Yambo Community Forum

Dear all,

I am dealing with a metallic magnetic 2d system for which I would like to calculate optical properties. Can I calculate optical properties using LRC and BSE?

Optics within RPA works fine, but I can not set up calculations for LRC [yambo -o c -k lrc] following tutorial 1d wire. For the 1d wire the input contains tddft environment as well as LRC_alpha parameter line. In my case the input contains just:

optics # [R OPT] Optics
chi # [R CHI] Dyson equation for Chi.
Chimod= "LRC" # [X] IP/Hartree/ALDA/LRC/BSfxc
NGsBlkXd= 1 RL # [Xd] Response block size
...

The same happens when I would like to set up for BSE kernel. There is in addition yambo crash when invoking [yambo -o b -k sex -V all -y h] with the following error:
At line 895 of file mod_itm.f90
Fortran runtime error: End of record
I'm using version 3.3.1 rev 2132 compiled with gfortran. Once the -V flag is omited [yambo -o b -k sex -y h] yambo provides input but without any further verbosity.

With Haydock I entered to the problem of "Fractional e/h occupations. Use coupling BSEmode or diagonalization solver". Using BSEmode="coupling" calculations end with '[WARNING]Fractional occupations still unsupported (Haydock)' and no optics files appear. Going with BSSmod= "t" [yambo -o b -k sex -y t] yields successful yambo end-up but there are also no optics files. The same happened for BSSmod= "d".

Best regards,
Martin

Dear Martin,
the tutoriasl relates to the new version of Yambo which is available in the SVN repository
http://www.yambo-code.org/download.php
since few days. Anyway it is a developing version, we are now in a testing phase, so use it at your own risk. Of course any comments and bug report is more than welcome.

Best,

Daniele

Dear Martin,
thank you for the feed back.

As Daniele pointed out you could check the last glp/devel version (3.4.0) which is available trough the svn repository. This is however a beta version. The last stable is 3.3.0 which is with the old input file synopsis.

I will check the points you addressed and I'll try to fix them there as soon as I can.

For the moment:
1- The lrc kernel and in general tddft does not work on magnetic systems, this is why you canno set-up calculations in this case
2 - I'll check the yambo crash with the -V all
3 - Fractional occupations presently works only with the diagonalization solver (we will move these in the haydock solver soon). The double message is not very smart sorry for that. We will improve it.
4 - yambo -y t is not expected to work in the GPL.
5 - Instead the run with yambo -o b -k sex -y d should work fine. Could you attach the report of this run to understand why yambo is not printing the output?

Dear Davide,

Thank you for your response. I have moved to recent svn 3.4.0 version and according to your suggestion [yambo -o b -k sex -y d] I successfully generated eps/alpha spectra. Concerning the not printed output, I have realized that this happened when I did not set BndsRnXs and let them default to BndsRnXs: 1|1| which might be actually physically a nonsense.

I would like to know what WehDiag and WehCpl variables control. As far I remember Claudio mentioned some approximations to e-h matrix (Tamm-Dancoff, etc.) during his talk in Yambo 2013 workshop in Lausanne, but I am missing connection to yambo input variables. Could somebody comment on it?

I found that yambo looks in SAVE folder for databases. Does it mean that it can in some sense reuse existing databases when, for instance, one does convergence study on range of BSEBands? Imagine, I have done calculations with BSEBands: 4 | 12| and I would like to go with BSEBands: 3 | 13 |. Should I remove some/all databases or is it legitime to keep some of them? Sorry for that silly question -- it has an extension -- what if one checks range of BndsRnXs. If yambo can reuse some of the databases could not someone provide a small guide which database depends on which variables?

Best wishes,
Martin

martin.gmitra wrote: I found that yambo looks in SAVE folder for databases. Does it mean that it can in some sense reuse existing databases when, for instance, one does convergence study on range of BSEBands? Imagine, I have done calculations with BSEBands: 4 | 12| and I would like to go with BSEBands: 3 | 13 |. Should I remove some/all databases or is it legitime to keep some of them? Sorry for that silly question -- it has an extension -- what if one checks range of BndsRnXs. If yambo can reuse some of the databases could not someone provide a small guide which database depends on which variables?

Hallo Martin

In general if yambo finds that the new parameters in the input are not compatible with one in the existing database, it will redo the calculation and write on top of them.
If you want to keep some database for later use, I recommend you to use the -J option when you call yambo to run your job.
e.g. 'yambo -J 2-12' wil write the db related with the run (in your case bse) in a 2-12 directory instead that on the SAVE

I would like to know what WehDiag and WehCpl variables control. As far I remember Claudio mentioned some approximations to e-h matrix (Tamm-Dancoff, etc.) during his talk in Yambo 2013 workshop in Lausanne, but I am missing connection to yambo input variables. Could somebody comment on it?

We are rewriting the documentation for the new variables. For the moment:
BSEmod controls the BSE Hamiltonian.
The default "causal" will give you ((R,0),(0,-R*)), the resonant option is the Tamm-Dancoff approx Claudio spoke about H=R, the coupling option is the full BSE H, i.e. ((R,C),(C,-R*)). The latter, as you may understand is the most expensive of those options, not only in terms of diagonalization, but also because you need to calculate matrix elements. The most expensive part are the eh interaction integrals, and WehDiag and WehCpl are logical flags that control the calculation of those integrals. WehDiag includes them just in the diagonal of H (quicker approximation); in the case of BSEmod=coupling if you want eh interaction part also in the C matrices you should uncomment WehCpl (full BSE).

I hope I managed to be clear enough.
Davide, please correct me if I did not get all details right.

Best
m

Hi all,

Thanks Mytra for your explanation, it is more clear now.
I would like to ask you:
(1) how is defined the dielectric tensor alpha when considering non periodic Z cartesian direction, setting NonPDirs= "Z"?
(2) Is there physical explanation why tensors' z-component [LongDrXd: 0 | 0 | 1] within RPA LF [yambo -o c -k hartree] drops significantly down in comparison to RPR no LF [yambo -o c -k IP] as well as to BSE calculations? Please see attached figure (2d_specs.png: green line in right top plot) to the post.

Best,
Martin

martin.gmitra wrote:Hi all,
(2) Is there physical explanation why tensors' z-component [LongDrXd: 0 | 0 | 1] within RPA LF [yambo -o c -k hartree] drops significantly down in comparison to RPR no LF [yambo -o c -k IP] as well as to BSE calculations? Please see attached figure (2d_specs.png: green line in right top plot) to the post.

Those are local field effects due to the strong confinement you have along z. Just look at it classically and put a plate in a perpendicular field E. The electrons will be rearranged because of the field and create a local field counteracting the external field. For a more quantum view of local field effects you can review Claudio's lecture or look at e.g. the Rev Mod Phys on TDDFT and Green's function methods.

cheers,
m