ISYM=0

[Neutrino]

Hi,

I wonder if some body has tested the effect of allowing symmetry breaking (ISYM=0) on the result of CI-NEB calculations. The parameter is computationally more expensive than the default ISYM=2 and I'm curious to see if it has any effect on the computed barriers.

[graeme]

It can be terrible if you initial guess passes through high-symmetry points. In this case, vasp will zero the forces at this points and prevent the CI-NEB from finding the saddle. I doubt there is any case where using symmetry makes sense in a reaction pathway calculation.

[Neutrino]

Thank you for your reply. Indeed I used to use ISYM=2 due to memory limitation in my computational resources. Apparently it is a dangerous setting for the NEB calculations. Recently I expanded my memory resources and switched to ISYM=0. I'm running some tests now and will post the comparison in few days.

[lgxyz]

[quote="Neutrino"]Thank you for your reply. Indeed I used to use ISYM=2 due to memory limitation in my computational resources. Apparently it is a dangerous setting for the NEB calculations. Recently I expanded my memory resources and switched to ISYM=0. I'm running some tests now and will post the comparison in few days.[/quote]

I waiting for the results. Indeed i use the default ISYM (i.e. 2) in NEB calculations.

[Neutrino]

So I did a comparison between ISYM=0 and ISYM=2 for oxygen vacancy hop in ThO2. I considered both neutral vacancy (F-center) and doubly charged. In both cases I get exactly the same migration barrier n regardless of the symmetry parameter. May be this problem is not the best test!

[graeme]

If there are no high symmetry points, the calculations should be identical. It is only if you have one of these points along the path, the symmetry will be enforced and the configuration will not be able to go to a lower energy state, if that breaks the symmetry.