Sensing disturbances in the force with AI: gravity and light in the cosmos

Abstract

Gravity is the strangest force known to mankind. It is weak enough to allow a toddler to stand, but strong enough to bind galaxies together over trillions of kilometres. Its ability to warp spacetime, altering the way we perceive the universe around us, is a powerful probe for so many great unknowns. From dark matter to the Hubble constant, when we seek the ephemeral, we turn to gravity. Black holes (BHs) and neutron stars (NSs) hold the key to a plethora of poorly understood phenomena. Despite thousands of years of study, we still fall short when attempting to explain supernovae. Although the successes of both quantum mechanics and general relativity have been celebrated, we know them to be incomplete as they have incompatible descriptions of BHs. If we are to acquire answers to these and more problems, we need to find more BHs and NSs. BHs and NSs are overwhelmingly isolated objects, with 70--90\% of BHs thought to be solitary. Because the vast majority do not emit significant electromagnetic radiation, our only hope of finding these objects is through observations of their gravity. Thus, we must look for signs of gravity's pull by searching for the warping of light travelling near to these objects: a phenomenon called gravitational lensing. This thesis explores how to distinguish the signs of gravitational lensing. Beginning with a review of the formation processes of BHs and NSs, it then continues into an overview of gravitational lensing, considering both macrolensing and microlensing. Following this is a review of semi-supervised learning methods, to arm the reader with the requisite knowledge to understand the methods presented later in the thesis.