NEB irregular intermediate image spacing

[miguel_san-miguel]

Dear VASP Team & Community,

I conducted a NEB run involving 8 intermediate images to study the dissociation of a single H2 molecule on a pristine In2O3 (111) surface. The stability of both the initial and final positions was verified through phonon calculations. I initially preconverged with 5 irreducible k-points (from a regular mesh) and then run again with 10 (i.e. the previously determined optimized minimum-size mesh).Below is the input configuration used for the simulation:

ISTART = 1
ISPIN = 2
LREAL = Auto
ENCUT = 500
GGA = PS
PREC = A
IVDW = 12
LWAVE = F
LCHARG = F
LDIPOL = T
IDIPOL = 3
DIPOL = 0.5 0.5 0.5 #surface is at the center of the simulation cell with 7.5 Å vacuum above and 7.5 Å below
ISMEAR = 0
SIGMA = 0.01
NELM = 600
EDIFF = 1E-06
NSW = 350
IBRION = 3
POTIM = 0.01
ISIF = 2
EDIFFG = -0.05
SPRING = -5.0
IMAGES = 8

This was the converged result:

Image Coordinate Energy
0 0.000000 0.000000 -0.005403 0
1 0.575538 0.013775 0.000000 1
2 1.155446 0.037795 0.000000 2
3 1.752447 0.044760 0.000000 3
4 2.393366 0.053194 0.000000 4
5 3.026822 0.060578 0.000000 5
6 5.285085 -0.283756 0.000000 6
7 5.904995 -0.428009 0.000000 7
8 6.519384 -0.606187 0.000000 8
9 7.129280 -0.571166 -0.024608 9

The intermediate images were generated by a linear interpolation with minor tweaks to correct unphysical distances between the H atoms. The initial adsorption position was stable and obtained by a Monte Carlo thermal annealing run.
I subsequently performed a dimer calculation using VTST script that interpolated between images 5 and 6 but it converged in a single step with no significative change.

I would greatly appreciate your assistance with the following concerns:

How confident can I be that there is no transition state of higher energy between images 5 and 6, given the substantial distance spanning them? This occurrence is the first of its kind for me, as my previous Nudged Elastic Band (NEB) runs consistently yielded somewhat regularly spaced images. Could I performed a "sub"-NEB with images 5 and 6 as initial and final positions, even though they are not minima ?

Another issue is that image 8 exhibits lower energy than image 9. However, I came across information in DOI:10.1063/1.3684549 suggesting that regular NEB calculations can lead to such behavior, in contrast to single-surface NEB (ssNEB). Considering the relatively low energy difference, I initially disregarded it. Should I reconsider? If necessary, I can provide a compressed file (.rar) containing the calculation data.

Appreciate your attention and eventual counsel,
Sincerely,

[svijay]

Dear miguel_san-miguel,

Suggestions to your comments:

How confident can I be that there is no transition state of higher energy between images 5 and 6, given the substantial distance spanning them? This occurrence is the first of its kind for me, as my previous Nudged Elastic Band (NEB) runs consistently yielded somewhat regularly spaced images. Could I performed a "sub"-NEB with images 5 and 6 as initial and final positions, even though they are not minima ?

Since the energy differences from the initial state (0) to the highest energy state (5) are rather small in comparison to the difference in energy between the initial and final state, it might be worthwhile investigating the possibility of a higher energy state between images 5 and 6. One approach would be to increase the total number of images. Since this approach is somewhat computationally cumbersome, an alternative would be to try a climbing image NEB provided by the VTST tools (which are not maintained by VASP). A good place to start to see how to perform CI-NEB calculations is here: https://theory.cm.utexas.edu/vtsttools/ ... bing-image

Another issue is that image 8 exhibits lower energy than image 9. However, I came across information in DOI:10.1063/1.3684549 suggesting that regular NEB calculations can lead to such behavior, in contrast to single-surface NEB (ssNEB). Considering the relatively low energy difference, I initially disregarded it. Should I reconsider? If necessary, I can provide a compressed file (.rar) containing the calculation data.

Given the relatively small energy difference between 8 and 9, it may be alright to neglect it (especially if you are after an energy difference between the initial state and the transition state, i.e. difference between state 0 and the state with the highest energy) as long as the structures do not change very much between 8 and 9. If you performed an independent relaxation for structures 8 and 9 - do they relax to the same structures with the same energy?

Sudarshan

[miguel_san-miguel]

Dear Sudarshan,

Thank you for your helpful insights and quick response.
I'll try the CI-NEB next, starting from the converged NEB run, and hope to soon report the results. As for your question, when I relaxed image 08, constrains off, it falls back into the image 09, with same energy.
One aditional info that I forgot to provide is that phonon calculation for the image 05 shows exactly one imaginary frequency, whose vibration is associated to the reaction coordinate, as one would expect for a transition state. So I'm confident that 05 is a TS for the reaction - whether is the only one I'll wait to the CI-NEB result.

Regards,
Miguel