How to choose converge criteria in GW calculations

[zsjan]

Dear all:
What is the conventional criteria for the converge of GW calculations? In the tutorials, the criteria is judged by the direct gap and indirect gap for Si. However, there is some metallic state in my system. After several test, I found that the correction in o.qp file still have big variation about 0.1 eV or 0.2 eV for some states at some K points. For example, see these values:
# K-point Band Eo E-Eo
1 65 -1.7761 0.91255 0.95572 0.84764
1 66 -1.68149 1.02837 1.07572 0.89015
1 67 -1.54092 1.07752 1.10869 0.92758
The last three columns are different corrections obtained after test and the parameters I have used is quite large (EXXRLvcs, NGsBlkXp…). I just cannot make a judge about the converge. Please give me some advices.

[Daniele Varsano]

Dear Shijun,
the criteria of convergence is alway related to the quantity you are interested in.
Usually energy difference converge faster than single bands qp energies.
Anyway, with the dew details you provided is impossible to help you, as I do not know which input variables you are changing on what amount etc...
If you provide more info we can try to advise you.
Best,
Daniele

[zsjan]

Dear Shijun,
the criteria of convergence is alway related to the quantity you are interested in.
Usually energy difference converge faster than single bands qp energies.
Anyway, with the dew details you provided is impossible to help you, as I do not know which input variables you are changing on what amount etc...
If you provide more info we can try to advise you.
Best,
Daniele

Thank you!
What is the the energy difference? Is the energy correction difference betwen state n and state n+1?
The input file is below:
em1d # [R Xd] Dynamical Inverse Dielectric Matrix
gw0 # [R GW] GoWo Quasiparticle energy levels
ppa # [R Xp] Plasmon Pole Approximation
HF_and_locXC # [R XX] Hartree-Fock Self-energy and Vxc
EXXRLvcs= 10827 RL # [XX] Exchange RL components
Chimod= "Hartree" # [X] IP/Hartree/ALDA/LRC/BSfxc
% BndsRnXp
1 | 600 | # [Xp] Polarization function bands
%
NGsBlkXp= 4000 RL # [Xp] Response block size
% LongDrXp
1.000000 | 0.000000 | 0.000000 | # [Xp] [cc] Electric Field
%
PPAPntXp= 27.21138 eV # [Xp] PPA imaginary energy
% GbndRnge
1 | 600 | # [GW] G[W] bands range
%
GDamping= 0.10000 eV # [GW] G[W] damping
dScStep= 0.10000 eV # [GW] Energy step to evalute Z factors
DysSolver= "n" # [GW] Dyson Equation solver (`n`,`s`,`g`)
%QPkrange # [GW] QP generalized Kpoint/Band indices
1| 8| 1|600|
%
%QPerange # [GW] QP generalized Kpoint/Energy indices
1| 8| 0.0|-1.0|
%
Actually, I want to calculate the optical spectra after GW corrections, so I have included all the bands in all the k points ( QPkrange).

[Daniele Varsano]

Dear Shijun,

What is the the energy difference?

For instance a gap, or an electron=hole energy difference.
Ok this is your input, but on what variable you are testing convergences? What is changing in the number you posted above?

Actually, I want to calculate the optical spectra after GW corrections, so I have included all the bands in all the k points ( QPkrange).

OK, but probably you do not need such a bing number of bands for the absorption sepctra.

Daniele