BSE calculations

[samaneh]

Dear Developers,
I have an error in my bse report file
"*ERR* Metallic bands : 97 "
" [ERROR] STOP signal received while in :[07.01] Main loop
[ERROR] Error in parallel wf distribution"
I've asked you before in another post which you told me to decrease the
electronic temperature, I decreased it but I have the error again!!!
and as you see in the input file (in the attached file) I mentioned the Drudeterm but
in the report file it says that this term is not included!!!
Also I have run RPA calculation with the G0W0 database
and as you can see in the report file I have
"*ERR* Metallic bands : 97 "
but the job was done!!!
I would be thankful if you help me to overcome this problem.
Best regards

[Daniele Varsano]

Dear Samaneh,

Linear Response:

Code: Select all

 *ERR* Metallic bands         :  96

This means that your are reading a database:

Code: Select all

 [RD./SAVE//db.dipoles]-

calculated in a previous run, and now thinks are changed: when applying QP corrections, the Fermi Energy change and you have 97 metallic bands:

Code: Select all

 [QP] Fermi level variation [ev]: 1.806227
 [QP] Last Filled/Metallic band :  93   97

Yambo does not stop, just recalculate them and substitute the databsae:

Code: Select all

[WR./SAVE//db.dipoles]-

BSE:
Same as before: you read the QP database, you have a fermi energy variations and the information in the dipoles database is not valid anymore.
The code stopped for another reason which now it is not totally clear to me, can you also post the standard output l_*?
Also it is not clear to me why DrudeWBS is not recognized: while of course DrudeWXs cannot be recognized as you are asking to read the screening from the Plasmon pole database (ndb.pp). As the em1s keyword was missing in your input, all the Xs keyword are ignored.
I suggest you to work with clean input, using the variable that you really need.

Finally please note that the orthonormalization is very loose, thinks about to rise the FFT Gvecs.:

Code: Select all

 [WF loader] Normalization (few states)  min/max  :0.591E-10 0.914

Best,
Daniele

[samaneh]

Dear Daniele,
Thanks for your reply
I've attached the output l_*
But for the FFTGvect I should note that it is a converged value!!
I mean I have done rpa spectrum without any database with different maxGvec (from 700rl to 2000rl)
to see if the shape of the spectrum does not change!!!
Best regards

[Daniele Varsano]

Dear Samaneh,
I had a look to your output and unfortunately it is not easy to understand in detail the failure.
The problem is related with the parallele structure. I suggest you to try to run your calculation with less than 100 cores, and see if it works.
If not, you can update to the new release, which has a different and more powerful parallel structure, even if I suggest you to wait few times as a new versione with many bug fixes will be released soon.

Best,

Daniele

[samaneh]

Dear Daniele,

Thank you for your reply.
I have another question
that why I have negative value of the imaginary part of dielectric function
in zero energy (E=0) in RPA calculations for some values of drude frequency (e.g. -9727.)
and the other data for other energy values are positive ??
Although for some other values of drude terms all the data is positive.
I think according to the theory it should be positive values!!
Best regards

[Daniele Varsano]

Dear Samaneh,
yes it should be positive, even if divergent. I think that some numerical oddity is happening there.
As it is not well defined, I would not trust it.

Best,
Daniele

[samaneh]

dear Daniele,
Thanks
But it should be noted that for the frequencies
which have negative values for imaginary part of dielectric function at E=0
the value for that but without LFE is positive!!!
"
# E/ev[1] eps /Im[2] eps /Re[3] eps0/Im[4] eps0/Re[5]
#
0.000 -345.8 0.3570E+06 0.1868E+08 0.1609E+10
0.101010 24.577442 -4.222519 24.573441 -3.859070
0.202020 51.830711 21.202522 51.900509 21.571245"

Best regards