Strange results of 3.2.3-r696

[hplan]

Dear Developers:

I performed self-energy calculations on graphene fluoride, but got strange results in different executing sequences.
I tried to test the convergent trend of the energy gap by increasing NGsBlkXp in sequence. However, if i regenerated a new database and ran a calculation with the same settings, i found the energy gap is different: the one using the old database gave 5.90 eV, the other from the new-generated database then gave 5.59 eV. Would you please give me some suggestions and comments ? The related outputs are attached. The first one is obtained using the database of the last calculation, and the 2nd used a new generated database.


Best,

Haiping

[Daniele Varsano]

Dear Hai-Ping,

the two calculations have different FFTGvecs setting.
This can be the source of the difference in the results.

Cheers,

Daniele

[hplan]

Dear Daniele,

In fact, i used the same input file with '-F' , and specified FFTGvecs as 2599 RL. I also confirmed it in their output files: all gave the number of RL as 2599 .

Thank you.

Best,

Haiping

[Daniele Varsano]

Dear Hai-Ping,

in the ouput you sent us, one of them (o.qp_13) has the FFTGvecs= 2599,
the other one has not, so I presume that all G vectors are used in the FFT
transform. Please check if this is the source of the differences. Look in the
two reports the HF value of the Sigma_X (<1|HF|1>) etc.
They do differs. In the r_HF_and_locXC_em1d_ppa_gw0_rim_cut_13 they are
calculated with FFTGvecs = 2599, while in the r_HF_and_locXC_em1d_ppa_gw0_rim_cut_12
they are read from a previous calculation.

Cheers,

Daniele

[hplan]

Thank you.

Does it mean that we need use a new database when testing several parameters like FFTGvecs and NGsBlkxlp et al ?
Because these two calculations were performed by reading the same input file, and FFTGvecs was also specified in the input settings .


Cheers,

Haiping

[Daniele Varsano]

Sorry for insisting,
yambo repeat the main input parameter at the end of the report and output files.
In one file you have the FFTGvecs fixed to a certain number in the other one does not
appear. Please be sure that you have not done some confusion with files. It can easily happen!!!
Anyway it is always good practice to look in the report file which database has been read and
which one was written, and if there are ERR and Warnings (WRN).

Best,

Daniele

[hplan]

Thank you.

yes, i noticed that the end of the outputs indicate FFGvecs was not specified in one of my calculations.
I try to figure out what happened.

Probably, i should do relevant test calculations using a new database for further verifications on different parameters.

Best,

Haiping