People have been constructing star formation models since the 1960's
at least, and it is still a subject of very active research. One
paper I just happened to find in a quick search is by Banerjee &
Pudritz (2006 ApJ 641, 949). One key sentence from the Abstract
reads "Here we report on our three-dimensional, adaptive mesh,
magnetohydrodynamic simulations of collapsing, rotating, magnetized
Bonnor-Ebert spheres, whose properties are taken directly from
observations." This will give some idea of the state of the art in
this field.

Yas, I have also seen models from the danish Niels Bohr Institute now, which
are impressive. Aparently now the models are at a place where they match
observations perfectly.
Is the term adaptive mesh also used for SPH, or do they in fact reconstruct
a tetrahedron mesh continously, and in that case, why?

Of course if all you want is an interesting computer science project,
it doesn't have to be useful for current research.

It is reallyjust an interesting computer science project. In general people
tend to simulate water and I thought it would be interesting to move away
from that and simulate nbody with the different considerations needed :-)


This (and much of the rest of your message) sounds much closer to
galaxy collisions than star formation. That is, if anything, an even
more active "industry" today than star formation modelling. As an
example of the sort of thing that might make a good project, Antunes
& Wallin (2007 ApJ 670, 261) constructed a model of a specific pair
of interacting galaxies. An extract of their Abstract reads "In
N-body/smoothed particle hydrodynamics (SPH) simulations of
AM 0644-741, we recreate the star formation features, as well as the
underlying kinematics." In other words, they are modelling
separately stars, gas, and (probably) dark matter, and they are
superposing on their model some prescription for how many stars form
at a given gas density. This is a very simple project by today's
standards.

Interesting. So they do not simulate the starformation as such but rather
the gas and then they impose some expectations about starbirth... that seems
managable. I was aware that for a plausible starbirth simulation I would
probably need to take quite a few extra factors into account.
I will look at the reference.

The classic paper on the subject is by Toomre & Toomre (1972 ApJ 178,
623). For more, you can do an ADS search either for papers citing
that one or for Abstract keywords. (ADS is at
http://adsabs.harvard.edu , and I think there's a mirror site in
Europe.)

Thanks!

The "softened" potential is important, but particles are ejected in
real galaxy collisions.

Yes, I realize that ejection is not an error in itself. The ejection I was
thinking about was the result of large timesteps in the integration combined
with very close passages of point masses. That will create ejections even
for a 2-body system.

Or go ask at the astronomy
department if your university has one. If not, you check whether
someone in the physics department is doing astrophysics.

Good idea. We have astronomy somewhere here. I should probably go talk with
them about the relevant factors to include in the model as well as things
such as support radius and good particle size etc for this kind of
simulation.

Good luck with your project.

Thanks :-)