more NEB

Message

mhyman · #1 Post by **mhyman** » Fri Aug 05, 2005 11:48 pm

First a general question-
This question has been addressed before, but I'm not clear on the answer. How can I check how close a calculation is to convergence?

More specifically-

I have run an NEB calculation with 3 images, and it converged. I added four additional images by linearly interpolating between the images and between the images and the end points. Image 01 has moved such that it is almost exactly image 00. Image 00 is more stable by approximately 0.04 eV. The calculation ran out of time at about 100 cycles and I want to restart it, but I want to know what to do about image 01. Should I remove it, or make a new one linearly interpolated between 00 and 02?

Thanks,
Matt

mhyman · #2 Post by **mhyman** » Mon Aug 08, 2005 5:24 pm

Follow up: I gave the calculation a few more cycles and image 01 became more stable than 00. I did a geometry optimization on that, and it radically reoriented the molecules and was much more stable than 00. It looks like this pathway isn't valid. If anyone can still answer my fist question, I'd appreciate it.

Thanks,
Matt

Brane · #3 Post by **Brane** » Wed Aug 10, 2005 12:23 pm

About your first question, I usually do as follows:
After finishing the calculation with 3 images, interpolate the two end points by 7 immediate POSCARs. Then replace the POSCARs in 02 04 06 with CONTCAR in the previous 3 images. That's say, using the converged 3 images for the continuing calculation.
As for the 2nd question, one possibility is that your image 00 is not the globle mimina, since image 01 is more stable than image 00. If the activation energy between 00 and 01 is very small, the MEP will go like that.

mhyman · #4 Post by **mhyman** » Thu Aug 11, 2005 2:44 am

Thanks for the reply, but I guess I wasn't all that clear in my question. If I have a NEB calculation running, how do I check to see how close it is to converging? Is there a line in the output that I can check?

Thanks,
Matt

vasp_user · #5 Post by **vasp_user** » Thu Oct 24, 2013 7:24 pm

To see how close your NEB run is to converging-- run "nebef.pl" from within the directory that contains the images for your job. The second column from the left [of the output of nebef.pl] can be compared to what you set for EDIFFG (I'm assuming a negative value in INCAR for EDIFFG).

The website for NEB (at least the NEB version I use and I assume you use) discusses nebef.pl at the below website
http://theory.cm.utexas.edu/vasp/scripts/

Also the above website has a link to the NEB forum, I have found posting NEB questions to that forum helps.