A Probabilistic Gaia DR3 Catalogue of Stars with Very Low Galactic Rest-Frame Speeds
Abstract
We present a Gaia DR3 catalogue of stars selected by very low Galactic rest-frame speed, VGRF < 25 km s−1. Using Lindegren et al. (2021) parallax zero-point corrections, Bailer-Jones et al. (2021) photogeometric distances, and Monte Carlo propagation of the Gaia parallax–proper-motion covariance, we identify a primary Tier A+B sample of 541 stars with P(VGRF < 25 km s−1) > 0.84 and a broader Tier A+B+C sample of 1,952 stars with P > 0.50. Summing 1 − P over the members, the internal Monte Carlo scores imply nominal score-implied membership purities of 94% ± 1% for Tier A+B and 73% ± 1% for Tier A+B+C. Orbit integrations in the static, axisymmetrized Hunter et al. (2024) Galactic potential map the broader sample onto compact, highly radial model orbits, with median eccentricity 0.949, median pericentre 154 pc, and median apocentre 7.35 kpc. Matched controls place the slow sample at the compact-pericentre, high-eccentricity end of a velocity-stratified sequence. As a separate empirical count diagnostic, a smooth Gaussian-mixture continuation of the matched 25–260 km s−1 controls gives an observed-to-continuation ratio of 7.4 for Tier A+B+C; this is not a selection-corrected intrinsic excess. The adjacent 25–50 km s−1 band also fails to reproduce the slow-sample eccentricity distribution (unweighted Anderson–Darling p < 0.001). A controlled GeDR3mock injection-recovery exercise recovers 60.4% of true sub-25 km s−1 stars end-to-end (87.0% conditional on entering the pipeline). The dominant selection-function systematic affects deprojection rather than catalogue membership: 83.2% of valid Tier A+B sources fall in low-parent-count (n < 10) GaiaUnlimited cells, and including prior-filled cells changes the Tier A+B effective-volume scale by a factor of 2.09. We release the catalogue as a probabilistic kinematic sample with model-dependent orbit diagnostics and subset-qualified chemical context.
