Abstract
The energy-versus-volume curve of the spin-density wave (SDW) in body centred-cubic Cr is calculated with the density functional theory/full-potential linearized augmented plane wave (DFT/FLAPW) method using the generalized gradient approximation (GGA). The predicted ground state is not the SDW, in contrast to an earlier FLAPW calculation. A conjecture is formulated that the widely varying results of the local density approximation (LDA) and GGA— and of different solution methods—can be scaled by the size of the calculated moment. As a consequence, experimentally relevant properties of the SDW can be calculated by tuning the moment. The implications of these results for the ability of DFT to describe Cr are discussed.
For the SDW, we use a supercell of dimensions (a, a, 20 a) (a is a bcc Cr lattice constant), which is close to the observed period of the SDW (20.83 a). First we check our supercell calculations against other methods. Figure 3 shows the energy difference ESDW −EAF for different values of the reduced wavevector q/a∗ (p = 1/(1−q/a∗) with p the period in bcc lattice constants, q is the wavevector of the SDW and a∗ = 2π/a) obtained with GGA/FLAPW (FLEUR implementation) [15] and GGA/PAW [17,18]. Both for a = 2.884 Å (squares) and a = 2.849 Å (circles) there is excellent agreement between the PAW data (white) and our FLAPW results (black), even on this mRy/atom scale (the deviating values [17] of the less accurate LMTO-ASA method will be discussed below). Also the FLEUR-FLAPW point at q/a∗ = 0.917 is in good agreement, the one at q/a∗ = 0.292 that favours the SDW is not. Because it can now be excluded that this is due to the FLAPW method, we can strengthen the conclusion of Hafner et al that the latter point is erroneous, and we show that there is no contradiction between PAW and FLAPW.
To download the article click on the link below:
0 Comments