Topological detection of stellar communities in the Magellanic Bridge using Machine Learning and Gaia DR3 data

Abstract

The increasing availability of massive astrometric and photometric datasets, such as those provided by the Gaia mission (DR3), demands the development of scalable analytical approaches. In this work, we present an automated pipeline for the detection and validation of stellar communities in the Magellanic Bridge region. We replace the traditional approach of searching for purely spatial (2D) overdensities with a topological search. To achieve this, we model the groupings by constructing an undirected graph over an expanded five-dimensional phase space, consisting of spatial coordinates (α, δ), parallax (ϖ), and proper motions (μα*, μδ). We apply the Leiden algorithm to this k-NN network to identify kinematically coherent communities. For astrophysical validation, we analyze Color-Magnitude Diagrams (CMD) by superimposing theoretical PARSEC isochrones calibrated for young populations. The methodology proves to be highly robust against galactic field star contamination, establishing an adaptable baseline for future deep catalogs, such as those from the Vera C. Rubin Observatory.