BSE spectra considering momentum

[Daniele Varsano]

Unfortunately not, as usually they are not needed, it is not a calculation done by yambo, as they are exported from QE.
Here in attachment a script to extract the dipoles, you need to adapt for the reading of the wfs.

Best,
Daniele

[Fadil]

I see, thank you!

[Fadil]

Dear Daniele,

I think the ordering in 'ns.wf_fragments_*' files should be
from left nband, spin, gvector, Re/Im(wf_coefficients),
right?

The number of g_vec obtained from './setup/G_shells_finder.F' are much larger than the above wave function coefficients.
Actually, the number of wave function coefficients are the same with the 'Max WF components' and the number of 'g_vec' is the same with 'RL vectors (CHARGE)'.

<---> Max WF components : 9843
<---> RL vectors (WF): 9843
<---> RL vectors (CHARGE): 79117

So how do we match 'g_vec' and wave function coefficients?

Best

[Daniele Varsano]

Dear Fadil,

I think the ordering in 'ns.wf_fragments_*' files should be
from left nband, spin, gvector, Re/Im(wf_coefficients),
right?

Yes sorry, I wrote from left instead then from right.

G_vec contains all the gvectors which are the ones of the charge, i.e. (G-G') where G are the ones of the wfs (wf_ng along the code).
About the matching, I would not want to send you on the wrong path. In the code the procedure to go from G to R space is alway done via FFT, someone else more expert in the ordering can give you more precise details.

Best,
Daniele

[andrea.ferretti]

Dear Fadil,

concerning your question about the matching of vectors in g_vec (density) with FFT grids and wave functions, you may want to have a look at the yambo-internal map:

ir=fft_g_table(ig,iGo)

which, given a G-vector index ig, maps it to the FFT grid. (Actually, it is G-Go which is mapped).

Coming to the connection of wfcs and rho G-vectors:
assuming wfc contains components up to G*, the charge density contains component up to 2G*, making the ecutrho value 4 times that of the wave functions.
As a rule of thumb, the G-vecotrs for wfcs are about 1/8 of those of the density (both are distributed in spheres, one with radius 1/2 of the other one).

HTH

Andrea

[Fadil]

Dear Daniele and Andrea,

Thank you for your help!

G_vec contains all the gvectors which are the ones of the charge, i.e. (G-G') where G are the ones of the wfs (wf_ng along the code).

Where can I find the 'wf_ng' ?
It is not in the './setup/G_shells_finder.F'.

Best

[Daniele Varsano]

Dear Fadil,

wf_ng is assigned by the p2y interface and stored in the ns.db1 database.

you can find it here:
/interfaces/p2y/p2y_db1.F

it can be changed by setting in input the FFTGvecs variable (see ./src/interface/INIT_load.F)

Best,

Daniele

[Fadil]

Dear Daniele,

The q=0 max strength is correct, while for finite q the pre-factor is wrong.
Note the values reported are normalised to the maximum value so the relative intensity for each q is correct.
The absolute value instead should be modified as S=S*10^-10/|q|^2

Is the above prefactor correct for q=/0 polarizabilities (\alpha)?
-q=0 seems correct but the polarizabilities for q=/0 are very high with versions 5.0.3 and 5.0.4.

[Daniele Varsano]

Dear Fadil,

yes, the prefactor now is correct.


Best,
Daniele

[Fadil]

It seems no.
Can you please check the attached outputs? It is the same with the 5.0.4.
Since I apply the Coulomb Cutoff, I need to rescale the results.
Is it \alpha=\alpha*10^-10/|q|^2 ?