BSE calculation error (BSE is skipped)

[pandachang97]

Hello Yambo experts,

Good morning/afternoon. I have a new error with my BSE calculation. The DFT is done by QE-7.0 with SOC on. The system is a 2D-MoS2 with 26 electrons. The static screening calculation is done without any problems. Then, the kernel calculation of BSE is skipped at all Q points. The error message is listed below:

Code: Select all

 
    [WARNING] Reading of io_X for iq=38 failed despite previous check.

    [WARNING] Missing all or part of PP/Em1s DB

    [WARNING] Bethe Salpter section skipped. Impossible to build the kernel.

I also posted up my r_setup, input and output files in the attachment. Please let me know if my parameters used have any problems

[pandachang97]

Hello Yambo experts,
It seems the problem is solved by increasing number of Coulomb cutooff to 10 Ry. But, please let me know how to estimate the number of this parameter?
Also, I have another question. I checked several posts in Yambo forum and I found that the number of G vectors can be estimated by the formula, G^2/2 < E_cutoff. I am a little confused. Last week, I have several trials with more G vectors ( G^2/2 > E_cutoff ) and the peaks from the vertical excitation (q=0) in the absorption spectrum disappeared. I think, even if we include more G-vectors in the calculation, they could be ignored with the coefficient equals to 0. Maybe I am wrong. Please correct me. Thank you and best regards,
Xin

[Daniele Varsano]

Dear Xin Chang,

it seems there is some problem in reading the screening. I noticed from the log file of the screening that it is a restart, starting for iq=38.
The yambo complains when reading this iq.

I suspect that something went wrong in the restart procedure (missing/corrupted or inconsistent file). My advise is to recalculate the screening from scratch.

* Next, it is not clear what you mean by number of Coulomb cutoff. The Coulomb cutoff you are using is analytical and you have all the components you requires.

*G^2/2 < E_cut is just a different way to count the G vector components in the various quantity (response function, bse kernel etc.), you can assign as a reciprocal lattice number (RL), or by defining an energy cutoff in energy units (Ha, Ry).

* It is not clear to me what do you exactly mean with trials with G^2/2 > E_cut . If you refer to the E_cut you used in your ground state calculation for the wfs and density, you cannot assign more than that as G components of the wfs/density are not present.

Best,
Daniele

[pandachang97]

Hello Daniel,
Good morning/afternoon. Thank you for your response. I changed the number of RandGvec to 5 Ry in my dynamical screening with SOC. Then, it worked. But, it is really weird that my BSE kernel calculation failed without SOC. That one used to be successful. When I use RandGvec with 5 or more Ry, I got the same error. Could you please let me know the essence of the RandGvec? Thank you again and best regards,
Xin