RPA_LF

[kxzhou]

Dear all,
1) The dielectric function has been computed from DFT energies in GW calculations. Can this be used to calculate the absorption spectrum for RPA_LF? How should this be done?
2) Is it possible to use the ndb.qp obtained from GW calculations to compute the dielectric function for RPA_LF and thereby obtain the spectrum? Could this be challenging due to the need for an excessively dense k-point grid and a large number of quasi-particle bands?

Looking forward to your reply! Thanks!
Kexin

[Daniele Varsano]

Dear Kexin,

1) No, as for the GW most probably you computed the dielectric screening at RPA level either using Plasmon Pole approximation (which requires just 2 frequencies) or MPA (limited number of frequencies). For a complete spectrum, you need to compute it for all the frequencies of the spectrum you are interested in.

2) Yes, it is possible using the keyword

Code: Select all

XfnQPdb= "E < ./$DIR/ndb.QP"

You do not need to calculated QP for all bands and k points. Yambo will read the calculated values and interpolate for the missing ones.

Best,
Daniele

[kxzhou]

Dear Daniele,

2) Yes, it is possible using the keyword

Code: Select all

XfnQPdb= "E < ./$DIR/ndb.QP"

You do not need to calculated QP for all bands and k points. Yambo will read the calculated values and interpolate for the missing ones.

Thanks for your reply!
For RPA_LF calculations, the number of BndsRnXd required can reach hundreds, but in G0W0, only the quasi-particle energies of two bands were calculated. Can interpolation still be performed under these conditions?
Can the parameters for RPA_LF, such as BndsRnXd, NGsBlkXd, and LongDrXd, be set the same as those for G0W0_ppa calculations?

Best,
Kexin

[Daniele Varsano]

Dear Kexin,
1) if you calculated the GW corrections for just two values, the effect of the interpolation will be just a scissor operator, i.e. a rigid shift opening the gap. If a scissor approximation is valid, that's ok, otherwise you need to calculate the correction for more bands and k points.
I suggest you to calculate the corrections for few bands and different k points and verify if a scissor/stretching approximation can be applied. See e.g. step 3 of this tutorial: https://wiki.yambo-code.eu/wiki/index.p ... tandalone)

2) BndsRnXd: These now depends on the range of the spectrum you are interested in.
NGsBlkXd: You can use the value obtained to converge the GW corrections.
LongDrXd: This gives you the polarization of the incident light, so it depends on the direction you want to look at.


Please note that RPA_LF can also be obtained in the eh basis set, doing a BSE like calculation setting BSKmod="Hartree". In this way you do not need to set the number of frequencies as you have the spectrum. What way is more convenient depends on the system (bands, k points, dimensions of the screening).

Best,

Daniele

[Debe]

Is it possible to use the scissor operator (and the stretching coefficients) for a system having some partially occupied states in the bandgap ?
I.e. ideally one need a stretching coefficient for the valence band, a stretching coefficient for the impurity states in the gap, a stretching coefficients for the conduction band.
Furthermore, the scissor operator should be applied NOT to separate the un-occupied bands (unoccupied impurity states+conduction band) from the occupied bands (occupied impurity states+valence band), but rather the impurity band from the valence bands and the conduction band from the valence band (or from the impurity band).

[Daniele Varsano]

Dear Alberto,

You can have a try, but I'm afraid this would not work.

The solution I can see for your needs is to use yambopy to read KS states, act on them for your specific case and create an ndb.QP database with the modified energy levels.

Best,

Daniele