Minimization in NEB

LAMMPS LAMMPS Mailing List Mirror
1

Hi,

I am studying adatom diffusion barriers using NEB in LAMMPS. For surface barriers such as hopping and exchange, I can find similar results with other calculations in literature. However, the energy barriers for steps and kink sites are much smaller than expected results.

I have followed the documentation and the exact script for the example of NEB. I have tried to use both quickmin and fire minimizer and also i have tested different spring constants. I have selected all atoms in my simulations for "fix neb" command (ex: fix 1 all neb 1.0). There are not any other constraints in the system. However, in the movie of diffusion process, the migration of adatoms occur perfectly but all atoms in the system never minimize even the material is a nanostructure. Therefore, I think that the problem in my system is caused by the minimization procedure in NEB. I have worked with different metals (such as Cu, Ag …) using both eam and meam potentials but the same problem exists.

How can I correct the minimization procedure for all atoms in every image of NEB?

Thanks.

Regards,

Mine Konuk

2

I don't know what you mean by "never minimize".

You can dump out snapshots of all your replicas as you
run NEB. So if you examine those snapshots you
should see that the coords of various atoms are changing,
even ones not directly involved in the barrier. Each
replica does not go to a minimum energy config b/c it
has additional spring forces acting on it.

Steve

3

I am sorry for the wrong expression about the "never minimize". I just tried to explain that atoms did not relax correctly in the minimization. I checked the snapshots of the replicas. The coordinates of atoms only change with order of 10^-6 but it is not an expected change in my simulation system.

Anyway, I will check my calculation one more time.

Thank you very much for your help.

Mine Konuk

4

There could be a bug with NEB in LAMMPS,
but just b/c it gives a different answer than you
expect isn't sufficient info to figure out what
it might be.

Steve