Dimer run converging to the initial state

[mmg016]

Dear Prof. Henkelman and users,

I am trying to find the transition state of a simple reaction such as CH --> C + H on a Ni(111) slab. For this, I have optimized CH on the slab ( attached as POSCAR_01.tar.gz to this post) and I also optimize co-adsorbed 'C + H' ( attached as POSCAR_02.tar.gz to this post.

Then, with the intention of running a dimer run, I created a MODECAR using these two POSCARs as initial and final position. I then placed the initial POSCAR in the calculation folder with the MODECAR and started the dimer run.

However, after running 200 odd ionic steps, dimer is exactly converging to the very initial state. I have attached the whole dimer calculation folder for you to take a look.

Similar situation is occurring also in the other reactions such as CO ----> C + O after around 400 ionic steps and 10 hours calculation time.

What am I doing wrong here?

Thanks,
MG

[graeme]

If you start a dimer calculation from a minimum, as you have done, it will search for a saddle in the general direction of your initial MODECAR file. The search that you posted has done just that. But, while you displaced the H atom, thinking of OH dissociation, the dimer calculation converged to a state with OH on a bridge site which I expect is a saddle for OH hopping from one hollow site to a neighboring hollow site.

If you want to find a specific reaction pathway to a specified final state, you should used the nudged elastic band method.

If you have a good guess for a saddle structure (which can be generated from a rough NEB calculation) then you can also use the dimer method to efficiently converge to that nearby saddle.

[mmg016]

Dear Prof. Henkelman,

Thank you for your advice. I'll try to implement it now. This is what I plan to do now:

- I'll run a rough NEB calculation between initial_position and final_position POSCARS.
- After running vfin.pl on the result, I'll use neb2dim.pl script to create a dimer run folder and run the Dimer hoping to converge to a transition state.

Could you please confirm if this is a right strategy.

Thanks again,
MMG