I posted a patch to allow the fire and quickmin minimizers to use
a non-zero energy tolerance. This uses an eariler suggestion of
Carolyn Phillips
to skip the energy tolerance check for the first few steps after velocities
are reset to 0. So your problem now runs, though it appears to be not
as good as steepest descent on this problem. That may be b/c you
are using it with ReaxFF with high tolerances and I believe ReaxFF does
not go perfectly to 0.0 at the cutoff.
Aidan may want to comment further.
Steve