2023A&A...676A..11L

Context. The link of the Gaia frame in terms of non-rotation with respect to the International Celestial Reference System (ICRS), which is realized via very long baseline interferometry (VLBI) at radio wavelengths, has to be conducted for the wide range of optical magnitudes in which the spacecraft observes. There is a sufficient number of suitable counterparts between the two measurement systems for optically faint objects. However, the number of common optically bright (G ≤ 13 mag) objects is sparse as most are faint at radio frequencies, and only a few objects suitable for astrometry have been observed by VLBI in the past. As a result, rotation parameters for the optically bright Gaia reference frame are not yet determined with sufficient accuracy.
Aims. The verification of the Gaia bright frame of DR2 and EDR3 is enhanced by the reevaluation of existing VLBI observations and the addition of newly acquired data for a sample of optically bright radio stars.
Methods. Historical data from the literature were reevaluated, ensuring that the calibrator positions and uncertainties (used for the determination of the absolute star positions in the phase-referencing analysis) were updated and homogeneously referred to the ICRF3, the third realization of the International Celestial Reference Frame. We selected 46 suitable optically bright radio stars from the literature for new radio observations, out of which 32 were detected with the VLBA in continuum mode in the X or C band, along with radio-bright calibrators in the ICRF3. Improved Gaia-VLBI rotation parameters were obtained by adding new observations and utilizing more realistic estimates of the absolute position uncertainties for all phase-referenced radio observations.
Results. The homogenization greatly improved the steadiness of the results when the most discrepant stars were rejected one after another through a dedicated iterative process. For Gaia DR2, this homogenization reduced the magnitude of the orientation parameters to less than 0.5 mas but increased that of the spin parameters, with the largest component being the rotation around the Y axis. An adjustment of the position uncertainties improved the reliability of the orientation parameters and the goodness of fit for the iterative solutions. Introducing the new single-epoch positions to the analysis reduced the correlations between the rotation parameters. The final spin for Gaia DR2 as determined by VLBI observations of radio stars is (-0.056, -0.113, +0.033) ± (0.046, 0.058, 0.053) mas yr^–1. A comparison of the new results with external, independently derived spin parameters for Gaia DR2 reveals smaller differences than when using the historical data from the literature. Applying the VLBI data to Gaia EDR3, which was already corrected for spin during Gaia processing, the derived residual spin is (+0.022, +0.065, -0.016) ± (0.024, 0.026, 0.024)mas yr^–1, showing that the component in Y is significant at the 2.4σ level.
Conclusions. Even though our analysis provides a more accurate frame tie, more VLBI data are needed to refine the results and reduce the scatter between iterative solutions.