I don't think ReaxFF should be "seeing" any of the atoms with NULL types.
I.e. the neigh
list it is passed should only contain atoms that are the ReaxFF types.
For both
the I and J indices. It should also never be doing a loop over all nlocal
atoms, only
over what is in its passed neigh list, for both the I,J indices.
Yes, that could be the problem in reax/c implementation, particularly in
reaxc_force.cpp that the hydrogen bond neigh list is looping over
nlocal+nghost.
If you do fix 1 solute npt, then the ReaxFF atoms will not be time
integrated, and thus
not move. So that is probably why the problem goes away - the config
never changes.
No, it is the opposite - ReaxFF atoms are in the group solute. So when
only ReaxFF atoms move there is no hbondchk error.
Ray