NEB computation for CH3O + H reaction

[wangjb]

Dear everyone,

I want to compute the reaction barrier of reaction CH3O + H --> CH3OH on Cu(111) surface in order to reproduce others work. I adopt 6 images and firstly converge with the setup:
IOPT=7, timestep=0.01, EDIFFG = -0.1, ENCUT = 300, PREC = normal (decrease computation cost)
after convergence, continue computation with IOPT = 1, MAXMOVE=0.1, but due to the slow convergence of image 4, the job is hang on, cannot proceed to the next step. The following is the nebef.pl results:

0 0.00000000 -122.40063000 0.00000000
1 0.57161200 -122.41354100 -0.01291100
2 2.22259100 -121.89552700 0.50510300
3 1.59835100 -121.19520000 1.20543000
4 1.77959600 -119.83122700 2.56940300
5 0.77448400 -122.22481500 0.17581500
6 0.47699900 -122.37618800 0.02444200
7 0.00000000 -122.52369200 -0.12306200
The attachment is the movie file named rtH3COH-1-281-0220.rar obtained by nebmovie.pl. Could anyone give comments to this path? How to modify the path? Any advice is appreciated, thank you very much.

Jingbo

[kambiz]

Hi,
Sorry, but just a question,
So after using IOPT=7, all forces are less than 0.1 and then after using IOPT = 1, the job is not converging??

Cheers

[graeme]

It is a very long path! It doesn't look too bad, but you might need a few more images, and/or use the climbing image technique so that one image goes to the saddle.

I also see quite a bit of movement in the top Cu layer. I would be a little worried about having only 3 layers -- 4 would be better. Also, it is probably essential to have at least a 2x2x1 k-point mesh.

You might also check to see if there are any intermediate minimum along the path. There could be, for example, a H diffusion process before the hydrogenation reaction. If so, you can minimize to find the intermediate minimum (or minima) and run separate smaller NEB calculations between those to better resolve the saddle.

[wangjb]

[quote="kambiz"]Hi,
Sorry, but just a question,
So after using IOPT=7, all forces are less than 0.1 and then after using IOPT = 1, the job is not converging??

Cheers[/quote]

Yes. I have done some NEB calculations, my experience is that It is hard to converge using IOPT=1 and EDIFFG=-0.03 after convergence with IOPT=1 and EDIFFG=-0.1. There may be some problems in this system configuration. As Prof. Graeme said, I should add some images, check the layer number and inspect the intermediate minimum.

The initial interpolation path looks good, however after NEB calculation, it is strange. Thanks for your attention.

[wangjb]

Dear Prof. Graeme,

I will try again based on your suggestions. Thank your very much.

Jingbo