Spin-polarized calculations: theory
Posted: Fri May 09, 2008 8:37 am
I have a general question about spin polarized calculations.
The thing that got me wondering is the fact that the exchange-correlation --- which is already present in the usual NON-spin polarized calculations --- should operate on both the positions and spins of the electrons. In other words, the spins of the electrons should already be accounted for in these cases. So then there would actually be two densities, "up" and "down", already treated in these cases. So how do truly spin polarized calculations (SPC) differ from these ?
Does the exchange-correlation in the NON-spin polarized calculations in fact only look at the positions, and not the spins, so that the interaction between spin up and spin down electrons is ignored ? This would lead to two uncoupled systems (up, down), and in effect only one of them needs to be treated (spin up), and a factor of two is added later (to account for spin down).
In SPC the interaction terms are then included, and using the user supplied difference in spin up and spin down (= MAGMOM), one of the spin bands are preferentially filled, to ensure a net magnetization ? This would explain why SPC starting with MAGMOM = 0.0 does not lead to net magnetization, although this should be correct (for iron, for instance), since the user has not specified any spin preference.
This is all just a lot of speculation, since I have not found any details on SPC in the VASP manual or in any other easily accessible document (perhaps because of sloppy searching ? :-)). Any clues/help would be nice to understand how SPC work in VASP or in general.
The thing that got me wondering is the fact that the exchange-correlation --- which is already present in the usual NON-spin polarized calculations --- should operate on both the positions and spins of the electrons. In other words, the spins of the electrons should already be accounted for in these cases. So then there would actually be two densities, "up" and "down", already treated in these cases. So how do truly spin polarized calculations (SPC) differ from these ?
Does the exchange-correlation in the NON-spin polarized calculations in fact only look at the positions, and not the spins, so that the interaction between spin up and spin down electrons is ignored ? This would lead to two uncoupled systems (up, down), and in effect only one of them needs to be treated (spin up), and a factor of two is added later (to account for spin down).
In SPC the interaction terms are then included, and using the user supplied difference in spin up and spin down (= MAGMOM), one of the spin bands are preferentially filled, to ensure a net magnetization ? This would explain why SPC starting with MAGMOM = 0.0 does not lead to net magnetization, although this should be correct (for iron, for instance), since the user has not specified any spin preference.
This is all just a lot of speculation, since I have not found any details on SPC in the VASP manual or in any other easily accessible document (perhaps because of sloppy searching ? :-)). Any clues/help would be nice to understand how SPC work in VASP or in general.