EPW Forum

Dear EPW users,

I find that the electron linewidth, comprising of both long-range and short-range terms, for equivalent k-points are not exactly the same. Sometimes, the difference can reach 50% in my calculation of a polar material- zinc blende GaAs. For example, the linewidth at (0, 0.02, 0) in crystal coordinate is ~1.2 meV which is lower than the value (~ 2.4 meV) at other degenerate points e. g., (0.02, 0, 0). I also checked that the energy and velocity at these points which are very close.

Should degenerate kpoints have the same linewidth? Have EPW averaged the values for degenerate points? Thank you for your help!
p.s. I use gamma-centered k- and q-meshes.

Tianshi Wang
Graduate student, University of Delaware

Dear Tianshi Wang,

Observable should indeed be the same for symmetry equivalent k-points.

However, due to numerical accuracies, it is not exactly true. It should be close if you are converged.

No average is performed on the k-points. However, for each k-point an average is made for the linewidths on degenerate bands.

The energies and velocities should be exactly the same (at least to a lot of digit). If not, you might want to raise ecut, coarse k-grid and check that your
Wannier interpolation is correct (i.e. that the Wannier interpolation respect the symmetry of your crystal).

Best,
Samuel

Dear Samuel,

Thank you for your reply! I will check the convergence.

Best
Tianshi

Dear Samuel,

Thanks for your reply! I would like to ask another question.
When I calculated the electron-phonon coupling of Si (space group 227), I used "vme = wannier" in EPW to calculate the energy, electron linewidth, and velocity of the k-point in the full BZ. In addition to the problem of symmetry equivalence between k-points (the problem encountered by the questioner), I also encountered a problem:

The velocity of the k-point does not satisfy the constraints of the k-point little co-group.

For example:
In primitive basis, the coordinates of the k-point are (0.1, 0.1, 0), and the little co-group is 4m.m, which will constrain the velocity of the point to (Vx,0,0). But the calculated result is (0.70260269, 0.01692294, -0.15118160). It can be seen that there is a significant deviation in the y and z directions, and other k-points may deviate more.

I tried a dense coarse grid (nk1=nk2=nk3= 10, nq1=nq2=nq3= 5), and I also tried "vme = dipole", and the problem still exists.

I don't know if this is also a problem with Wannier interpolation. How can I avoid this problem?

Best,
Yan
Graduate student, Nanjing University