NEB calculation

[Mahesh_chandran]

Dear All,
I am trying to calculate the migration energy for an atom to move into a vacant site using NEB. Which of the following gives the physically correct minimum energy parth, and hence the migration energy:
(1) Moving only that particular atom and freezing the rest using "selected dynamics"
(2) Relaxing all the atoms (keeping ISIF=2).

Regards
MC

[admin]

all atoms should be relaxed (do not set 'selective dynamics' in POSCAR). Please also interpolate the positions of all atoms in the starting guesses of the POSCARs you give in the respective subdirectories to achieve better convergence.

[Mahesh_chandran]

The calculation with all atoms allowed to relax does not converge (the initial guesses were the interpolated images between the end points). I used IBRION=1 and POTIM=0.2.
I also tried another route: first ran the calculation with only the target atom relaxing and freezing the rest using "selective dynamics". The converged minimum energy path was then used as starting guess for relaxing all atoms. Even this calculation does not converge. I find that the crystal structure is significantly distorted if all atoms are allowed to relax.

Will freezing certain degrees of freedom of some atoms help in convergence? Or is there any alternate way to get faster convergence.

Rgards
MC

[tjf]

What do you mean "does not converge"? Are your energies oscillating or what?

[admin]

1) please make sure that the geometries of initial and final state were fully converged.
2) please check if the ionic steps themselves are electronically converged or if the ionic step simply finished after NELM electronic steps without having reached convergence.
3) the cell shape and volume have to be kept constant in a NEB calculation
4) setting 'selective dynamics' is counter-acting the NEB relaxation constrictions when you set it in the NEB - run.

[des]

Dear Admin,
I have question about NEB ‘selective dynamics' in POSCAR. You said we don't have to use it in the POSCAR files. But I have seen from some people on surface calculations, they have used this parameter to fix at least one or two or more atomic layers and I use this method in my simulation. Now I am confused about it. I am now wondering which method I have to use. Which method do you recommend it?
I thank you

[des]

Dear Admin,
<span class='smallblacktext'>[ Edited Sun Jul 09 2006, 11:32PM ]</span>

[Mahesh_chandran]

I finally finished the NEB calculation. I first started with 5 images and got convergence with full relaxation of atoms (ISIF=2). Latter added intermediate images to improve the minimum energy path (MEP).

Regards
Mahesh

PS: I agree with des that some people do fix one, more or all atoms in the NEB calculations.

[lanjh]

I am confronted with the question, too. We just want to investigate the surface migration. But the crystal structure is significantly distorted if all atoms are allowed to relax. Is this reasonal ?

[admin]

If you calculate the equilibrium structures of the final and initial state of your reaction path (the scf-calculations of the end-points), you can of course choose 'selective dynamics' to keep the positions of some of the atoms frozen.
For the NEB method, the method itself restricts the relaxation of the atoms, because the spring force is decoupled from the relaxation of the images perpendiculaar to the path (please read some literature on the NEB method). It is therefrore not reasonable to give additional restrictions in POSCAR.

[tjf]

It seems to me to be entirely appropriate to freeze certain atoms in certain situations. For example, when looking at adsorbate behaviour on a metal surface, to reduce truncation effects in the rigidity of your surface.

Let the chemistry guide you, and be aware of the approximations you're applying!