8.2.1 Selection of objects for Astrometry, Photometry, and Spectrophotometry

The selection parameters are given in Table 8.1. The list has been created in three main steps:

• •

  Prediction of all the possible transits for a set of SSOs over the Gaia DR3 time span using the Gaia orbit, the scanning law, and a numerical integration of the SSO motion starting from the osculating elements and osculating epoch given in Astorb (Bowell et al. 1994).

• •

  Cross-matching the transits with the Gaia IDT detections to ensure that a potential candidate SSO had been actually seen by Gaia and observed during the predicted transit. Match first by the crossing time (computed vs. IDT to 0.1 s) and then by the sky coordinates within a window basically of 1${.}^{\mathrm{\prime \prime }}$5.

• •

  Iterative filtering on the tentative list to ensure that the selection criteria were met and no contaminants were left.

These conditions were applied iteratively until a satisfactory selection meeting all the requirements was achieved. The final input selection had 3 513 248 transits for 156 837 known and numbered asteroids. The still unnumbered asteroids were not considered in this selection due to their expected poorly determined orbits due to insufficient observational data. The coverage in orbital semimajor axis displayed in Figure 8.4 shows that SSOs within each of the broad categories are found in the selection. Pluto may include the resolved pair Pluto and its largest satellite Charon matched to different transits.

The sky distribution of the selected observations is shown in Figure 8.5 on a density plot in equatorial coordinates. The SSOs are all found in the vicinity of the ecliptic plane, but the distribution in longitude is markedly non-uniform with periodic features of higher and lower concentration. This is a known artefact resulting from the Gaia scanning over a relatively short duration of 34 months compared to the nominal mission of five years. The features seen every 60${}^{\circ }$ have an annual shift of 72${}^{\circ }$, so that the coverage becomes more uniform after five years.

An important feature for the orbit reconstruction is the number of times an asteroid is observed during the 34 months of the Gaia DR3 data set, and even more crucial is the number of different periods of time, well separated from each other. The distribution of the number of transits is shown in Figure 8.6. The fainter the asteroid, the smaller is the number of detected transits, since at large elongation the apparent brightness drops below the detection limit. This is a selection bias that impacts the quality of the orbits for the asteroids with the IAU number above $\sim$50 000. The censoring is better seen with the bottom plot which shows the same histogram but for the first 20 000 numbered asteroids, which are on the average brighter and seen without loss at any solar elongation that Gaia can explore.

The observations of the set of natural satellites observable with Gaia have been searched in the IDT data like the asteroids. Their passages were first predicted using the ephemeris of the satellites provided by V. Lainey (personal communication) in 2016, the Gaia orbit, and the scanning law. The predicted list is then compared to the IDT transits and a match is selected if there is an agreement in the transit time and position. Due to the proximity of the planet, the number of contaminants is much higher than usual and sometimes the selection of a single match may become somewhat arbitrary. This is the case in particular for the Galilean satellites. Unlike for the asteroids, there has been no filtering for the number of detected transits, given the fact that even one very accurate observation of a satellite has scientific value.

Out of the 44 possible satellites (2 for Mars, 18 for Jupiter, 14 for Saturn, 7 for Uranus, and 3 for Neptune), 42 have been matched to at least one IDT transit. The two missing ones are the jovian satellites Amalthea and Thebe. We have found 999 transits associated with natural satellites. On one hand, some satellites, like the saturnian satellites Hyperion and Dione, are well observed with more than 50 transits. On the other hand, the jovian satellite Callirrhoe, the saturnian satellites Mimas and Albiorix, and the neptunian satellite Proteus are matched only once as shown in Table 8.2.

For Gaia DR3, it has been decided to select transits of potential observations of new asteroids based on an independent search carried out in the IDT data. As this search relies heavily on the along-scan motion ($V_{\mathrm{AL}}$) estimates provided by the IDT, the search is reliable only after 1 December 2016 (IDT Run 1214), when an updated version of the IDT code has considerably improved the quality of this estimate. Therefore, for Gaia DR3, there were altogether only 6 months of data exploitable for this search.

This is a search without matching in position but by the systematic recognition that an IDT transit with a significant along-scan motion ( $>1.5$ mas s${}^{-1}$) can be paired with another one after an interval of time corresponding to the passages PFOV–FFOV and FFOV–PFOV, and the same separated with at least one satellite revolution (6 h). The actual interval of time must be compatible with the estimated $V_{\mathrm{AL}}$ or, in other words, the great-circle motion between the two epochs projected along-scan should be close to the estimated $V_{\mathrm{AL}}$. Likewise, the two $V_{\mathrm{AL}}$ values of the potential pair must be also compatible. In a second step, all the pairs are examined to search for a longer chain, with 3 or more observations in a sequence found in consecutive IDT runs.

The present search is run with two different modes, differing by the acceptance criterion applied to validate the detection:

• •

  In the first mode, the search is limited to asteroids detected only in two consecutive passages, usually the preceding and following fields of view, but sometimes with a larger interval, but always less than two satellite revolutions. The checks that these two passages are from the same moving source are rather limited with only two transits. Essentially, the positional change should agree with the AL motion determined by IDT and the two values of magnitude and motion should be similar.

• •

  The second mode is more relaxed regarding the acceptance thresholds since one has at least three transits available presumably from the same object, implying more constraints to establish the consistency expected in the case that they are from the same source. We have identified sources with $4,5,\mathrm{\dots },42$ chained transits. In this case, the safety margin from the available constraints is very high and the risk of an erroneous identification goes down quickly with the number of observations. It is obvious that a real moving source has been found and properly chained. Whether it is a new asteroid will remain unknown until its positions and even its computed orbit could be identified with the latest data from the IAU Minor Planet Center (MPC).

Obviously, in the search, one finds also all the known asteroids, and this would be the great majority if the corresponding transit identifiers were not stored in ancillary files, read in parallel to discard these matches. Actually, this test can be disabled, allowing to evaluate and improve, when feasible, the efficiency and the reliability of the algorithm.

The list for DR3 comprises 4522 transits of unmatched asteroids (some new at this date, some with too poor an orbit to be matched to a transit, see Section 8.5) corresponding to 1531 groups of chained transits. This may correspond to a maximum of 1531 new asteroids, but it cannot be excluded that the same sources have been detected at two different epochs, but their bundled transits could not be chained together. This will be done at a later stage when a good preliminary orbit is computed. On one hand, about 60 % of the asteroids fall in the group of only 2 transits and, on the other hand, there are 60 with at least 7 transits. The detailed distribution is listed in Table 8.3. It is important to keep in mind that there is a delay of about 4 years (for Gaia DR3) between the production of this list and the actual release of the results. So the flag for a matched or unmatched SSO is based on the existing orbital data available at the end of 2017 (see Section 8.5).