[QE-users] forc conv thr not converging force components?

[Tom Demeyere]

If my memory serves me correctly “forc_conv_thr" does not correspond to the max forces on any atoms, I don’t remember exactly how it is calculated but it is a more complex expression.

By experience it is not always possible to reach very low forces on all atoms, especially if you are using some functional/corrections where the forces are not always perfectly consistent with the energy.

If you made sure that conv_thr is small enough so that non-scf correction to the forces is extremely small you might want to call pw.x as an energy/forces evaluator and interface it with a more specialised optimiser. Scipy(L)BFGS often does the trick for me.

-----------------------------------------
Tom Demeyere MSc
PhD student
Skylaris Research Group
School of Chemistry, University of Southampton

On 28 Dec 2023, at 16:39, Adam Michalchuk <a.a.l.michalchuk at bham.ac.uk> wrote:

You don't often get email from a.a.l.michalchuk at bham.ac.uk<mailto:a.a.l.michalchuk at bham.ac.uk>. Learn why this is important<https://aka.ms/LearnAboutSenderIdentification>
CAUTION: This e-mail originated outside the University of Southampton.
Dear all,

I appreciate this topic has been discussed a number of times in the forum, with the answer that all components of all forces should satisfy the forc_conv_thr. This leads me to my probably very simple query that likely stems from a misunderstanding.

when I run a vc-relax calculation with e.g. a very tight forc_conv_thr=1.0D-5, my calculation finishes 'successfully' with e.g.:

    atom    1 type  3   force =     0.00000250    0.00000250    0.00000250
     atom    2 type  3   force =     0.00000432    0.00000559    0.00000432
     atom    3 type  3   force =     0.00000432    0.00000432    0.00000559
     atom    4 type  3   force =     0.00000559    0.00000432    0.00000432
     atom    5 type  2   force =    -0.00000391    0.00000100    0.00000100
     atom    6 type  2   force =     0.00000361   -0.00001336   -0.00001336
     atom    7 type  2   force =     0.00000100   -0.00000391    0.00000100
     atom    8 type  2   force =     0.00000100    0.00000100   -0.00000391
     atom    9 type  2   force =    -0.00001336   -0.00001336    0.00000361
     atom   10 type  2   force =    -0.00001336    0.00000361   -0.00001336

even if I put rather ridiculous thresholds of forc_conv_thr=1.0D-7 it makes no real difference to the force components:

     atom    1 type  3   force =    -0.00000456   -0.00000456   -0.00000456
     atom    2 type  3   force =     0.00000832    0.00001867    0.00000832
     atom    3 type  3   force =     0.00000832    0.00000832    0.00001867
     atom    4 type  3   force =     0.00001867    0.00000832    0.00000832
     atom    5 type  2   force =    -0.00000414   -0.00000409   -0.00000409
     atom    6 type  2   force =    -0.00000040   -0.00004042   -0.00004042


If I understand the output correctly, it appears as though some force components have converged fine, though quite a few components have not reached the threshold, and changing the threshold makes little difference. As we do a lot of work on vibrational dynamics in dispersion bound crystals, tight forces are usually quite important. We've recently moved into using QE and I'm trying to get to grips with how to control the forces in QE optimisations but am clearly missing something. Could anybody please point me in the right direction?

Many thanks in advance,
Adam


 Adam Michalchuk
University of Birmingham UK

_______________________________________________
The Quantum ESPRESSO community stands by the Ukrainian
people and expresses its concerns about the devastating
effects that the Russian military offensive has on their
country and on the free and peaceful scientific, cultural,
and economic cooperation amongst peoples
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu<http://www.max-centre.eu/>)
users mailing list users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20240104/cef3dc8f/attachment.html>