Exoplanet detection with precise radial velocity (RV) observations is currently limited by spurious RV signals introduced by stellar activity (i.e. faculae, starspots). Here we show that machine learning techniques such as linear regression and neural networks can significantly remove the activity signals (primarily starspots/faculae) from real center-of-mass RV shifts. Previous efforts have focused on carefully filtering out activity signals in time using Gaussian process regression (e.g. Haywood et al. 2014). Instead, we separate activity signals from true center-of-mass RV shifts using only changes to the average shape of spectral lines, and no information about when the observations were collected.
We have trained our machine learning models on both simulated data (generated with the SOAP 2.0 software; Dumusque et al. 2014) and observations of the Sun from the HARPS-N Solar Telescope (Dumusque et al. 2015; Phillips et al 2016; Collier Cameron et al. 2019). ). We find that these techniques can successfully predict and remove stellar activity from approximately 700 observations taken nearly daily over three years with the HARPS-N Solar Telescope, reducing the RV scatter from 1.47m/s to 0.78 m/s (a 47% or factor of ~ 1.9 improvement). For the simulated data, these techniques can predict the stellar activity signal nearly exactly, reducing the RV scatter from 82.0 cm/s to 3.1 cm/s which suggests that we may be able to achieve more precise velocities when applied to data from newer stabilized spectrographs like ESPRESSO. In the future, these or similar techniques could remove activity signals from observations of stars outside our solar system and eventually help detect habitable-zone Earth-mass exoplanets around Sun-like stars.
Acknowledgement - ZLD acknowledges the generous support from the UT Office of Undergraduate Research, the TIDES Advanced Research Fellowship and the Junior Fellows Honors Program. ZLD and AV acknowledge support from the TESS Guest Investivator Program under NASA grant 80NSSC19K0388.