[ase-users] NEB not exiting while maximal force is less than fmax

[Boubacar Traore]

Thanks Marcin,

Sure, I will report to the list if I find anything interesting!






Le Jeudi 6 mars 2014 13h09, Marcin Dulak <Marcin.Dulak at fysik.dtu.dk> a écrit :
 
Hi,

On 03/05/2014 11:10 PM, Boubacar Traore wrote:



Dear ase users,

As I'm currently working on ASE neb with siesta, I'm observing something that I can't figure out.
In my neb.py, I'm using an fmax = 0.04
that is : 
--------------------------------

qn = MDMin(neb, trajectory='neb.traj', logfile='neb.log')

qn.run(fmax =0.04) 

--------------------------------

When I check the evolution of the calculation using :
ase-gui -t -g e,fmax neb.traj  > result.txt

I get an fmax less than 0.04 in result.txt, which mean the calculation should exit, but this is not the case.

when you use ase-gui to print fmax you get fmax for all images in optimizers every iteration,
note that the fmax for image[0] and image[-1] are constant (tha'ts
    how neb works).
From ase-gui you get what's returned by "f =" in https://trac.fysik.dtu.dk/projects/ase/browser/trunk/ase/gui/images.py
On the other hand optimizer is clever and knows it uses neb, so it
    returns what's given by
"get_forces" in https://trac.fysik.dtu.dk/projects/ase/browser/trunk/ase/neb.py
Before doing any neb work please watch http://www.youtube.com/watch?v=wKJLRBsVt7Q
neb is not a black-box method, so don't focus too much
on getting low fmax convergence, but rather try to figure out how
    important is fmax
in your case, and how stable are the neb results.
If you find anything interesting please report to the list.

Best regards,

Marcin




>
>On the other hand, in the neb.log file, fmax is completely different and give something around 0.06 and the calculation almost gets stuck around this 0.06 without exiting.
>
>
>The total energy in both cases are the same.I checked the output file of the transition point and its fmax is less than 0.04 as in result.txt but the calculation is still running.
>
>
>I don't understand why the calculation is still on while fmax is less than 0.04 in result.txt and output file (also siesta .FA file) while in the neb.log file, the force is still at ~ 0.06.
>
>
>I see the same thing in both MDMin and FIRE optimizers!
>
>
>Could anyone enlighten me about this? Is there something else other than fmax as an exit criteria?
>
>
>
>Best regards,
>Boubacar
>
>
>
>_______________________________________________
>ase-users mailing list
>ase-users at listserv.fysik.dtu.dk
>https://listserv.fysik.dtu.dk/mailman/listinfo/ase-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.fysik.dtu.dk/pipermail/ase-users/attachments/20140307/ccba4a17/attachment.html>