A data-driven approach to understanding supernovae as astrophysical probes

Abstract

Forthcoming time-domain surveys, such as the Rubin Observatory Legacy Survey of Space and Time, will vastly increase samples of supernovae (SNe) and other optical transients. This large stream of photometric data will allow the development and improvement of machine-learning techniques for analysing their light curves and provide a better understanding of these phenomena as astrophysical probes. In this thesis, I present PISCOLA, an open-source data-driven/machine-learning light curve fitter. Although PISCOLA can be used to estimate the rest-frame light curves of any transient, without the need for an underlying light-curve model, here I present an application to type Ia supernovae (SNe Ia) as distance indicators. I tested PISCOLA on simulations of SNe Ia to validate its performance, showing it successfully retrieves rest-frame peak magnitudes for average data qualities of current cosmological surveys. When compared to the existing light-curve fitter SALT2 on real data, I find small differences in the estimated light curve parameters. I introduce an original analysis by decomposing the estimated rest-frame light curves of SNe Ia from the Pantheon sample with Non-Negative Matrix Factorization. This decomposition is used as a new way of standardising SNe Ia that provides similar scatter in the measured distances as SALT2. Additionally, I use PISCOLA to study the wavelength-dependent variation of colour in SNe Ia to have a better understanding of their intrinsic variation, finding no disagreement with the SALT2 model. I also present the study of SN 2016aqf, a low-luminosity type II SN (LL SN II) with extensive photometric and spectral coverage. I use nebular (late-time) spectra to estimate a progenitor mass of 12 ± 3 M , and measure the [Fe II] λ7155 and [Ni II] λ7378 lines, mainly found in LL SNe II, to estimate its Ni/Fe abundance ratio, a parameter sensitive to the inner progenitor structure and explosion mechanism dynamics. Placing this measurement in the context of a sample of SNe II, I find that the measured distribution of Ni/Fe abundance ratio does not agree with those predicted by theoretical modelling.

More information

Identifiers

ORCID for Mark Sullivan: orcid.org/0000-0001-9053-4820

Catalogue record

Export record