EPW Forum

Dear All,

I have a following question

Is there a way to specify coarse k-point mesh (the one one used in nscf run) explicitly in epw.in
instead of using nk1 nk2 nk3 ?

Otherwise if one uses nk1 nk2 nk3 for nscf.in one might end up with negative values in kx, ky, kz for specific examples, which is not excepted later in epw run
"coarse k-mesh needs to be strictly positive in 1st BZ".

Many thanks in advance,
Andrij

Dear Andrij

Indeed the coarse k-mesh needs to be a uniform mesh with coordinates [0:1) (crystal coordinates). This is a requirement for the wannierization (it would be the same if you were to run just wannier90). Therefore in the nscf.in you are expected to input such uniform positive grid.

Hope this helps,
Carla

Dear Carla

Many thanks for your reply. I understand that requirement for "wannierization" and I can use a uniform positive grid in nscf.in by
(instead of K_POINTS automatic)

K_POINTS crystal
nk
k1x k1y k1z
...

However my question is whether it is possible to do the same in epw.in instead of using nk1,nk2,nk3

Many thanks,
Andrij

carla.verdi wrote:Dear Andrij

Indeed the coarse k-mesh needs to be a uniform mesh with coordinates [0:1) (crystal coordinates). This is a requirement for the wannierization (it would be the same if you were to run just wannier90). Therefore in the nscf.in you are expected to input such uniform positive grid.

Hope this helps,
Carla

Dear Andrij,

Its currently not implemented.

Keep in mind that this coarse grid needs to be commensurate with the phonon grid as well.

In principle you could implement it yourself if you want the coarse k/q grids to be non-homegneous. You would also need to correctly compute the associated weights.

You can take a look at the way "filkf" is implemented and try to implement something like "filk" I guess.

Best,

Samuel