Time-Domain Studies in the New Eras of Multi-Messenger Astrophysics and Big Data

astrophysics;supernovae;big data;machine learning;kilonovae

Villar, Victoria Ashley ; Conroy, Charlie,Berger, Edo

University:Havard University

Department:Astronomy

关键词: astrophysics;    supernovae;    big data;    machine learning;    kilonovae;   

Others  :  https://dash.harvard.edu/bitstream/handle/1/37365823/VILLAR-DISSERTATION-2020.pdf?sequence=1&isAllowed=y

美国|英语

来源: Digital Access to Scholarship at Harvard

【 摘 要 】

Time-domain astrophysics probes high-energy phenomena, the end-lives of massive stars and the creation of new elements which enrich the cosmos. Wide-field, untargeted photometric surveys have drastically increased the quantity and variety of known extragalactic transients, continuously challenging our understanding of late-stage stellar evolution. At the same time, gravitational wave detectors have ushered in a new era of multi-messenger astrophysics. This thesis presents a series of theoretical and observational studies which address the questions associated with a new era of data-driven, multi-messenger time-domain astrophysics.First, we explore the breadth of engines and progenitor systems which lead to extragalactic transients. Through a systematic census of optical light curves, we explore the spread of stellar eruptions, explosions and collisions in simple feature spaces. We then take a detailed look into the eruption of a massive star with a neutron star companion. Second, we focus on the collisions of compact objects: neutron star mergers and black hole-neutron star mergers. We present a detailed analysis of the complete photometric dataset covering the first two months of the kilonova associated with the gravitational event GW1701817, resulting in the strongest constraints on the kilonova properties. We then present the first Spitzer Space Telescope infrared observations of the kilonova. Our results highlight the discrepancy between theory and the observed late-time behavior of kilonovae. Finally, we focus on the upcoming age of large data streams. In late 2022, the Vera C. Rubin Observatory will begin its Legacy Survey of Space and Time (LSST), which will increase our annual discovery rate of extragalactic transients by two orders of magnitude. Transients which are currently rare will become commonplace. In this context, we first present a case study of the rare class of superluminous supernovae (SLSNe) in the new era of LSST. We find that the number of LSST SLSN discoveries will rival the current literature sample in less than a week, and we will be able to recover physical parameters for most events without multi-wavelength follow up. However, no population studies of transients will be possible without classification of LSST light curves. We present two data-driven machine learning methods, one supervised and one semi-supervised, which classify supernovae from the Pan-STARRs Medium Deep Survey.