Speaker
Description
A significant fraction of stars in the Galactic field reside in multiple systems, yet most of these remain spatially unresolved in Gaia observations. Current state-of-the-art stellar parameter inference tools deliver exquisite multi-dimensional maps of the Milky Way but typically neglect unresolved binary companions, potentially biasing derived parameters when the secondary contributes significantly to the overall flux.
We previously inferred stellar masses and binary mass ratios for Gaia DR3 open cluster members (Malhotra et al. 2026), and now extend the method to the Galactic field while explicitly accounting for unresolved companions. Using a simulation-based inference (SBI) framework, we combine multi-band photometry, spectroscopic labels from high-resolution surveys, and Gaia binarity indicators (e.g., RUWE) to jointly constrain stellar age, mass, distance, line-of-sight extinction, and binary mass ratio (q), including regimes compatible with dark, compact object companions. Our tests on synthetic populations show that accounting for the unresolved companion reduces both bias and scatter in the recovered ages. At the same time, the inference runs about two orders of magnitude faster than traditional isochrone fitting. This framework is also flexible enough to include new observables as astrometric and spectroscopic surveys continue to grow.