Problem with NEB calculation

[xsf007]

Dear all,

I used the improved NEB method (G. Henkelman, et al. JCP 113 (2000) 9978) and the quick-minimum to calculate a foreign interstitial atom diffusing from a octahedral site to neighbor one in a bcc lattice with a empirical potential calculation. This is a very simple diffusion path, and the saddle point is at tetrahedral site. However, there is a problem confusing me. In my calculation,when the value for force convergence is set to 0.001eV/A, the result seems resonable.

image energy force(J/m)
2 -1743.6695 1.78918E-13
3 -1743.6521 2.98117E-13
4 -1743.6212 4.17872E-13
5 -1743.5815 4.52411E-13
6 -1743.5273 5.38239E-13
7 -1743.4540 7.12308E-13
8 -1743.3574 1.97888E-13
9 -1743.2327 6.18925E-13
10 -1743.0905 2.41768E-13
11 -1742.9550 9.44083E-13
12 -1742.8593 5.02526E-13
13 -1742.8239 2.09141E-14
14 -1742.8593 5.13605E-13
15 -1742.9550 9.31966E-13
16 -1743.0904 2.63389E-13
17 -1743.2326 5.94224E-13
18 -1743.3573 2.05628E-13
19 -1743.4540 6.90701E-13
20 -1743.5273 5.43056E-13
21 -1743.5815 4.5571E-13
22 -1743.6211 4.12946E-13
23 -1743.6520 2.90183E-13
24 -1743.6695 1.74062E-13

While I set the force convergence as 1.0E-4ev/A, the energy of the images drop quickly to approach that of endpoints and the diffusion path is crash, even some points on diffusion path run out of the range between the endpoints. I don't know if this phenomenon is reasonable or not. Please give me some advice.

Another two question: (1) Can the NEB method be directly used to calculate the diffusion path and energy barrier when an external force is imposed on the host lattice along a certian direction . (2) Under the effect of elastic force, is every image stable, can the force of every atom be close to zero during the minimization process?

Thanks!