The dependence of AGN activity on stellar and halo mass in semi-analytic models

Active galactic nuclei (AGN) feedback is believed to play an important role in shaping a variety of observed galaxy properties, as well as the evolution of their stellar masses and star formation rates. In particular, in the current theoretical paradigm of galaxy formation, AGN feedback is believed to play a crucial role in regulating star formation activity in galaxies residing in relatively massive haloes, at low redshift. Only in recent years, however, has detailed statistical information on the dependence of galaxy activity on stellar mass M_*, parent halo M_DM mass and hierarchy (i.e. centrals or satellites) become available. In this paper, we compare the fractions of galaxies belonging to different activity classes (star forming, AGN and radio active) with predictions from four different and independently developed semi-analytical models. We adopt empirical relations to convert physical properties into observables (Hα emission lines, Oiii line strength and radio power). We demonstrate that all models used in this study reproduce the observed distributions of galaxies as a function of stellar mass and halo mass: star-forming galaxies and the strongest radio sources are preferentially associated with low-mass and high-mass galaxies/haloes, respectively. However, model predictions differ from observational measurements in many respects. All models used in our study predict that almost every M_DM > 10¹²M_⊙ dark matter halo and/or M_* > 10¹¹M_⊙ galaxy should host a bright radio source, while only a small (few per cent) fraction of galaxies belong to this class in the data. In addition, radio brightness is expected to depend strongly on the mass of the parent halo mass in the models, while strong and weak radio galaxies are found in similar environments in data. Our results highlight that the distribution of AGN activity as a function of stellar mass provides one of the most promising discriminants between different gas accretion schemes.