first off, please always copy the the list on your replies.
thanks.
Thanks for that Axel.
Actually we are using the fix nve/sphere routines for point dipoles, which I still can't seem to find a reference for.
why did you ask about the rigid integrator, if you wanted
to know how fix nve/sphere is working? it makes it easier
to give an accurate answer, if the question is precise. 
I have the references for a complex rigid body, but what about a rigid body sphere with orientation properties such as a point dipole? There appear to be no ref's for this. Looking at the documentation for a complex rigid body (fix_rigid), I seem to infer this would handle composites of particles that would include a dipole, however stated is the following:
"Each body must have two or more atoms"
yes. that is what in "MD slang" a rigid body is, i.e. an object build
from multiple points that are integrated together and don't change
their relative geometry.
So my question is, how is the fix nve/sphere integrated for the dipole? And is it possible to use the "fix rigid" code to simulate single spherical dipoles...i.e. is the above statement hard and fast, which would appear on the surface to be a more rigorous integration protocol.
have you _looked_ at the code in fix_nve_sphere.cpp
the three steps of the velocity verlet integrator are in
::initial_integrate (step 1 & step 2)
and ::final_integrate(step 3). the former updates c.o.m. velocities
from forces for a
half step and similar the rotational velocities (omage) from the
torque. then positions
are updated from velocities and mu from omega (through d_mu/dt = omega cross mu)
and finally another half step for c.o.m. and rotational velocity updates.
so the major difference to fix rigid would be how the moment of inertia
is determined (which is rather simple for spherical particles).
does that answer your question?
cheers,
    axel.