[ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Questions and doubts about features of non linear optic in Yambo (yamb_nl)

Moderators: Davide Sangalli, claudio, myrta gruning

Dean
Posts: 98
Joined: Thu Oct 10, 2019 7:03 am

[ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Dean » Tue Apr 11, 2023 11:25 am

Dear all,
I am trying to perform RT-BSE calculations for a 2D semiconductor following the tutorial: https://www.yambo-code.eu/wiki/index.ph ... ion_(TDSE).
I got the error "[ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!" in the second step (Time dependent Bethe Salpeter equation). The first step (The Real Time Collisions) is performed successfully.
And I have reproduced the RT-BSE spectrum of 2D h-BN successfully.
More details can be found in the attachments.

Any suggestion will be appreciated.
You do not have the required permissions to view the files attached to this post.
Dr. Yimin Ding
Soochow University, China.

User avatar
Davide Sangalli
Posts: 610
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: [ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Davide Sangalli » Tue Apr 11, 2023 4:09 pm

Dear user,
the problem is due to a fail of the subroutine eval_G_minus_G called by DIPOLE_overlaps

Code: Select all

ng_tmp=eval_G_minus_G(wf_ng_1st_BZ,iG0)
Not sure why. You have a system with more than half a million of G-vectors.
There could be some rounding error ...

You can try to lower the cutoff on the wave-functions from input

Code: Select all

FFTGvecs= XX Ry
and see if this helps.
The maximum meaningful value would be what you set for ecut in the pw.x run (i.e. the energy cutoff on the wfs.)

Best,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

Dean
Posts: 98
Joined: Thu Oct 10, 2019 7:03 am

Re: [ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Dean » Wed Apr 12, 2023 4:06 am

Dear Davide,
Thanks for your reply.
It works when I set FFTGvecs= 10 Ry.
But I wonder how can I set the values of FFTGvecs? Is there any reference?
And what is the meaning of datas in the file, o-xxxx-YPP-eps_along_E, such as Im/eps_d1/d2/d3?
Last, in this website of "Prerequisites for Real Time propagation with Yambo", https://www.yambo-code.eu/wiki/index.ph ... with_Yambo
it is advised to reduce the number of G-vectors (MaxGvecs) to 1000 for 2D h-BN (about 50% of the initial ones) to speed up nonlinear calculations. So, for other 2D materials, how can I set the values of MaxGvecs?
Thanks in advance.
Dr. Yimin Ding
Soochow University, China.

User avatar
Davide Sangalli
Posts: 610
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: [ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Davide Sangalli » Wed Apr 12, 2023 8:49 am

Code: Select all

FFTGvecs
Controls the energy cutoff on the WFs.
In general you can keep it to the default value, i.e. the one used for ecutwf qith pw.x

However, in many cases it is convenient to lower it. In your case you are forced to do so by the error message.
In these cases, you need to check how much the results are affects, i.e. suppose the maximum is 40 Ry, and you want to lower it.
Then, you should try to perform the same run at different values (10, 15, 20, etc ... ) and see when and if it converges.

Best,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

Dean
Posts: 98
Joined: Thu Oct 10, 2019 7:03 am

Re: [ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Dean » Wed Apr 12, 2023 11:18 am

Dear Davide,
Thanks for your reply. It seems that it has no choice but converge test.
By the way, from my test, FFTGvecs can be set as large as the value of ecutrho in pw.x not ecutwfc. Is that right?
Dr. Yimin Ding
Soochow University, China.

User avatar
Davide Sangalli
Posts: 610
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: [ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Davide Sangalli » Wed Apr 12, 2023 12:57 pm

By the way, from my test, FFTGvecs can be set as large as the value of ecutrho in pw.x not ecutwfc. Is that right?
You can go up to ecutrho, but, above ecutwf, the result should only be that you may have a more demanding simulation, without any gain in the final results.

D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

Dean
Posts: 98
Joined: Thu Oct 10, 2019 7:03 am

Re: [ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Dean » Thu Apr 13, 2023 4:09 am

Dear Davide,
Many thanks for your reply.
After test, I also find some strange results.
Firstly, for some values of FFTGvecs (such as 10 Ry or 25 Ry) in the second step (Time dependent Bethe Salpeter equation), the calculated eps. looks reasonable. Please see the attachments.
But foe some values such as 50 Ry, the calculated eps. looks unreasonable with negative data. So why is the reseaon? Should the values of FFTGvecs in the second step (Time dependent Bethe Salpeter equation) be smaller or equal to that in the first step (The Real Time Collisions)?
Secondly, the above results are calculated along (010) direction for my system. However, for (100) or (101) directions, the calculated eps. looks strange with very large negative data. Please see the attachments.
Lastly, I perform the nonlinear calculations by executable yambo_nl compiled without --enable-dp. The log file shows double Precision. Is that OK?
Thanks in advance.
You do not have the required permissions to view the files attached to this post.
Dr. Yimin Ding
Soochow University, China.

User avatar
Davide Sangalli
Posts: 610
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: [ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Davide Sangalli » Thu Apr 13, 2023 11:32 am

FFTGvecs wan not meant to increase the cutoff above ecutwf. I think in this case something bad is happening.
I will investigate this issue.

For yambo_nl, yeah, it is always compiled in double precision, since in single precision results are not good.

Best,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

Dean
Posts: 98
Joined: Thu Oct 10, 2019 7:03 am

Re: [ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Dean » Thu Apr 13, 2023 12:32 pm

Dear Davide,
When I set UseDipoles in the second step (Time dependent Bethe Salpeter equation), the calculated time-dependent polarization and eps looks reasonable. If I do not set UseDipoles, the result looks unreasonable. Please see the attachment.
And this issue occurs in the calculations for (100) or (101) directions, but not (010).
The out files shows "Fixed dipoles: only linear response properties are correct!" . Next, I want to perform non-linear calculations (such as SHG), then the results will be reliable?
Thanks in advance.
You do not have the required permissions to view the files attached to this post.
Dr. Yimin Ding
Soochow University, China.

User avatar
Davide Sangalli
Posts: 610
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: [ERROR] STOP signal received while in[06] Dipoles Error search for G=G1-G2 !!

Post by Davide Sangalli » Thu Apr 13, 2023 12:36 pm

With "UseDipoles" it works fine because it does not go through the subroutine which has the issue with the G-G.
However it works only for linear response. Beyond the approach is not valid.

Best,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

Post Reply