2017A&A...606A..90L

Aims. We compared positions of the Gaia first data release (DR1) secondary data set at its faint limit with CCD positions of stars in 20 fields observed with the Very Large Telescope (VLT) FORS2. We aim at an independent verification of the DR1 astrometric precision taking advantage of the FORS2 position uncertainties which are smaller than one milli-arcsecond (mas).
Methods. In the fields that we observed with FORS2, we projected the Gaia DR1 positions into the CCD plane, performed a polynomial fit between the two sets of matching stars, and carried out statistical analyses of the residuals in positions.
Results. The residual root mean square (rms) roughly matches the expectations given by the Gaia DR1 uncertainties, where we identified three regimes in terms of Gaia DR1 precision: for G~=17-20 stars we found that the formal DR1 position uncertainties of stars with DR1 precisions in the range of 0.5-5mas are underestimated by 63±5%, whereas the DR1 uncertainties of stars in the range 7-10mas are overestimated by a factor of two. For the best-measured and generally brighter G~=16-18 stars with DR1 positional uncertainties of <0.5mas, we detected 0.44±0.13mas excess noise in the residual rms, whose origin could be in both FORS2 and Gaia DR1. By adopting Gaia DR1 as the absolute reference frame, we refined the pixel scale determination of FORS2,leading to minor updates to the parallaxes of 20 ultracool dwarfs that we published previously. We also updated the FORS2 absolute parallax of the Luhman 16 binary brown dwarf system to 501.42±0.11mas.