Improving the Predictive Power of Density Functional Theory Calculations

Ann Mattsson
Sandia National Laboratories

The behavior of materials in the full range of temperature-­-density phase space is not dominated by one single physical process but instead many different processes compete and create a complicated multi-­-phase phase-­-space structure. We can usually describe each individual phase, dominated by one type of physics, by effective Hamiltonians and quasi-­-particle/perturbation theories (defined in a general sense). However, in order to make predictive calculations it is imperative to start from the laws of nature, the Dirac Equation (Dirac), or its non-­-relativistic limit, the Schrödinger Equation (SE), equations that are valid for any temperature and density. There are very accurate and predictive capabilities, based on the SE, in use in Quantum Chemistry calculations, such as the Coupled Cluster expansion. Even with extremely large super computers, however, the computational cost of these methodologies is prohibitive. This is an illustration of the constant competition between accuracy and computational cost in the Computational Science field. Density Functional Theory (DFT), its time dependent implementation (TDDFT), and the even more efficient Orbital-­-Free approach (OFDFT), all have the needed formal theoretical foundation and, in addition, can be computationally efficient. However, presently available approximations for the exchange-­-correlation functional limit the predictive power of these approaches. In this talk, I will discuss my systematic method for designing new functionals, the subsystem functional approach. In particular I will discuss materials containing elements with d-­- and f-­-electrons, such as actinides and transition metal systems. I will also touch upon how this approach can be used for creating kinetic energy functionals for OFDFT, the importance of including relativity for actinide systems, and how limitations of functionals for van der Waals’ systems can be circumvented and DFT provide extremely important information for molecular crystals, such as explosives. Sandia National Laboratories is a multi-­-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-­-AC04-­-94AL85000.

Presentation (PDF File)

Back to Semiclassical Origins of Density Functional Approximations