[Wien] Problem in SP+SO calculation

[Peter Blaha]

> More that one month, my problem in SP+SO calculation (spin-polarization +
> spin orbital coupling) still presents, e.g. in Fe6C1 (SG_162). I tried to
> change all the possible input parameters, but no help.
>
> My problem is, once the SO coupling included, in each SCF cycle WIEN2K says,
> "Energy to separate semicore and valencestates:-999.00000", regardless the

I could verify your problems. Aparently it is another example for a bug with
the relativistic LOs.

Compare the eigenvalues in case.scfso  with and without RLOs (defined in
case.inso):

Without RLO: Aparently there are 12 Fe-3s (-5.xRy) and 36 Fe-3p states, which
             is ok.
     -5.4933772   -5.4898256   -5.4882286   -5.4882285   -5.4877765
     -5.4877765   -5.3030341   -5.2993449   -5.2976915   -5.2976914
     -5.2972203   -5.2972202   -3.0581789   -3.0573933   -3.0571405
     -3.0570311   -3.0456359   -3.0419113   -3.0304325   -3.0193464
     -3.0189809   -3.0171594   -3.0167227   -3.0097486   -2.9723582
     -2.9723318   -2.9722416   -2.9715760   -2.9579251   -2.9550374
     -2.8430812   -2.8423116   -2.8418980   -2.8417820   -2.8310462
     -2.8261074   -2.8173465   -2.8050342   -2.8041627   -2.8020765

     -2.8016378   -2.7954859   -2.7851243   -2.7850551   -2.7840346
     -2.7838253   -2.7701415   -2.7671925   -0.2897566   -0.0880226

With RLO we have 2 "ghost-states (spin up and dn)" at -7 and -4 Ry:
     -7.3502943   -5.4933772   -5.4898256   -5.4882286   -5.4882285
     -5.4877765   -5.4877765   -5.3030341   -5.2993449   -5.2976915
     -5.2976914   -5.2972203   -5.2972202   -4.1732755   -3.0583184
     -3.0575541   -3.0573025   -3.0571910   -3.0456838   -3.0420292
     -3.0303027   -3.0194769   -3.0191123   -3.0172911   -3.0168631
     -3.0093392   -2.9724072   -2.9723803   -2.9722859   -2.9716236
     -2.9579179   -2.9549944   -2.8431740   -2.8424106   -2.8420047
     -2.8418884   -2.8311200   -2.8262007   -2.8173299   -2.8051049

     -2.8042306   -2.8021470   -2.8017145   -2.7954897   -2.7851708
     -2.7851007   -2.7840741   -2.7838731   -2.7700058   -2.7671423



At present I can only repeat my warnings: Use the relativistic LOs with care,
and check/compare your results without them. They should be hardly needed
for light elements like Fe.

PS: Since these ghost states do not appear for all k-points, "wide semicore
bands" occur without any gap between them, so ESEPER could not be found.

                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------