It is well know that GGA+U calculations can end up in different states (see previous posts).

If you have used the same parameters, you should use the total energies of the two calculations and the one with lower energy should be the better one (but not necessarily the best !)

In any case, I always recommend:

First do a standard GGA calculation (also check forces, if they are small).
Then add SO and do a GGA+SO calculation (if you really have heavy elements?)
x lapwdm -up -so      to generate density matrices

and finally add the "U" and do   GGA+so+U

There is, however, no guarantee that you find the lowest energy ground state ! Sometimes it is good to start with a smaller U-value than anticipated. Sometimes you may do a structure relaxation at the GGA-level and use this structure,
or a structure relaxation at GGA+U (no SO).

If SO is not a major effect, I'd first do everything without SO.

From these advices you can see that there is not a unique and save way, which works in all cases.

On 11/23/2015 12:10 PM, Santu Baidya wrote:
Dear Prof. Blaha and wien2k users,

      I use wien2k version WIEN2k_14.1. Recently I did spin-orbit
coupling calculation under GGA+U+SOC approximation with two different
approaches- one starting from collinear ground state and another
starting from scratch (random wavefunction). Two approaches give me
completely different results -metal and insulator.

Please tell me which approach should I accept as solution.

I look forward to get reply.

Thanking you,

Santu Baidya
University of Duisburg-Essen

Wien mailing list


Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300             FAX: +43-1-58801-165982
Email:    WIEN2k:
Wien mailing list

Reply via email to