The importance of non-photon noise in stellar spectropolarimetry - The spurious detection of a non-existing magnetic field in the A0 supergiant HD?92207

摘要

Context. The low-resolution, Cassegrain mounted, FORS spectropolarimeter of the ESO Very Large Telescope is being extensively used for magnetic field surveys. Some of the new discoveries suggest that relatively strong magnetic fields may play an important role in numerous physical phenomena observed in the atmospheres as well as in the circumstellar environments of certain kinds of stars. Aims. We show in detail how small instabilities or data-reduction inaccuracies represent an alternative explanation for the origin of certain signals of circular polarisation published in recent years. Methods. With the help of analytical calculations we simulate the observation of a spectral line in spectropolarimetric mode, adding very small spurious wavelength shifts, which may mimic the effects of seeing variations, rapid variations of the stellar radial velocity, or instrument instabilities. As a case study, we then re-visit the FORS2 measurements that have been used to claim the discovery of a magnetic field in the A0 supergiant HD?92207. In addition, we present new observations of this star obtained with the HARPSpol instrument. Results. Both calibration and science data show compelling evidence that photon-noise is not the only source of error in magnetic field measurements, especially in sharp spectral lines. Non-photon noise may be kept under control by accurate data reduction and quality controls. Our re-analysis of FORS2 observations of HD?92207 shows no evidence of a magnetic field, and we are able to reproduce the previous FORS detection only by degrading the quality of our wavelength calibration. Our HARPSpol spectropolarimetric measurements show no evidence of a magnetic field at the level of 10?G. Conclusions. Our work contributes to a better understanding of the importance of accurate data treatment and instrument characterisation, and demonstrates that ultra-high signal-to-noise ratio measurements do not automatically translate into ultra-high accuracy.