[Wien] Possible bugs in Mixer

[Guoping Zhang]

Dear Prof. Blaha,

Thank you so  much for your further great help!

I followed your suggestions and found a possible reason. The problem is
with oneAPI (2024-2025) which is very buggy. Its setvar.sh cannot detect
errors when siteconfig has link errors. Although the wien2k compilation has
no errors, when the mixer calls its  routines, it produces two different
results on the :INFO lines, which is the first warning, after the 4th
iteration.  I used two different accounts on the same machine, but it gave
different results as well. I found an older compiler which does not need
oneAPI. Things are back to normal.

Thank you so much once again!

Best wishes,

Guoping



On Sat, Dec 20, 2025 at 4:01 AM Peter Blaha <peter.blaha at tuwien.ac.at>
wrote:

> Clearly, :DIS seems to converge. So "mixer" cannot be the problem, since
> it produces a density which is identical to the new density in the next
> iteration (within 0.0000001).
> In any case, on machine2 you still have a non-zero :dis, while on
> machine 1 it is zero to all digits. So scf quality is not EXACTLY the
> same (although for ALL practical purposes it is the same).
>
> Sure, you should check the convergence of the total energy (and other
> quantities like  :FGL, :EFG   :VCOUL :FER ...) as well.
>
> > Eigenenergies are
> >
> >      57    0.555640059910   0.55363321799308D-03
> >      58    0.555640061603   0.55363321799308D-03
> >      59    0.740271026751   0.00000000000000D+00
> >      60    0.740271028515   0.00000000000000D+00
> >      61    0.748205005635   0.00000000000000D+00
> >      62    0.748205006555   0.00000000000000D+00
> > -------------------------------------------------------------------
> >      57    0.555640058339   0.55363321799308D-03
> >      58    0.555640061273   0.55363321799308D-03
> >      59    0.740271025639   0.00000000000000D+00
> >      60    0.740271027711   0.00000000000000D+00
> >      61    0.748205004954   0.00000000000000D+00
> >      62    0.748205005071   0.00000000000000D+00
>
> I do not understand these numbers. What should they be ???
>
> Check convergence of eigenvalues with :EIGxxxx
>
>
> Momentum matrix elements: They are determined only up to a phase (same
> with the eigenvectors). So you cannot compare plain numbers between 2
> machines for these quantitites.
>
> Regards
> --
> -----------------------------------------------------------------------
> Peter Blaha,  Inst. f. Materials Chemistry, TU Vienna, A-1060 Vienna
> Email: peter.blaha at tuwien.ac.at
> WWW:   http://www.imc.tuwien.ac.at      WIEN2k: http://www.wien2k.at
> -------------------------------------------------------------------------
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:
> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20251222/11a6293d/attachment.htm>