Convergence and Fermi level/Wannierization/...

[gbrunin]

Hi everyone,

I have a few questions about the convergence of EPW calculations.
1) If I have a doped semiconductor, and I run convergence tests for a given Fermi level, do I need to redo the convergence tests if I change the Fermi level ? Or can I simply use the same input as before with another fermi_energy ?
2) If I do convergence tests on a material, and then change the wannier functions, are the tests still valid ? I think not because the integrals will depend on the wannierization as well as on the fine grids, but I'm not sure.

I guess my questions can be summed up as "do convergence tests on fine k/q-points grids and smearings depend only on the material and the properties that we are interested in ?"

Thanks for your help,
Guillaume

[carla.verdi]

Hi Guillaume,

1) If your Fermi level is still inside the gap (nondegenerate semiconductor) then I don't think the convergence will change. On the other hand, if your Fermi level moves very close to the band edges, or your semiconductor is degenerate and it moves inside the conduction/valence band, then the convergence might change a bit - just think about converging a k integration when the Fermi pocket is quite large (easier) or when it is very small (more k points / low smearing required).

2) What do you mean exactly by changing the Wannier functions? If you mean taking another material, with different band structure etc, then yes, in principle you will have to check your results. If you change the wannierization for the same material I am not sure what you mean - there should be ideally one global minimum for your wannierization.

Best
Carla

[gbrunin]

Hi,

Thank you for your quick reply. For my second question : I did my convergence tests with Wannier functions obtained from defined projections, and the spreads are quite good (around 2) except for one function, for which the spread is around 6 (this function comes from a projection on an s orbital centered between two close atomic layers, which is supposed to be delocalized, so to me it was okay to have a larger spread). This still gives me a good band structure and good decays of H and the dynamical matrix. I was wondering if I could find better projections though. And if I do, I was wondering if I was going to have to redo convergence tests.

Thanks,
Guillaume

[carla.verdi]

Hi Guillaume,

If you find better projections but the decays look similar, and the band structure is already well interpolated, I don't think you need to redo convergence tests; it would be good to make just a couple of tests to make sure everything is ok.

Carla