Chemo-kinematic analysis of metal-poor stars with unsupervised machine learning

Metal-poor stars play an import role on the understanding of near-field cosmology. Old and long-lived late-type stars provide important clues to the early galaxy chemical evolution. In particular, these stars may help to clarify one of the open problems in nucleosynthetic theory:the \textit{locus} of the r-process. In this work, we report on preliminary results of a chemo-kinematic analysis of a sample of metal-poor stars observed by the GALAH spectroscopic survey. GALAH provided chemical abundances and radial velocities for 340\,000 stars. We cross-matched this sample with the \textit{Gaia} DR2 catalog, which includes accurate and precise measurements of proper motions and parallaxes for more than 10$^9$ stars.Our final sample of objects with metallicities [Fe/H] $\leq$ $-$1.0 dex has 1\,146 stars. From these, 1\,072 stars have radial velocities determined by the GALAH survey. With this selection, we used \textit{galpy} to integrate stellar orbits adopting a Milky Way gravitational potential and derive parameters such as orbital eccentricity, farthest distance reached from the plane, action angles and angular momentum. We explored the chemical and orbital data with unsupervised machine learning (Hierarchical clustering, k-means cluster analysis and correlation matrices). Our final goal is to find an optimal way to separate different Galactic stellar populations and stellar groups originating from merging events, such as Gaia Enceladus and Sequoia.

Theme – Machine Learning, Statistics, and Algorithms