Extending the Reach of Gravitational Wave Detectors and Probing the Isotropic Stochastic Background

Abstract

Beginning with the first detection of gravitational waves in 2015 by the Advanced Laser Interferometer Gravitational Wave Observatory (aLIGO) a new era of astrophysics emerged. Within 5 years, aLIGO has detected 50 mergers from binary black hole (BBH) and binary neutron star (BNS) systems and kick started the field of multi-messenger astrophysics with the measurement of an electromagnetic counterpart to the BNS merger GW170817. A detection on the horizon for LIGO is that of the stochastic gravitational wave background (SGWB). The detection of such a background would have far reaching consequences in astrophysics and cosmology as these measurements can probe the first fractions of a second after the Big Bang, revealing insights and parameters of proposed and possibly undiscovered cosmological models. Even a SGWB formed by BBH and BNS mergers near to us would provide valuable information about star formation rates, the formation of large scale structure, as well as the populations of these compact objects. There are two main topics in this dissertation: detector characterization/data quality methods and characterization of the SGWB. The first chapter provides a background into General Relativity and derives import dynamics relevant to LIGO that are heavily used throughout the thesis. Chapter 2 delves into detector characterization and understanding the noise which enters into the detector output data stream. This is accomplished through the development of a coherence calculation package, \texttt{STAMP-PEM}, which creates a hash table lookup for quick followup analysis and a user friendly API. In Chapter 3 I discuss the sky-averaged SGWB, search methods and present the most recent results combined over aLIGO's first three observing runs. Chapter 4 will extend the isotropic SGWB search to Cosmic Explorer sensitivities and attempt a novel solution to subtract the foreground compact binary coalescences (CBCs) in the $f-t$ space. This mock data analysis sets a benchmark for future search methodologies and sensitivities. Finally in Chapters 5 and 6 I discuss data quality filtering methods. The former will employ a new deep learning architecture known as \texttt{DeepClean} to identify and subtract noise couplings of arbitrary order without introducing artifacts or phase misalignment of the output signal. The final chapter is dedicated to the construction of analytic filters used for removing linear, nonlinear and non-stationary noise. These analytic filters are useful as they are lightweight and can brute force search through the auxiliary channel combinatorics and aid in identifying relevant physical couplings.