Abstract
A linear scaling approach for general and accurate pseudopotential density functional theory calculations is presented. It is based on a finite difference discretization. Effective O(N) scaling is achieved by confining the orbitals in spherical localization regions. To improve accuracy and flexibility while computing the smallest possible number of orbitals, we propose an algorithm for adapting localization regions during computation. Numerical results for a polyacetylene chain and a magnesium oxide ring are presented.
Original language | English |
---|---|
Article number | 294210 |
Journal | Journal of Physics Condensed Matter |
Volume | 20 |
Issue number | 29 |
DOIs | |
State | Published - Jul 23 2008 |
Externally published | Yes |