[Wien] Inconsistency between energy and forces ?

Peter Blaha pblaha at zeus.theochem.tuwien.ac.at
Tue Nov 25 15:24:11 CET 2003


> > Just a guess: Please make a short test with LDA (5) instead of GGA.
> > I'd like to see if the total energy differences between the two
> > structures comes from GGA-grids.
>
> Indeed, with LDA the energies are only 0.0008 apart. And also the forces are
> almost identical now in the small and the large cell. This feels like
> magic...!
>
> Is it possible to explain the background of this? And why is the problem
> apparantly related to particular cell symmetries (and not to particular
> atoms, or atom-atom distances, ...)? Is it known which space groups are
> 'vulnerable' for GGA? Can it be corrected with e.g. a larger NPT? ... ?

It is related to the FFT transformation in lapw0 and the calculation of
gradients for GGA in the interstital region. Thus it concerns only CXY
lattices and only GGA calculations.

Sorry, aparently an already known bug was not fixed for CXY lattices.

When for GGA in lapw0 the Gradients are calculated on the real-space
FFT mesh, we need to know if this mesh is karthesian or with some odd
angles. This is provided with the switch ORTHO and for some reasons
this is set to TRUE in CXY case (although I've already fixed this for
CXZ and CYZ lattices, even with some comments).

Please change this in SRC_lapw0/latgenf.f

      RVFAC=2.D0
!      ORTHO=.TRUE.                           ! comment this line
      ORTHO=.false.                           ! insert this line
      GOTO 100
!
!.....CXZ CASE (CXZ LATTICE BUILD UP)
 51   CONTINUE
!.....CXZ ORTHOROMBIC CASE



Stefan, please let me know if with this fix you get comparable energies/forces.
I'll put it then as update on the web.


PS: One will NEVER get identical numbers, because of sqrt(2) scaling an
exactly "equivalent" k-mesh is never possible.


PPS: This bug concerns ALL CXY calculations with GGA !!!! So it is really
serious.

                                      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/
--------------------------------------------------------------------------




More information about the Wien mailing list