[Wien] freezing of shells, and DOS

[Alessandro Mirone]

Hello Wien community,

   Which is the best way to freeze electrons in the shell
and what are the effects of freezing on the DOS?

I am calculating hematite, a spin polarised calculation,
the calculation converges well and I can see in 
the partial density of states the 3d resonances not far from Fermi energy

Then I freeze the 3d shell for one Fe atom ( the unit cell counts 4 Fe and
6 O), for this atom I set 5 3d up electrons in case.incup and 0 3d down
electron in file case.incdn with a shift of 1 Ry. In case.in2c I change 
the number of free electrons from 92 to 87. Last but not least the
linearisation of the apw 3d wave is set by hand to 20 Ry in case.in1c 

With this setup calculation converges well, the final energy is 0.5Ry
higher than in the unfreezed case.

Apart this difference in total energy I get a very strange behaviour
when calculating the unfreezed DOS in the field of the freezed hematite :

 -- I use clmsum clmup and clmdn files of the freezed case,
   run lapw0, lapw1 ( up/dn ) and finally the dos, but the
   3d resonance is about 8-10ev lower !!!


Why does the resonance shifts so low?
To calculate the dos

  -- I reset case.in1c resetting the 3d linearization energy and letting
lapw1 to search the optimal
  -- I even worked in a new independent directory on an unfreezed case:
     when using the clmsum/up/dn freezed  files the 3d resonance
still shifts
  -- 10eV is a lot ( one expects U of the order of 5-8 eV in general )

  -- Fearing a charge leack I tryed also with a 3Ry shift in .inc, .incup
, .incdn, but that does not help. When I use 5Ry lcore for some reason
   complains  that the shift is too small 


  -- I am 100% sure the all the effect is contained in clmsun clmup clmdn

So to resume, I cannot explain such a big shift. I can expect a small
shift because the number of 3d electrons in the unfreezed case
can be different from 5, but such a big shift would correspond to a big
difference.
Do you have any idea?

Regards

                  Alessandro Mirone