8.5 Unmatched Solar System objects
Author(s): François Mignard, Thierry Pauwels
When the list of unmatched objects was set up in early January 2017, not every transit of sources displaying a significant sky-plane motion could be identified with known SSOs by position matching. An additional population of unmatched sources was created in the input list. These sources were selected primarily on the basis that there were Gaia transits clearly associated with moving objects but not paired to computed positions. In general, only a couple of transits within a single day could be linked together, providing a strong confirmation of the detection of an unmatched source. In a few cases, a sufficient number of successive transits could be associated with the same object allowing to compute a preliminary orbit. However, even with a preliminary orbit, it was not possible to match the transits to a known source, at least with an orbit reliable enough to match the few positions with certainty. Whether the moving bodies were truly new, detected for the first time with a favourable geometry, or simply too poorly known, was impossible to settle in 2017. However, it was important for us to add these transits into the astrometric basket of Gaia DR3. Otherwise this opportunity would have been lost, at least for the present release.
Since then, a lot of new ground-based observations have been treated by the MPC, and many new orbits have been published, while existing orbits have been improved. Poor orbits in 2017 turned into reliable ones in 2021, allowing to better match the computed position to Gaia observations. In June 2021, the list of unmatched threads was checked against the list of orbits then available. At that time, already 675 out of the 1320 threads could be identified with known asteroids. This list is continuously increasing, while new orbits are being published at the MPC. This exercise will be done again in a later stage, closer to the data release, and the full list of identifications will be published in a supplementary paper.
Two other independent experiments pointed exactly in the same direction. First, repeating the match of January 2017 for a trial sample, but by using an orbit file of 2020 instead of 2017, increased the number of direct position matches, thus depleting the population of unmatched transits. The second test was even more significant: for the set of unmatched sources having at least 10 transits in near succession, it was possible to run an initial orbit determination. Then a match was searched between these crude orbits and the most recent Astorb release (in this case October 2021), by also propagating the crude orbit to the same epoch as Astorb. This is a match in the 6D space of orbital parameters instead of a match in position between an observed and computed place. The results were clear, since in this limited test on about 100 orbits, only a couple of cases did not come with an unambiguous association to an Astorb orbit. Repeating this with an older version of Astorb (late 2018) was not so conclusive: either Gaia detections at that time did not correspond to an asteroid in the Astorb file or the orbital elements were not close enough to make the match reliable.
No measures can be taken to correct for these features that are deeply rooted in the Gaia timeline, with a period of at least four years between the start of the preliminary operations of a processing cycle and the release of the final products. We will know later how many of these secondary detections (those not matched to computed positions) were really unknown asteroids at the time they were caught by Gaia and pinpointed by the CU4 SSO search algorithm.