Convergence study issue

[Laura Caputo]

Good afternoon,

I am currently studying the GW corrections to the band gap of a graphene layer doped with a BN ring.

After calculating the WFK.nc file with Abinit, I am doing the convergence test with Yambo. However, I obtain a strange result in particular when I plot the band gap values in function of the NGsBlkXp value with different BndsRnXp.
In attachment, I have uploaded the graphs. I was wondering if the valence/conduction band graphs are correct and, if so, why I obtain such a strange convergence for the gap.

If it can help, I have uploaded also an example of input and report file from the convergence.

With regards,
Laura Caputo

PS I have not finished yet, that's why the '400' line is interrupted.

Edit: I can't upload the report/input. I have the files in .txt format but even though I upload them, they don't show up in the post. I'll put the input here:
Input:
#
# ooooo oooo .. ooo ooo ooooooooo. .oooo.
# `88. .8" .88. `88. .88 `88" `Y8b dP" `Yb
# `88. .8" .8"88. 888b d"88 88 888 88 88
# `88.8" .8" `88. 8 Y88. .P 88 88oooo888" 88 88
# `88" .88ooo888. 8 `888" 88 88 `88b 88 88
# 88 .8" `88. 8 Y 88 88 .88P `8b d8"
# o88o88o 888o8 88 o88bood8P" `Ybod8P"
#
#
# Version 5.0.1 Revision 19547 Hash e90d90f2d
# Branch is
# MPI+HDF5_IO Build
# http://www.yambo-code.org
#
HF_and_locXC # [R] Hartree-Fock
gw0 # [R] GW approximation
ppa # [R][Xp] Plasmon Pole Approximation for the Screened Interaction
dyson # [R] Dyson Equation solver
em1d # [R][X] Dynamically Screened Interaction
EXXRLvcs= 50 Ry # [XX] Exchange RL components
VXCRLvcs= 362775 RL # [XC] XCpotential RL components
Chimod= "HARTREE" # [X] IP/Hartree/ALDA/LRC/PF/BSfxc
% BndsRnXp
1 | 400 | # [Xp] Polarization function bands
%
NGsBlkXp= 2 Ry # [Xp] Response block size
% LongDrXp
1.000000 | 1.000000 | 1.000000 | # [Xp] [cc] Electric Field
%
PPAPntXp= 27.21138 eV # [Xp] PPA imaginary energy
XTermKind= "none" # [X] X terminator ("none","BG" Bruneval-Gonze)
% GbndRnge
1 | 400 | # [GW] G[W] bands range
%
DysSolver= "n" # [GW] Dyson Equation solver ("n","s","g")
%QPkrange # [GW] QP generalized Kpoint/Band indices
13|13|64|65|
%
X_Threads=24 # [OPENMP/X] Number of threads for response functions
DIP_Threads=24 # [OPENMP/X] Number of threads for dipoles
SE_Threads=24 # [OPENMP/GW] Number of threads for self-energy
RandQpts = 1000000 # [RIM] Number of random q-points in the BZ
RandGvec = 100 RL # [RIM] Coulomb interaction RS components
GDamping= 0.100000 eV # [GW] G[W] damping
dScStep= 0.100000 eV # [GW] Energy step to evaluate Z factors

Report:

[Daniele Varsano]

Dear Laura,

the convergence does not see really strange, but just very slow (looking at VBM and CBM, then the gap is a difference and it should converge faster (note the different scale in the plots).
My advise, if you do not have already done, when adding bands in NGsBlkXp increase also GbndRnge to the same value and use the GTermKind= "BG" .

Best,
Daniele

[Laura Caputo]

Dear Daniele,

Thanks for the very quick reply. What seems strange to me is that a higher number of bands used in the calculations sometimes have a higher value of band gap compared to fewer bands (for example, the 200 line is higher compared to the 300 line).

Concerning your advice, I may have a question: for NGsBlkXp + BndsRnXp convergence, I set a fixed number of GbndRnge (a high one in order to ensure ita convergence). Is it better to always keep BndsRnXp and GbndRnge equals and set the termination as you said?

Thanks,
Laura Caputo

[Daniele Varsano]

Dear Laura,
this happens because the VBM and CBM are converging with a different slope. The gap anyway seems converged with respect the number of bands while the single states they are not.
My advice was intended to eliminate one parameter of convergence, but of course they can also be studied separately.
The terminator on G speed up the convergence without loosing too much the performance, at difference of the terminator of X which results to slow down the calculation and it is memory intensive.

Best,
Daniele

[Laura Caputo]

Dear Daniele,

I understand. Thanks a lot for the clarification and for the support.

With regards,
Laura Caputo

[Laura Caputo]

Dear Daniele,
I have a last question, hope it's not a problem to reply here.
If I use the BG terminator in the convergence, shall I use it also in the final GW calculation?

Thanks,
Laura

[Daniele Varsano]

Dear Laura,
yes sure, you need to be consistent.

Best,
Daniele

[Marzieh]

Hi Developers,

I am doing GW convergence test for LiNbO3.

I have ploted and attached band gap, valence/conduction energy values in function of the NGsBlkXp value with different BndsRnXp.

It is strange the valence/conduction do not go to be converged!!!

I was wandering which values is better to take as converged valus.

Some calculations are still running and since it takes time Could you please tell me if I can take NGsBlkXp=7 and BndsRnXp=300 as converged values for continuing the calculation ?



One of the input:

# Version 5.0.1 Revision 19547 Hash e90d90f2d
# Branch is
# MPI+OpenMP+SLK+HDF5_MPI_IO Build
# http://www.yambo-code.org
#
HF_and_locXC # [R] Hartree-Fock
gw0 # [R] GW approximation
ppa # [R][Xp] Plasmon Pole Approximation for the Screened Interaction
dyson # [R] Dyson Equation solver
em1d # [R][X] Dynamically Screened Interaction
X_Threads=0 # [OPENMP/X] Number of threads for response functions
DIP_Threads=0 # [OPENMP/X] Number of threads for dipoles
SE_Threads=0 # [OPENMP/GW] Number of threads for self-energy
EXXRLvcs= 100 Ry # [XX] Exchange RL components
VXCRLvcs= 23291 RL # [XC] XCpotential RL components
Chimod= "HARTREE" # [X] IP/Hartree/ALDA/LRC/PF/BSfxc
% BndsRnXp
1 | 300 | # [Xp] Polarization function bands
%
NGsBlkXp= 7000 mRy # [Xp] Response block size
% LongDrXp
1.000000 | 1.000000 | 1.000000 | # [Xp] [cc] Electric Field
%
PPAPntXp= 27.21138 eV # [Xp] PPA imaginary energy
XTermKind= "none" # [X] X terminator ("none","BG" Bruneval-Gonze)
% GbndRnge
1 | 500 | # [GW] G[W] bands range
%
GTermKind= "none" # [GW] GW terminator ("none","BG" Bruneval-Gonze,"BRS" Berger-Reining-Sottile)
DysSolver= "n" # [GW] Dyson Equation solver ("n","s","g")
GWoIter=0 # [GW] GWo self-consistent (evGWo) iterations on eigenvalues
GWIter=0 # [GW] GW self-consistent (evGW) iterations on eigenvalues
%QPkrange # [GW] QP generalized Kpoint/Band indices
1|1|34|35|
%


Best,
Marzieh

[Daniele Varsano]

Dear Marzieh,

if you are interested in the gap, which converge faster, I would say it is very well converged. If you are interested in the valence/conduction bands some check is still needed, especially for the valence bands.

Anyway, I suggest you to use

Code: Select all

GTermKind= "BG" 

this algorithms speed up the convergence with respect unoccupied bands.

Best,
Daniele

[Marzieh]

Dear Daniele,

Thanks for your reply.
I wanted to use gw correction for BSE and SHG calculation. Then which one you suggest me to consider as convergence criteria? valence or gap?

Best,
Marzieh