>You don't need huge numbers of particles to get realistic looking results.
^^^^^^^
>With 2000 particles and a split hamiltonian to reduce the number of force
>calculations a simulation need only take a hour or two on a 100 mips
>workstation. I attend lectures in between the simulations to fill up the time.
"looking" - that's the problem. I don't want to criticize your work,
maybe just add a little cold water from the astronomer's perspective.
There are many nice pictures of simulations that look like interacting
galaxies (dating back to Toomre & Toomre 1972). But an observer can
remain suspicious - there are enough inputs into any interaction
simulation that you can tweak to eventually make a nice picture,
but by then it's hard to say what that tells us about what's really
out there.
One big problem is that any kind of simulation using particles is
essentially a Monte Carlo technique. (effectively, you are monte carlo
sampling the distribution function, then integrating the particle paths,
which are the characteristic curves.) This is intrinsically
inefficient, and the biggest problem is the sqrt(N) particle noise.
This can make it real hard to resolve perturbations on a background
density, such as spiral arms. N-body simulations are basically still
not sufficiently fine to settle questions of spiral arm formation ...
which ties into things like disk stability, disk heating etc.
(See those articles by Sellwood which Chris Mihos referenced ...)
Then one takes two "disk galaxies" (neither of which quite behaves like
a real disk galaxy) and whacks them together - particles everywhere;
the problem isn't the simulations, it's interpreting them.
I guess my opinion boils down to this: we don't fully understand the
dynamics of isolated galaxies yet, interacting galaxies are yet
another can of worms. Worth studying but not yet clear what we're
learning.