Gaia Early Data Release 3. Updated radial velocities from Gaia DR2

Context. Gaia's Early Third Data Release (EDR3) does not contain new radial velocities because these will be published in Gaia's full third data release (DR3), expected in the first half of 2022. To maximise the usefulness of EDR3, Gaia's second data release (DR2) sources (with radial velocities) are matched to EDR3 sources to allow their DR2 radial velocities to also be included in EDR3. This presents two considerations: (i) a list of 70 365 sources with potentially contaminated DR2 radial velocities has been published; and (ii) EDR3 is based on a new astrometric solution and a new source list, which means sources in DR2 may not be in EDR3.
Aims: The two aims of this work are: (i) investigate the list in order to improve the DR2 radial velocities being included in EDR3 and to avoid false-positive hypervelocity candidates; and (ii) match the DR2 sources (with radial velocities) to EDR3 sources.
Methods: Thetwo methods of this work are: (i) unpublished, preliminary DR3 radial velocities of sources on the list, and high-velocity stars not on the list, are compared with their DR2 radial velocities to identify and remove contaminated DR2 radial velocities from EDR3; and (ii) proper motions and epoch position propagation is used to attempt to match all sources with radial velocities in DR2 to EDR3 sources. The comparison of DR2 and DR3 radial velocities is used to resolve match ambiguities.
Results: EDR3 contains 7 209 831 sources with a DR2 radial velocity, which is 99.8% of sources with a radial velocity in DR2 (7 224 631). 14 800 radial velocities from DR2 are not propagated to any EDR3 sources because (i) 3871 from the list are found to either not have a DR3 radial velocity or it differs significantly from its DR2 value, and five high-velocity stars not on the list are confirmed to have contaminated radial velocities, in one case because of contamination from the non-overlapping Radial Velocity Spectrometer windows of a nearby, bright star; and (ii) 10 924 DR2 sources could not be satisfactorily matched to any EDR3 sources, so their DR2 radial velocities are also missing from EDR3.
Conclusions: The reliability of radial velocities in EDR3 has improved compared to DR2 because the update removes a small fraction of erroneous radial velocities (0.05% of DR2 radial velocities and 5.5% of the list). Lessons learnt from EDR3 (e.g. bright star contamination) will improve the radial velocities in future Gaia data releases. The main reason for radial velocities from DR2 not propagating to EDR3 is not related to DR2 radial velocity quality. It is because the DR2 astrometry is based on one component of close binary pairs, while EDR3 astrometry is based on the other component, which prevents these sources from being unambiguously matched.