7.4.2 Properties of the input data

The input data are made of a list of RVs (and their 1$\sigma$-uncertainties) that are corrected to make reference to the barycentre of the solar system. It is to be noted that the error on the barycentric correction is not propagated to the error on the barycentric RV. The barycentric RVs are accompanied by a time of observations expressed in barycentric Julian days with respect to the Gaia DR3 reference epoch. The number of points of the time series corresponds to the number of epochs (transits) for which the radial velocity has been successfully measured. The data could present several weaknesses such as outliers and wrong radial velocities due e.g. to template mismatch errors. These problems are documented in the spectroscopic processing Chapter 6 and will not be further detailed here. The outliers, if existing, were supposed to be removed by the spectroscopic processing and no further treatment is applied in the present processing. The post-filtering of the RVS spectroscopic processing pipeline (Section 6.5.2) has not been fully applied on the data processed by the spectroscopic binary processing for reasons of priorities.

Some selections are also made on the objects. In the case of Gaia DR3, the kind of stars is restricted to the range of effective temperatures 3 875 K to 8 125 K (mainly ((M))-K-G-F-(A) stars). No emission-line object have been selected and in principle only normal stars are entering the game. The objects with a rv_renormalised_gof larger than 4 were rejected (Katz et al. 2023). The measurements of the epoch (per transit) radial velocities are restricted to the range –1 000 to +1 000 $\mathrm{km}{\mathrm{s}}^{-1}$. This measurement of the individual RVs on individual spectra is only possible for bright objects with a $G_{\mathrm{RVS}}$ mag adopted threshold. This strongly reduces the amount of selected objects. The stellar RVs have not been processed if the number of transits (data points) was less than 10.

The nature of the measurement itself (cross-correlation with templates) implies that it is not possible to discriminate between a change in RV due to a global shift of the line (as expected for SB1 objects) and a shift due to a line-profile variation (as expected e.g. for intrinsic variables). Although possibilities to get rid of this problem exist, they have not been implemented for Gaia DR3 and this certainly constitutes a limitation of the spectroscopic binary pipeline. In particular, fake SB1 could persist in the sample corresponding to never-deblending SB2.

When the SB2 channel (Section 7.5 and Damerdji et al. 2022) analyses an SB2 time-series and the amount of RVs associated with the secondary object is too low, the object is considered as an SB1 and is forwarded to the present channel. This is a second possibility of entering the SB1 channel.