metastable state vs. convergence speed

[gump_813]

Hi, if a metastable state shows up in an NEB run, is it going to speed down the convergence?
if yes, any suggestions when doing such calculation?
Thank you.

[dano]

By metastable do you mean the NEB found an intermediate minima?
- If this is the case I would stop the calculation, converge the intermediate, sand then start two NEBs connecting state 1 to the intermediate and then the intermediate to state 2.

If by metastable you mean you have found a shoulder and along this shoulder you have low curvature modes the NEB will converge slower (especially with quick-min). This would be a case in which I would make use of some higher order optimizers that make use of the curvature. (IOPT=1, IOPT=2)

Dan

[gump_813]

I mean the first, thank you.