Local correlation approaches for ground and excited states of periodic systems

点击翻译按钮获取中文摘要

Computer simulations of microscopic properties are key to the design and understanding of new materials, and the aim of the proposed research is to develop a new set of tools to perform accurate simulations for solids, which can be greatly expanded on in the future by us or other groups, in Canada or elsewhere.The past decade has seen major progress in the development of highly accurate, systematic, and efficient wave function based quantum chemistry methods for molecular systems. A major innovation has been the development of local correlation coupled cluster methods using a very compact representation using so-called pair natural orbitals (PNOs). While calculations for molecular systems are accurate, versatile and efficient, this is currently not true for periodic solids. Density functional theory, often in conjunction with plane waves and pseudo potentials are most widely used, but such calculations for solids are not nearly as accurate or reliable as wave function methods used for molecules. In this proposal the aim is to generalize the proven methodology for molecules to the domain of periodic solids, up to full 3-dimensional crystals. The novelty of our strategy is the use of translational symmetry in conjunction with local PNO correlation approaches, in complete analogy to molecular systems. The key idea is to define a set of localized occupied orbitals that reflect the symmetry/periodicity of the system, but which will in general be linear dependent. This new approach requires non-trivial modifications to conventional methodologies (that have been partially tested in a pilot implementation for molecules). A new automatic code generator will be developed and this will greatly facilitate effcicient computer implementations. The long-range part of the Coulomb potential plays an essential role for solids, in particularly for metals, and the use of plane wave techniques have been proven particularly effective. In a second line of research we therefore plan to consolidate the best parts of solid state and molecular calculations by partitioning the full Coulomb potential in a short-range and an essentially pure longrange part. The short-range part is treated by direct space (molecular-like) techniques, while the long-range part is treated in a compact reciprocal space, using Fourier transforms, i.e. plane waves. By judicially combining the best of both worlds we anticipate we can develop a wave function based computer program to calculate structure, energetics, material and spectroscopic properties of solid state systems at the same level of accuracy and efficiency as state-of-the-art molecular codes. The research is open-ended as the use of a code generator will enable smooth implementation of methodology that currently still lies in the future, and will provide very high level training for students.