[Wien] Three doubts

[Peter Blaha]

> 1) My first question is rather technical. It might sound a bit strange,
> but for technical reasons: since I only have to diagonalize the
> Hamiltonian, and all I have to do is to run x lapw1 -up/dn, I would like
> to write the eigenvalue files on the scratch directory, together with
> the vector files. Is it too complicated to modify the appropriate script
> to write the eigenvalue files to create a up(dn)lapw1_xxx.def file with
> the eigenvalue files on scratch? If so, could you indicate me the
> appropriate script, and which lines to change?

Either you change the x_lapw script, or, more simple,
create the uplapw1.def and dnlapw1.def files once   (x lapw1 -up/-dn -d)
edit up/dnlapw1.def ach change the path to the output1up/dn file
run:  lapw1(c) uplapw1.def

>
> 2) My second question is about the execution of initso_lapw. Since I
> have to run the diagonalization for different numbers of k-points, I'd
> like to make a job in which klists with increasing numbers of k-points
> are generated automatically using initso_lapw, and then the calculation
> is run until a desired number of k-points is reached. Is there any way
> of running initso_lapw in batch mode, just as it is possible with init_lapw?

You don't need to run initso for k-convergence tests again. Just do it 
once, and afterwards do:  x kgen -so


>
> 3) My third question is conceptual. I have asked it before on the list,
> but there was no answer, and I still could not find the answer anywhere,
> so here it goes again. I am doing spin-orbit coupling calculations on a
> spin-polarized system, and I am still in doubt as to the criteria to
> choose the magnetization axis for the SOC calculation. Should the
> magnetization axis be chosen parallel to the actual magnetization of the
> system? I have seen in the list that if we perform calculations for
> different axis, we can have a measure of the magneto-crystalline
> anisotropy of the system, so I am a bit unsure of how to actually
> perform the calculation. For the moment, I have chosen the default
> configuration proposed by initso_lapw, but for the moment I am not
> willing to calculate the magneto-crystalline anisotropy, just the
> eventual differences in the results when SOC is added. Now, is there
> such a thing as "the correct magnetization axis for spin-orbit
> coupling"? If yes, how should I choose it? This is a rather new subject
> for me, so I'd appreciate any very patient answers :)

If you know the magnetization direction from exp, then use it in your 
calc. Otherwise you should in principle test various possibilities.
Fortunately, when the elements are not too heavy and the structure is 
not very anisotropic, you may not see much difference in a 
"bandstructure", since the magnetocristalline anisotropy usually ranges 
from a few  mRy to even only a few microRy !!

-- 
Peter Blaha
Inst.Materialchemie, TU Wien
Getreidemarkt 9
A-1060 Vienna
Austria