Partial Self Consistency in G

Run-time issues concerning Yambo that are not covered in the above forums.

Moderators: Daniele Varsano, andrea marini, Conor Hogan, myrta gruning

Partial Self Consistency in G

Postby luca.montana » Fri Jun 13, 2014 9:10 pm

Dear all,

Does any one know how to use partial self consistency in G (scf-GW0) in yambo ?

Regards
Luca Montana
PhD student
University of York, UK
luca.montana
 
Posts: 27
Joined: Fri Jun 13, 2014 6:52 pm

Re: Partial Self Consistency in G

Postby Daniele Varsano » Fri Jun 13, 2014 9:22 pm

Dear Luca,
Self consistency in yambo it has not yet released in the gpl version.
We are working on that and hopefully it will be included in the next release.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
User avatar
Daniele Varsano
 
Posts: 2058
Joined: Tue Mar 17, 2009 2:23 pm

Re: Partial Self Consistency in G

Postby luca.montana » Sat Jun 14, 2014 9:53 am

Dear Daniele,

one more question :

Concerning coulomb cutoff should one take into account this technique in all 3 steps : namely : 1) GW + 2) W (static screening calculation) + 3) BSE kernel calculation
or is it enough to include it only in GW and W (static screening calculation) ?

Bests
Luca Montana
PhD student
University of York, UK
luca.montana
 
Posts: 27
Joined: Fri Jun 13, 2014 6:52 pm

Re: Partial Self Consistency in G

Postby Daniele Varsano » Sat Jun 14, 2014 1:00 pm

Dear Luca,
I use to calculate once for all after the setup, and then use the coulomb cutoff both in GW and in BSE kernel (including also in the screening). When calculating the excitonic matrix you can concatenate runlevels, so you can calculate W and the kernel in the same run (yambo -o eh -k sex), or if you prefer calculate the screening in a separate run. For my experience the volume effects are greater in the quasiparticle energies calculation (GW) being charged excitations.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
User avatar
Daniele Varsano
 
Posts: 2058
Joined: Tue Mar 17, 2009 2:23 pm

Re: Partial Self Consistency in G

Postby luca.montana » Sat Jun 14, 2014 8:48 pm

Dear Daniele,

my problem is : when using yambo -g s -d -k hartree -c and yambo -b -c (both including coulomb cut-off) and yambo -o b -k sex -y d (without cutoff) i get proper results.
But when i also include the cut-off in BSE kernel, namely : yambo -o b -k sex -y d -c results become wrong !

How do you see this ?

Bests
Luca Montana
PhD student
University of York, UK
luca.montana
 
Posts: 27
Joined: Fri Jun 13, 2014 6:52 pm

Re: Partial Self Consistency in G

Postby Daniele Varsano » Sat Jun 14, 2014 9:34 pm

Dear Luca,
I cannot see this, as I do not know what system are you dealing with, what kind of cutoff you are using, and what mean for you right and wrong.
The reason could be severals. I can see that you are using real integration method to calculate the GW qo energies, this method it is very hard to converge.
If you want you can post your input/reports/outputs and will have a look.
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
User avatar
Daniele Varsano
 
Posts: 2058
Joined: Tue Mar 17, 2009 2:23 pm

Re: Partial Self Consistency in G

Postby Daniele Varsano » Sat Jun 14, 2014 10:31 pm

Dear Luca,
I just did some checks, and I noticed something inconsistent in the BSE when using the coulomb cutoff.
I do not know if the problem is in the construction of the excitonic matrix or after the diagonalization, I need to check it carefully.
I will look at it in details in order to spot the problem and solve it, but I cannot say you exactly when, as I have some urgent deadlines in the next period.
In the meanwhile I suggest you to *not* use the coulomb cutoff in the last step (kernel and diagonalization) and check the convergence with respect the supercell volume. As I told you, big volume effects exists in the GW part, while in the absorption they are minor and hopefully the spectra can be converged with respect the supercell volume without using the cutoff.
Thanks for noticing the problem.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
User avatar
Daniele Varsano
 
Posts: 2058
Joined: Tue Mar 17, 2009 2:23 pm

Re: Partial Self Consistency in G

Postby luca.montana » Mon Jun 16, 2014 8:00 am

Dear Daniele,

Some days ago i accidentally detected this paper : http://journals.aps.org/prb/abstract/10.1103/PhysRevB.72.245425 in which on page 14 you see the spectrum
of water monomer. I can confirm their result and even better it using a 30 bohr big box, only, as i said before, when i use these commands : yambo -g s -d -k hartree -c and yambo -b -c (both including coulomb cut-off) and yambo -o b -k sex -y d (without cutoff) , otherwise using coulomb cutoff in BSE part gives wrong excitation energies.

big volume effects exists in the GW part, while in the absorption they are minor and hopefully the spectra can be converged with respect the supercell volume without using the cutoff

it means that volume effects for GW part are of minor importance when using coulomb cutoff since the advantage of it, is to boost the convergence of the qp energies with respect to the super-cell size ?

For ethene molecule unfortunately the first excited state is of double excitation character , no possibility to have dynamic kernel in yambo?


Best wishes
Luca Montana
PhD student
University of York, UK
luca.montana
 
Posts: 27
Joined: Fri Jun 13, 2014 6:52 pm

Re: Partial Self Consistency in G

Postby Daniele Varsano » Mon Jun 16, 2014 9:09 am

Dear Luca,
it means that volume effects for GW part are of minor importance when using coulomb cutoff since the advantage of it, is to boost the convergence of the qp energies with respect to the super-cell size ?

Yes, when using the coulomb cutoff it speed up the convergence of the qp energies with respect the supercell size, that otherwise it would be very slow. For the absorption spectra, these effects should be minor.

About double excitations, there is an implementation of dynamical kernel, but it si not in the GPL release. It still need to be updated and carefully tested. Hopefully it will be included in the next release, but I cannot say you exactly when it will happen.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
User avatar
Daniele Varsano
 
Posts: 2058
Joined: Tue Mar 17, 2009 2:23 pm

Re: Partial Self Consistency in G

Postby luca.montana » Tue Jun 17, 2014 5:06 am

Dear Daniele,

should quasi particle energies in GW part be converged with respect to the volume of the super-cell AND coulomb cutoff radius R_c
Or one can relax a bit the condition of convergence of QP energies with respect to the coulomb cutoff radius R_c and only take into account the convergence regarding super-cell volume ?

I red your paper : "An exact Coulomb cutoff technique for super-cell calculations"
Unfortunately i see no where in the paper explicitly which one is more important, i assume QP convergence with respect to super-cell volume is much more important, but i would like to be sure of this fact.

when i increase the super-cell size, should i also include more unoccupied bands into the calculation or is it enough to use that number of bands for which the convergence of qp energies
for one specific representable box size and R_c has been reached already ?

With best wishes
Luca Montana
PhD student
University of York, UK
luca.montana
 
Posts: 27
Joined: Fri Jun 13, 2014 6:52 pm

Next

Return to Other issues

Who is online

Users browsing this forum: No registered users and 1 guest