Hello all,
I want to optimization a hexagonal structure actually SiC-4H with 8 atoms(4 c and 4 Si), Based on this tutorial, for doing it in exciting I need to install “sgroup” to do it, but the page to download “sqroup” isn’t available anymore, somebody has any suggestion to how optimize my structure by running different code an a different way of tutorial? or somebody can help me to find “sgroup” and install it on “exciting”.
http://exciting.wikidot.com/oxygen-general-lattice-optimization#toc9
many thanks,
Best,
Nasrin
Hi Nasrin,
Thank you you very much for noticing that the old link is not working anymore.
We have updated the link. Please, try again to download sgroup and tell us if everything is working fine now.
Best,
Pasquale Pavone
(exciting team)
Dear Prof. Pasquale Pavone,
Zes, Thank you very much for your quick reply, I could optimize my structure and it worked, for the next part I have a question, I am new in solid state and in exciting too. I want to calculate the G0W0 my structure but I don’t know how should I select these parameters, I read the papers and I know what is these as a parameter but in numerical computational I don’t have any idea to what should I do:
<gw
taskname=“g0w0”
ngridq=“3 3 3” ???
nempty=“22”
ibgw=“1” ???
nbgw=“10” ???
coreflag=“xal”
>
<mixbasis
lmaxmb=“3” ???
epsmb=“1.d-4” ???
gmb=“1.0” ???
>
<freqgrid
nomeg=“12” ???
>
For ibgw I select 1 and for nbgw I select 30 because the number of my valence band from the SCF calculation obtained 22 and I select this number since for plotting the band structure be able to see more conduction bonds.
My structure is SiC-4H with 8 atoms in unit cell and after plotting with xcrysden I had 56 atoms.
Can you please guide me how should I select these items based on different structures? I am so curious to understand deeply.
Best Regards,
Nasrin Farahani
Dear Nasrin,
It would be helpful to us if you could post your complete input file.
Furthermore I can recommend you to check the following links:
Best regards,
Pasquale Pavone
(exciting team)
Hi Nasrin, for GW you need to converge the number of local orbitals per l-channel in the species file. For example, I’ve posed l=0 for Zr below. You’d do the same for every l-channel, of each species.
Please see this paper for an explanation of the physics.
trialEnergy values can be recommended by LOrecommendations (see the exciting input reference) during a ground state run - and make sure to run with one MPI rank when generating that file.
<lo l="0">
<wf matchingOrder="0" trialEnergy="3.43" searchE="false"/>
<wf matchingOrder="1" trialEnergy="3.43" searchE="false"/>
</lo>
<lo l="0">
<wf matchingOrder="0" trialEnergy="11.87" searchE="false"/>
<wf matchingOrder="1" trialEnergy="11.87" searchE="false"/>
</lo>
<lo l="0">
<wf matchingOrder="0" trialEnergy="23.56" searchE="false"/>
<wf matchingOrder="1" trialEnergy="23.56" searchE="false"/>
</lo>
<lo l="0">
<wf matchingOrder="0" trialEnergy="38.23" searchE="false"/>
<wf matchingOrder="1" trialEnergy="38.23" searchE="false"/>
</lo>
Once done, you can converge your GW settings:
<gw
taskname="g0w0"
nempty="3000" Pick some large value in excess of the total number of states
You can reduce it once the basis is converged
ngridq="2 2 2" Start small, converge last.
skipgnd="false" Important! Else the ground state won't be re-run with the new basis
>
<mixbasis ! Fine to use defaults
lmaxmb="4"
epsmb="0.001"
gmb="1.0"
></mixbasis>
<freqgrid
nomeg="32" ! Converge from say, 10 to 30 once basis is converged. Do so for low-k-sampling
freqmax="1.0"
></freqgrid>
<barecoul ! Fine to use defaults
pwm="2.0"
stctol="1e-16"
barcevtol="0.1"
></barecoul>
<selfenergy ! Fine to use defaults
actype="pade"
singularity="mpb"
></selfenergy>
</gw>
Cheers, Alex
Also note, all-electron GW is extremely expensive. Don’t expect to be able to do 56 atoms. Even 20 is resource-intensive.
Dear Alex,
It was really helpful, many thanks for your quick response 
I have another question from this part “Excited States from BSE”,Excited States from BSE - exciting
from the convergence part, I need to converge" ngrid-k, ngrid-q, nempty , gqmax, and nstlbse" , I am a little confused how should I do it.
For example, I did the convergence for the GW for the nempty and ngrid-q, I did it by obtaining the band gap for different number of empty bands and when it converges, I converged the band gap respect to ngrid-q, sooo simple. But, here I don’t know how start and must converged based which parameter?
Can you please guide me,
Best,
Nasrin
Dear Prof. Pasquale Pavone,
Many thanks for recommending these links, there were so useful specially the last one, and helped me to find the correct way.
Best Regards,
Nasrin
Hi Nasrin, have you tried looking at any papers on the BSE with exciting?
I’m not an expert in the BSE but the main steps are:
- converge gqmax
- converge ngridk
- converge nempty
Species files are typically not modified to include extra local orbitals.
W.r.t. GW however you need to converge the local orbitals too (before nempty and the q-grid), as I’ve described above, and as explained in the paper I linked above. Else your results will not be converged. This requires adding to the species files.
Can you please also put questions not related to the original comment in a new question thread, such that it’s easier for other people looking at the forum to find relevant information.
Cheers,
Alex
Dear Alex,
Thank you very much,
Best,
Nasrin