7.1.2 Input filtering
As mentioned above, the basic NSS processing selects its input lists by fixing an input threshold to the constancy level of the source measurements. As the various NSS pipelines do not use the same fitting procedure as the upstream processing, in particular concerning the definition of the goodness-of-fit and the robustness of the solution, the non-constancy definition may not be fully consistent with the upstream definition.
This section describes the adopted input criteria for each NSS channel. It is worth noting that if a source fulfills these criteria while not belonging to the corresponding NSS table, this means that the source has been rejected during the NSS processing.
Astrometric binaries
In order to select the sources with the best the best signal to noise, only sources with magnitude were kept for Gaia DR3. The detection of a non-constancy of the astrometry followed the indication in the Gaia DR2 documentation, a value of the ruwe “significantly greater than 1.0 (say, ) could indicate that the source is non-single or otherwise problematic for the astrometric solution”. Consequently the first criterion adopted to select candidates among the Gaia DR3 solutions as potential NSS candidates was ruwe .
However, this selection (mostly) contains contaminants originating from double or binary stars partially resolved double or binary stars rather than unresolved binaries. For these resolved sources, the position of the photocentre changes depending on the scan-angle and a period analysis may find spurious periodic motions actually related to the scanning law, as described in Holl et al. (2023a). Several criteria were then added to filter out these outliers, and keep only the unresolved sources, such as ipd_frac_multi_peak to avoid the doubles with a large separation or ipd_gof_harmonic_amplitude to reject pairs with smaller separations. The visibility_periods_used criteria was also added, in order to avoid spurious solutions, as a too small degree of freedom may have given a reasonable solution by chance only. Beside this filtering based on astrometric criteria, a photometric criterion was also added to avoid a potential flux contamination from a neighbour. For this purpose, the corrected BP and RP flux excess factor as defined by Riello et al. (2021, Eq. 6) has been used, together with its uncertainty (Riello et al. 2021, Eq. 18).
In summary, the sources analysed by the NSS astrometric channel were thus selected using the following criteria: phot_g_mean_mag & ruwe & ipd_frac_multi_peak & ipd_gof_harmonic_amplitude & visibility_periods_used &
Spectroscopic binaries
The sources were selected either among the sources having enough epoch radial velocities, rejecting too hot or too cold stars, i.e. using rv_renormalised_gof & rv_nb_transits & rv_template_teff or if the source had been detected as SB2 by the spectroscopic processing.
Eclipsing binaries
The input list for candidate eclipsing binaries contains about two millions sources and can be found in the Gaia DR3 table vari_eclipsing_binary. The construction of this list can be found in Section 10.7.1 and Mowlavi et al. (2023).