Superluminous supernovae in large astronomical surveys

Abstract

This thesis focuses on the photometric classification and the properties of superluminous supernovae (SLSN), as observed in large astronomical surveys. When working with large samples of transients and a limited spectroscopic follow-up campaign, photometric classifications are the only tool available to select objects which can subsequently be used to study their statistical properties including the rate and its evolution. I begin by introducing the surveys which produce the transients archive used in this work. I discuss the effect of their properties and design on the work performed in this thesis. I also introduce the sources of auxiliary data, including the spectroscopic followup facilities as well as summarise the Dark Energy Survey (DES) spectroscopic sample of SLSNe. Next, I develop a number of models and techniques used to simulate both core-collapse supernovae (CCSN) and SLSNe. I use the spin-down of a magnetar model in conjunction with spectroscopic UV absorption templates to build SLAP, a tool for simulating SLSN at any redshift, in any arbitrary photometric system. Similarly, I develop CoCo which can be used to simulate and generate the templates for CCSNe. I then use the tools developed in this thesis to build a definition of SLSNe in terms of the spin-down of a magnetar model and apply it to a sample of transients detected by the Supernova Legacy Survey (SNLS), uncovering one previously unclassified SLSN. I use this and two previously spectroscopically confirmed objects to calculate the rate of SLSNe at z∼ 1 with the help of a Monte Carlo simulation of the survey. I find the rate to be 91+76 −36 SNe Yr−1 Gpc−3, equivalent to 2.2+1.8 −0.9 × 10−4 the rate of CCSN at the same redshift. Finally, I use the models of CCSN and SLSNe developed in this work as well as the simulations of SN Ia and AGN to build a large artificial training sample of DES-like transients to be used in the photometric selection of SLSNe. Based on this, I build a two-stage machine learning photometric classification tool. In the first step I separate SN from other contaminating transients with an accuracy of 99.8%. Then, I separate the sample of SNe into its individual subclasses, achieving an overall accuracy of 97.85%. Using complementary selection techniques, I identify 26 new SLSNe candidates in DES.

More information

Identifiers

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

Catalogue record

Export record