Evaluating Strategies for Extremely Precise Radial Velocity Measurements with the Sun

Extremely Precise Radial Velocity (EPRV) observations of the Sun provide the best opportunity for tuning, verifying, and validating strategies for mitigating the impact of stellar variability on RV observations, a prerequisite for detecting rocky planets in the habitable zones of Sun-like stars. We present EPRV observations of the Sun using NEID and compare the effectiveness of multiple strategies for mitigating the effects of stellar variability, such as using curated line lists and data-driven models. Among EPRV spectrographs, NEID is unique thanks to its combination of instrumental stability and large wavelength grasp, allowing for detailed comparisons of line-shape metrics from the NUV to NIR. Further, the extremely high SNR of daily averaged spectra makes it practical to measure RVs from individual spectral lines much more precisely and/or to make EPRV measurements using many fewer lines than for typical night-time observations. We use this capability to evaluate the utility of individual spectral lines and to show the potential for measuring EPRVs via curated line lists. A new generation of EPRV spectrographs is advancing the state-of-the-art towards the detection of rocky planets in the habitable zones of Sun-like stars. The practical limit appears to be set by intrinsic stellar variability. Several strategies for mitigating the effects of stellar variability have been proposed, but it is not yet clear which will prove most practical, powerful, or reliable. Previous comparisons based on EPRV observations have found significant disagreements across methods, but have been unable to identify which methods are most accurate or reliable since true stellar velocity is unknown. The high environmental stability of NEID results in a stable line spread function (LSF) that allows us to use subtle line shape diagnostics as features for training data-driven models to mitigate stellar variability. Between early 2021 and 2023, NEID documented changes in solar spectra, as the Sun transitions from being magnetically quiet towards solar maximum. We evaluate the RV accuracy of models trained during one phase of the solar cycle work when applied during other phases.