Berlin 2012 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

O: Fachverband Oberflächenphysik

O 20: Theoretical methods I

O 20.5: Talk

Monday, March 26, 2012, 18:45–19:00, A 060

Analysis and elimination of the linearization error within the FLAPW method — •Gregor Michalicek, Markus Betzinger, Christoph Friedrich, and Stefan Blügel — Peter Grünberg Institut and Institute for Advanced Simulation, Forschungszentrum Jülich and JARA, 52425 Jülich, Germany

The full-potential linearized augmented-plane-wave (FLAPW) method is widely considered to be one of the most accurate electronic structure methods for solids, providing the density functional answer to high precision. Within the atom-centered muffin-tin (MT) spheres, its basis consists of solutions to the radial scalar-relativistic Dirac equation evaluated at a predefined energy parameter and – as is characteristic of linearized methods – also includes their first energy derivatives to allow for variations around the energy parameter. In some cases, however, the flexibility of the basis is insufficient, and the remaining linearization error noticeably affects calculated physical quantities, e.g., equilibrium lattice constants and band gaps. Then, the results may also depend on parameters that are not convergence parameters, such as the MT radii and the energy parameters. To remove the dependence and improve the numerical precision, we analyze in detail how the basis can be extended in a systematic manner by adding local orbitals, defined with either energy parameters at higher energies or second-order energy derivatives. Furthermore, improving the MT part of the basis in this way leads to a faster basis-set convergence so that the resulting basis yields a higher precision than the popular APW+lo method at equal basis-set sizes.