TESS data show that less than 7 % of Roche-allowed Gaia DR3 SB1 orbits with M2 sin i < 80 MJ around dwarf stars are real (mainly P < 4 d; desert mass prior), against 23-34 % for 80-150 MJ

Gaia DR3 gives many single-lined spectroscopic (SB1) orbits with short periods and brown-dwarf (BD) minimum masses. We used public TESS light curves to measure the fraction f of these orbits that are real. A real orbit gives eclipses at the Gaia conjunction times with a known probability, and ellipsoidal variation at half the Gaia period with a known amplitude. We compared the observed signals with the signals that we expect if all orbits are real. Injections into the raw light curves gave the recovery, and control periods gave the false-alarm rate. We tested 11,726 SB1 and SB1C orbits of dwarf stars with M2 sin i < 150 MJ (eclipse test) and 3,377 orbits with P < 2 d (ellipsoidal test). Orbits that fail a periastron and Roche-lobe test are a control. For Roche-allowed orbits with M2 sin i < 80 MJ (desert companion mass function, 95 % intervals), the eclipse test gives f = 0.009 [0, 0.026] (4,254 stars; 28 detections against 386 expected and 14.9 false alarms; at most 0.105 without a false-alarm subtraction), and the ellipsoidal test gives f = 0.052 [0.039, 0.067] (1,381 stars; 56 detections against 980 expected). Most of the information comes from P < 4 d. For M2 sin i = 80-150 MJ, f = 0.23 [0.17, 0.30] (eclipse) and 0.34 [0.30, 0.38] (ellipsoidal). The impossible BD-class orbits give 0.013 and 0.022, which is the level of signals that do not come from the Gaia orbit. f increases with brightness, K/sigma_K and P, and decreases with RUWE. A logistic model of these Gaia quantities, fitted to the ellipsoidal stars, predicts the eclipse detections of the other stars (BD class: 28.0 predicted, 23 observed) and gives a score for 10,336 allowed stars. The ellipsoidal amplitudes of the BD-class detections are small (median 0.30 of the sin i = 1 value, against 0.54 for the star class; p = 0.0005). After a calibration on 10 eclipsing systems, which include WASP-18 b, 22 of 25 BD-class detections give a companion mass above 80 MJ. Thus most real BD-class orbits at P < 2 d are probably stellar companions at low inclination, and short-period BD candidates from Gaia DR3 SB1 orbits need independent confirmation. Main limitations: the companion mass function, an eclipse efficiency of 0.39 [0.28, 0.54] relative to the ellipsoidal test, a small number of control false alarms, possible blends, and at most two TESS sectors per star. The package has the code, per-star tables without Gaia catalogue columns, figures and reproduction steps.

Gaia DR3 · spectroscopic binaries · SB1 orbits · brown dwarf desert · brown dwarfs · TESS · eclipses · ellipsoidal variation · false positives · catalogue validation · injection-recovery · transit probability · Roche lobe · companion mass function · reliability model · WASP-18 · low-mass companions · QLP light curves