Increasing images requires more computer resources.
If we have multiple TS between locally stable images, we may split the one whole NEB run into separate NEB runs.
However, how can I increase the resolution if I need more images, i.e., the high resolution?
Can I run NEBs as if the following?
1. Run NEB with 4-8 images.
2. Separate paths if there are more than two locally stable images because the initial and final images are at least locally stable.
3. For each separate image, insert interpolated images between the locally stable and TS images (not between two locally stable images). And run NEB without the climbing method.
Is step 3 possible?
I found some comments about the steepest descent method, but how can I apply it?
Moreover, the steepest descent method in the official VTST site seems to purpose on finding TS, not the resolution.
How to increase resolution of images in NEB method?
Moderator: moderators
Re: How to increase resolution of images in NEB method?
You can certainly add more images by interpolation. It is best if you can try to keep equal spacing, so for example, going from a 4 image to a 9 image calculation. You can always keep the climbing image on.
Re: How to increase resolution of images in NEB method?
graeme wrote:
> You can certainly add more images by interpolation. It is best if you can
> try to keep equal spacing, so for example, going from a 4 image to a 9
> image calculation. You can always keep the climbing image on.
My purpose is to get the proper structures from the high-resolute images during the transition instead of the energy barrier.
I'm using NEB with VASP program.
I thought that each image will be distributed in parallelized nodes to get the MEP with increasing memory.
My cluster is limited in computer resources for more than 32 images, so can I do the NEB without parallelization to save the memory?
I mean NPAR=1 tag in INCAR file. I don't know how parallelization works with NEB.
Moreover, can I believe the calculated intermediate structures describe the real intermediate structures? Because I hope to investigate other intermediate properties instead of the intermediate energies.
> You can certainly add more images by interpolation. It is best if you can
> try to keep equal spacing, so for example, going from a 4 image to a 9
> image calculation. You can always keep the climbing image on.
My purpose is to get the proper structures from the high-resolute images during the transition instead of the energy barrier.
I'm using NEB with VASP program.
I thought that each image will be distributed in parallelized nodes to get the MEP with increasing memory.
My cluster is limited in computer resources for more than 32 images, so can I do the NEB without parallelization to save the memory?
I mean NPAR=1 tag in INCAR file. I don't know how parallelization works with NEB.
Moreover, can I believe the calculated intermediate structures describe the real intermediate structures? Because I hope to investigate other intermediate properties instead of the intermediate energies.