A large difference between with and without rim_cut in bse absorption

[lyzhao]

Dear all,
I found a large difference in peak positions between with and without rim_cut in the bse calculations.
As a test, I run bse calculations on the 2D monolayer MoS2 with and without rim_cut.
The line shape of spectra are very similar, but a large shift (~1.23 eV) of peak positions are found (see figure).

fig.png

Notes:
1. scissor correction of 0.60 eV, PWSCF 7.2, yambo 5.2.1 are used for both.
2. Based on the experimental results, the result without rim_cut should be more acceptable.
3. It seems to be a rigid shift, is anything wrong in my input file?

Input and output files of pw and bse are attached.

Any help will be appreciated.

[Daniele Varsano]

Dear Youzhao,

the difference is due most probably to the fact in one case you are truncating the coulomb potential and looking at the polarizability, in the other case
the dielectric function of a diluted system, which depends on the volume. Moreover, as you have a very elongated cell needed to isolate the 2D system, even if you do not use the cutoff coulomb potential, the RIM should be used to evaluate accurately the integrals near q=0.

Best,
Daniele

[lyzhao]

Dear Daniele,
Thanks for your quick reply.
I'm not sure if I understood your reply.
At present, I mean that the positions of the absorption peaks caused by the two methods are very different,
not the intensity of the peaks which is related to the cell volume.
Becasue the line shapes with and without rim_cut are very similar, a difference would be related to a rigid shift.

More helps are appreciated,
Best regards

[Daniele Varsano]

Dear Youzhao,
when not using the cutoff, if you do not apply the stochastic evaluation of Coulomb integrals, the head of the screened potential it is not accurately described. Moreover, in the whole calculation the two potential are different. A different head in the screened coulomb potential has a net effect of shifting your absorption peak.

For a comparison of the two calculations, you would need to use the RIM at least, and then converge the calculation without the cutoff potential with respect to the supercell volume.

Best,
Daniele

[lyzhao]

Dear Daniele,
Thanks for your help again.
As a feedback, I ran the calculation based on the rim without Coulomb cutoff.
Comparing with the results without both the rim and Coulomb cutoff, I get the following figure.
The Coulomb cutoff should lead to the shift of peaks.

rimnocut.png

Best regards.

[Daniele Varsano]

Dear Youzhao,

the only concern I have looking at your input/output is the size of the k grid that seems to me very coarse.
Larger grid are usually needed to get converged results in GW/BSE.

Best,
Daniele