convergence tests

[samaneh]

Hi there,
The convergence tests to do "yambo -g n -p p"
are :
1- k-point
2- EXXRLvcs
3-BndsRnxp
4-NGsBlkxp
5-GbndsRange
am I right?
2-How about convergence tests to do "yamo -o b -k sex -y h"
3-Is there any difference between the parameters "FFTGvcs" and "EXXRLvcs"
Thanks
S. Ataei, PhD student of Condensed matter Physics, Tehran

[Daniele Varsano]

Dear S. Atei,
1) Right.
2) Yes there is difference: EXXRLvcs are the Gvectors entering in the summation for the Hartree part in the BSE kernel, FFTGvecs, control the number of G in the Fourier Transform for the calculation of the dipoles. You can decrease it for memory reason, but not too much, you can check the orthonormalization in the report file.
Moreover you need to converge the direct term: BSENGblk, and the dielectrix matrix as in the GW calculation.
Best,
Daniele

[samaneh]

Dear Daniele,
Thanks, I have some more questions:
1- First I relax my structure (vc-relax with QE) and then perform yambo convergence tests, is that right?
2-The value of EXXRLvcs parameter for my structure is around 105000 RL therefore the calculations for k-point test
is long, can I reduce it to do these calculations, if yes how much is enough?
Best
S. Ataei

[Daniele Varsano]

Dear Smaneh,

1) Yes it is, you can do the excited states calculations on what you want, usually it is meaningful to do it on a relaxed ground state structure.
2) As usual, convergcence tests are system dependent, so it is impossible to answer to your question, just test it. Next note that, the exchange part of the Self Energy is the less expensive in a GW calculation.
Best,
Daniele