The ABINIT Users Mailing List ( CLOSED )

[Ivan Rungger <runggeri@tcd.ie>]

Dear Xavier and ABINIT users,

Xavier Gonze wrote:

> It is an interesting problem. I am not very familiar with the coupled 
> effect
> of spin-orbit coupling and the DFT approximation for the d-band of 
> transition metals.
> Actually, it might be
> that what you observe is not a problem of the pseudopotential. Just to 
> be sure,
> have you tried to reproduce the atomic splitting, and compare with the 
> results
> of an all-electron LDA code ?

No, I did not do this (yet).

> Are you aware of other calculations using other codes
> (may be all-electron using WIEN ?), that would give a correct (or a 
> bad !) answer ? 

There are some papers on abinitio LDA calulations including spin-orbit 
coupling of MnBi in the NiAs structure I am aware of, e.g. J. Koehler 
and J. Kuebler, J. Phys.: Cond. Mat. 8 (1996) 8681-8694,  R. Coehoorn 
and R. A. de Groot, J. Phys. F: Met. Phys. 15 (1985) 2135-2144.
The effect of including spin-orbit coupling is only small, the main 
features of the bandstructure (especially the splitting of bonding and 
antibonding d bands) result already from a spin polarized calculation 
without spin-orbit.
I did calculations with ABINIT without spin-orbit using the HGH and the 
TM pseudopotential for Mn from the abinit web site. With the TM pseudo 
the bandstructure was comparable to the ones found on the papers, while 
with the HGH pseudo the d bands were wrong.  I did also a calculation 
with SIESTA (no spin-orbit), and the bandstructure was similar with the 
one obtained with ABINIT+TM pseudopotential.
Therefore I wanted to try another pseudopotential than the HGH also for 
spin-orbit calculations. I will now perform some tests on the HGH 
pseudopotential for Mn, and I will also try the HGH pseudopotential from 
the web page of ABINIT, which includes semi-core states.

> The pseudopotential might be problematic, but also, it might be that 
> there is a
> problem with the implementation in ABINIT (that is why the atomic test 
> is interesting).

In the ABINIT documentation magnetic calculations including spin-orbit 
are described as develpment features. Are these features well tested by now?
Can I get the spin and the orbital components to the magnetic moment 
from ABINIT? Is there any further documentation on the subject (ABINIT + 
magnetism + spin-orbit coupling)?

> You can contact stephane.bernard@cea.fr in this respect.
>
> You should contact Andrew Rappe rappe@sas.upenn.edu . 

OK, I will do this.


Thanks for the help,
   Ivan