Dear All,
I am having problem in predicting the TS using the dimer method. Because the criteria in dimer method is to check the calculation for zero force, the calculation is always giving a TS state irrespective of the final state. I have different final states, but the TS what I am getting is same for all the final states which is not true. The calculation is forcing the structure to go to the same TS structure, because it is the structure with zero force and its a TS.
How to solve this? The starting energy in the dimer calculation is similar to the one which is obtained in NEB but as the calculation progresses the structure goes to the same structure in all the cases.
Please, any clues or any help.
Thank you
Dimer method and its convergence
Moderator: moderators
You need to say a little more about how you are starting the dimer calculations. The dimer calculation only requires an initial configuration (POSCAR) and an initial guess at the lowest mode (MODECAR). The final state configuration is not an input, unless you are using it to generate one of these two files - are you?
I'm guessing that you are interpolating between an initial state and various possible final states to start a dimer calculation, and all such calculations are converging to the same saddle point.
If you want to find a specific path between known initial and final states, the NEB is a good method. This will tell you at least one pathway that ends up at your final state, and if such paths go through the saddle point that you are finding with the dimer method. The dimer method is more appropriate for converging upon a nearby saddle point, or exploring variable possible mechanisms leading away from an initial state. It seems that you want to try the later, and are starting searches in directions towards known final states. Instead, you can search more widely by making random displacements from your initial state, and use these configurations as initial configurations in the dimer searches. A sensible choice for the modecar is along the direction of the displacement.
One last possibility is to do these searches in a systematic way. We are developing a script for finding the possible mechanisms and rates for leaving an initial state, and then doing kinetic Monte Carlo to simulate the state-to-state dynamics. More information about this can be found at: http://theory.cm.utexas.edu/vtsttools/akmc/
I'm guessing that you are interpolating between an initial state and various possible final states to start a dimer calculation, and all such calculations are converging to the same saddle point.
If you want to find a specific path between known initial and final states, the NEB is a good method. This will tell you at least one pathway that ends up at your final state, and if such paths go through the saddle point that you are finding with the dimer method. The dimer method is more appropriate for converging upon a nearby saddle point, or exploring variable possible mechanisms leading away from an initial state. It seems that you want to try the later, and are starting searches in directions towards known final states. Instead, you can search more widely by making random displacements from your initial state, and use these configurations as initial configurations in the dimer searches. A sensible choice for the modecar is along the direction of the displacement.
One last possibility is to do these searches in a systematic way. We are developing a script for finding the possible mechanisms and rates for leaving an initial state, and then doing kinetic Monte Carlo to simulate the state-to-state dynamics. More information about this can be found at: http://theory.cm.utexas.edu/vtsttools/akmc/
First I have done NEB and found a structure which is close to the saddle point. I took this POSCAR as the starting point for the dimer run. And I have done two types of calculations from this point:
(1) Started the dimer calculation without mentioning MODECAR, and the program assigned itself a MODECAR, as it is suggested by your website
(2) I have generated MODECAR file using the script modemake.pl and giving input files as the InitialState & FinalState poscars.
In case (1) calculations, irrespective of the final state, all the calculations leeds to the same transition state.
In case(2), presently these calculations are running, but looks like they are all directed towards the same TS as it was in case (1).
No, I have not done any interpolation as described by you in your reply.
For your information Iam studying the C-C bond breakeage of ethylene and forming two CH2 species on the surface.
Thank you
(1) Started the dimer calculation without mentioning MODECAR, and the program assigned itself a MODECAR, as it is suggested by your website
(2) I have generated MODECAR file using the script modemake.pl and giving input files as the InitialState & FinalState poscars.
In case (1) calculations, irrespective of the final state, all the calculations leeds to the same transition state.
In case(2), presently these calculations are running, but looks like they are all directed towards the same TS as it was in case (1).
No, I have not done any interpolation as described by you in your reply.
For your information Iam studying the C-C bond breakeage of ethylene and forming two CH2 species on the surface.
Thank you