[Pw_forum] NEB Convergence Problem
victor.bermudez at nrl.navy.mil
Mon Nov 7 17:28:44 CET 2011
Thanks for your very helpful comments. I think that there may be a fundamental problem
with what I'm trying to do. What I'm looking at is a "simple" surface reaction where an HF
molecule reacts with an Si-OH on the surface of hexagonal SiC to form an Si-F site and
release H2O. In other words, HF + Si-OH --> Si-F + H2O, where Si-OH is a surface species.
Beginning with the IS, which is a free HF molecule a few Angstroms from the surface, the
HF first goes through a weakly-physisorbed state and then the TS. Once the reaction has
occurred the H2O goes through a weakly-physisorbed state before leaving as a free molecule
(the FS). There may also be more than one physisorbed state for the HF and the H2O. These
are very likely real (not artifacts of the calculation) and produce the structure that you
noticed to either side of the TS in the reaction path I sent earlier. The potential energy
surface I'm faced with can't really be modeled as a simple IS --> TS --> FS progression. I
don't think that it would be acceptable to force it to fit such a simple scheme, even if
that were possible.
In my initial runs, my IS was a guess for the first physisorbed state of HF, and my FS
was a guess for the last physisorbed state of the H2O product. This was done to avoid
exactly the problems that you've described. The problem is that, if these are not good
guesses then, even if these are optimized, the reaction path is forced to begin and/or
terminate incorrectly. Choosing the IS and FS as the free molecules avoids this.
So, at this point, my question is how can I deal with this complex energy surface in the
context of NEB, as implemented in QE ? If I understand correctly, the only solution is to
"zero in" on the TS. This means using as the IS the last physisorbed state before the TS
and for the FS the first physisorbed state after the TS. I would obtain these using the
appropriate images in my present reaction path, so they should be good guesses. This would
in effect isolate an IS --> TS --> FS segment centered on the true (highest-energy) TS in
the path. Is this the correct thing to do ? My activation energy would then be the
difference between this TS and a system consisting of isolated free HF and bare
OH-terminated SiC (which I can easily compute).
Victor M. Bermudez
U.S. Naval Research Laboratory
4555 Overlook Ave., S.W.
Washington, DC 20375-5347
E-mail: victor.bermudez at nrl.navy.mil
More information about the users