The Properties of Galaxies and Active Galactic Nuclei Using Panchromatic Fitting

This thesis investigates the evolving nature of galaxy populations from the early Universe to the present day using broadband spectral energy distribution (SED) fitting of both real and simulated datasets, incorporating photometry and spectroscopy. Our analysis is performed with the newgeneration energy-balance Bayesian forward-modelling SED fitting code PROSPECTOR, which treats photometry and spectroscopy on an equal footing and includes the capability to model AGN contributions to galaxy SEDs, both of which are critical to this work. First, we investigate the nature of the faint radio source population. Recent efforts have combined fits from multiple SED-fitting codes to account for the host galaxy and any active nucleus that may be present. We show that it is possible to produce similar-quality classifications using a single energy-balance SED fitting code, PROSPECTOR, to model up to 26 bands of UV– far-infrared aperture-matched photometry for ∼31,000 sources in the ELAIS-N1 field from the LOFAR Two-Metre Sky Survey (LoTSS) Deep fields first data release. One of a new generation of SED-fitting codes, PROSPECTOR accounts for potential contributions from radiative active galactic nuclei (AGN) when estimating galaxy properties, including star formation rates (SFRs) derived using non-parametric star formation histories. Combining this information with radio luminosities, we classify 92 per cent of the radio sources as a star-forming galaxy, high- /low-excitation radio galaxy, or radio-quiet AGN and study the population demographics as a function of 150MHz flux density, luminosity, SFR, stellar mass, redshift and apparent �-band magnitude. Finally, we use PROSPECTOR SED fits to investigate the SFR–150MHz luminosity relation for a sample of ∼133,000 3.6 �m-selected � < 1 sources, finding that the stellar mass dependence is significantly weaker than previously reported, and may disappear altogether at log10(SFR/�⊙ yr−1) > 0.5. Second, we develop a fully Bayesian spectrophotometric classification framework that combines LoTSS DR1 photometry with DESI EDR spectroscopy for 6,038 radio sources. Leveraging the full posterior distributions from PROSPECTOR, we derive probabilistic classifications based on mid-IR excess, emission-line diagnostics, and radio-excess criteria, enabling both maximumlikelihood and confidence-threshold classifications. Comparing our 90 per cent reliability classifications (P90) with the photometric maximum-likelihood classifications and the spectroscopic probabilistic classifications from previous studies demonstrates the advantages of a joint photometric– spectroscopic approach. In particular, the inclusion of emission-line and mid-IR information reveals populations of radiative-mode AGN that are missed by photometric or spectroscopic methods alone. While this thesis presents initial results from this framework, forthcoming spectroscopic datasets from surveys such as WEAVE-LOFAR and ORCHIDSS will allow for a substantial expansion of this approach, enabling a more comprehensive understanding of the faint radio source population. Finally, we explore the recovery of burstiness, �, defined as the log of the ratio of the short-term star formation rate (averaged over the last 10 Myr) to the intermediate-scale star formation rate (averaged over the last ∼100 Myrs) using UV-to-mid IR photometry from an idealised data set based on Hubble plus James Webb Space Telescope NIRCam and MIRI imaging. We generate synthetic photometry for 1,380 galaxies spanning 4.64 ≤ � ≤ 10 from the SPHINX20 cosmological radiation-hydrodynamic simulations and use PROSPECTOR to infer physical galaxy properties, including �, under two non-parametric star formation history (SFH) priors: a continuity prior and a bursty-continuity prior. While the former prefers constant SFHs, the latter is more flexible, and is capable of modelling short-term variability in SFHs. We find that the choice of SFH prior exerts a strong and often undue influence on the inferred parameters, with individual � estimates remaining unreliable irrespective of the prior used. While there is some success in recovering the average � for samples of galaxies, our results reveal strong redshift-dependent biases and systematic effects arising from photometric band selection and wavelength coverage. These limitations highlight the difficulty of identifying galaxies undergoing bursts of star formation or temporary quenching events using photometry alone, underscoring the need for spectroscopic information to break degeneracies and improve the fidelity of burstiness recovery. Together, the results presented in this thesis demonstrate the power and flexibility of modern Bayesian SED fitting frameworks for studying galaxy populations across cosmic time. By unifying photometric and spectroscopic information, developing probabilistic classification methods, and stress-testing the recovery of recent star formation activity, this work lays the foundation for exploiting forthcoming spectroscopic datasets to obtain a more complete and physically motivated understanding of galaxy evolution, the nature of the faint radio source population, and the role of AGN across the history of the Universe.