rim_cut and rimw

[xjxiao]

Dear all,
When performing BSE (Bethe-Salpeter Equation) calculations in Yambo, in which of the three steps—calculating the static dielectric function, calculating the BSE kernel, and diagonalization—should rim_cut and rimw be applied?
Thanks!
Xiao

[Daniele Varsano]

Dear Xiao,

it applies when calculating the BSE kernel, anyway please note that rimw for BSE is experimental, so I suggest you to make appropriate checks.

Best,
Daniele

[xjxiao]

Dear Xiao,

it applies when calculating the BSE kernel, anyway please note that rimw for BSE is experimental, so I suggest you to make appropriate checks.

Best,
Daniele

Dear Daniele,
Thanks a lot for your help!
After performing GW corrections using rim_cut and rim_w, I used XfnQPdb to read the GW-corrected bands and recalculated ndb.em1s. If the number of bands obtained from the GW calculation is relatively small (only sufficient for subsequent BSE calculations), is the resulting ndb.em1s still trustworthy?
Next, I read the ndb.em1s generated above and attempted two approaches to compute the kernel and solve the BSE:
①With both rim_cut and rim_w.

r-bse_44_qp_rim_cut_rim_w_optics_dipoles_bss_bse.txt

②With only rim_cut.

r-bse_44_qp_sj_rim_cut_optics_dipoles_bss_bse_02.txt

The calculated exciton excitation energies differed significantly—the second approach yielded results ~0.4 eV lower than the first.

exciton.rar

I'm shocked by such a large disparity. Is there any good way to determine which result is more accurate?

[Daniele Varsano]

Dear Xiangjun,

If the number of bands obtained from the GW calculation is relatively small (only sufficient for subsequent BSE calculations), is the resulting ndb.em1s still trustworthy?

Yes, it should be as Yambo perform an interpolation for the missing values. Anyway, please note that using GW corrected energy for the calculation of the screening it is not the standard procedure you find in most of the BSE calculations. KS energies are commonly used as usually it provides more accurate results as somehow these mimics the vertex corrections absent in the GW band structure.

Is there any good way to determine which result is more accurate?

Perform a convergence test with respect to the k point grids and verify the convergences. It is possible that your k point grid is too coarse for the BSE.

Best,
Daniele