[Wien] Bug in kgen?

[Peter Blaha]

Thank's for pointing out this problem.

The specification of three "division numbers" is apparently very dangerous and
if you "break" the symmetry with these numbers, it will produce nonsense.
This is also true for P lattices (try P cubic and 3,4,5 mesh).

The subroutine  reduz  is not prepared to handle "wrong" divisions. It creates
"integer" indices for a k-vector (like (2 0 0) and (0 0 2)) and if a symmetry
operation transforms x into z, these k-points are "equivalent", because it does not take
into account that the divisors in x and z is different. (actually, these k-points would be
(2/3 0 0) and (0 0 2/5) ad thus are of course NOT equivalent.

I do not have a quick solution. At the moment make sure, that your mesh is
compatible with lattice symmetry.

mazin schrieb:
> I found strange behavior in the kgen module.
> 
> In the outputkgen file there is a list of all k-points with the 
> corresponding inequivalent points ("relations"). This starts with the 
> 37th line in the case.outputkgen file. This information is quite useful 
> for postprocessing. I found that this list is always correct if you use 
> a non-centered lattice, that is, not body-centered, face-centered, 
> base-centered etc, and it is always correct when the number of divisions 
> is the same for all reciprocal lattice vectors. However, if you use a 
> centered lattice (try fcc, for instance), and division numbers as, say, 
> 5,5,2, the resulting list is plainly incorrect. It gives "relations" 
> with numbers that exceed the number of inequivalent points. Yet the 
> actual list of inequivalent points (case.klist) is correct. Any comments?
> 
> Thanks
> Igor

-- 
-----------------------------------------
Peter Blaha
Inst. Materials Chemistry, TU Vienna
Getreidemarkt 9, A-1060 Vienna, Austria
Tel: +43-1-5880115671
Fax: +43-1-5880115698
email: pblaha at theochem.tuwien.ac.at
-----------------------------------------