# [lammps-users] question on minimization

Dear All,

I am minimizing a gold nanometer using Foiles (1986) potential and
Haftel’s potential (Molecular-dynamics description of early film deposition of Au on Ag(110), M.I.Haftel and Mervine Rosen, PRB 1995).

I tried with three combinatios of " maxiter maxeval " options of minimize command

1) 1e-14 1.0e-16 10000 10000
2) 1e-14 1.0e-16 100000 100000
3) 1e-14 1.0e-16 1000000 1000000

while for Foiles potential, the three combinations give the same pxx and total energy values after minimization, it is not the case for Haftel’s potential. I feel these values should not change with the maxiter and maxeval values.

Does anybody have any clue.

I am pasting the results for three combinations and for both potentials after minimization.

Foiles potential
Step Pxx TotEng
Combination 1 2460 -8.3423 -25861.848
Combination 2 2460 -8.3423 -25861.848
Combination 3 2460 -8.3423 -25861.848

Haftel’s potential
Step Pxx TotEng
Combination 1 1834 -22523.715 -25284.786
Combination 2 17631 -4145.109 -25585.735
Combination 3 42168 -135.0508 -25725.542

sivashankar
|

Dear All,

I am minimizing a gold nanometer using Foiles (1986) potential and
Haftel's potential (Molecular-dynamics description of early film deposition of Au on Ag(110), M.I.Haftel and Mervine Rosen, PRB 1995).

I tried with three combinatios of " maxiter maxeval " options of minimize command

1) 1e-14 1.0e-16 10000 10000
2) 1e-14 1.0e-16 100000 100000
3) 1e-14 1.0e-16 1000000 1000000

while for Foiles potential, the three combinations give the same pxx and total energy values after minimization, it is not the case for Haftel's potential. I feel these values should not change with the maxiter and maxeval values.

Does anybody have any clue.

for a system with a substantial number of degrees of freedom,
a minimizer can get easily stuck in a local minimum. which
minimum that is can depend on the functional form of the potential,
or its implementation, cutoff, minimizer algorithm and minimizer
settings. sometimes the tiniest differences make the minimizer
go "left or right" and then you may, or may not get different results.

cheers,
axel.

Hi Axel

Please correct me if i am wrong:
The system is said to be minimized if the stopping criterion is linesearch alpha is zero or if the system reaches etol or ftol. The system is not minimized if the stopping criterion is max force evaluations or max iterations.

Thanks
sivashankar

— On Sun, 25/7/10, Axel Kohlmeyer [email protected] wrote:

> From: Axel Kohlmeyer <akohlmey@…43…4…>
> Subject: Re: [lammps-users] question on minimization
> To: “siva shankar” <shankara546@…545…>
> Cc: lammps-users@…430…orge.net
> Date: Sunday, 25 July, 2010, 6:40 PM
>
> On Sun, Jul 25, 2010 at 8:38 AM, siva shankar <shankara546@…545…> wrote:
> >
> > Dear All,
> >
> > I am minimizing a gold nanometer using Foiles (1986) potential and
> > Haftel’s potential (Molecular-dynamics description of early film deposition of Au on Ag(110), M.I.Haftel and Mervine Rosen, PRB 1995).
> >
> > I tried with three combinatios of " maxiter maxeval " options of minimize command
> >
> > 1) 1e-14 1.0e-16 10000 10000
> > 2) 1e-14 1.0e-16 100000 100000
> > 3) 1e-14 1.0e-16 1000000 1000000
> >
> > while for Foiles potential, the three combinations give the same pxx and total energy values after minimization, it is not the case for Haftel’s potential. I feel these values should not change with the maxiter and maxeval values.
> >
> > Does anybody have any clue.
>
> for a system with a substantial number of degrees of freedom,
> a minimizer can get easily stuck in a local minimum. which
> minimum that is can depend on the functional form of the potential,
> or its implementation, cutoff, minimizer algorithm and minimizer
> settings. sometimes the tiniest differences make the minimizer
> go “left or right” and then you may, or may not get different results.
>
> cheers,
> axel.
>
> >
> > I am pasting the results for three combinations and for both potentials after minimization.
> >
> > Foiles potential
> > Step Pxx TotEng
> > Combination 1 2460 -8.3423 -25861.848
> > Combination 2 2460 -8.3423 -25861.848
> > Combination 3 2460 -8.3423 -25861.848
> >
> > Haftel’s potential
> > Step Pxx TotEng
> > Combination 1 1834 -22523.715 -25284.786
> > Combination 2 17631 -4145.109 -25585.735
> > Combination 3 42168 -135.0508 -25725.542
> >
> > sivashankar
> >
> >
> > ------------------------------------------------------------------------------
> > This SF.net email is sponsored by Sprint
> > What will you do first with EVO, the first 4G phone?
> > Visit sprint.com/firsthttp://p.sf.net/sfu/sprint-com-first
> > _______________________________________________
> > lammps-users mailing list
> > [email protected]
> > https://lists.sourceforge.net/lists/listinfo/lammps-users
> >
>
> –
> Dr. Axel Kohlmeyer [email protected]
>
> Institute for Computational Molecular Science
> Temple University, Philadelphia PA, USA.

|

Yes.

Specifically:

The minimization procedure stops if any of several criteria are met:

• the change in energy between outer iterations is less than etol
• the 2-norm (length) of the global force vector is less than the ftol
• the line search fails because the step distance backtracks to 0.0
• the number of outer iterations exceeds maxiter
• the number of total force evaluations exceeds maxeval

So, if the maximum number of iterations or evaluations you’ve specified is exceeded and the minimizer is consequently halted, then no, you shouldn’t consider your system as fully minimized.

Paul

This is odd. Do the 3 runs with the Haftel potential reproduce
each other up to the iteration they stop on?

Steve