ABINIT Discussion Forums

Hello,

I've noticed that during a contour deformation screening calculation, prior to a GW calculation, the files containing the dielectric function *_EM1_LF and *_EM1_NLF don't appear to contain intraband contributions. It looks like calculating the intraband contribution is forcefully disabled in the source file screening.f90 . Is there a way to extract the full dielectric function with intraband contributions ? To what extent are intraband contributions to the dielectric function accounted for when calculating quasi particle energies?

Many Thanks,
Patrick

Intraband contributions are important in case of metals for which you need a very very dense k-grid (18x18X18) to obtain converged results.
For semi conductors and insulators intraband transitions are absent.

LUCA

Thanks for the reply Luca. I am looking at metals and metal-insulator transitions, so it would be good to know if the screening file (calculated using the contour deformation method: optdriver=3, gwcalctyp=2) already does contain intraband contributions. Are you able to confirm it does contain intraband contributions when using metallic occupations? If so, how is this calculated within the code and is the band structure dependent, effective mass of the electrons included?

Line 1122 of the Abinit 8.0.7 screening.f90 file: " add_chi0_intraband=.FALSE." would appear to override intraband contributions and I am reluctant to change it to ".TRUE." if the associated code isn't ready yet.

It is easy to verify whether the intraband contributions are contained in the screening or not by just comparing to the results obtained for a well known metal example.
Just try "add_chi0_intraband=.TRUE." and look what happens. I never check it.

Intraband contributions are always band structure dependent. If you are lucky, these contributions could be small in your case. But to know this you simply have to plot the band structure with the Fermi energy. Typically the intraband contributions are calculated in a semi-empirical way by use of a DRUDE complex plasma frequency from electron energy loss spectrum. Ab-initio calculation of intraband contributions requires a huge huge number of k-point sampling of the screening.

LUCA

Hi All,

I was just wondering isthere's an update regarding this topic? I'm planning to do a simple RPA with/without local field effects but with intraband term included.

So far, the forums don't really have any other answer except this one. I'm using a cluster so it's a bit tricky for me to modify the code manually.

Cenna