[Wien] lcore change from wien2k_02 to wien2k_03

[Natalie Holzwarth]

Thanks very much for your helpful comments.  I enclose a few more details
below and welcome additional comments if you have time.

On Tue, 15 Jul 2003, Peter Blaha wrote:

> >       Perhaps this is an old question, but we were unable to locate a
> > discussion of it in the archives.
> >       We were trying to check our implementation of wien2k_03 wrt runs
> > using the wien2k_02 code and found there is a shift in the total energy
> > (~0.04 Ry) which we traced to a small difference in the lcore pgms -- in
> > particular the insld.f pgm:
> >
> >   version_02:
> >     271       DO 10 I=1,NP+NPXX
> >     272       IF(I.GT.NP)DV(I)=DV(NP)
> >     273   10  DV(I)= DV(I)/2./DR(I)
> >
> >   version_03:
> >     271       DVR=DV(NP)
> >     272       DO 10 I=1,NP+NPXX
> >     273       IF(I.GT.NP)DV(I)=DVR
> >     274   10  DV(I)= DV(I)/2./DR(I)
> >
> >   What this means is that DV(I) was generally discontinuous at RMT in
> > version_02 but not in version_03.  This effected only some of the atoms in
> > our unit cell; those which have a small tail beyond RMT.   Our questions
> > are:
> >
> > 1.  Should we assume that the version_03 code gives a better
> > approximation?
> 
> Yes of course.
> 
> > 2.  It would be reassuring if one could argue that the energy shift from
> > version_02 to version_03 is reasonably constant wrt the valence properties
> > of the system.  For example, hopefully it would not effect the valence DOS
> > or the lattice relaxation??  Does anyone have experience with this?
> 
> I would expect such more or less constant E shift.
> 
> However, if there is a large shift, I would be concerned about "charge leakage"
> which might affect a lattice relaxation (probably not the DOS, which is quite
> insensitive). Ideally, core states have "zero"! wavefunctions outside RMT and
> thus the bug of the old version should not make any difference.

I have seen in the literature some treatments of core tails.  Are these
not implemented in wien2k?

> 
> How large is the difference in :NEC01 of case.scf ?

For example, we have
:NEC01: NUCLEAR AND ELECTRONIC CHARGE  304.00000 303.97497   1.00008

This is mostly due to a phosphate group (PO4) which we have treated by
choosing RMT=(P-O bond)/2.  In a pseudopotential treatment, we would
choose the equivalent of RMT to be larger for P than for O in order to
reduce the core "leakage", but my reading of your userguide and other
sources encouraged me to choose RMT to be equal for P and O.  Was that a
bad idea?  If we had chosen a larger RMT for P than O, we would probably
have needed to choose a larger value for RKMAX and/or worry about ghosts?

     Thanks again for your insight.    Natalie Holzwarth
 
> Could you include more states as valence ? Could one change RMT ?
> 
> Regards
>                                       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/
> --------------------------------------------------------------------------
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> 

N. A. W. Holzwarth			email: natalie at wfu.edu
Department of Physics			www: http://www.wfu.edu/~natalie
Wake Forest University		        voice: 336-758-5510	
Winston-Salem, NC 27109-7507            fax: 336-758-6142 
U. S. A.