# 6.1.1 Gaia EDR3: radial velocities propagated from Gaia DR2

Gaia DR2, based on 22 months of data, was the first release to include radial velocities from the Radial Velocity Spectrometer (RVS). These were for 7.2 million stars brighter than $G_{\rm RVS}=12$ and with effective temperature $T_{\rm eff}$ in the range from about 3550 to 6900 K. The overall radial-velocity precision is around 200-350  m s${}^{-1}$at the bright end and approximately 1.8  km s${}^{-1}$ at the faint end, with significant dependence on $T_{\rm eff}$ (Section 6.5.2, Katz et al. 2019).

Gaia DR3, based on 34 months of data, has been split into two releases to make data products available as soon as they are ready. Gaia EDR3, which comes first, does not include any new spectroscopic data, which will be published in the full Gaia DR3. For convenience, the 7.2 million radial velocities from Gaia DR2 have been copied to Gaia EDR3, except for 14 800 sources for the following reasons:

• The re-publication of Gaia DR2 radial velocities provided the opportunity to clean up spurious cases: the 70 365 potentially contaminated sources reported in https://www.cosmos.esa.int/web/gaia/dr2-known-issues#RadialVelocitiesCrowdedRegions were analysed. A few extra sources with highest radial velocities not included in this potentially contaminated sample, but also candidate to be spurious, were analysed individually. In the end, 3876 radial velocities from Gaia DR2 were identified as spurious and not copied to Gaia EDR3. Among them are all sources with radial velocities greater than about 630  km s${}^{-1}$. This analysis, which made use of the radial velocity information already available from the data processing for Gaia DR3, is described in detail in Seabroke et al. (2020).

• The propagation of radial velocities from Gaia DR2 to Gaia EDR3 required cross-matching the sources between the two Data Releases. In practice, this cross-match could not be based exclusively on the nominal source identifier evolution (Section 3.4.13), because about 0.5% of the Gaia DR2 sources with radial velocities (i.e. $\sim 35\,000$ sources) could not be matched to the correct astronomical sources in Gaia EDR3 (see Torra et al. 2020 for issues pertaining to the source identifier evolution). As a result, a dedicated cross-match was performed using both the astrometric position and radial velocity information already available from the data processing for Gaia DR3, (Seabroke et al. 2020). In the end, only 10 924 Gaia DR2 radial velocities could not be associated with any Gaia EDR3 source.

It is recommended to treat the Gaia EDR3 radial velocity catalogue as independent from Gaia DR2.