Page 1 of 1

qe-occupation

Posted: Thu Apr 02, 2026 9:32 am
by sunxl
Dear all,

1、I calculated a system using VASP, and the band gap was approximately 0.5 eV. When I used QE with
occupations = 'smearing'
smearing = 'gauss'
degauss = 0.01
after the conversion from the input form to that compatible with Yambo is completed, and after executing yambo, a prompt indicating a metallic system is obtained.
[02.05] Energies & Occupations
==============================

[X] === General ===
[X] Electronic Temperature : 0.258606E-1 300.100 [eV K]
[X] Bosonic Temperature : 0.258606E-1 300.100 [eV K]
[X] Finite Temperature mode : yes
[X] El. density : 0.35884E+23 [cm-3]
[X] Fermi Level : -3.570586 [eV]

[X] === Gaps and Widths ===
[X] Conduction Band Min : -3.570586 [eV]
[X] Valence Band Max : -3.570586 [eV]
[X] Filled Bands : 123
[X] Metallic Bands : 124 126
[X] Empty Bands : 127 600

[X] === Metallic Characters ===
[X] N of el / N of met el : 248.0000 2.0000
[X] Average metallic occ. : 0.166667

[WARNING] [X] Metallic system
However, when I used ypp to obtain the interpolated band structure, there was a gap. What is the reason for this? Will this warning have any impact on the subsequent calculation of the gw band?

2、When I set occupation = 'fixed', it seems that this problem won't occur. We know that for semiconductors and insulators, "fixed" should be used. But what about when the PBE band gap of a system is approximately 0.03 eV, how should the occupation be chosen? When the PBE band gap is zero and the energy bands have a slight overlap, that is, when it is a pseudo-metal, using the hse and gw methods will increase the band gap. In this case, how should the occupation be selected?


Best,
sunxl

Re: qe-occupation

Posted: Tue Apr 07, 2026 10:44 am
by Daniele Varsano
Dear Sunxl,

if yambo reads the electronic structure as metals it means there are fractional occupation. You can check if in QE it happens the same.
The reason is probably due to the large smearing value you are adopting. Next note that Yambo uses Fermi Dirac distribution instead of Gaussian and this
can provide some differences that reduce for smaller smearing values. In general, you want to use a small value of the smearing provided you reach convergence wrt the k points.

Best,
Daniele