MPA error in yambo 5.3

[csk]

Dear all!

I can successfully use the MPA with yambo 5.2, however when using the same input file (generated with yambo 5.3, similar syntax as in the tutorial example), I get the error "Damping = 0.0 not compatible with MPA dielectric constant". Now this can be mitigated by setting a small number in the damping, e.g.
% DmRngeXm
0.0001 | 2.721139 | eV # [Xm] Damping range
%

However, then I still get an error in the error file of the slurm controller "Caught signal 8 (Floating point exception: integer divide by zero)" when yambo later tries to compute X from X0 at the first Q-Pt.

How can this different behaviour be understood? I generate the input file with yambo 5.3 and I also attach the input, log, and report files of the run with yambo 5.2 in the attachment. Should I rather report an issue on git?

Thanks for your help,
Christian

[Daniele Varsano]

Dear Christian,

yes, assigning a small value in the Damping for the X is the way to proceed.

Code: Select all

% DmRngeXm
0.0001 | 2.721139 | eV # [Xm] Damping range
%

It is not clear to me what0s the source of the error. Can you please post the input file of the 5.3 run?
In the manwhile can you try to run the 5.4 beta branch and see if the problem perists.

Best,

Daniele

[Giacomo Sesti]

Dear Christian,

I rerun your calculation on yambo 5.3 to reproduce your error, using your input file. For the DFT, I used the same lattice parameters and atomic positions, but different pseudos.

The MPA calculation works correctly for me, provided that you set a finite DmRngeXm as you did:
% DmRngeXm
0.0001 | 2.721139 | eV # [Xm] Damping range
%
I attach the report and log file of my calculation.

Btw I noticed a typo in your input file, it should be NGsBlkXm instead of NGsBlkXp for MPA. Otherwise, the calculation just runs just on 1 RL ( it did not signal any error also in that case)

Best,
Giacomo

[csk]

Dear Giacomo and Daniele!

Thanks for your help! I think this typo was indeed the problem: "it should be NGsBlkXm instead of NGsBlkXp for MPA", thanks for spotting it!

Now the calculation runs fine. That you don't get an error with 1 RL block size and I do could maybe depend on the machine and/or parallelization...?

Anyway, thanks a lot!
Christian