Saddle point frequencies using IBRION = 5

[KNG208]

Hi,
I'm studying the vacancy assisted diffusion in the Ni system using first principles for the solutes Re, W and Ta. To get the diffusion coefficients, I also need to calculate the effective frequency v* for the solute - vacancy exchange, which is the product of all the N frequencies at the starting point of the transition to the product of N-1 frequencies at the saddle point (1 frequency being imaginary at the saddle point).

However, I'm getting an imaginary frequency for Ta at the saddle point, but all real frequencies for W and Re. The INCAR used is same in all the cases:-

SYSTEM = Ni
ENCUT = 400
EDIFF = 0.000001
ISMEAR = 1
SIGMA = 0.1
ISIF = 3
IBRION = 5
EDIFFG = -0.01
NSW = 60
ISPIN = 2
IALGO = 48
NSIM = 2
NELM = 150
POTIM = 0.001
LREAL = Auto
NFREE = 2

The saddle points have been determined using the Nudged Elastic Band calculations using the same values for EDIFF, EDIFFG. I use a 3x3x3 supercell containing a solute atom, a vacancy and 106 Ni atoms.

What should I do?

[alex]

Well, that would depend on the underlying theory. E.g. TST says omit the imaginary mode.
If you are lacking the imag. mode for W and Re you have to look harder, I'm afraid.

Hth,

alex

[KNG208]

The step size (POTIM) should be as large as possible (to get a better signal to noise ratio) without leaving the confines of the harmonic approximation.
Is the POTIM value of 0.001 too low??

Has anyone worked on a similar system before?