Dimer method questions

Message

gjenness · #1 Post by **gjenness** » Tue Feb 07, 2012 9:53 pm

I am trying to find the transition state and barrier heights for a couple reactions occurring on a metal surface (eventually I would like to do this on a nanoparticle), and from a couple papers I have read and my experiences with NEB, the dimer method of Henkelmann and co-workers seems to be the ideal method.

I am noticing that there several papers where this method is done by only taking the initial and final states, and by running the dimer method, the transition state then pops out. As an example, the recent paper by Kapur et al. (J. Phys. Chem. C 114, 10171 (2010)) states:

"The dimer method [44] was used to ﬁnd the transition state (TS) structures for reactions examined in this study, and the initial direction along the dimer was generated based on the reaction products."

My questions are:
1) In doing this, does one start at the initial state, and then just simply generates the vector (i.e. using the VTST script modemake.pl)?

2) Or does one instead generate an initial guess (by say, interpolating between the initial and final states), and then generating the vector from that guess to the final state (or initial to final)?

Thanks for any help!

graeme · #2 Post by **graeme** » Tue Feb 07, 2012 10:29 pm

If you have an idea of the reaction mechanism, for example, if you know the product state, it would be best to estimate the saddle geometry using interpolation. Setting the initial mode along the vector between reactant and product is also a good way to initialize the search.

Starting the dimer calculation at the reactant state minimum is not a good idea if you want to look for a particular reaction mechanism. It is appropriate if you want to search the potential energy landscape for many different possible reactions.

gjenness · #3 Post by **gjenness** » Wed Feb 08, 2012 1:48 am

For the reactions I am doing, I have both an initial and final state-the only thing I do not have is the transition state. For example, I have reproduced NEB results for CO dissociation on the Rh (111) surface by interpolating between the initial and final as an initial guess, and then taking my vector to be along my guess to the final state (using the VTST modemake.pl script).

However, when I tried the same trick for the CH+H --> CH2 reaction on Rh (111), I get nonsense-it converges to an unrealistic saddle point.

Would it be more optimal to define my vector from initial --> final, or from TS guess --> final?