zero barrier from NEB calculations

Dear lammps users:

I have a question about minimization implemented in lammps. I have a defective configuration with an interstitial and a vacancy in it. After I relax it with minimization command using either cg or sd method, the defect pair finds an optimal location. Then I use NEB calculations to calculate the barrier when moving the interstitial to the vacancy. However, I always get a zero barrier like this:

3156 9.9485711e-09 7.6750495e-10 4.5723397e-12 9.9485711e-09 4.5723397e-12 0 5.4487092 3.8110102 0 -16642.612 0.066666667 -16642.655 0.13333333 -16642.906 0.2 -16643.173 0.26666667 -16643.483 0.33333333 -16643.992 0.4 -16644.533 0.46666667 -16645.057 0.53333333 -16645.579 0.6 -16646.098 0.66666667 -16646.641 0.73333333 -16647.232 0.8 -16647.695 0.86666667 -16647.951 0.93333333 -16648.046 1 -16648.061

I have tried all kinds of convergence criteria during minimization but the configuration seems stable. So why there is 0 barrier but the system cannot relax to the low energy state? What’s wrong here?

If you are interested, I can provide my input files.

Thank you very much!