The converged attitude, calibration and global parameters obtained in the primary solution
were used to perform a source update on all sources which had at least one AstroElementary with valid cross matched AF
transits. The solution was done performing a five-parameter update and the full covariance matrix
was used to compute the formal astrometric errors and correlations.
As for the primary source update all the secondary sources where treated as single stars and the reference epoch of
the solution was J2015.5. The primary sources where also updated in the secondary update as
this is easier for practical reasons and also to ensure that all sources in the catalogue were treated in the same way.
The attitude used for the secondary update was already in the ICRF3 reference frame which automatically ensures that all
the source positions for the secondary data set are also in the same reference frame.
Sources failing to meet the acceptance criteria for a full five-parameter solution
were relegated to a two-parameter solution instead (so just the position was determined), as
described in Lindegren et al. (2018) section 4.3. Five-parameter and two-parameter solutions
in the Archive can be distinguished using the field astrometric_params_solved which
will take the values 31 (a 5 parameter solution) and 3 (a 2 parameter solution) respectively. This step included a computation
of the pseudo-colour for all sources using the chromaticity calibration from the
primary run. The pseudo-colour, is much less precise than
the calculated from and photometry, but has the
advantage that it can be obtained for every source. In the secondary update
we used this calculated pseudo-colour (given in the Gaia Archive as
astrometric_pseudo_colour) as a proxy for for all sources.
We adopted this scheme to ensure the most uniform astrometric treatment of the
sources; moreover, it mitigates the astrometric impact of the BP, RP flux excess
issue Evans et al. (2018), which tends to make faint sources in crowded areas
too blue as measured by photometry. Normally, however, the pseudo-colour is not an
astrophysically useful estimate of the colour due to its
low accuracy compared with the photometric estimates.