Sunday, October 30, 2011
New Parameters for Galacticus v0.9.1
A model such as Galacticus requires several input parameters to be set - these control everything from the cosmological model (which is well known) to the details of feedback from supernova explosions (which isn't well known). Choosing the best values for these parameters isn't easy - their effects on model results are non-linear and correlated. The best approach is to survey the parameter space using a Bayesian approach to constrain the model to fit various data sets. That work is under way, but it's a slow process. Until then, I've found a set of parameters which works reasonably well at matching a variety of data sets. These have been coded into v0.9.1 of Galacticus as the defaults, and are available as an input file here.
The stellar mass function looks like this:
Not too bad - the "bump" around M* is a limitation of the feedback model used, so there's scope of improvement there - it could be better, but I wanted to get a closer to the K-band luminosity function and HI mass function at z=0:
The model does very well at matching the color-magnitude relation of galaxies measured from the SDSS:
Other constraints that I looked at were the Tully-Fisher relation and galactic disk sizes (where this model has the common problems of producing disks which rotate somewhat too fast and having massive disks that are too small - problems that are likely related), the black hole mass-spheroid mass relation (which works quite well), the luminosity function of different morphological classes (which also works well), and the star formation history of the Universe (not bad, although somewhat too high at low redshifts).
There's room for improvement in all of these results - a systematic search of parameter space will either find parameters that give much better fits, or it will show that no parameter values give a good fit. In that case there'll be work to do figuring out what physics is missing or needs to be modelled more accurately.