Impurity scattering for dense grids

[Vahid]

Dear EPW community,

I have a system which requires a 250x250x250 k/q grid to converge the conductivity without impurity. I had to break up this big job into k-segments and q-segments to finish it.

To add impurity, I tried to break up the q-grid into segments but the code crashed. Does the impurity routine only work for k/q grids defined in the full BZ using nkfs and nqfs (because of IBTE)? I know that mp_mesh_k reduces the BZ points to the wedge but still that is a huge job. Is there a way to break up this large impurity scattering calculation?

Thanks,
Vahid

Vahid Askarpour
Department of Physics and Atmospheric Science
Dalhousie University
Halifax, Canada

[Vahid]

I did manage to break up the calculation by setting IBTE to .false. and removing the magnetic fields.
Vahid

[Vahid]

In the c-BN example for impurity scattering, degaussw=0 (adaptive smearing). I have two quick questions:
1. Does the impurity routine work only with degaussw=0?
2. Does adaptive smearing work for segments of fine k or q?

Thanks,
Vahid

[Vahid]

After running many tests, I conclude that the impurity scattering in EPW requires iterative BTE, adaptive smearing and a regular fine k/q mesh. The calculation cannot be broken into smaller segments.

I would appreciate a simple yes or no (hopefully someone other than me) as to whether the above conclusion is correct. I am only interested in the RTA impurity scattering rate not mobility calculations.

Thanks,
Vahid

[Nina Girotto]

Hello!

I would suggest looking a thte most recent tutorial on transport calculations (Friday 19th June: Hands-on tutorial: 2D/3D transport and LSDA ) that can be found on this website: https://epw2026school.dryfta.com/74-schedule. Note the use of the flag mp_mesh_k that can reduce the number of k points, and you can use degaussw. Also, in the tutorial you can notice the use of 'restart' flag, that enables your calculation to restart from the last computed step.

Hope this helps,
Nina

[Vahid]

Hi Nina,

Thank you for responding. I have already seen that tutorial. There are several issues:

1. If the k/q grid is 250^3/250^3, the use of mp_mesh_k which reduces the BZ points to wedge points is still a very large number (115211 to be exact). Even if you continue to restart, it will take at least a month if not longer to finish this impurity scattering job.

2. Impurity scattering is tied with IBTE and cannot be done by itself.

3. I usually break up a scattering rate calculation into several k-segments. For each k-segment, I break up the q-grid into segments. So a job that could take a month to finish, can be finished in several days. This I am afraid cannot be done with the current code. I may have to make several modifications to the code to make it run independent of IBTE.

Best,
Vahid