Hi, on the GW first tutorial, it is stated at the start that:
Note that, unlike standard band structure calculations, here the KS states must be computed on a regular grid of k-points. (This limitation is also present with hybrid functional calculations).
Does that mean it is mandatory to use the ngkpt input variable to define the kpt grid? Or it is still possible to use kptrlatt?
Hi,
Have you tried?
I would say it should work. kprlatt produces regular k-point grids, which should be ok.
However, I would advise you not to use it.
The GW calculations scale nastily wrt to the k-point grid.
So it is in general a good policy to keep full control of the k-point grid.
Best,
Fabien
Hi Fabien, thanks for the reply.
I am having a segfault in the screening calculation and I am investigating why. While revisiting the tutorial to check if I didn’t miss anything I noticed this warning and thus wondered if there was specific instructions relating to the grid definition. I was using kptrlatt indeed but I tried with ngkpt and still getting the segfault therefore the problem is elsewhere.
I’ll close this as GW should work with kptrlatt and open a new topic on the problem I have.