I would like to calculate the energy and force acting exclusively between two atoms, not the effective force acting on the two atoms separately. I’m using EAM potential for my simulation. For doing so, if I define two groups that contain the two atoms, and then if I use group/group pair yes command, would that make sense?

This will work, but as documented on the compute group/group doc
page the EAM potential only calculates the pairwise portion
of its interaction in this mode. The embedding term
is effectively a many-body term, so it does not make
sense to partition it into a force between only 2 atoms.

The compute pair/local command will do the same
thing for lots of pairs of atoms at the same time,
depending on how you define the group argument to the
command.

This will work, but as documented on the compute group/group doc
page the EAM potential only calculates the pairwise portion
of its interaction in this mode. The embedding term

i don't think that the documentation is correct. the way i read the
source, the Pair::single() function of PairEAM will use previously
computed embedding information for the whole system.

You’re right. I forgot that someone added the embedding
terms to the single() method for pair EAM. The
doc page should be changed. However it looks
like the returned energy from single is only
the pairwise part, no embedding term.

I think what I said in the previous email may apply
to a pairwise energy (but not forces). I.e.the embedding function
is a nonlinear function of the total density summed over many
neighbors. So you can’t really extract one atom’s
contribution as a pairwise energy. But maybe you can
do that for pairwise forces, and that was the code
added to single(). Would have to think about it a bit more.

A check (via compute pair/local) would be that
the sum of pairwise forces over neighbors will sum
to the total force on the central atom. But I don’t think
that will be the case for summing pairwise energies.

Not sure about whether the implementation for eam in question has a nonlinear embedding function for the electron densities; seems to add the phi on a per atom basis where the eflag block is. As long as the energy function is additive one can in theory define a site energy, like the typical 1/2 convention for pairwise forces, and potentially get away with it.