A team of astronomers examined how well the Gaia DR3 catalog treats close binary stars and report a persistent ≈7% “floor” of systems that look like single stars in Gaia’s astrometry but show signs of a hidden companion in other data. They cross-matched an initial list of candidate systems with infrared (2MASS) and optical (Pan‑STARRS) surveys to look for small, consistent photometric offsets that would not be expected for true single stars. The authors interpret this 7% as a systemic sensitivity limit of the Gaia single‑star model in DR3.

The study started from a large catalog called WDSS and reduced roughly 16 million raw entries to 2.36 million unique candidate systems. After strict spatial and quality filters and cross‑matching to Gaia DR3, 2MASS, and Pan‑STARRS, the authors kept a “Gold Standard” sample of 120,418 sources (about 5.1% of the parent set). They built an empirical polynomial ridge line for the combined catalogs and measured magnitude residuals (ΔG) relative to that ridge line. Those residuals, together with other diagnostics, were used to search for unresolved or partially resolved binary systems.

The paper uses a “Triple Constraint” to flag suspect binaries: astrometric noise measured by RUWE (Renormalised Unit Weight Error), photometric excess measured as a magnitude residual (ΔG), and the absence of an official Gaia Non‑Single Star (NSS) classification. RUWE is an internal Gaia metric for how well a single‑star model fits the astrometric data; DR3 commonly uses RUWE>1.4 to mark overt astrometric problems. The authors find a tri‑modal “Detection Gap”: 14,705 stars (≈12%) show overt astrometric discordance with RUWE>1.4, while many others have low RUWE (<1.4) yet show a consistent ≈0.75 magnitude offset and infrared excess — a sign that a second, cooler star is present but has been “absorbed” into Gaia’s single‑star solution. They also used Gaia image flags (IPD multi‑peak) and 2MASS JHK plus Pan‑STARRS y‑band measurements to confirm cooler companions that are faint in Gaia’s G band.