New astrometry confirms transverse accelerations of dark comets 2003 RM, 2012 UR158 and 2005 UY6; out-of-plane terms do not improve later position predictions for 1998 KY26, 2016 NJ33 and 2005 VL1

Question: Do the non-gravitational accelerations reported for 14 'dark comets' (Seligman et al. 2023, 2024) predict where these objects appear in later astrometry? Approach: We re-fitted all objects with an independent orbit code (ASSIST/REBOUND, JPL DE440 with 16 massive asteroids, MPC ADES astrometry, Eggl 2020 star-catalogue debiasing, own weights). Then we fitted gravity-only, A2, A2+A3 and A1+A2+A3 models only to the data that existed before each claim, or before the last apparition. We predicted the newer MPC positions (2021-2026) with each model. As a calibration, we applied the same test to 19 random near-Earth asteroids for which JPL fits only Yarkovsky A2. Results: Our full-data fits reproduce the published A2 and A3 values within about 2 sigma for most objects. Out-of-plane A3 is 7-21 sigma for 7 inner objects and for the linked arcs 2010 VL65 + 2021 UA12 (chi2 56 with A3, 17,169 with gravity only). No control has A3 at 3 sigma or more. For 2003 RM, 2012 UR158 and 2005 UY6, the transverse A2 fitted before the claim predicts the new positions to 0.3-0.5 arcsec rms. Gravity only misses by 2.9-11.3 arcsec. The new data show 0.91-1.00 of the predicted shift. For 1998 KY26 (2024 data), the A2-only orbit predicts better than the A2+A3 orbit (2.8 versus 3.3 arcsec rms). For 2016 NJ33 and 2005 VL1, the gravity-only orbit predicts the last apparition best (0.7 and 7 arcsec); all non-gravitational orbits miss by 8-51 arcsec. A3 improves forecasts only for 2016 GW221 and 2012 UR158, where the predicted A3 shift is about 0.2 arcsec. Even the full solar momentum flux would give the A2 of 2003 RM and 2012 UR158, and the A3 of 2016 NJ33, 2013 XY20 and 2016 GW221, only for bulk densities below about 230-730 kg/m3. Thus radiation forces cannot explain them. Conclusion: The large transverse accelerations of the outer dark comets are real and steady. The out-of-plane accelerations of the best-observed inner dark comets do not act as constant forces across apparitions. Limitations: Only 9 objects have a usable test, 2 of them on data used by the original claims, and some tests have 2-9 new observations. Our weights are simpler than JPL's, and our formal errors underestimate real forecast errors. 2006 RH120 and 2010 RF12 were not fitted. The primary analysis does not use radar.

dark comets · non-gravitational acceleration · near-Earth asteroids · orbit determination · Yarkovsky effect · astrometry · forecast test · 1998 KY26 · 2003 RM · ASSIST