Improved precision of radial velocity measurements after correction for telluric absorption^★

¹ LATMOS, Université Versailles-Saint-Quentin, 11 Bd D’Alembert, 78280 Guyancourt, France
e-mail: anastasia.ivanova@latmos.ipsl.fr
² Space Research Institute (IKI), Russian Academy of Science, Profsoyusnaya street, 84/32, Moscow 117997, Russia
³ GEPI, Observatoire de Paris, PSL University, CNRS, 5 place Jules Janssen, 92190 Meudon, France
e-mail: rosine.lallement@obspm.fr
⁴ LATMOS, Sorbonne Université, 4 place Jussieu, 75005 Paris, France

Received: 29 September 2022
Accepted: 5 February 2023

Abstract

Context. The detection of planets around other stars by measurement of stellar radial velocity (RV) variations benefits from improvements to dedicated spectrographs that can lead to the achievement of a precision of 1 m s⁻¹ or better. Spectral intervals within which stellar lines are contaminated by telluric lines are classically excluded from the RV processing.

Aims. We aim to estimate the potential improvement of the precision of radial velocity measurements from telluric-absorption removal and subsequent extension of the useful spectral domain.

Methods. We developed a correction method based on the online web service TAPAS that allows a synthetic atmospheric transmission spectrum to be determined for the time and location of observations. This method was applied to telluric H₂O and O₂ absorption removal from a series of 200 consecutive ESPRESSO exposures of the K2.5V star HD40307 available in ESO archives. We calculated the radial velocity using the standard cross-correlation function (CCF) method and Gaussian fit of the CCF, with uncorrected spectra and the ESPRESSO standard stellar binary mask on one hand, and telluric-corrected spectra and an augmented binary mask with 696 additional lines on the other.

Results. We find that the precision of radial velocity measurements is improved in the second case, with a reduction of the average formal error from 1.04 m s⁻¹ down to 0.78 m s⁻¹ in the case of these ESPRESSO data and this stellar type for the red arm. Using an estimator of the minimal error based on the photon noise limit applied to the full CCF, the error is reduced from 0.89 m s⁻¹ down to 0.78 m s⁻¹. This corresponds to a significant decrease (by about 35%) in the observing time required to reach the same precision in the red part.