Modeling Molecular Hydrogen and Star Formation in Cosmological Simulations

We describe a phenomenological model for molecular hydrogen formation suited for applications in galaxy formation simulations, which includes non-equilibrium formation of H2 on dust and approximate treatment of both its self-shielding and shielding by dust from the dissociating UV radiation. The model is applicable in simulations in which individual star forming regions – the giant molecular complexes – can be identified (resolution of tens of pc) and their mean internal density estimated reliably, even if internal structure is not resolved. In agreement with previous studies, calculations based on our model show that the transition from atomic to fully molecular phase depends primarily on the metallicity, which we assume is directly related to the dust abundance, and clumpiness of the interstellar medium. The clumpiness simply boosts the formation rate of molecular hydrogen, while dust serves both as a catalyst of H2 formation and as an additional shielding from dissociating UV radiation. The upshot is that it is difficult to form fully-shielded giant molecular clouds while gas metallicity is low. However, once the gas is enriched to Z ∼ 0.01 − 0.1Z⊙, the subsequent star formation and enrichment can proceed at a much faster rate. This may keep star formation efficiency in the low-mass, low-metallicity progenitors of galaxies very low for a certain period of time with the effect similar to a strong “feedback” mechanism. The effect may help explain the steep increase of the mass-to-light ratio towards smaller masses observed in the local galaxy population. We apply the model and star formation recipes based on the local amount of molecular gas to an output from a cosmological simulation of galaxy formation and show that resulting global correlations between star formation and gas and H2 surface densities are in good agreement with observations. Subject headings: cosmology: theory – galaxies: evolution – galaxies: formation – stars:formation – methods: numerical