k points: PWscf vs Linear Response
Hello,
I am using QE 6.6. I noticed the following issue. I consider TiO2 without symmetries (nosym=.true. and noinv=.true. in the PWscf input) and I have 8 k points (this corresponds to the 2x2x2 k grid centered at Gamma). Then, when I run one of linear response codes (e.g. the HP code) I get the following issues:
- If 8 k points are listed directly in PWscf, then the routine LR_Modules/setup_nscf.f90 reads these k points from the XML file (because nks_start > 0). But then this same routine calls irreducible_BZ, and as a result 4 new k points are found and added to the list. So instead of 8 k points I end up with 12 k points. I do not expect to have 12 k points but the original set of 8 k points (which corresponds to the 2x2x2 0 0 0 k grid). Any ideas why setup_nscf finds more k points than there must be?
- If I use the automatic generation of k points for PWscf (i.e. 2x2x2 0 0 0), I obtain the same 8 k points in PWscf. However, when I use the HP code on top, then the routine LR_Modules/setup_nscf.f90 calls the routine kpoint_grid which generates 6 k points and then calls the routine irreducible_BZ which finds 2 more k points, so 8 k points in total. All seems good, however not all k points are exactly the same as those in the PWscf calculation, i.e. the ordering of k points is not the same and even the sign of some components of k points is not the same. I would need exactly the same k points and the same ordering as in PWscf. Why this is happening?
Thanks a lot, Iurii