GW PPA termination with no error

[shobhit21287]

Dead Daniele,
We installed the new version of yambo, here are the config logs for the same.
We reran the same GW calculations after producing the save file again, however this time, we encountered a different error. We are attaching the report, input and the log files here again.

report.txt

l-gw_ppa_forum_30_1_HF_and_locXC_gw0_dyson_em1d_ppa_el_el_corr_CPU_1.txt

config-5.2.0.log

Thanks a lot for your time.
Best,
Shobhit

[shobhit21287]

Dear Daniele

I am attaching the rest of the files in this post

conifg-5.1.2.log

gw_ppa.txt

Best,
Shobhit

[Daniele Varsano]

Dear Shobit,

in our system you have 46 occupied bands, ess the report file:

Code: Select all

[X] === Gaps and Widths ===
  [X] Conduction Band Min                           :  1.280137 [eV]
  [X] Valence Band Max                              :  0.000000 [eV]
  [X] Filled Bands                                  :   46
  [X] Empty Bands                                   :   47  105

and you are trying to calculate the electronic screening using only 30 bands:

Code: Select all

% BndsRnXp
   1 | 30 |                         # [Xp] Polarization function bands
%

In the polarizability expression Xo (see https://www.yambo-code.eu/wiki/index.php/Cheatsheets) you have a sum-over-state so you need to include at least one empty state. Then the number of empty states is a parameter that has to be converged.

Best,
Daniele

[shobhit21287]

Dear Daniele,
While increasing the gw bands fixed our issue, we have a new one at hand, again in the dipole matrix calculations.

Upon rerunning the gw calculations, it stopped without giving an error at the 130th iteration. We tried to change the number of bands being considered at bottom as well the issue persisted. I am attaching the log file for your reference.

l-gw_ppa_160_8_allbands2_HF_and_locXC_gw0_dyson_em1d_ppa_el_el_corr_CPU_1_01.txt

Best
Shobhit

[Daniele Varsano]

Dear Shobhit,

these are not iteration, but it is the index of q points you are calculating the screening. I suggest you to have a look at what yambo is actually calculating, for instance the Yambo cheatsheet that you can find here:
https://www.yambo-code.eu/wiki/index.php/Cheatsheets

You can try to restart your calculation, ie just running again yambo, with a bettere distribution of CPU over bands:

Code: Select all

X_and_IO_CPU= "1 1 1 4 16"                 # [PARALLEL] CPUs for each role
X_and_IO_ROLEs= "q g k c v"    

balancing the CPU assigner to "c" and "v" according to how may empty bands you have included in the calculation.

Best,
Daniele

[shobhit21287]

Dear Daniele,
After rerunning DFT calculations to check for convergence of parameters, we are now receiving the following error in setup.

[ERROR] STOP signal received while in[03] Transferred momenta grid and indexing
[ERROR] [RL indx] error in mapping

We have referred to other forum posts as well and tried multiple k-point grids, removing time inversion symmetry (for an asymmetric material) and tried the setup with NoDiagSC but none of it works.
Please guide us for the same.

Best,
Shobhit

[Daniele Varsano]

Dear Shobhit,
can you post your nscf QE input file and r_setup file?

Best,
Daniele

[shobhit21287]

Dear Daniele,
Here are the files.

MoSe2_nscf.txt

r_setup.txt

Best,
Shobhit

[Daniele Varsano]

Dear Shobhit,

I do not what is going wrong here,
I can see you are now using a version of the code which is older than the one you were using before (5.1.2).
Next, besides the spin treatment (no spin-orbit coupling), and a different structure (alat), what has changed in your k point grid?
In previous calculation you had 154 k points and now you have 151.
Can you try to repeat the calculation you did before (without NoDiagSC) using the release of the code, maybe you can spot what does differ comparing the two r_setup file.

Best,

Daniele

[shobhit21287]

Dear Daniele,

We ran the files without and with the setup

setup
NoDiagSc

The report files and the nscf input file are attached below

nscf.txt

r_setup_nodiagsc.txt

r_setup_plain.txt

With the default yambo initialisation, we are getting an empty hf.in file where as with nodiagsc, the initialisation isn't completing as can be seen in the r_setup file

Best,
Shobhit