Yambo Community Forum

Hi all:
I am calculate the fulerene Sc4C2@C80. I found the band gap is much higher than I want. For example experiment is 1.56 eV. GW calculation gives 3.3 eV. I think GW may not work for my system with Sc atoms. Can I not use plasmon pole approximation? so I still can calculate GW approximation fot the system Sc4C2@C80.

Thanks

Zhifan Chen
Clark Atlanta University
USA

You may try the static COHSEX in the latest svn.

Dear Zhifan,
as stated in the mail, I hope you received, these are the
new features you can find downloading the last svn:

* COHSEX self-energy
* real-axis GW (beyond the Plasmon-pole approximation)
* lifetime calculations

So, you can try a GW calculation performing a real-axis integration.
Of course, this calculations is much more cumbersome than using
the plasmon-pole approximation.

Cheers,

Daniele

Well, I have still not added to the GPL repo the real-axis GW and the lifetimes I plan to do it ASAP, anyway

zchen wrote: I am calculate the fulerene Sc4C2@C80. I found the band gap is much higher than I want. For example experiment is 1.56 eV. GW calculation gives 3.3 eV. I think GW may not work for my system with Sc atoms. Can I not use plasmon pole approximation? so I still can calculate GW approximation fot the system Sc4C2@C80.

Anyway, my point is: why do you think PPA may be no working in your specific case ? The gap is not small, as far as I can see. What is the measured plasma frequency of your system. I mean, try first to understand why the PPA may not work, trace back the problem and try to solve it using other approaches. Note the COHSEX is an even more drastic approx, as it is not a dynamical self-energy. So use it carefully.

For example experiment is 1.56 eV.

This is the measured photomession or optical gap?
Note that GW quasiparticle gap have to be compared with photoemission experiments.

Daniele

Thank all of you
The experimental band gap is electrochemical band gap( same as optical band gap?). I do not have measured data of photoemmision band gap. Roughly say how much difference are two band gaps? I am sorry I do not have measured plasma frequency. In the calculation I used PPAPntXPp=27.21138 eV. I read the yambo paper it said Plasmon-pole approximation can be questionable when some of valence orbitals are spatially localized (like in d or f metals). It is also said the metallic screening (noble and transition metals) does not make the RPA meaningful. I think Sc may be belong to it. However, Sc in Sc4C2@C80 donate 3 electron/per Sc atom. Therefore Sc become ion with no d electron outside. So I am not sure whether Sc is the reason caused the trouble.

Another posibility is to do coulomb cutoff. The molecule is about 8A. Sc4C2@C80 has a core Sc4C2 with +6 electron charge inside C80 which has -6 electron charge. I used supercell 12A x 12A x12A. I do not have the experience whether I need consider to do Coulomb cutoff for GW calculation.

I will give a try to use COHSEX. However, I am not a developer. I cannot use : svn checkout svn+ssh://developername@scm.qe-forge.org/svn/yambo.


Thanks

Zhifan Chen
Dept. of physics
Clark Atlanta University
Atlanta, GA 30314
USA