Highly correlated systems

Research project P 14699 Highly correlated systems Peter BLAHA 27.11.2000 The tremendous increase in computational power during the last years together with the development of new algorithms has allowed to efficiently solve complicated mathematical problems. One of the most significant conceptual developments in computational physics and chemistry was the formulation (and continuous improvement) of "density functional theory" (DFT). This was honored recently by awarding the Nobel price in chemistry to W. Kohn. On the other side, computer packages have been developed to solve the resulting "Kohn- Sham equations" as accurate and efficient as possible. Over the last 15 years we have implemented the full- potential LAPW method into the WIEN97 computer program, which is now used worldwide by almost 400 groups. One of the remaining challenges is that the exact functional form to be used in DFT is unknown and one has to make approximations at this level. The widely used "local density approximation" (LDA) and, even more, its recent improvement by the "generalized gradient approximation" (GGA) allows to apply theoretical solid state models to many real material science problems. However, for certain classes of materials, usually summarized as "highly correlated systems", LDA and GGA calculations may fail. For those systems, in particular for late- transition metal oxides or lanthanide and actinide compounds, the known experimental facts cannot be reproduced and thus reliable predictions of new materials or unknown properties are not possible. In the present project we like to study such highly correlated systems by means of the LDA+U method. We will implement 1DA+U (and also the "orbital polarization method" (OP)) into our WEEN97 package and then study transition metal compounds like the high-Tc superconductors as well as magnetic properties of compounds including very heavy elements (like U), where also relativistic effects (spin-orbit coupling) play an important role. We will focus on ground state properties like charge and spin-densities and the resulting electric field gradent, which can be compared to available accurate NMR and Mössbauer data.

Modern computers with their large computational power allow theoretical simulations of more and more complex materials. Theoretical simulations became a vital tool in materials design and while they certainly cannot replace experiments, they can guide and supplement them. Such simulations are now possible for many materials, in particular also because of the development of versatile, robust and user-friendly computer codes (one of them is the WIEN2k code used and further enhanced in the present project). However, materials containing elements with 3d, 4f or 5f electrons are sometimes not well described by standard methods based on density functional theory and in particular the local density approximation. Such systems, often termed "highly correlated materials" need improved methods for the description of their electronic structure and this was the topic of the present project. We implemented the so called "LDA+U" method into our WIEN2k code which usually gives a proper description of these systems without increasing the computational demand significantly. In addition we developed a "non- collinear-magnetism" (NCM) code, which allows that the magnetic moments of the atoms in the unit cell point into different directions. These new features were tested on various examples and will be made available to the more than 700 research groups worldwide which have presently licensed the WIEN2k code.