2015A&A...576A..32G

Using an advanced semi-analytical model (SAM) for galaxy formation, we investigated the statistical effects of assuming two different mechanisms for triggering AGN activity on the properties of AGN host galaxies. We considered a first accretion mode where AGN activity is triggered by disk instabilities (DI) in isolated galaxies, and a second feeding mode where galaxy mergers and fly-by events (interactions, IT) are responsible for producing a sudden destabilization of large quantities of gas, causing the mass inflow onto the central supermassive black hole. The effects of including IT and DI modes in our SAM were studied and compared with observations separately to single out the regimes in which they might be responsible for triggering AGN activity. We obtained the following results: i) the predictions of our model concerning the stellar mass functions of AGN hosts point out that both DI and IT modes are able to account for the observed abundance of AGN host galaxies with M_*≲10¹¹M_☉; for more massive hosts, the DI scenario predicts a much lower space density than the IT model in every redshift bin, lying below the observational estimates for redshift z>0.8. ii) The analysis of the colour-magnitude diagram of AGN hosts for redshift z<1.5 can provide a good observational test to effectively distinguish between DI and IT mode, since DIs are expected to yield AGN host galaxy colours skewed towards bluer colours, while in the IT scenario the majority of hosts are expected to reside in the red sequence. iii) While both IT and DI scenarios can account for AGN triggered in main sequence or starburst galaxies, DIs fail in triggering AGN activity in passive galaxies. The lack of DI AGN in passive hosts is rather insensitive to changes in the model describing the DI mass inflow, and it is mainly caused by the criterion for the onset of disk instabilities included in our SAM. iv) The two modes are characterized by a different duration of the AGN phase, with DIs lasting even on time scales of ∼Gyr, much longer than in the IT scenario, where the galaxy interaction sets the duration of the burst phase around ∼10⁷-10⁸yr. v) The scatter of the star formation rate SFR-L_bol relation could represent another crucial diagnostics to distinguish between the two AGN triggering modes, since the DI scenario predicts an appreciably lower scatter (especially at low-to-intermediate AGN luminosities) of the relation than the IT scenario. vi) Disk instabilities are unable to account for the observed fraction of AGN in groups for z≲1 and clusters for z≲0.7, while the IT scenario matches observational data well.