SIMBAD references

We present an investigation of the stability of the two chips in the FORS2 camera CCD mosaic on the basis of astrometric observations of stars in 20 sky fields, some of which were monitored for four to seven years. We detected a smooth relative shear motion of the chips along their dividing line that is well approximated by a cubic function of time with an amplitude that reaches ∼0.3-pixels (px) or ∼38mas over seven years. In a single case, we detected a step change of ∼0.06px that occurred within four days. In the orthogonal direction that corresponds to the separation between the chips, the motion is a factor of 5-10 smaller. This chip instability in the camera significantly reduces the astrometric precision when the reduction uses reference stars located in both chips, and the effect is not accounted for explicitly. We found that the instability introduces a bias in stellar positions with an amplitude that increases with the observation time span. When our reduction methods and FORS2 images are used, it affects stellar positions like an excess random noise with an rms of ∼0.5mas for a time span of three to seven years when left uncorrected. We demonstrate that an additional calibration step can adequately mitigate this and restore an astrometric accuracy of 0.12mas, which is essential to achieve the goals of our planet-search program. These results indicate that similar instabilities could critically affect the astrometric performance of other large ground-based telescopes and extremely large telescopes that are equipped with large-format multi-chip detectors if no precautions are taken.