KfnQP_DbGd_INTERP_mode vs XfnQP_DbGd_INTERP_mode

[milesj]

Hello,

I'm performing a BSE calculation on with a G0W0 corrected quasiparticle energies, but my G0W0 grid is smaller than my BSE grid (6x6x6 vs 12x12x12). I assumed that, since the variable to reference the relevant G0W0 quasiparticle energies is KfnQPdb, the program must interpolate the G0W0 bands as determined by KfnQP_DbGd_INTERP_mode (i.e. either NN or BOLTZ). However, I tried adding the lines

Code: Select all

KfnQPdb= "E < ./test_666/ndb.QP"                  # [EXTQP BSK BSS] Database action
KfnQP_DbGd_INTERP_mode= "BOLTZ"     # [EXTQP BSK BSS] Interpolation DbGd mode

as well as the lines

Code: Select all

KfnQPdb= "E < ./test_666/ndb.QP"                  # [EXTQP BSK BSS] Database action
KfnQP_DbGd_INTERP_mode= "NN"     # [EXTQP BSK BSS] Interpolation DbGd mode

to my input to test the difference between the two interpolation methods, but it seems to have interpolated in exactly the same way in both calculations. I read in another post viewtopic.php?t=2493 that the variable to change is instead XfnQP_DbGd_INTERP_mode. Is this true? I thought all the XfnQP variables were only for if you're doing a self-consistent GW calculation? Or does YAMBO just interpolate all coarse GW grids in the same way now? To be clear, in both calculations it is interpolating the G0W0 values, but both calculations turn out exactly the same (when I plot the GW-corrected bands using ypp, using NN gives a very jagged graph whereas BOLTZ gives smooth bands, so I don't an interpolation for BSE should be giving the exact same results)

The version of yambo is yambo-5.1.1

Best,
Miles

[Daniele Varsano]

Dear Miles,

the right variable is KfnQP as you are using. XfnQP is the interpolation used for the calculation of RPA screening, so as you argued it is useful for sc-GW.
Can you check in your report file if the interpolation is done accordingly to the input variable? You should have some info in the report file when the electronic structure (gap, Fermi energy) is recomputed, indicated with E+QP.

Best,
Daniele

[Daniele Varsano]

Dear Miles,
please forget my previous answer.
The fact you obtain the same result is correct. When you interpolate qp energies for BSE what it is done is a linear interpolation of E_qp vs E_ks. This is different from the interpolation to obtain a band structure, as in that case you need to interpolate along a given path, and you can choose between two different algorithms.ù

Best,
Daniele

[milesj]

Hi Daniele,

In that case do these variables do anything? If by default it always linearly interpolates it seems like these variables both wouldn't change anything, since they only pop up for BSE calculations.

Best,
Miles

[Davide Sangalli]

Dear Miles,
I've recently created a wiki page with few info about the way in which QP corrections are interpolated.
https://wiki.yambo-code.eu/wiki/index.p ... orrections

The variables "DbGd_INTERP_mode" instead are not working at present. Please ignore them.

Best,
D.