ZTF, Gaia, OGLE and ASAS-SN data confirm steady eclipse-depth changes in 3 of 10,385 detached Gaia DR3 eclipsing binaries; most other ZTF depth trends do not continue

We searched 20,615 detached eclipsing binaries from the Gaia DR3 catalogue (dec > -28 deg, G 13-16.5, P 0.4-6 d) for eclipse-depth changes in ZTF DR24 light curves from 2018 to 2025. A steady depth change is the usual sign of orbital-plane precession by a third body. We fitted the eclipse depth in each season, combined independent light curves, and used a fixed candidate rule. 10,385 targets had usable data. 46 targets have two independent light curves with same-sign trends at |z| >= 2.5, against 3 with opposite signs. But the Gaia DR3 catalogue depths from 2014-2017 do not follow the ZTF trends of the population (regression coefficient 0.03, 95% interval 0.00 to 0.07; real linear trends give -1). Thus most significant ZTF depth trends are not long-lived geometric changes. An injection test gave 0 false candidates in 515 targets and 12% (50%) recovery for depth changes of 2%/yr (5%/yr). 20 targets passed the rule. We fitted a two-sphere eclipse model with only the inclination free per epoch, and we used independent data. Independent data confirm a steady eclipse-depth change in three systems: Gaia DR3 1945251062721688960 (twin eclipses, P = 5.2561 d, +0.67 +/- 0.06 deg/yr; ZTF g and Gaia DR3 epochs agree), Gaia DR3 4069286250900737920 = OGLE-BLG-ECL-209157 (P = 1.6345 d, +0.47 +/- 0.10 deg/yr; OGLE-IV 2010-2015 and Gaia epochs continue the trend for 15 years), and Gaia DR3 2204624756208243712 = ASASSN-V J221453.49+612643.2 (P = 1.2753 d, -0.57 +/- 0.06 deg/yr; ZTF g and ASAS-SN g agree, but the TESS eclipse shape does not show the inclination change, so its cause is less certain). The other candidates are spotted or contact-like systems, a 1-day sampling alias, or trends that a second filter or Gaia does not repeat. The rates agree with nodal precession by tertiaries with outer periods of roughly months to years, but the data do not measure the third bodies. We give model depth forecasts for 2026-2030. Limitations: a simple spherical-star model, a non-blind choice of the confirmation checks, low completeness for changes below about 30%, and strongly diluted TESS data that confirm none of the changes.

eclipsing binaries · orbital precession · triple stars · eclipse depth variation · inclination change · ZTF · Gaia DR3 · OGLE · ASAS-SN · TESS · time-domain survey · injection-recovery