Dear Stephen and LAMMPS crew,
I have been using PRD and TAD quite succesfully for some cases, but I'm a little concerned that some events are being selected which do nothing and should not be there. Let me be clear:
a) suppose a simulation is stopped and energy minimization is undertaken to compare with the original configuration. This min is stopped by the stopping criteria maxiter or maxeval.
It is thus not a truly minimized state and can yield an event which would not actually take place.
b) If it is a TAD simulation, the NEB calculation carries the final state into the initial state yielding an energy barrier very close to zero, and it is therefore accepted.
Of course the obvious suggestion is to increase both maxiter and maxeval to ensure this does not happen. I have several situations where increasing both to 10000 timesteps or larger still caused spurious zero energy barrier events.
I would like to suggest a change to the code, so that event which do not exit the min step via etol or ftol are rejected from event checking.
I'm looking for this in the code but it may take some time for me to figure out our to implement this idea, and you guys are much more familar with the internal working of the code.
What do you guys think? Should this be considered?
Cheers, Mathieu