negative self-energy

[Weiqing Zhou]

Hi all,

In full-frequencies GW calculation, I find imaginary part of self-energy is negative at some states. I think it is a numerical problem instead of a physical one.
As try, I play with GDamping and dScStep parameters, and find it is hard to eliminate the negative self energy exactly.
So what is problem here? How to ensure that all self-energy is positive?

Best,
Weiqing

[Daniele Varsano]

Dear Weiqing,
I do not know if this reflects your case, but for conventions lifetime of electrons(hole) is set positive(negative).
See here for an example:
http://www.yambo-code.org/tutorials/Rea ... /index.php

Best,
Daniele

[Weiqing Zhou]

Dear Daniele,

Many thanks!

Acutally, I have noticed the Lifetime Tutorials in Yambo main page before. What I use is Version 4.3.0 Revision 129, and I think there are some updates about self-energy compared to the version used in Tutorials. Different from the result shown in Tutorials, the self-energy from Yambo 4.3.0 is all positive for electron and hole. I post my Al result in attachment.
And From Yambo 4.3.0, I still got some negative self-energy, which is my problem.

Best,
Weiqing

[Daniele Varsano]

Dear Weiqing,
I see, can you post the report file of your calculation where you find the problem, I can't see it from the file you post or am I missing something?

Best,
Daniele

[Weiqing Zhou]

Dear Daniele,

Sorry for my negligence. Last post, I just upload Al result to compare with Tutorials.

The negative self-energy appears in a semi-metal I study recently. Actually, self-energy is all positive and look nice when GDamping is default value, namely 0.1 eV. I think GDamping, as a small value, should be as small as possible theoretically and self energy should get converged as GDamping decreases. Then I do GW calculation with different GDamping (0.1 eV, 0.01 eV and 0.001 eV ) and the self-energys with different GDamping as a function of KS energy are plotted in attachment. It is clear that self-energy approach convergence as GDamping decreases, while negative self-energy appear when GDamping is smaller than 0.01 eV. How to eliminate the negative self-energy with a small GDamping?

Best,
Weiqing

[andrea.ferretti]

Dear Weiqing,

not sure this is at the origin of the issue, but typically lowering the broadening parameters should go together with an increase in the kpt grids.
I think that if you just lower the dumping, you reach a point when your calculation is not converged wrt kpts.
Personally, I would follow an algorithm where, given a broadening value one makes sure to converge kpts, then one can lower the broadening and re-check for kpts and so on

Andrea