[ase-users] Optimization - fmax in NEB

[Sungsik Kim]

Hello, Jens

Thank you for your help.

I used QuasiNewton optimizer with fmax=0.05 for initial and final surface
potential energy.

And I used FIRE optimizer with fmax=0.05 for NEB calculation with DACAPO
calculator.

However, the calculation had not been completed, I changed fmax to 0.5 to
reduce computing time.

30000th does not mean iteration. It is the number of calculation in
optimization on python shell.

I tested path and energy barrier with EMT calculator by your comment
because DACAPO calculator take lots of time.

In this run, I used Quasinewton optimizer for initial and final state and
FIRE for NEB, all with EMT calculator.

First I ran it with different fmax such as 1.0, 0.5, 0.1, 0.05, and 0.01 in
NEB calculation.

I got 0.625ev for activation energy for those fmax = 0.05.

Second I run it with fmax = 0.05 with different iteration such as 5, 10,
20, 30, 40, 50.

In this run, I still got the same activation energy, 0.625eV.

Path is getting more refined until 30 iteration. But after 40 it is a
little changed. it is slightly fluctuated initially.

Finally, even though I used different fmax such 0.5 and different
iteration, I got the same activation energy.

Does it mean that I can used fmax= 0.5 to get activation energy for
dissociative adsorption of methane on Pt with DACAPO calculator?

Thank you for your help again.

Sincerely

Sungsik





2013/4/2 Jens Jørgen Mortensen <jensj at fysik.dtu.dk>

>  Den 31-03-2013 17:21, Sungsik Kim skrev:
>
>  Hello,
>
> This is Sungsik.
>
> Thank you for your help last time. I got correct NEB images now.
>
> I have a question about fmax in optimaztion.
>
> I need activation barrier for dissociative adsorption of methane on
> platinum.
>
> I already get surface potential energy with fmax 0.05 which is default by
> DACAPO.
>
>  However, in getting the activation barrier with DACAPO calculator with
> fmax=0.05,
>
> Calculation has not been completed even though I run it over 10days.
>
> And my computer looks like being exploded after 30000th calculation.
>
> I ran it by NEB with EMT calculator with different fmax in NEB code.
>
> Even though surface potential energy is a little different, but I got the
> same activation
>
>  energy in different fmax value such as 0.5 , 0.1, and 0.05 in optimize in
> NEB.
>
> It is okay to use fmax value. 0.5 in optimazation in NEB to get activation
> barrier ?
>
> I would like to know if this affects activation barrier too much?
>
>
> I think one can often get away with quite large fmax, but it will depend
> on the reaction so you will have to do some tests yourself.  What do you
> get if you plot the activation barrier as a function of iteration number
> for your 30000 iterations?  Does the barrier change?  Does the path change?
>
> Maybe you can do some tests for a small system or if you have a realistic
> model potential you can use that for some quick tests - like you did with
> EMT.
>
> What optimizer did you use?
>
> Jens Jørgen
>
>
> Thank you for your time.
>
> Sincerely
>
> Sungsik
>
>
> _______________________________________________
> ase-users mailing listase-users at listserv.fysik.dtu.dkhttps://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/20130402/0bb6a523/attachment.html>