Page 1 of 1
Pre-factor computation problem
Posted: Tue Nov 08, 2011 6:15 pm
by Neutrino
I tried to do a convergence study for the value of the pre-exponential for a diffusive jump as a function of the parameter POTIM. I varied POTIM from 0.001 to 0.012 with a step of 0.001. For all calculations: EDIFF = 1E-08, PREC=High. Out of 12 simulations only 4 with POTIM values: 0.002,0.003,0.011,0.012 , gave me a saddle point that has one imaginary vibrational frequency. The other 8 cases indicated that the saddle configuration has two imaginary frequencies. I have obtained the saddle point configuration from a well converged (EDIFFG=-0.01) CLIMBING IMAGE NEB. Any thoughts on why this happened? Thanks!
Re: Pre-factor computation problem
Posted: Tue Nov 08, 2011 10:34 pm
by graeme
Are there any zero modes in your system, such as translation or rotation?
Try converging the force at the saddle a bit more, e.g. to 0.002 eV/Ang. You can do this with your original NEB, a single image NEB, or the dimer method using the neb2dim script, (in case these calculations are expensive).
Re: Pre-factor computation problem
Posted: Wed Nov 09, 2011 1:03 pm
by Neutrino
It is a diffusive jump in bulk and hence there are three imaginary frequencies corresponding to translation. So in the initial state I obtain 3 imaginary frequencies and in the saddle there are 4 or 5 depending on POTIM.
I will try to follow your suggestion and see what happens when I repeat with one value of POTIM that gave 2 imaginary frequencies. Thank you!