CHARACTERIZING OPEN CLUSTERS AND SPECTROSCOPIC ECLIPSING BINARIES WITH MACHINE LEARNING FRAMEWORKS

Abstract

In recent years there has been an explosion of data from space-based and ground-based astronomical surveys. This has led to the necessity of moving away from single-use, boutique analyses to flexible automated data analysis frameworks. In this dissertation two specific projects are carried out with this aim in mind. The first project concerns groups of stars known as open clusters. The Gaia satellite has revolutionized our understanding of open clusters with the high volume of high precision astrometric measurements. However, not many analyses of these objects have fully included the high precision uncertainties of the Gaia DR2 data. Using Extreme Deconvolution Gaussian Mixture Models, an automated framework is constructed and utilized to recover individual members of 426 open clusters. It’s performance was validated with previous literature results, finding a 98.1% recovery rate with automated model selection and open cluster identification. The second project focuses on pairs of bound stars known as spectroscopic eclipsing binaries (SEBs), and inferring the complete orbital parameters of these systems using joint time-series analysis in a Bayesian framework. A sample of 12 SEBs with both TESS photometric light curves and APOGEE radial velocity measurements were selected. Using a gradient based Markov Chain Monte-Carlo algorithm, a binary orbital model was constructed to infer full orbital parameter solutions. This framework was able to flexibly find solutions for 11.5/12 SEBs, laying the ground work for an automated framework to be used to develop a benchmark catalog of SEBs.