Electron linewidth for degenerate bands

[Luis]

Hi Samuel,

I've followed the EPW tutorial and reproduced your Fig.5a (arXiv:1604.03525v1) for the valence band.

Since the calculation of the linewidth (Im[Sigma_nk]) does not sum over the band index n, that quantity should not invariant when the band n happens to be degenerate.

For B-doped diamond, the valence band is doubly degenerate along the L-Gamma direction. I was expecting that in such cases the user should manually average the EPW linewidths of all the states in the degenerate subspace, analogous to what is mentioned in the paper:
"The square moduli of electron-phonon matrix elements are averaged over degenerate bands and modes."

However, I'm finding that the linewidths for the two highest occupied bands (n=3,4) that come from EPW (purple and blue in the attached image) are already equal to each other (and invariant between different runs) along the L-Gamma direction. Is EPW already printing out the average linewidth when degeneracy is detected?

Thanks in advance and regards,

Luis

[sponce]

Dear Luis,

Thank you for posting your question here.

Degenerate band should not be equal but could by chance. If you look more closely, it seems that the blue and purple are not exactly equal to each other.

In the new version of EPW that will be release in September 2016, I explicitly make the average in case of degeneracy directly in the code.
I must say that the average will be based solely on a test of eigenenergies values (quite stringent though). It potentially could be that two bands that cross at the same energy would then be wrongly
averaged. This should be quite unlikely.

Best,

Samuel