Stellar activity correction using PCA decomposition of shells

Context. Stellar activity and instrumental signals are the main limitations to detection of Earth-like planets using radial velocity (RV) technique. Recent studies show that key mitigating those perturbing effects might reside in analysing spectra themselves, rather than RV time series a few proxies. Aims. The goal this paper is demonstrate we can reach further improvement precision by performing principal component analysis (PCA) decomposition shell series, with as projection spectrum onto spacenormalised flux versus gradient. Methods. By PCA it possible obtain basis first-order spectral variations not related Keplerian motion. coeffcients associated then be used correct for non-Dopplerian signatures RVs. Results. We applied new method on YARARA post-processed HD10700 HD128621. On HD10700, demonstrate, thanks planetary signal injections, approach successfully disentangle real Dopplerian from systematics. application methodology HD128621 shows strong stellar seen at rotational period one-year aliases becomes insignificant periodogram analysis. root mean square five-year data reduced 2.44 m/s down 1.73 m/s. This allows us strongly mitigate activity, however, noise injections tests indicate high signal-to-noise ratio (S/N>250) required observed