BSE vs. HF

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

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

BSE vs. HF

Postby martinspenke » Thu Apr 14, 2016 3:27 pm

Dear Daniele,

I assumed that when i set BSKmod= "SEX" , BSENGBlk= 1 , and NGsBlkXs= 1 in BSE calculation,
i should obtain the same result as when i set BSKmod= "HF" , BSENGBlk= 1 , and NGsBlkXs= 1.
But i obtain 2 against each other shifted same spectra. Am i doing something wrong here ?

I think that by turning off the screening in BSE, i have to obtain the same result as when i run a time dependent HF calculation.

Best wishes
Martin
Martin Spenke, PhD Student
Theoretisch-Physikalisches Institut
Universität Hamburg, Germany
martinspenke
 
Posts: 149
Joined: Tue Apr 08, 2014 6:05 am

Re: BSE vs. HF

Postby Daniele Varsano » Fri Apr 15, 2016 3:35 pm

Hi Martin,

I think that by turning off the screening in BSE, i have to obtain the same result as when i run a time dependent HF calculation.


That's tight but by setting BSENGBlk= 1 you are not turning off the screening in BSE, but just considering 1 component (G=G'=0) in the direct term: in HF and SEX they do differs as in BSE it is screened and in HF it is not.

If you discard it totally (BSENGBlk= 0 eV) than the two calculation will be equivalent, but it is not anymore HF or BSE< but simply RPA in both case.

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: 1978
Joined: Tue Mar 17, 2009 2:23 pm

Re: BSE vs. HF

Postby martinspenke » Sat Apr 16, 2016 5:58 pm

Dear Daniele,

many thanks.

I am interested in exchange effects of BS kernel on small organic molecules (methanol, etc).
However, i have difficulties to extract exchange component of the BS kernel. I thought just by setting BSENGBlk= 0 eV and NGsBlkXs= 0 eV, it is possible to obtain a very similar spectrum as IP-GW-spectrum.

I know this because in yambo 3.3.0, you can set the kernel explicitly to BSresKmod= "x" ,and therefore obtaining directly the exchange part of BS kernel.
And the resulting spectrum is very similar to that of IP-GW, but only decreased a little in intensities.

However, if i set BSENGBlk= 0 eV and NGsBlkXs= 0 eV in yambo 3.4.1, i get a red shifted spectrum with respect to IP-GW-spectrum, and this is not correct.

Of course i simply could use yambo 3.3.0, and proceed further but from whatever reason p2y of yambo 3.3.0 does not generate for me the non local pp contributions in my SAVE directory.

with all that being said, my question is how to extract the exchange part of BS from yambo 3.4.1 ? I add here that in yambo 3.4.1 it seems to be mandatory to include em1s in BSE input file, whereas in yambo_3.3.0 i can run BSE exchange only calculations without em1s keyword, and i think the inclusion of em1s causes the red shift, even if NGsBlkXs= 0 eV !


Best wishes
Martin
Martin Spenke, PhD Student
Theoretisch-Physikalisches Institut
Universität Hamburg, Germany
martinspenke
 
Posts: 149
Joined: Tue Apr 08, 2014 6:05 am

Re: BSE vs. HF

Postby Daniele Varsano » Sun Apr 17, 2016 1:20 pm

Dear Martin,
as you said, setting
set BSENGBlk= 0 eV (or 0 RL) , it is equivalent to include exchange only in the BSE.
Please note that this is equivalent to do am RPA calculation including Local Field.
This is not an IP-GW calculation as you are including local filed effects. The entity of the LF effects it is system dependent, anyway you should not have a
red-shifted spectra.

Can you please post your report?

You can also fo a TDHF calculation setting BSENGBlk= 0 eV (or RL), and this should do the same job.
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: 1978
Joined: Tue Mar 17, 2009 2:23 pm

Re: BSE vs. HF

Postby martinspenke » Mon Apr 18, 2016 9:51 pm

Dear Daniele,

Attached the report file of yambo_3.3.0.
In yambo_3.3.0, i need to set BSENGBlk= 1 RL, in order to get the code worked. But since SresKmod= "x" is set, i think even if BSENGBlk= 1 RL , i should get the exchange contributions only.
The obtained GW+BSE-exchange spectrum is only slightly blue shifted with respect to IP-GW spectrum and decreased in intensities. Results obtained from yambo_3.3.0 are reasonable for me.

In the attached report file you see the line " [WARNING]Wrong serial number for ./SAVE/db.QP" which is okay, since i copied pasted.

In yambo_3.4.1 i set BSENGBlk= 0 eV, Chimod = "IP", and NGsBlkXs= 1 eV producing the IP-RPA without local field effects which is equivalent to IP-GW spectrum.
However, IP-GW+BSE-exchange spectrum should be different than IP-GW spectrum, as in the case of yambo_3.3.0.

I would like to obtain the same IP-GW+BSE-exchange spectrum from yambo_3.4.1, too .

Bests
Martin
You do not have the required permissions to view the files attached to this post.
Martin Spenke, PhD Student
Theoretisch-Physikalisches Institut
Universität Hamburg, Germany
martinspenke
 
Posts: 149
Joined: Tue Apr 08, 2014 6:05 am

Re: BSE vs. HF

Postby Daniele Varsano » Tue Apr 19, 2016 4:37 am

Dear Martin,
However, IP-GW+BSE-exchange spectrum should be different than IP-GW spectrum, as in the case of yambo_3.3.0.

Yes, sure, can you post the report of the 3.4 run. Can you also post the outputs?

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: 1978
Joined: Tue Mar 17, 2009 2:23 pm

Re: BSE vs. HF

Postby martinspenke » Wed Apr 20, 2016 1:29 am

Dear Daniele,

I did a new clean re-run of the calculation, and now i can reproduce the results of yambo 3.3.0 with yambo_3.4.1. Problem solved. Many thanks.

A further point :
when i in the COHSEX approximation force yambo to include empty bands by UseEbands, i see considerable change of the band gap.
Questions :
1) Does forcing of empty bands in the green's function take place in the SEX or in the COH term ? Since in both terms there are KS wave functions. This is to me not clear from the tutorial.

2) I see that in the tutorial and also in my system the change of the electronic band gap, once empty bands are included,
is NOT negligible, therefore, i do not have a clear idea of why people call the elimination of empty bands in COHSEX as an advantage, if results without empty bands are not converged ?

Question regarding yambo_3.4.2
In Yambo_3.4.2, when i set NGsBlkXs= 2000 RL , or to higher values, i end up with a netcdf error. How can i compile yambo_3.4.2 without netcdf, as it happens automatically during compilation.
So far i could n't disable netcdf.


Best wishes
Martin
Martin Spenke, PhD Student
Theoretisch-Physikalisches Institut
Universität Hamburg, Germany
martinspenke
 
Posts: 149
Joined: Tue Apr 08, 2014 6:05 am

Re: BSE vs. HF

Postby Daniele Varsano » Wed Apr 20, 2016 2:32 pm

Dear Martin,
1-2) the cohsex self energy terms can be calculated as done in GW, i.e. using empty states, or the summation can be avoided and occupied only states are used.
These are two different algorithms. Some documentation can be found here and in the cited references.
In GW instead it is not straightforward to avoid empty states and in Yambo you are obliged to converge wrt the empty states.
The use of empty bands in COHSEX is yambo mostly for debugging purpose, and the correction using empty states should converge to the one obtained using occupied states only. So I strongly suggest you to not use empty states.

3) In Yambo 3.4.2 the netcdf libraries are mandatory and cannot be disabled: what kind of error do you have? Netcdf error are usually just consequence of other kind of problem arising in the run.

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: 1978
Joined: Tue Mar 17, 2009 2:23 pm

Re: BSE vs. HF

Postby martinspenke » Wed Apr 20, 2016 5:51 pm

Dear Daniele,

1- I see. However the electronic band gap changes once i force the empty bands in the
Green's function of COHSEX:
% GbndRnge
1 | 10 | # [GW] G[W] bands range
%

"The use of empty bands in COHSEX is yambo mostly for debugging purpose, and the correction using empty states should converge to the one obtained using occupied states only"
This is clearly not the case neither in the tutorial you are referring to nor for my system !

One further point : for instance a system has 100 occupied states, meaning :
% GbndRnge
1 | 100 | # [GW] G[W] bands range
%
However, Yambo_3.4.1 automatically adds another 100 empty states to it, resulting in
% GbndRnge
1 | 200 | # [GW] G[W] bands range
%

even, if UseEbands in DISabled !

2- I try to reproduce that error, and will share it here, since i switched back to yambo_3.4.1 and deleted that error.

Bests
Martin
Martin Spenke, PhD Student
Theoretisch-Physikalisches Institut
Universität Hamburg, Germany
martinspenke
 
Posts: 149
Joined: Tue Apr 08, 2014 6:05 am

Re: BSE vs. HF

Postby Daniele Varsano » Wed Apr 20, 2016 5:59 pm

Dear Martin,

1. Again. If you use empty bands or not they are two totally different implementations. If you do not use empty bands there are not sum-over-states.
If you use Empty bands you need to converge the gap with respect empty bands, if you do not use them, these are not required anymore as only occupied are used.
The converged value using empty bands and the one without using empty bands should be compatible.

2. This does not matter, as if you do not use empty bands this variable it is not needed anymore.



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: 1978
Joined: Tue Mar 17, 2009 2:23 pm

Next

Return to Other issues

Who is online

Users browsing this forum: No registered users and 1 guest

cron