well, i deliberately didn't mention pair_airebo.cpp,
since it is quite convoluted and very difficult to
read for somebody that has yet to learn the basics
about programming pair styles in LAMMPS.
whether an existing potential file is a good template
is less defined by the material it is used for, but
rather the logical flow of data.
cheers,
axel.
Thank you for the replies.
I will first read the article by Plimpton, to get a more basic understanding of the implementation.
As for the EAM approach, I think it is very different because it seems something like a mean-field approximation, but I am not sure, I will have a look at it later.
LCBOPII has indeed many similarities with the AIREBO potential, so I think also the data-flow should be similar.
Both have different interaction ranges, short-range bond order vs. long-range LJ-like.
Both are bond order potentials, which depend on local environment (e.g. the number of neighbors of each atom).
But again, I should read up on the basics before I continue.
I started implementation without worrying about speed or parallelization, just to get something to work first, but it will probably pay off in the end, as you said, to go back and think first 
Many thanks for the suggestions,
Jaap
The simplest bond-order potential in LAMMPS is Tersoff.
Steve