, but also physical insight and local and hence more economical approach to electron correlation. Our research will build on the current development in the field and should contribute to the progress of theoretical and computational chemistry in becoming an equal partner with other more traditional fields of chemistry.
Research Directions: Quantum Mechanical Studies of Biological Systems
We plan to extend and apply our linear-scaling semiempirical methods to study mechanism of enzyme reaction systems. Our continuing collaboration with biochemists and biophysicists at the University of North Carolina-Chapel Hill is essential to this effort.
We are also developing novel methods to describe the active site of enzymes with quantum mechanics and the rest of the enzyme with empirical molecular mechanics. Our development will be focused on the optimal modeling of the boundary between the quantum mechanical part and the molecular mechanics part. The proposed method will be applied to study the catalytic reaction mechanisms. This will lead to accurate and efficient methods of molecular modeling for enzyme and proteins. It will foster our understanding of biological catalytic processes at the molecular level, and facilitate the design of drugs and inhibitors.