HI-selected galaxies in models of galaxy formation and evolution

In this work, we study the basic statistical properties of HI-selected galaxies extracted from six different semi-analytic models, all run on the same cosmological N-body simulation. One model includes an explicit treatment for the partition of cold gas into atomic and molecular hydrogen. All models considered agree nicely with the measured HI mass function in the local Universe, with the measured scaling relations between HI and galaxy stellar mass, and with the predicted 2-point correlation function for HI rich galaxies. One exception is given by one model that predicts very little HI associated with galaxies in haloes above $10^{12}{M}_{\odot}$: we argue this is due to a too efficient radio-mode feedback for central galaxies, and to a combination of efficient stellar feedback and instantaneous stripping of hot gas for satellites. We demonstrate that treatment of satellite galaxies introduces large uncertainties at low HI masses. While models assuming non instantaneous stripping of hot gas tend to form satellite galaxies with HI masses slightly smaller than those of centrals with the same stellar mass, instantaneous gas stripping does not translate necessarily in lower HI masses. In fact, the adopted stellar feedback and star formation affect the satellites too. We analyze the relation between HI content and spin of simulated haloes: low spin haloes tend to host HI poor galaxies while high spin haloes are populated by galaxies in a wide range of HI mass. In our simulations, this is due to a correlation between the initial gas disk size and the halo spin.