Three years of HARPS-N high-resolution spectroscopy and precise radial velocity data for the Sun

Abstract

Context. The solar telescope connected to HARPS-N has been observing the Sun since the summer of 2015. Such a high-cadence, long-baseline data set is crucial for understanding spurious radial-velocity signals induced by our Sun and by the instrument. On the instrumental side, this data set allowed us to detect sub-m s −1 systematics that needed to be corrected for. Aims. The goals of this manuscript are to i) present a new data reduction software for HARPS-N, ii) demonstrate the improvement brought by this new software during the first three years of the HARPS-N solar data set, and iii) release all the obtained solar products, from extracted spectra to precise radial velocities. Methods. To correct for the instrumental systematics observed in the data reduced with the current version of the HARPS-N data reduction software (DRS version 3.7), we adapted the newly available ESPRESSO DRS (version 2.2.3) to HARPS-N and developed new optimised recipes for the spectrograph. We then compared the first three years of HARPS-N solar data reduced with the current and new DRS. Results. The most significant improvement brought by the new DRS is a strong decrease in the day-today radial-velocity scatter, from 1.27 to 1.07 m s −1 ; this is thanks to a more robust method to derive wavelength solutions, but also to the use of calibrations closer in time. The newly derived solar radial-velocities are also better correlated with the chromospheric activity level of the Sun in the long term, with a Pearson correlation coefficient of 0.93 compared to 0.77 before, which is expected from our understanding of stellar signals. Finally, we also discuss how HARPS-N spectral ghosts contaminate the measurement of the calcium activity index, and we present an efficient technique to derive an index free of instrumental systematics. Conclusions. This paper presents a new data reduction software for HARPS-N and demonstrates its improvements, mainly in terms of radial-velocity precision, when applied to the first three years of the HARPS-N solar data set. Those newly reduced solar data, representing an unprecedented time series of 34550 high-resolution spectra and precise radial velocities, are released alongside this paper. Those data are crucial to understand stellar activity signals in solar-type stars further and develop the mitigating techniques that will allow us to detect other Earths.

FIESTA – disentangling stellar variability from exoplanets in the Fourier domain

“…For example, for a typical spectrum CCF from the HARPS-N solar telescope (Dumusque et al 2021, see Section 5), we find that using up to k = 6 ΦESTA basis functions 1 reaches the modelling error of 6.7×10 −4 (Fig. 1), corresponding to a SNR 2 = 1.4 × 10 3 of the CCF it is able to represent.…”

“…not significantly affected by clouds), the differential extinction for each observation less than 0.1 m/s, and the data reduction software (DRS) reports good quality results. The same selection criteria were also used in de Beurs et al (2020) and Dumusque et al (2021). The HARPS-N CCFs and RVs used in Dumusque et al (2021), Collier Cameron et al (2021 and in this paper (daily binned as in first row of Fig.…”

“…The same selection criteria were also used in de Beurs et al (2020) and Dumusque et al (2021). The HARPS-N CCFs and RVs used in Dumusque et al (2021), Collier Cameron et al (2021 and in this paper (daily binned as in first row of Fig. 7) were processed by the new HARPS-N DRS as opposed to those used in de Beurs et al (2020) processed by an older version of the HARPS-N DRS.…”

“…3. CCD detector warm-ups: Instrumental features can be seen most clearly as the sudden jumps of ∆RV 1 and ∆RV 2 that coincide with times of interventions to the CCD detector (Dumusque et al 2021) . While not strictly periodic, these this effect likely leads to the periodogram peaks near ∼ 210 d and ∼ 285 d.…”

Stellar activity correction using PCA decomposition of shells

“…Hence, a Doppler-invariant average line profile is obtained, followed by a modelling of the variations seen in this line profile using a PCA. Results on the HARPS-N solar data set (Dumusque et al 2021) show that planetary signals with amplitudes as small as 0.4 m s −1 can be recovered. By using a convolution neural network (CNN) to model the variations seen at the CCF level, de Beurs et al (2020) show that an 3.5 Earth-mass planet at 365 days, with an amplitude of 0.3 m s −1 , could be recovered on the same solar data set.…”