Preparing for a cosmological analysis of next generation radio observations

In order to prepare for a cosmological analysis of the next generation of radio continuum surveys such as the Evolutionary Map of the Universe (EMU), we here investigate the galaxy properties and cosmological clustering of radio sources in a number of existing large area surveys with high sensitivity. In order to understand the types of galaxy likely to be observed at radio frequencies, we investigate the host galaxy properties of radio sources as a function of redshift and mass out to a redshift of z = 2.25 and use the resulting relations to produce a new simulated radio continuum sky. We compare the measured angular clustering of galaxies in this simulation to both real observations (via two of the largest μJy surveys to date) and more traditional numerical models of angular clustering at 1.4 GHz. Using cross identification of deep Very Large Array observations at 1.4 GHz with two deep Near Infrared (NIR) surveys, the Four Star Galaxy Evolution redshift survey (ZFOURGE)and the Newfirm Medium Band Survey (NMBS), we determine the galactic properties of 424 radio sources at flux densities greater than 50μJy in the COSMOS and CDFS fields. Both of these NIR surveys use medium band infrared filters to increase the accuracy of photometric redshifts at much earlier cosmic times than optical surveys. The wealth of ancillary data in these two fields is substantial, with up to 39 bands of photometry spanning from the ultraviolet to the far-infrared. This allows us to accurately estimate galaxy stellar-masses, star formation rates and dust contents. Using this data we identify radio-loud AGN using their 1.4 GHz and NIR+UV based star-formation rates. Objects whose radio emission is consistent with the star-formation rate are classified as star-forming dominated sources and those whose emission at 1.4 GHz is greater than three times what is expected from the NIR+UV star-formation rate are classified as radio-loud AGN. We compare the host properties of these objects in a mass (M ≥ 1010:5 M) and luminosity (L1:4 > 1024 W Hz-1) complete sample against the galaxy properties of non-AGN hosts finding them to be in agreement across abroad range of redshifts (0.25 ≤ z < 2.25). We determine the mass and redshift dependant radio-loud AGN fraction across the redshift range (0.25 ≤ z < 2.25) finding that while radio-loud AGN show very little change in rarity between these epochs, they exhibit a strong dependence on the stellar-mass of their host galaxy. Extending this work further, we utilise new Very Long Baseline Interferometry observations to study if these relations hold for a sample of AGN selected on the basis of their surface brightness temperature. We find that the relationship between VLBI detected sources (which provides an unambiguous confirmation of AGN activity) and the mass and redshift similar sample of non-AGN sources is identical to that of radio-loud AGN, and that VLBI surveys selected are not only a part of the radio-loud AGN population, but a representative sample. To prepare for the EMU cosmology analysis we measure the clustering of bright (S1:4 >15 mJy) radio sources in the NRAO Very Large Array Southern Sky Survey, using both the angular power spectrum and two point angular correlation function and find excellent agreement with previous work on this survey. In addition to this, we measure the observed clustering in the Jansky Very Large Array observations of Stripe-82 and the Australian Telescope Compact Array, Large Area Survey of the South Pole Telescope Field at 300μJy. We find that the radio sky at these lower flux limits is far less dominated by extended radio sources than at higher flux densities. Fitting the correlation function with a powerlaw at scales between 0.1 and 10 degrees yields results inconsistent with previous work at flux both above and below our current limits. This is due to a substantial over density of sources measured on scales of 0.1-0.3 degrees and removing these scales from the fit results in a measured powerlaw of amplitude A = 1.6 ± 0.24 x 10 -3, with a slope of -0.8 ± 0.3 equivalent to aγ =1.8. Using an inversion of the cosmological Limber equation we convert this measurement into the spatial correlation length r=00 = 2.2+3:41:7 h1 Mpc and find this to be in good agreement with previous studies. The substantial dependence of this cosmological analysis on a small number of elevated data points, shows just how important the need for accurate modelling of observational systematics can be. We use the relationship between radio flux and star-formation rate as well as the relations found above, detailing the prevalence of radio-loud AGN as a function of stellar mass and redshift, to convert simulated catalogues based on the dark matter halos of the semianalytic Millennium Simulation into an expected radio sky. We compare our observations against both this newly created simulation and more traditional numerical simulations of the expected auto correlation function. Both show moderate agreement with the corrected observations and we hope that in future work the easy addition of noise, bright side-lobes and extended sources to our new simulation will allow us to better understand and remove the effects of these issues from observations.