We explore the cosmic evolution of radio AGN with low radio powers (L 1.4GHz 5 × 10 25 W Hz −1 ) out to z = 1.3 using to-date the largest sample of ∼ 600 low luminosity radio AGN at intermediate redshift drawn from the VLA-COSMOS survey. We derive the radio luminosity function for these AGN, and its evolution with cosmic time assuming two extreme cases: i) pure luminosity and ii) pure density evolution. The former and latter yield L * ∝ (1+z) 0.8±0.1 , and Φ * ∝ (1+z) 1.1±0.1 , respectively, both implying a fairly modest change in properties of low radio-power AGN since z = 1.3. We show that this is in stark contrast with the evolution of powerful (L 1.4GHz > 5 × 10 25 W Hz −1 ) radio AGN over the same cosmic time interval, constrained using the 3CRR, 6CE, and 7CRS radio surveys by Willott et al. (2001). We demonstrate that this can be explained through differences in black hole fueling and triggering mechanisms, and a dichotomy in host galaxy properties of weak and powerful AGN. Our findings suggest that high and low radio-power AGN activity is triggered in different stages during the formation of massive red galaxies. We show that weak radio AGN occur in the most massive galaxies already at z ∼ 1, and they may significantly contribute to the heating of their surrounding medium and thus inhibit gas accretion onto their host galaxies, as recently suggested for the 'radio mode' in cosmological models. 16 The 'radio mode' is defined here as given in Croton et al. (2006).

We derive the cosmic star-formation history out to z = 1.3 using a sample of ∼350 radio-selected star-forming (SF) galaxies, a far larger sample than those in previous, similar studies. We attempt to differentiate between radio emission from active galactic nuclei and SF galaxies, and determine an evolving 1.4 GHz luminosity function (LF) based on these VLA-COSMOS SF galaxies. We precisely measure the high-luminosity end of the SF galaxy LF (star-formation rate 100 M yr −1 ; equivalent to ultra-luminous IR galaxies) out to z = 1.3, finding a somewhat slower evolution than that previously derived from mid-infrared data. We find that more stars are forming in luminous starbursts at high redshift. We use extrapolations based on the local radio galaxy LF; assuming pure luminosity evolution, we derive L * ∝ (1 + z) 2.1±0.2 or L * ∝ (1 + z) 2.5±0.1 , depending on the choice of the local radio galaxy LF. Thus, our radio-derived results independently confirm the ∼1 order of magnitude decline in the CSFH since z ∼ 1.

No abstract