G0W0 Calculation using Truncated Coulomb Potential

[Bapan]

Dear Daniele,
I would like to perform G0W0+BSE calculations to compute the optical properties of 1D systems with 30 atoms in a supercell ( with CUTGeo = "box XZ", as my system is periodic in the y-direction). Therefore, I have to start with G0W0 calculations. For the purpose, in the following options, I can understand all of them, except the last terminology "sex". Can you please elaborate on it? I am not able to find any description of it in your Wikipedia page.
With 30 atoms, the QP energy will be calculated by employing the COHSEX scheme.
-k <opt> :Kernel [opt=hartree/alda/lrc/hf/sex]

I would also like to generate the input file using the following command: (to use the COHSEX scheme) after generating a truncated Coulomb potential:

yambo -J 1D_WR_WC -F yambo_G0W0.in -p c -g g -r -k sex -V qp


2. Which Dyson equation solver should I use for more accurate calculations? I think the "green" option?

3. In the section "Avoid numerical divergences using the Random Integration Method (RIM)" of Yambo Wikipedia (http://www.yambo-code.org/wiki/index.ph ... al_systems) page, how did you decide that "4 (5) is the highest occupied (lowest unoccupied) band"?

You can provide some extra suggestions if you think.
Please also let me know if I am doing anything wrong. Thanks in advance.




Thanks and regards,
Pritam Bhattacharyya,

[Daniele Varsano]

Dear Pritam,
here some comments and suggestions to your questions:

I would like to perform G0W0+BSE calculations to compute the optical properties of 1D systems with 30 atoms in a supercell ( with CUTGeo = "box XZ", as my system is periodic in the y-direction).

This is ok, but I suggest you to use:

Code: Select all

CUTGeo = "ws y"
CUTwsGvec= 0.700000  

This is a coulomb truncation on the Wigner size cell and it turns to be more efficient and stable than "box".

For the purpose, in the following options, I can understand all of them, except the last terminology "sex".

sex stands for screened exchange, anyway this keyword (-k sex) it is meant for building the kernel of BSE calculation so you do not need it for GW.

2. Which Dyson equation solver should I use for more accurate calculations? I think the "green" option?

The Newton solver is the standard one which is very robust.

So in summary for a GW calculation, using plasmon pole approximation you can build the input file by:

Code: Select all

yambo -p p -g n -r -J dir

if you want to compute quasiparticle energies by using the COHSEX approximation you can use:

Code: Select all

yambo -p c -r -J dir 

Here you do not need Dyson solver as the self-energy it is static.

how did you decide that "4 (5) is the highest occupied (lowest unoccupied) band"?

If you inspect your report setup you will find something like:

Code: Select all

[X]States summary         : Full        Metallic    Empty
                          0001-0008               0009-0060

where you can see the last occupied and first empty band.

Best,
Daniele

[Bapan]

Dear Daniele,
Thank you very much for your valuable information. Just, I have started learning about GW+BSE calculations using YAMBO and its related theory as much as possible. Actually, I want to perform G0W0+BSE calculations to compute the optical absorption spectra.

Due to the dynamic screening, PPA expected to produce more accurate results for semiconductors as well as insulators, as compared to the COHSEX scheme which includes static screening. My system is a semiconductor.
1. Thus, should I continue with the PPA for more accurate performance?

................ anyway this keyword (-k sex) it is meant for building the kernel of BSE calculation so you do not need it for GW.

After performing the G0W0 calculation, I will perform BSE calculations. Therefore, I need to include the keyword to build the kernel of BSE calculation.
2. Which option should I select among the Kernels [opt=hartree/alda/lrc/hf/sex] again for accuracy? (I think it should also depend on the selected screening PPA/COHSEX)



Thanks & regards,
Pritam.

[Daniele Varsano]

Dear Pritam,

1. Thus, should I continue with the PPA for more accurate performance?

Yes.

2. Which option should I select among the Kernels [opt=hartree/alda/lrc/hf/sex] again for accuracy?

The BSE kernel is sex, it is independent from the way you calculate the QP correction.
In order to familiarize with Yambo and the theory, you can have a look to GW and BSE tutorials present in the wiki page where you can also found theory and documentation.
http://www.yambo-code.org/wiki/index.ph ... =Main_Page

Best,
Daniele

[Bapan]

Dear Daniele,
At first, I am trying to reproduce the result of the following paper, where the authors have used the YAMBO code:
https://journals.aps.org/prb/abstract/1 ... .94.035437

They truncated the Coulomb potential using box shape, and used the COHSEX scheme. To reproduce the results, I am following their work. In this system, I am getting the following warnings (in large fonts) when I was generating a truncated Coulomb potential using "box XZ":

<---> [WARNING]Impossible to define the grid unit vectors
<---> [WARNING]Trying to expand the k-grid
<---> [02.05] Energies [ev] & Occupations
<---> [03] Transferred momenta grid
<---> [04] Coloumb potential Random Integration (RIM)
<---> [04.01] RIM initialization
<---> Random points | | [000%] --(E) --( <01s> Random points |########################################| [100%] --(E) --(X)
<01s> [04.02] RIM integrals
<01s> Momenta loop | | [000%] --(E) --(X <08s> Momenta loop |############# | [033%] 06s(E) 20s <14s> Momenta loop |########################## | [066%] 13s(E) 19s <21s> Momenta loop |########################################| [100%] 20s(E) 20s(X)
<21s> [WARNING]Non periodic geometry and unit cell are incompatible
<21s> [05] Coloumb potential CutOff :box



Do you think, it may produce the wrong results?



Thanks & regards,
Pritam.

[Daniele Varsano]

Dear Pritam,
the warning you get is not related to the coulomb cutoff but most probably with your ground state setup, but I cannot say much without inspecting your report file. If you want you can upload your r_setup file (you need to rename with an allowed suffix e.g. .txt).

Best,
Daniele

[Bapan]

Dear Daniele,
I have uploaded the r_setup file which is attached with this email. Here the issues are:
1. "Unit cell is Unknown"
2. "[WARNING]Impossible to define the grid unit vectors"
3. "[WARNING]Trying to expand the k-grid"
4. "[WARNING]Non periodic geometry and unit cell are incompatible"



Thanks & regards,
Pritam.

[Daniele Varsano]

Dear Pritam,
Most probably the problem here is your choice of the unit cell for your system, which seems to be very strange.
Your unit cell choice, even if correct, which I do not know, makes anyway the calculations very inefficient considering you have 3D point sampling for a 1D system. I suggest you check it carefully and compare the DFT results with the literature before passing to GW calculations. Note that in the paper you mention tha BZ is correctly sampled with a 2D sampling for phosophorene and 1D sampling for the ribbons.

Best,
Daniele

[Bapan]

Dear Daniele,
I also can't understand why Yambo is considering my 1D system as 3D. I have already performed a series of DFT calculations. All these systems are well verified. Actually, we have written a manuscript on armchair phosphorene nanoribbon and would like to add the GW+BSE results. I can assure you that there is no issue with the geometry. Below, I have written the input file of SCF calculation which can be used in the Quantum Espresso computer program:



&control
calculation='scf',
prefix='5-APNR',
restart_mode='from_scratch',
verbosity = 'high'
wf_collect=.true.
pseudo_dir = './',
outdir='./',
/
&system
ibrav = 0,
nat= 14, ntyp= 2,
ecutwfc = 75.0, ecutrho = 300.0,
force_symmorphic=.true.
/
&electrons
conv_thr = 1.0d-8
electron_maxstep = 100
! mixing_beta = 0.3
/
ATOMIC_SPECIES
P 30.97 P.pbe-n-nc.UPF
H 1.00 H.pbe-n-nc.UPF
CELL_PARAMETERS angstrom
23.4934696332240982 -0.0296640892802778 0.2532170991463525
-0.0053938151898170 4.5758037663241291 -0.1289205185452811
0.2135436529804321 -0.6405976298089230 17.7148296300377410
ATOMIC_POSITIONS angstrom
P 7.987350 1.629322 3.395612
P 1.391946 1.613651 3.444926
P 4.689002 1.594141 3.413563
P 3.011332 3.879792 1.321066
P 6.324071 3.887437 1.297318
P 3.036064 0.119680 3.426646
P 6.350086 0.128139 3.403022
P 7.966361 2.392510 1.274782
P 1.373859 2.378840 1.324795
P 4.671166 2.413169 1.307258
H 0.387034 0.617661 3.203193
H 8.973025 3.386677 1.514237
H 8.993993 0.639545 3.137910
H 0.367009 3.367695 1.584148
K_POINTS {automatic}
1 16 1 1 1 1



Thanks & regards,
Pritam.

[Daniele Varsano]

Dear Pritam,
I'm not saying there is something wrong with your input, but that it is not the simplest and probably most efficient way to define a 1D system. If you check the k point sampling in the Yambo report file they are indeed 1D in rlu units, most probably the warning are harmless and you can continue the calculations. Anyway, as you can see from the report file, and also from the QE output, using the current setup no symmetries are taken into account, while I believe that for this system, symmetries (e.g. inversion?) can be exploited reducing the computational cost.

Best,
Daniele