UT theoretical chemistry code forum

Hello, everyone,

I am running an NEB calculation using the Austin versioned quick min (default setting); while it runs reasonably, but i am puzzled when i looked at the progression of the force using nebconverge.pl script and saw a very periodic force spikes of every7-8 ionic steps. I am wondering what is the reason for these spikes and how could i get around it by redefining some of the parameters?

I was digging into old posts under the impression that there was a discussion about this issue during an NEB calculation; but i could not find it. Any insight would be helpful.

thanks

Kai

It would be interesting to see how large these spikes are and a plot of convergence (force vs iterations).

Quickmin uses a molecular dynamics algorithm so that atoms accelerate along the force. The algorithm can overshoot the minimum, at which point the velocity is zeroed. There can be an increase in the force at this point, but not a spike.

It would also be good to check if the spikes go away when the time step (POTIM) is reduced.

Could you elaborate a bit more regarding making smaller dynamic step in the IOPT=3 setting? I belieave you are referring to the tag of Timestep? Do you think it is a good idea to reduce Maxmove?

This is a typical curve i got using nebconverge.pl; with a spike initially at 0.5 eV/A, and later to 0.03 eV/A. The calculation converged well till force down to 0.05 ev/A, but i am worried the spike issue could cause further problem if i am aiming at 0.02 eV/A

The parameter i used as follows, the Austin versioned quick min with default setting

IOPT = 3 #Parameter Default Description
MAXMOVE = 0.2 #Maximum allowed step size for translation
TIMESTEP = 0.1 #Dynamical time step

I would be very surprised if your step size was hitting the MAXMOVE parameter. In the Austin version of quick-min the TIMESTEP is the correct flag to change. Making this smaller will keep your images from building up velocity as quickly and they will have a smoother interaction with the potential.