Total energy of the saddle image sensitive to K sampling

[cenwanglai]

I have converge my CNEB to 0.05 with a single Gamma K-point.

When I switched the saddle image to SCF total energy calculation with more K points (3 * 3 * 1), the latter is higher at ~ 0.1 eV in energy.

And, the total force of scf calulation is seem to be unconverged (for some free atom, the force is ~ 0.1 ).

Should I run the CNEB with KPOINTS as the same as SCF calculation?

[graeme]

yes