number of band (nbnd) in the calculation and projection

[jir520]

Dear EPW users

I am trying to understand parameters used in EPW correctly since I am quite new to the code. It is a great help if one can answer for my two questions below.

01
I kindly request your clarification regarding the correct total number of bands (Bloch states) for the nscf and wannier calculation. In nscf of diamond example, nbnd=4 (which is total bands). But since there are 2 atoms in the cell and there are 5 bands per each atom ( 2 x s and 3 x p shells), do we need to use 10 bands there ? Or can we use any number greater than to our required wannier bands there? ( Simply what is the correct method of getting nbnd for nscf?) Is it important to use only primitive cell and is it wrong if one use conventional cell?

02
Also I have no idea how to get the correct position for projections. In diamond case, the center is 0,0,0. is it the same if I take it as "C:sp3" (since there is a C atom at 0,0,0 ). When I go through this forum I saw some people are taking projection positions exactly like f=0.32541,0.45324,0.45678. How do we get it that way? Can we first check the position of the center of wannier function in case.wout by considering the shape ( as Ga:d N:sp3 for GaN case ) and then putting it in a second try ? DOES the position affect to the final answer ? What is the accepted way of doing this?

I appreciate your answer very much.

Thank you
Jaya

[sponce]

Dear Jaya,

1) It all depends on what you want to calculate. If you are interested in valence and conduction bands, then you need to include all the bands up to the one you are interested in the nscf.
So in the case of diamond, if you want only the valence, you can use nbnd=4 in nscf.in and if you are interested in conduction bands, you can use nbnd=8 for example.
Then the number of Wannier band you are going to wannierized will be a subset of that total number of bands from nscf.in
You can use the cell you like but primitive cells are always smaller or equal to conventional one so its usually wiser to use primitive cells for computational reasons.

2) I personally never use centre positions. You can indeed use "C:sp3" to get 4 bands wannierized.
I think the centre position is for complicated systems with strong hybridizations where the normal atomic shape does not work.
Yes the position will affect your spread and therefore the quality of the interpolation.

PS: Can you modify your affiliation in your profile? "University:" should not be "PhD student" but the name of your university or research centre. Thank you.

Best,

Samuel

[jir520]

Dear Dr. Samuel

Thank you and I appreciate your support sending the reply. Don't we need to get nbnd=8 ( two times as for one atom) for only valance case since there are two atoms in the cell and since there are 4 valance electron shells per each atom.

I tried to change the affiliation. But it says I am not authorized to edit it

Thank you

Jaya


University of Saskatchewan
Saskatoon
Canada

[sponce]

Hello,

You can put two electrons on each bands (except if you have spin enable).

I changed your settings.

Best,

Samuel