NEB noob seeks tips...

Vasp transition state theory tools

Moderator: moderators

Post Reply
Posts: 6
Joined: Wed Mar 23, 2011 9:22 pm

NEB noob seeks tips...

Post by Adam »

Hello all. Kudos to whoever had the idea for this forum. It seems like a great resource.

I'm new to NEB calculations, and I'm using the NWCHEM suite to perform calculations on transition metal clusters. I'm trying to determine barriers and reaction paths for adsorption of atoms/molecules on the clusters, and I'm trying to figure out how best to set the parameters.

In case no one here has tried to use NWCHEM's NEB stuff, the things you have to set are:

Initial geometry
Final geometry

set neb:nebnew .true. #if .false. then neb is restarted from previous furn
set neb:nbeads 50 #number of beads in chain
set neb:steps 100 #number of optimization steps
set neb:verlet .false. #using crude conjugate gradient optimizer
set neb:stepsize 0.1 #steepest descent stepsize. make this 1/5th of the maximum change in a Pd coordinate/
task dft neb #task dft neb; task mp2 neb

These last parameters were grabbed from the web, and I haven't been able to get anyone from NWCHEM to explain them to me (the documentation isn't always very complete for NWCHEM).

So, my questions are:
1)I have the structure of the optimized products in hand, but is there any constraint on the way that I choose my reactants? Must they have the same symmetry as the final products? I know that for some methods of finding transition states it's necessary to constrain the calculation to have SOME symmetry and that one of the two must at least belong to a subgroup of the other, but I have no idea if that's necessary or helpful in this case.

2)What is the relationship of the stepsize to the total energy difference (or geometry difference) between products and reactants and the total number of beads?

3) Is there anything I can do to prepare the initial and final geometries to make the calculation run smoother or anything I should avoid to make it go badly?

Thanks for any help you can give me,

Post Reply