Atmospheric dispersion correction: model requirements and impact on radial velocity measurements

Observations with ground-based telescopes are affected by differential atmospheric dispersion when seen at a zenith angle different from zero, a consequence of the wavelength-dependent index of refraction of the atmosphere. One of the pioneering technology in detecting exoplanets is the technique of radial velocity (RV), that can be affected by uncorrected atmospheric dispersion. The current highest precision spectrographs are expected to deliver a precision of 10 cm s^-1 (e.g., ESPRESSO). To minimize the atmospheric dispersion effect, an Atmospheric Dispersion Corrector (ADC) can be employed. ADC designs are based on sky dispersion models that nonetheless give different results; these can reach a few tens of milli-arcseconds (mas) in the sky (a difference up to 40 mas); a value close to the current requirements (20 mas in the case of ESPRESSO). In this paper we describe tests done with ESPRESSO and HARPS to understand the influence of atmospheric dispersion and its correction on RV precision. We also present a comparison of different sky models, using EFOSC2 data (between 600nm and 700nm), that will be used to improve on the design of ADCs.