Page 1 of 1

NEB and IBRION

Posted: Tue Jan 16, 2007 3:19 am
by gutk
I found that NEB method maybe not convergence when using the CG method(IBRION=2) to relax the atom. In the vasp manual, they also tell us that the CG algorithm usually fails to converge. I want to know this is true or not.

Posted: Tue Jan 16, 2007 3:44 am
by graeme
It is true that the vasp CG method does not work (well) with the NEB. The problem is that it uses an energy criterion in its line optimization. The force projects in the NEB mean that the energy is not consistent with the force, so this aspect of the method fails.

It is, however, possible to implement the CG method using only forces. We have done this in our developer code (available through CVS). We are in the final stages of testing it, and should release it within the week. This code allows for the use of CG with the NEB.