skip to main content

gaia early data release 3 documentation

9 Cross-match with Gaia DR2

9.3 Quick inspection

A basic analysis of the table contents is presented in this section. Some key numbers are provided below. The corresponding ADQL queries can be found in Section 9.4.2.

  • Gaia DR2 gaia_source length: 1,692,919,135

  • Gaia Gaia DR2 sources in Gaia DR2 neighbourhood table: 1,689,802,302 (99.82%)

  • Gaia Gaia DR2 sources in Gaia DR2 neighbourhood table, one neighbour closer than 100 mas: 1,685,437,881 (99.56%)

  • Gaia Gaia DR2 source_id not in Gaia EDR3 but in Gaia DR2 neighbourhood table: 49,600,212 (2.93%)

It is apparent that most Gaia DR2 objects have a close Gaia EDR3 counterpart. The persistence of source_id between releases is now much higher than for Gaia DR2. However, there is still a small but significant fraction of objects for which the source_id changes. Some cases could probably be explained by multiplicity or crowding. However, it seems there is still some room for improvement.

The closest neighbours for a random sample of 10 million Gaia DR2 sources have been plotted in Figure 9.1. Gaia EDR3 sources with and without proper motions are shown separately. It is clear that matches are much better for objects with proper motion. The existence of these two distinct populations explains why no single rank column has been created, as for Gaia DR2. In any case, small radii such as 100 mas seem to collect most true counterparts between both releases.

Figure 9.1: Closest Gaia EDR3 neighbour to each Gaia DR2 source. A random sample of 10 million Gaia DR2 sources has been used. Gaia EDR3 sources with and without proper motions are plotted separately. The corresponding ADQL query can be found in Section 9.4.3.

If the same random sample is split between objects preserving the source_id between releases and not, similar results are obtained at small angular distances. See Figure 9.2. Spurious matches seem to dominate at distances around a few hundred mas.

Figure 9.2: Gaia DR2 source_id within and without Gaia EDR3 distance histogram. A random sample of 10 million Gaia DR2 sources has been used. Three curves are overimposed: Gaia DR2 source_id matches to Gaia EDR3, all Gaia EDR3 neighbours within 2” to each Gaia DR2 source_id not in Gaia EDR3 and only the closest pairs in the previous sample. The corresponding ADQL queries can be found in Section 9.4.4.

G band magnitude differences for source pairs closer than 100 mas have been represented in Figure 9.3. At this close range, most matches are expected to be due to the same physical object. The histogram shows that indeed most magnitudes are very similar.

Figure 9.3: G magnitude difference for neighbours within 100 mas. The corresponding ADQL query can be found in Section 9.4.5.

The validity of the approximate linear correction for high proper motion stars has been verified in Figure 9.4. There is no clear trend of errors getting larger for higher proper motion. The average angular distance between closest pairs is below a mas, most of them being below 100 mas. Histograms for the angular distance and G band magnitude difference are shown after applying this cut in Figure 9.5. The good behaviour of the cone search in this regime is evident. The distribution of magnitude errors seems slightly narrower than for the whole sample.

Figure 9.4: High proper motion stars: distance vs proper motion. All objects with proper motion larger than 100 mas/yr have been included. Two data sets are displayed: all DR2-Gaia EDR3 neighbours and those with the minimum angular distance. The corresponding ADQL queries can be found in Section 9.4.6.
Figure 9.5: High proper motion stars: angular distance (top) and magnitude difference (bottom) histograms. All objects with proper motion larger than 100 mas/yr have been included. The corresponding ADQL query can be found in Section 9.4.7.