I want to print out the force (fx fy fz or overall force) of several atom groups during MD simulation. I know there is a good command for this “compute group/group”, but I can’t use this command since I use multi-body potential (reax/c) My atom groups consist of different types of atoms, and each groups have different size and number of atoms.

Is there any compute or fix command to calculate the fx, fy, and fz of group of atoms? I already have per-atom fx fy and fz by dump command, but this is not what I want to get.

In addition to that, you can look into this recent work: http://arxiv.org/pdf/1503.06565.pdf . Authors derive a universal expression of the pairwise force between two atoms, for a general classical many-body potential (see equation 57 in particular), validated numerically. However, that’s not implemented in LAMMPS and you may want to double check if it actually applies to ReaxFF.

Is it possible to run the main simulation by Reax FF, and then rerun it by defining another FF (a pairwise additive potential) to calculate the interaction energy?? At least this can give an approximation of the energy.

Is it possible to run the main simulation by Reax FF, and then rerun it by
defining another FF (a pairwise additive potential) to calculate the
interaction energy?? At least this can give an approximation of the energy.

yes, if there is a separation between the two groups, the
approximation can be quite good.
if there are bonds between the two groups, it may be tricky.

1. The reason why there is no group/group feature for non-pairwise
additive potentials is that the many-body interaction energy between
groups is not a well-defined mathematical quantity.
2. Since you started out asking about forces and later switched to
energy, you are clearly not looking for a specific mathematical
quantity. Rather, you seem to after the subjective concept of whether
two groups are interacting or not.
3. Interaction between two groups is strongly correlated with spatial
proximity of member atoms.
4. So what you really need is a way to measure whether or not two
groups have member atoms that are near each other, a kind of minimum
distance measure.
5. This is easily calculated in a post-processing step. There might be
a way to do it in LAMMPS on the fly, but I'm not sure it is worth the
trouble.