Saturday, March 13, 2010

Galaxy formation as a condensation phenomena

Yesterday there was an interesting colloquium, Simulating star cluster evolution on high-end graphics cards, by a new UQ staff member, Holger Baumgardt.

The fact I am writing a blog post about it should be taken as a compliment and the comments below need to be taken with a large grain of salt, since the subject goes way beyond my expertise. But they may be interesting in terms of how an outsider sees things...

First, a few things I learnt.
Globular clusters are 10-12 billion years old, much older than typical galaxies.
Black holes may be at the centre of most galaxies.
UCDs are ultra-compact dwarf galaxies somewhere between star clusters and galaxies.

The figure below shows a plot of the radii of different objects versus their mass. It is taken from this paper.

Scaling relations for low-mass, hot stellar systems: half-light radius plotted against total mass.The dashed line shows the fitted relation for elliptical galaxies, while the solid lines indicates the median for Galactic globular clusters ( $r_{\rm h}=3.2$ pc) that do not follow a mass-radius relation.

The claim appears to be that there is a qualitatively different behaviour between galaxies (thousands of stars) and globular clusters (millions of stars).

Ben Powell asked an important question: does the data shown in the talk (similar to that above) justify this claim?

It is not clear to me that it does.

If there is a qualitative change when one has more than a million solar masses, than an important theoretical question to answer is why?

A challenge to computer simulations is to then try and reproduce this change. On a regular PC one can simulate the classical dynamics of thousands of stars. But what about a million? Hence, the desire to use graphics cards...

A couple of questions I had (as an ignorant condensed matter physicist):

What happens if one attacks this problem in the continuum limit? (i.e., rather than having a million point like particles one has a continuous mass distribution).
Can one write down a "density functional" type theory?

Are there any analogies to other problems concerning condensation of liquid droplets whether in nuclear physics or low density gases?

Alan Mark asked a good question about the dependence of the simulation results on the choice of initial configuration. No doubt inspired by his experience with molecular dynamic simulations of large biomolecular systems, where I believe this can be problematic.


  1. This is why it's fun to give talks in front of people from different disciplines - when it works, you get really good ideas!

  2. I still don't understand why the expansion of a star cluster would be adiabatic. Firstly they radiate heat away. Secondly they lose mass, which is energy. The process of losing mass would increase entropy, I think.