Spin polarized calculation in EPW

[chenhanghui]

Dear EPW developers,
I know from the EPW documentation that spin polarized calculations are not implemented in EPW. However, both phonon calculations and Wannier90 can handle spin polarized calculations, therefore I am wondering whether each mode-resolved electron_phonon coupling can be considered the sum of two spin contributions: i.e. lambda_{qv} = lambda_{qv}_up + lambda_{qv}_dn. Both lambda_{qv}_up and lambda_{qv}_dn are calculated using "g2" matrices for the given spin. Is this the right way to implement it or there is something fundamentally difficult that I naively missed, which prevents the implementation of spin-polarized calculations in EPW?
Thank you very much for your help.

Dr. Hanghui Chen
Department of Physics
NYU Shanghai and New York University

[hlee]

Dear Dr. Hanghui Chen:

As you know, in spin-polarized calculations (nspin=2 in the QE convention; see the subroutine of set_spin_vars: https://gitlab.com/QEF/q-e/-/blob/devel ... n_vars.f90) the spin-up and the spin-down components are decoulpled. So I think that within this scheme your definition of lambda is valid.

However, both phonon calculations and Wannier90 can handle spin polarized calculations, therefore I am wondering whether each mode-resolved electron_phonon coupling can be considered the sum of two spin contributions: i.e. lambda_{qv} = lambda_{qv}_up + lambda_{qv}_dn.

I am not 100% sure, but I think that such definition is adopted in the e-ph related calculations in QE.

I know from the EPW documentation that spin polarized calculations are not implemented in EPW.

I think that there is no fundamental difficulty with implementing it as long as we are ok with the above definition of e-ph related quantities.

Sincerely,

H. Lee

[chenhanghui]

Thank you very much, Dr. Lee.