Re: Spin dependent LDOS calculations

[Alex Ankudinov <alex@leonardo.phys.washington.edu>]

Hi David,
> I'm comparing my LDOS and XANES with a literature reference, carried out by 
> a collaborator of our group using band structure (FP-LMTO methods).
>   Yes, I do use the SPIN card to generate spin-dependent DOS. I get Cr-spin 
> up DOS AND Cr-spin down DOS from one use of the card, say SPIN 2 (one 
> calculation). Use of SPIN -2  generates the same plots in reverse.

Then definitely try calculations with NOHOLE card and increase cluster
size to 200 - 300 atoms.

> 
> 
> >  Also FEFF8 is not self-consistent code
> >with respect to spin on Cr. You can change that manually, by editing
> >subroutine getorb and changing the number of the variable ispn(24,i).
> >there is only one nonezero value in that array - 4. corresponding to
> >4 spin-up 3d electrons  on Cr (iph=2) and 4 spin-down on Cr (iph=3).
> >This number should not be integer.
> 
> 
> Could you expand on this a bit? I don't quite understand what this variable 
> is, and if its not to be an integer, what would be a sensible entry?

This variable sets the spin on each Cr atom. ispn = x means there are
y electrons with spin down and y+x spin-up electron. I need this
to calculate spin dependent electron density for use with LSDA
functionals. Thus total (up+down) density FEFF8 calculates
self-consistently. The difference (up-down) = x* \phi_i(r)**2
where e.g. for Cr \phi_i(r) is 3d orbital. The correspondence
between index i in ispn(iz, i) and orbital character is given at
the top of getorb subroutine. (iz- is the nucleus charge or element
number in periodic table). Thus by changing 'x' you modify the spin-up
- spin-down density. I put 4 in the code assuming d^4 configuration on
Cr and according to Hunds rules all of them are spin-up.
The better choice for 'x' is to set it to the value which you
probably have already from FP LMTO calculations.

Hope that clarifies the ispn(iz,i) variable for spin-dependent
calculations.

Best wishes
Alex Ankudinov