Virtual Observatory and machine learning for the study of low-mass objects in photometric and spectroscopic surveys

Abstract

Low-mass objects are ubiquitous in our Galaxy. Their low temperature provides them with complex atmospheres characterised by the presence of strong molecular absorption bands which, together with their faintness, have made their accurate characterisation a great challenge for astronomers over the last decades. M dwarfs account for 75% of the census of stars within 10 pc of the Sun, and their suitability as targets in the search for Earth-like planets has led many research groups to focus on the study of these objects, which is crucial for the understanding of the structure and kinematics of our Galaxy. Very low-mass stars and substellar objects with spectral types M7 or later, including the extended L, T, and Y spectral types, constitute the domain of ultracool dwarfs. The study of these objects, discovered definitively in 1995, is key for understanding the boundary between stellar and substellar objects and promises to experience a quantum leap thanks to the characteristics of new-generation surveys such as Euclid or LSST. Data analysis in the field of observational astronomy has undergone a paradigm shift during the last decades driven by an exponential growth in the volume and complexity of available data. In this revolution, the Virtual Observatory has become a cornerstone providing a system that fosters interoperability between astronomical archives around the world. In response to this growth in data complexity, the astronomical community has increasingly adopted machine learning techniques for the development of scalable, automated solutions. This thesis explores the discovery and characterisation of M dwarfs and ultracool dwarfs, using data-driven approaches supported by Virtual Observatory technologies and protocols. We rely on a variety of machine and deep learning techniques to develop flexible methodologies aimed at advancing our understanding of low-mass objects.