Supernovae and star formation in luminous infrared galaxies

The rate at which massive stars explode as core collapse supernovae (CCSN) can act as a tracer of the cosmic star formation history, independent from the conventional galaxy luminosity methods. However, estimates of the CCSNe rate suffer from significant statistical and systematic uncertainties. While upcoming next-generation optical survey telescopes will dramatically improve the statistics for determining the CCSN rate, they will still be prone to poorly understood systematic effects due to dust extinction and the restricted spatial resolution of seeing-limited observations. In this thesis I present the methodology and results of project SUNBIRD (Supernovae UNmasked By Infra-Red Detection), in which we aim to characterize the population of CCSNe that remain hidden in the nuclear regions of galaxies due to bright background emission and significant dust extinction. Improved limits of this missed fraction will be crucial in reducing the systematic uncertainties of current and future CCSN surveys. We have observed a sample of luminous infrared galaxies (LIRGs), which have high star formation rates and host bright and complex nuclear regions, but so far have shown a prominent shortfall of CCSN discoveries. To uncover CCSNe in this regime, we observed in the near-infrared (near-IR), which is less affected by dust extinction compared to the optical, using state-of-the-art laser guide stars adaptive optics (LGS-AO) imagers GeMS/GSAOI on the Gemini South telescope and NIRC2 on the Keck II telescope. These combined capabilities provide diffraction-limited image quality on 8-10m class telescopes, and at wavelengths with an order of magnitude less dust extinction than in the optical, allowing us to sensitively probe the complex star-forming and nuclear regions of these dusty galaxies in a completely new way. Over the course of project SUNBIRD we discovered four photometrically confirmed CCSNe and an additional five CCSN candidates. This has doubled the sample of AO-assisted CCSN discoveries in LIRGs, and includes one of the most nuclear CCSNe ever discovered in a LIRG (SN 2013if at 0.5" or 0.2 kpc projected radial offset) and one of the most dust-extincted CCSN discovered in any galaxy (AT 2017chi with ~12 magnitudes of extinction in V-band). Additionally, I discovered an extremely near-IR bright transient (AT 2017gbl) superimposed on its host's nucleus, believed to be a tidal disruption event. By comparing the total sample of AO CCSNe discovered in LIRGs with all documented seeing-limited optical and near-IR discoveries, I show that our method is singularly effective in uncovering CCSNe in the nuclear regions. Whereas seeing-limited optical and near-IR discoveries drop off in the central kpc of LIRGs, AO-assisted CCSN discoveries increase dramatically, as expected from their centrally peaked star formation. Finally, we have observed one of the SN factories in our sample, IRAS 18293-3413 with three CCSNe discoveries, with the cutting-edge AO-assisted MUSEintegral field spectrograph at the ESOVLT. I show that, contrary to expectations, the CCSNe in this LIRG do not trace star formation well, but instead align with 'raw' Hα. This is likely due to the effects of dust extinction, implying that even when using near-IR observations, detailed knowledge about the distribution of star formation and dust will be a crucial factor when converting observed CCSNe to a supernova rate.