Star formation in semi-analytic galaxy formation models with multiphase gas

We implement physically motivated recipes for partitioning cold gas into different phases (atomic, molecular, and ionized) in galaxies within semi-analytic models of galaxy formation based on cosmological merger trees. We then model the conversion of molecular gas into stars using empirical recipes motivated by recent observations. We explore the impact of these new recipes on the evolution of fundamental galaxy properties such as stellar mass, star formation rate (SFR), and gas and stellar phase metallicity. We present predictions for stellar mass functions, stellar mass versus SFR relations, and cold gas phase and stellar mass-metallicity relations for our fiducial models, from redshift z similar to 6 to the present day. In addition we present predictions for the global SFR, mass assembly history, and cosmic enrichment history. We find that the predicted stellar properties of galaxies (stellar mass, SFR, metallicity) are remarkably insensitive to the details of the recipes used for partitioning gas into H I and H-2. We see significant sensitivity to the recipes for H2 formation only in very low mass haloes (M-h less than or similar to 10(10.5)M(circle dot)), which host galaxies with stellar masses m* less than or similar to 10(8) M-circle dot. The properties of low-mass galaxies are also quite insensitive to the details of the recipe used for converting H-2 into stars, while the formation epoch of massive galaxies does depend on this significantly. We argue that this behaviour can be interpreted within the framework of a simple equilibrium model for galaxy evolution, in which the conversion of cold gas into stars is balanced on average by inflows and outflows.