The host galaxies of luminous quasars : a spectrum decomposition view

Actively accreting supermassive black holes, known as Active Galactic Nuclei (AGNs), are often implicated in the cessation of star formation in massive galaxies. Observational evidence for this, however, is mixed and difficult to interpret. The imprints of galaxy quenching and AGN fuelling may be embedded in the star-formation histories (SFHs) of their host galaxies, which are typically challenging to measure for the most luminous AGNs, known as quasars. In this thesis, I introduce a machine learning technique, Mean-Field Independent Component Analysis (MFICA), to decompose quasar spectra into their host-galaxy and central-engine components and to estimate the recent (<1 Gyr) star-formation histories of their host galaxies. We applied MFICA to quasar spectra from the Sloan Digital Sky Survey (SDSS) and found that most quasars reside in massive, star-forming galaxies. We see a 28x excess of post-starburst galaxies (PSBs) - galaxies that have rapidly quenched after a burst in star formation - compared to their number in ordinary galaxies, which we interpret as the efficiency of major mergers in fuelling quasar activity. We find, however, that most quasars are fuelled by secular processes and minor mergers. I then use MFICA to compare the SFHs of Type-1 and Type-2 AGNs, finding higher specific star-formation rates (sSFRs) for Type-2 AGN host galaxies, which is at odds with the unified model of AGNs. We find, however, that these sSFRs cannot be explained by evolution after a major merger, and instead suggest that a common gas and dust supply in galaxies fuels star formation and provides additional obscuration for AGNs beyond the dusty torus. Finally, we use MFICA to compile the largest sample of clean PSBs from the Dark Energy Spectroscopic Survey (DESI). We also use MFICA to enable the first statistical study of the circum-galactic media (CGMs) of PSBs (in Harvey et al. 2025).