|
We infer the star formation rates in dark matter halos at different redshifts from halo merger histories expected in a Lambda CDM cosmology using Bayesian inference and Bayes ratios to restrict the model complexity, constrained to match the observed stellar mass/luminosity functions of galaxies at different redshifts and the local cluster galaxy luminosity function, which has a steeper faint end than that of field galaxies. The only other assumptions that we make are that the star formation rate of central galaxies depends on the halo mass and redshift and that when a galaxy becomes a satellite its star formation rate is quenched exponentially and it can eventually merge with the central galaxy on a dynamical friction timescale. We find that 1) the star formation in the central galaxies of high mass halos (>10e12) has to be boosted at high redshift beyond what is expected from a simple scaling of the dynamical time; 2) below z=2 the star formation in halos below 1e11 must be quenched, which is not directly expected in standard stellar feedback models and is most easily explained by some form of preheating, and implies that there is a significant old stellar population in present-day dwarf galaxies with M_star < 10e8 and steep slopes for the high redshift stellar mass and star formation rate functions 3) the stellar mass of galaxies assembles in one of three ways depending on halo mass: > 1e12 the galaxies assemble through mergers and should hence have a spheroidal morphology and between 1e11 and 1e12 (e.g MW) it assembles slowly and at z>2 has less than 5% of its mass in place, which has extreme observational consequences. |
|