Dear Sophie,

One reason why the intraband and interband velocities are computed in separate parts of the code may be that, in the specific projects for which those parts of the code were originally written, the full velocity matrix was not needed. For example, in the paper you mention the intraband velocity is used for the low-field Hall coefficient in Sec. III, and the interband velocity is used for the interband magnetic circular dichroism in Sec. IV.

The other justification may be that the full optical conductivity is usually separated into intraband and interband parts, which are then computed separately.

But I agree that it might make sense to give the option to to compute the full velocity matrix. And if you have an application in mind that would benefit from that option, that could be a good reason to do it.

Perhaps the discussion of how to go about implementing this could take place on the GitHub page of the code, e.g., by raising an issue or making a pull request.


Ivo Souza

Centro de Fisica de Materiales
San Sebastian, Spain

On Thu, 8 Jul 2021, Sophie Beck wrote:

Dear All,

I would like to compute the elements of the full velocity matrix (Eq. 18 in YWVS07), i.e. including contributions of both 1) diagonal intraband and 2) off-diagonal interband on a given list of k-points. I have seen that both contributions are already implemented: 1) is easily available on a given k-mesh from geninterp.f90 and 2) is used in berry.f90.

I was wondering if there is a particular reason that I am overlooking as to why there is no option to compute the full velocity matrix. If this is not the case, I suppose that geninterp.f90 could be a reasonable place to add this?

Thanks in advance for your feedback.

Sophie Beck

Flatiron Research Fellow
Center for Computational Quantum Physics
Flatiron Institute, Simons Foundation
162 5th Ave, New York, NY 10010

Wannier mailing list

Wannier mailing list

Reply via email to