The physics involved in simulating galaxy collisions can be extremely complex. The most accurate simulations take into account not just points representing stars, but also magnetic fields and invisible dark matter, as well as n
-body interactions allowing the individual stars to interact with each other. These complex simulations are usually carried out on large-scale supercomputers over long periods of time. One of the more interesting aspects of galaxy collisions is that they can create density variations resulting in all kinds of emergent structure. Density waves can develop that lead to star formation from compressed gas clouds.
A couple of years ago, I wrote a Demonstration
that provides a simplified solution to galaxy collisions. This Demonstration is designed to run in real time inside a Manipulate
, so the problem has been simplified by removing n
-body interactions, dark matter, magnetic fields, and so on. Basically, it treats the two galaxies as large point masses with lots of massless test particles orbiting them. The test particles respond only to the two galaxy "centers." In a real galaxy collision, the chances of two stars getting close enough to each other to interact directly is very remote, so it's not too far of a stretch to ignore this effect for a first-order approximation. The more stars that are included in the simulation (by minimizing the star separation parameter), the more intricate the results (and the more computationally intense). In fact, as more stars are added, it becomes easier to see density variations where many test masses cluster together, but it still looks very discrete. Real galaxies, like the Milky Way, can have hundreds of billions of stars. Trying to carry out a point simulation with that many stars becomes a bit taxing on most home systems, and definitely exceeds the time constraints of a real-time dynamic tool like Manipulate. So how can we better visualize these density variations? I decided to try to modify my Demonstration to use one of the new features in Mathematica
9, namely volumetric rendering
. This way, we can simulate the galaxy collisions with fewer numbers of points, but render the results as if there were billions of stars, resulting in a more realistic and informative visualization.