[Pw_forum] k-points mismatch

[Paolo Giannozzi]

Eyvaz Isaev wrote:

> [...]  first I did phonon calculations for a base centered 
> orthorhombic cell (ibrav=9) using 4x2x2 mesh (10 q-points) 
> and got an error message from q2r.x "q not allowed" [...]
> Then I tried to generate q-points using kpoints.x which stated 
> that 4x2x2 mesh "has not the correct symmetry" and gave only 
>  6 q-points. OK, I used 4x3x3 to generate q-points and got 10
> q-points without any message from kpoints.x. But when I started
> phonons with 4x3x3 q-mesh, the ph.x program generated 14 q-points.
> 
> Can someone explain such kind discrepancy between two programs 
> generating q-points? 

I cannot (yet) explain why the two programs behave in a wrong
but different way, but I have half an idea of what triggers this
bug. With this Bravais lattice (which, by the way, kpoints.f90
incorrectly reports as "orthor 1-(fco)"), axis 1 and 2 have the
same length, different from the length of axis 3. So you should
use a 2x2x4 mesh, or a 3x3x4, not the ones you used. I don't see
why this confuses the logic of the symmetry check, though, since
there are no symmetry operations sending axis 1 into 2.

Paolo
-- 
Paolo Giannozzi, Democritos and University of Udine, Italy