Measuring precise radial velocities on individual spectral lines I. Validation of the method and application to mitigate stellar activity

Context. Stellar activity is the main limitation to the detection of an Earth-twin using the radial-velocity (RV) technique. Despite many efforts in trying to mitigate the effect of stellar activity using empirical and statistical techniques, it seems that we are facing an obstacle that will be extremely difficult to overcome using current techniques. Aims. In this paper, we investigate a novel approach to derive precise RVs considering the wealth of information present in high- resolution spectra. Methods. This new method consists of building a master spectrum from all available observations and measure the RVs of each individual spectral line in a spectrum relative to this master. When analysing several spectra, the final product of this approach is the RVs of each individual line as a function of time. Results. We demonstrate on three stars intensively observed with HARPS that our new method gives RVs that are extremely similar to the one derived from the HARPS data reduction software. Our new approach to derive RVs demonstrates that the non-stability of daily HARPS wavelength solution induces night-to-night RV offsets with an standard deviation of 0.4 m s(-1), and we propose a solution to correct for this systematic. Finally, and this is probably the most astrophysically relevant result of this paper, we demonstrate that some spectral lines are strongly affected by stellar activity while others are not. By measuring the RVs on two carefully selected subsample of spectral lines, we demonstrate that we can boost by a factor of two or mitigate by a factor of 1.6 the red noise induced by stellar activity in the 2010 RV measurements of alpha Cen B. Conclusions. By measuring the RVs of each spectral line, we are able to reach the same RV precision as other approved techniques. In addition, this new approach allows us to demonstrate that each spectral line is differently affected by stellar activity. Preliminary results show that studying in details the behaviour of each spectral line is probably the key to overcome the obstacle of stellar activity.