Multi-frequency matching, classification, and cosmic evolution of radio galaxy populations

This thesis combines the third data release of the Australia Telescope Large Area Survey (ATLAS) with complimentary DES photometry and OzDES targeted spectroscopy. ATLAS 1:4 GHz is a deep (< 20 ˜í¬¿Jy/beam) survey of the combined 6:3 deg\\(^2\\) sky area of CDFS and ELAIS-S1. Cross-matching of 4903 ATLAS sources to photometric and spectroscopic counterparts is performed visually using custom interactive software: MCVCM. I compare the results from this visual matching to results from simple coordinate-based nearest-neighbour matching. The visually constructed catalogues are substantially more robust due to a better accounting of chance alignments and dense field mismatches. Both cross-matching techniques are performed directly from radio-to-optical, and indirectly by matching radio sources to an infrared-optical paired catalogue. There is a definite improvement for both visual and nearest-neighbour cross-matching when performed via SWIRE infrared intermediary catalogue. I use the visual catalogue to build a reliable sample of ATLAS sources matched to SWIRE infrared photometry, DES optical photometry, and OzDES spectra. This catalogue was developed carefully and is intended to for use as a comparison and training catalogue for the development of intelligent automated cross-matching algorithms. Galaxies are segregated into star-forming and AGN populations using a combination of visual spectral classification, morphology, multi-frequency diagnostics, and optical spectral line diagnostics. For these populations, 1:4 GHz radio luminosity functions (RLFs) are constructed incrementally out to a redshift z < 1.0. These results improve upon ATLAS-DR1 RLFs and agree with established literature results. Evidence of the space-density evolution of radio source is apparent for both populations towards higher redshifts, indicating increased star-formation and AGN luminosity in earlier cosmic epochs. I construct RLFs for both visual galaxy classification and a combined classification based on a novel numerical combination of classification diagnostics. The combined classification provides higher reliability and completeness resulting in less confounded and deeper RLF. Cross-matching techniques capable of identifying extended and confounded radio sources combined with a robust galaxy classification scheme are therefore essential for quantifying the evolution of radio galaxy populations.