2019A&A...629A..27C

Context. Doppler spectroscopy has been used in astronomy for more than 150yr. In particular, it has permitted us to detect hundreds of exoplanets over the past 20 yr, and the goal today of detecting Earth-like planets requires a precision around 0.1m/s or better. Doppler spectroscopy has also been and will be of major importance for other studies such as the variability of fundamental constants and cosmological studies. For all these applications, it is crucial to have the best possible wavelength calibration. Despite the fact that the HARPS spectrograph has been operational at the 3.6-m ESO telescope for more than 15yr, and that it provides among the most precise Doppler measurements, improvements are still possible. One known problem, for instance, is the non-fully regular block-stitching of the charge-coupled devices (CCDs), which in some cases introduces one-year period parasitic signals in the measured radial velocity.
Aims. The aim of the presented work is to improve the wavelength calibration of the HARPS spectrograph to push further its planet-detection capabilities.
Methods. The properties of the CCD stitching-induced pixel-size anomalies were determined with light-emitting-diode (LED) flat-field frames, and then a physical, gap-corrected map of the CCDs is used for the fitting model of the spectral orders. We also used a new thorium line list, based on much higher-accuracy measurements than the one used up to now. We derive new wavelength solutions for the 15yr of HARPS data, both before and after the fibre upgrade that took place in 2015.
Results. We demonstrate that we do indeed correct the gap anomalies by computing the wavelength solutions of laser frequency comb exposures, known to have a very low dispersion, both with and without taking the gap correction into account. By comparing the rms of the most stable stars of the HARPS sample, we show that we globally decrease the radial velocity (RV) dispersion of the data, especially for the data acquired after the change of fibres of 2015. Finally, the comparative analysis of several individual systems shows that we manage to attenuate the periodogram power at one year in most cases. The analysis of the RVs derived from individual stellar lines also shows that we indeed correct the stitching-induced RV variation.
Conclusions. This improved calibration of the HARPS spectrograph allows to go deeper in the search for low-amplitude radial-velocity signals. This new calibration process will be further improved by combining the thorium calibration spectra with laser frequency comb and Fabry-Perot calibration spectra, and this will not only be used for HARPS but notably also for HARPS-N and the new ESPRESSO spectrograph.