Reference files for a successful NEB calculation

Vasp transition state theory tools

Moderator: moderators

Post Reply
Ricardo_R_JR
Posts: 3
Joined: Mon Mar 19, 2018 1:58 pm

Reference files for a successful NEB calculation

Post by Ricardo_R_JR »

Like others here I have problems with NEB convergence. I have been following the forum for a long time and have tried all advices given here. Even so, I never got a satisfactory result. If possible, I would like to have a reference file with a successful NEB calculation.

Thank you very much.

Ricardo.
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: Reference files for a successful NEB calculation

Post by graeme »

I have given many examples of how to converge NEB calculations as well as specific instructions about how to adjust settings to improve convergence. If you have a specific example of a problem calculation, you can post it. I will assert, however, that in the majority of such posts, I show that my standard advice leads to convergence.
Ricardo_R_JR
Posts: 3
Joined: Mon Mar 19, 2018 1:58 pm

Re: Reference files for a successful NEB calculation

Post by Ricardo_R_JR »

One conclusion from "J. Chem. Phys. 128, 134106 (2008); doi: 10.1063/1.2841941" is:

"[...] global implementation of the L-BFGS method, in which interimage curvatures are included in the memory of the optimizer, is the most efficient approach. Also, we show that the FIRE optimizer tends to be more efficient than the quick–min and is the preferred optimizer that does not (explicitly) rely on curvature information.".

Global L-BFGS setings:
IOPT = 1
LGLOBAL = .TRUE.

FIRE setings:
IOPT = 7
TIMESTEP = 0.1

When I tested, I found FIRE more efficient than global L-BFGS (using default parameters). Ok, maybe I needed to chose better parameters but I only converged one image. When I tried with 5-7 images NEB runs do not converged even after 1000 steps. In this forum it is comum to find the advice to use IBRION=3 (Damped molecular dynamics) with "TIMESTEP = 0.01" for VASP without VTST Tools. With VTST Tools in the figure 3 we conclued (doi: 10.1063/1.2841941) the QM is the worst one. It is comum to find here input files with QM:

# NEB:
IMAGES = 6
LCLIMB = .TRUE.
SPRING = -5
IOPT = 3 # QM = Quick-Min
TIMESTEP = 0.01

I read the topics in this forum and the articles on NEB and other methods but I continue without knowing which strategy works. Because of this I asked for reference files. It could be a simple example, some file from a calculation that has already been published or a short tutorial.

Thanks for your time.
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: Reference files for a successful NEB calculation

Post by graeme »

Here are some posts where users have had problems with convergence of NEB calculations. In these cases I have provided suggestions and also changed settings to improve convergence. The calculations are posted - you can consider any of them as reference examples that you can learn from.

http://theory.cm.utexas.edu/henkelman/f ... f=2&t=3407
http://theory.cm.utexas.edu/henkelman/f ... f=2&t=3357
http://theory.cm.utexas.edu/henkelman/f ... f=2&t=3176
http://theory.cm.utexas.edu/henkelman/f ... f=2&t=2887
http://theory.cm.utexas.edu/henkelman/f ... f=2&t=1644
http://theory.cm.utexas.edu/henkelman/f ... f=2&t=1516
Ricardo_R_JR
Posts: 3
Joined: Mon Mar 19, 2018 1:58 pm

Re: Reference files for a successful NEB calculation

Post by Ricardo_R_JR »

Thanks for your time!
Post Reply