Due to the superb astrometric accuracy and the small epoch difference between Gaia DR1 and Gaia DR2 (half a year), proper motion corrections are not needed for most stars for the purpose of neighbour identification, as opposed to the full treatment used in Section 2.4.9. For these slow moving objects, a search radius of 0.5” has been deemed bigger enough than the typical mas level positional precision for faint stars. Note their associated distances in dr1_neighbourhood are thus computed between non-coeval catalogues. From the implementation point of view, it is a close-neighbours search between DR2 and DR1. This table has been created directly on the database using the q3c_ join operator, and a multi-thread approach as described in González-Núñez et al. (2017).
High proper motion stars have been analysed using the epoch propagation routines implemented within the archive. For simplicity, they have been divided into two bins in proper motion: [0.1,1) and [1,10.4] arcsec/yr. They contain approximately half a million and a thousand objects, respectively. The maximum proper motion corresponds to Barnard’s star. Fixed cone search radii of 2 and 20 arcsec have been used for each bin. They correspond to two years times the maximum proper motion. The larger time lapse compared to the catalogue difference is needed to allow the inclusion of badly cross-matched DR1 observations. That is, high proper motion stars have some times generated multiple DR1 sources aligned in the proper motion direction, due to the limitations of the cross-match algorithm used at that time.
Regarding the implementation, cross-match identifications for high proper motions sources on the main DR2 to DR1 table are replaced by the more accurate and corrected output of the new generated tables.