how to avoid dimer search leading to TS with high barrier

Vasp transition state theory tools

Moderator: moderators

Post Reply
licondam
Posts: 1
Joined: Tue Apr 09, 2019 6:00 am

how to avoid dimer search leading to TS with high barrier

Post by licondam » Tue Apr 09, 2019 6:02 am

Hello guys,

In a adaptive KMC step, a number of dimer searches are launched to find possible events.
But, I find that some dimer searches converge to improbable transition states with very high barriers (tens of eV).
Even worse is that these dimer searches often far from the initial state and also take a lot of computational time to converge.
I am wondering how to avoid those searches if I don't know the final states.
Thank you.

graeme
Site Admin
Posts: 1832
Joined: Tue Apr 26, 2005 4:25 am
Location: University of Texas at Austin
Contact:

Re: how to avoid dimer search leading to TS with high barrier

Post by graeme » Tue Apr 09, 2019 1:25 pm

There are a few strategies to focus on low barriers.

One is to make small local displacements around atoms that are likely to have low barriers. In materials, this can be surface or under-coordinated atoms. In other systems, you might need to use chemical intuition to help focus on the most reactive environments.

Second, you can specify a maximum energy at which saddle searches are aborted. This is not a great solution in that it will turn high energy saddle searches into aborted searches, but it can save computational time.

But fundamentally, we do not have good systematic ways to control the energy of saddle found with the dimer method. That said, there are complementary method that do.

One is the biased gradient squared descent method, http://henkelmanlab.org/pubs/duncan14_194102.pdf , in which you can systematically look for saddles with increasing energy above a local minimum.

Second is to use high-temperature molecular dynamics (MD) to find reaction mechanisms, and then a single or double ended saddle search method (or both) to find saddles corresponding to those reaction mechanism. While MD will be more expensive for finding any one saddle, as compared to the dimer method, it can be more efficient for finding low energy saddles because MD identifies mechanisms with a probability proportional to the rate of the reaction. A description of this method can be found at http://henkelmanlab.org/pubs/chill14_214110.pdf .

A general discussion of the tradeoff between methods can be found in review papers including http://henkelmanlab.org/pubs/henkelman17_199.pdf and http://henkelmanlab.org/pubs/trochet19.pdf .

Post Reply