Page 1 of 1

NEB and Dimer

Posted: Fri Nov 03, 2006 4:20 am
by lbo
Hi VASP users,

I have two questions.

1. NEB requires two minina for input. Can I use two geomery close to minina instead of the exact optimized geometies?

2. For Dimer method. which directory helds the output of the TS structure, 01 or 02?

Thanks.

Posted: Fri Nov 03, 2006 4:54 am
by graeme
1. Yes, you can use endpoints which are not fully minimized for the NEB. Just remember that the energy barrier you find should be measured with respect to the minimized endpoints.

2. The two images are on either side of (but very close to) the saddle. Practically, you could take either image as the saddle, but to be precise, you can use the interpolate.pl script (from our vtstscripts):

interpolate.pl 01/CONTCAR 02/CONTCAR 0.5

to get an image 1/2 way between the two dimer images.

Posted: Fri Nov 03, 2006 3:58 pm
by hunterdai
I am a new one here, could you please tell me where I can download the the interpolate.pl script from your vtstscripts?

Thanks

Posted: Fri Nov 03, 2006 4:24 pm
by graeme
The scripts can be obtained from:

http://theory.cm.utexas.edu/vtsttools/scripts/

Posted: Mon Nov 13, 2006 3:51 am
by dwyuan
Dear graeme

Can I use the the unstable stucture (or close to TS) as a endpoints in NEB
run? Now, I want to map a reaction path raleted to two complicate molecules on surface using NEB. The stuctures change strongly after reacting. So, it is difficult to use NEB to identy the barrier directly.
I can guess the TS with fixing bond length method, and then I want
to identy the TS using NEB. I want to use dimer method, but
maybe, it is hard to converge. If I can use an unstable structure as
a endpoint, the problem may be done easily.


Thanks !

[quote="graeme"]1. Yes, you can use endpoints which are not fully minimized for the NEB. Just remember that the energy barrier you find should be measured with respect to the minimized endpoints.

2. The two images are on either side of (but very close to) the saddle. Practically, you could take either image as the saddle, but to be precise, you can use the interpolate.pl script (from our vtstscripts):

interpolate.pl 01/CONTCAR 02/CONTCAR 0.5

to get an image 1/2 way between the two dimer images.[/quote]

Posted: Mon Nov 13, 2006 4:16 am
by graeme
I can't see how this is a good idea.

If you have a guess at the transition state, and the reaction mechanism is complicted, I would try first to use the dimer method. If there is no nearby saddle for the dimer to find, you could turn back to the NEB, but I would not suggest using the TS guess as an endpoint. Better would be to find the endpoints of the reaction via a careful minimization. For example, if the bondlength is a fairly good measure of the reaction pathway, you could increase it and minimize to find one endpoint, then decrease it and minimize to find the other. One you have these endpoints, an NEB calculation should be able to find the TS.

To be even more careful, you could create an initial NEB from two linear paths, one from the initial state to the TS and the other from the TS to the final state. This way, you ensure that the initial pathway goes through your guess at the TS. Furthermore, if you use the climbing image, it is not essential that you choose an equal spacing for the initial paths on each side of the TS.

Posted: Mon Nov 13, 2006 5:23 am
by dwyuan
Thanks for you reply!