Astronomical Catalogue Cross Identification for Data Mining and Statistical Analysis of the Infrared and Faint Radio Sky
Cross-identifying complex radio sources with optical or infrared counterparts in surveys such as the Australia Telescope Large Area Survey (ATLAS) has traditionally been performed by a visual inspection of individual sources. However, with new surveys from the Australian Square Kilometre Array Pathfinder detecting many tens of million of radio sources such an approach is no longer feasible. The Likelihood Ratio (LR) allows the use of additional data about survey objects to be cross matched rather than just position and proximity, such as flux, probability distribution and catalogue surface density. This thesis presents a new software algorithm (LRPY - Likelihood Ratio in PYthon) to automate the process of cross-identifying radio sources with catalogues at other wavelengths using the LR. I demonstrate LRPY by applying it to the ATLAS Data Release 3 and a Spitzer-based multi-wavelength catalogue, identifying 3,848 cross matched sources using my LR-based selection criteria. I found that LRPY could be extended to identify radio sources with multiple infrared counterparts many of which I identify as interacting galaxy pairs. In addition I also investigated if LRPY could be used to select radio sources with more complex morphology, such as double lobe radio sources. This extension to the algorithm is shown to work well in identifying the host of many double lobe radio sources. A subset of 1987 cross matched sources in this thesis have flux density values for all four bands in the Spitzer/IRAC instrument, which allowed me to use various criteria to distinguish between active galactic nuclei (AGN) and star-forming galaxies (SFG). I found that 936 radio sources (≈47%) meet both of the Lacy and Stern AGN selection criteria. Of the matched sources, 295 have spectroscopic redshifts and we examine the radio to infrared flux ratio vs redshift, proposing an AGN selection criterion below the Elvis radio-loud (RL) AGN limit for this dataset. Taking the union of all three AGN selection criteria I have identified 956 cross matched sources as AGN (≈48%). From this dataset, we find a decreasing fraction of AGN with lower radio flux densities consistent with other results in the literature. I found there is a strong power law correlation demonstrated using the complete Fusion dataset seen in the AGN arm of the MIR [3.6]-[5.8] vs [4.5]-[8.0] plot for the LACY AGN selection wedge, but there was no such strong correlation for the Stern AGN wedge of the MIR [5.8]-[8.0] vs [3.6]-[4.5] colour-colour plot. Due to this I have analysed the cross matched sources in this thesis for a power law relationship, it was apparent that for the MIR [3.6]-[5.8] vs [4.5]-[8.0] colour-colour plots used for Lacy AGN selection the correlation between the AGN candidate infrared flux ratios to the power-law locus is very strong in the AGN arm. The Stern MIR colour-colour criteria [5.8]-[8.0] vs [3.6]-[4.5] shows no strong correlation for the AGN candidate infrared flux ratios to the power-law locus for the cross matched sources. I have also looked at the relationship between the MIR 8.0µm to radio 1.4GHz relationship, there is a clear distinction between AGN and SFG's. I have also looked at the relationship between the Radio L₁.₄GHz and MIR luminosities L₃.₆µm and L₄.₅µm, as prior work had provided existence of a correlation. What was found is that the higher redshift sources (z>0.3) lie within the top right quadrant of the plots, when the Stern and Lacy AGN are identified we see the majority of these AGN also lie within this same quadrant. Looking at the AGN selected sources we see a slope close to unity at both MIR wavelengths, for 3.6µm the slope is m = 1.004 ± 0.155 and at 4.5µm it was found to be m = 0.980 ± 0.153. I conclude the thesis by summarizing the work and discussing future work to be undertaken as a result of the research.