Dear all,
I have performed the CI-NEB calculations to find out the barrier for oxygen diffusion. At first, I did the calculation using 4 images, and the results seem reasonable, and convergence very well. Next, I want to refine and confirm the results via 8 images. I expected the calculations should be smooth, and straight. However, the procedure is very strange. Most time the last image could not find out a good solution, and the energy rises intensively (100eV compared to the final state). I am wonder why. Please give me some information. Thanks in advanced.
If need more information about my setups, please contact me.
with best regards
xianfeng
NEB calculation problem
Moderator: moderators
Re: NEB calculation problem
If you have a converged climbing image with a 4-image band, there is no need to use more images. You might want to minimize from the climbing image to make sure that it connects directly to your initial and final states, but a converged saddle will not change with resolution of the band.
If you add images 1/2-way between those of the 4-image band, I would not expect any problems converging this (9-image) band.
If you are having problems starting from a new linear band with more images, it is likely related to high initial forces. Start with a conservative optimizer, using for example, IBRION=3 with a small POTIM=0.01 until the forces get below 1 eV/Ang on each image. Then you can switch to your favorite optimizer to efficiently converge the band. But again, if you have used a climbing image to find a saddle, there is no need to check for convergence with more images.
If you add images 1/2-way between those of the 4-image band, I would not expect any problems converging this (9-image) band.
If you are having problems starting from a new linear band with more images, it is likely related to high initial forces. Start with a conservative optimizer, using for example, IBRION=3 with a small POTIM=0.01 until the forces get below 1 eV/Ang on each image. Then you can switch to your favorite optimizer to efficiently converge the band. But again, if you have used a climbing image to find a saddle, there is no need to check for convergence with more images.
Re: NEB calculation problem
Dear graeme,
Thanks for your kind help and your quick response.
Yes, I do the more images calculations using the similar procedure you suggested: at first, using IBRION =3 with a small POTIM = 0.01 until the forces get below 0.5 eV/Ang, and then switch to IBRION = 1 with POTIM = 0.20. However, for more images case, the first step could not find out a good solution. The energy of the last image rises intensively (100eV compared to the final state).
I don't know what is going on. Please give me more information.
cheers
xianfeng
Thanks for your kind help and your quick response.
Yes, I do the more images calculations using the similar procedure you suggested: at first, using IBRION =3 with a small POTIM = 0.01 until the forces get below 0.5 eV/Ang, and then switch to IBRION = 1 with POTIM = 0.20. However, for more images case, the first step could not find out a good solution. The energy of the last image rises intensively (100eV compared to the final state).
I don't know what is going on. Please give me more information.
cheers
xianfeng
Re: NEB calculation problem
if you post it, I'll take a look