CI-NEB calculation with breaking of bond and spin

[wbird]

Dear all,

I have met a problem when doing a CI-NEB calculation upon a dynamical course in which some weak valence bond breaks. The calculation couldn't converge in more than 200 steps with the last nebef.dat like that:

0 0.03846600 -936.56766000 0.00000000 4.8992
1 0.15014900 -936.41586300 0.15179700 4.8669
2 0.09884700 -936.36377600 0.20388400 2.8271
3 1.01469400 -936.04884600 0.51881400 3.9168
4 0.15002900 -936.04535000 0.52231000 3.9338
5 0.04215000 -936.29533100 0.27232900 2.9095

The last column is the mag value in OSZICAR appended by me. As in a course that breaks some valence bonds, it's not strange to see some varience of the mag value, however, the trend of the mag value who goes down and up and down again seems weird. The more strange thing is that the energy of the images seems to have some relationship with the mag value. Comparing the right-most two columns, it can be seen that when the mag is about 2.8, the energy will be about 0.2eV, and when the mag value is 3.9, the energy will be about 0.5eV.

These results make me doubt all the energy values at all, and maybe the changes of the energy values are caused by the calculated mag values, rather than the configuration change of the cluster. I have heard that sometimes the electronic iteration would converge at some exited state in the spined system, so I tried to do some no-spin calculations to examine, but one of the endpoints of this system will become no longer a local minimal site when spin=1, so this method seems to be not fit. Is there any other way to solve this problem?

Thanks very much for the suggestions!