Optimized basis sets for the collinear and non-collinear phases of iron

V M García-Suárez, C M Newman, C J Lambert, J M Pruneda, J Ferrer
2004 Journal of Physics: Condensed Matter  
Systematic implementations of density functional calculations of magnetic materials, based on atomic orbitals basis sets, are scarce. We have implemented in one such code the ability to compute non-collinear arrangements of the spin moments in the GGA approximation, including spiral structures. We have also made a thorough study of the degree of accuracy of the energy with the size of the basis and the extent of the orbitals. We have tested our results for the different phases of bulk iron as
more » ... s of bulk iron as well as for small clusters of this element. We specifically show how the relative stability of the different competing states changes with the degree of completeness of the basis, and present the minimal set which provides reliable results. (Some figures in this article are in colour only in the electronic version) Molecular dynamics packages based on density functional theory (DFT) [1] represent a specially useful set of tools in the theoretical analysis of materials. Most approaches use plane waves as a basis set, which allows a great degree of accuracy provided the number of plane waves is large enough, and use all the electrons in the atom to describe core and valence states. However, these approximations are numerically very expensive and do not scale linearly with the number of atoms. Other approaches, such as SIESTA, implement the tight-binding philosophy [2], using norm-conserving pseudopotentials [3], to integrate away core energy levels, and very flexible basis sets (BS) made up of numerical atomic-like wavefunctions to handle valence electrons. Assessment of the degree of reliability of those BS might be essential, since competing would-be ground states may in some instances have small energy differences. Such analyses have already been performed for selected molecules and solids [4, 5] . Those studies show how both the number of wavefunctions (WF) used as well as their extent are variational parameters, providing therefore a path for systematic improvements
doi:10.1088/0953-8984/16/30/008 fatcat:d7nu2y66afag7lbsi6tp2y3blq