Three Years of HARPS-N High-Resolution Spectroscopy and Precise Radial Velocity Data for the Sun

The solar telescope connected to HARPS-N has observed the Sun since the summer of 2015. Such high-cadence, long-baseline data set is crucial for understanding spurious radial-velocity (RV) signals induced by our Sun and by the instrument. This manuscript describes the data reduction performed to obtain unprecedented RV precision for the three years of solar data released along with this paper. The nearly continuous observation of our Sun has allowed us to detect sub-\ms\,systematics in the HARPS-N solar data reduced by the current HARPS-N data reduction software (DRS). To improve the RV precision of the solar data, we reduced them using the new ESPRESSO DRS and developed new recipes to mitigate the detected systematics. The most significant improvement brought by the new data reduction is a strong decrease in the day-to-day RV scatter, from 1.28 to 1.09\ms; this is thanks to a more stable method to derive wavelength solutions, but also to the use of calibrations closer in time. We also demonstrate that the current HARPS-N DRS induces a long-term drift of $\sim$1.2\ms, due to the use of non-stable thorium lines. As a result, the old solar RVs are weakly correlated to the solar magnetic cycle, which is not expected. On the contrary, the newly derived RVs are much more correlated, with a Pearson correlation coefficient of 0.93. We also discuss a special correction for ghost contamination, to extract a calcium activity index free from instrument systematics. Our work leads toward a better understanding of the instrumental and data reduction systematics affecting the HARPS-N spectrograph. The new solar data released, representing an unprecedented time-series of 34550 high-resolution spectra and precise RVs will be crucial to understanding stellar activity signals of solar-type stars, with the goal of enabling the detection of other Earths.