Influence of Coulomb Truncation on BSE results

[xueshanxihua]

Dear developers,

I performed with version 4.4.0 a HSE-BSE calculation (following viewtopic.php?f=16&t=1641&p=7946&hilit=shan+dong#p7946). The system calculated here is a 10-angstrom-thick 2D semiconductor with a HSE gap of 0.5 eV and a in-plane lattice constant of 4.2 angstrom. In the calculation, I used 10-angstrom vacuum and 6x6x1 kpoint grid. However, I find that if I don't use the Coulmob truncation, the calculated binding energy for lowest exciton state is about 0.3 eV. While with a Coulomb truncation, the binding energy turns to 1.3 eV. I understand a 10-angstrom vacuum is unambiguously too small for convergence in non-truncated case, but a difference of 1 eV is still a bit surprising.

Hence, I want to know if this difference is still reasonable for yambo, and how can I determine which result to use?

Thank you,


BEST,

ZEYU JIANG

[Daniele Varsano]

Dear ZEYU JIANG,
it seems a very large variation indeed.
What kind of truncation are you using? Note that k point sampling size affects the binding energy in a different way when using the truncated Coulomb potential. What you can do is to perform more calculations with a bare potential for a different amount of vacuum and see if you observe large variations towards the value found with the cutoff.
Best,
Daniele