Description
Granulation in the stellar photosphere is a major limitation in achieving precise radial velocity (RV) measurements. Granulation-induced signals occur on timescales of tens of minutes, with amplitudes of ~1 m/s for Sun-like stars and tens of cm/s for later-type stars. We use synthetic spectra of stellar atmospheres generated by MPS-ATLAS from MURaM simulations of the photosphere, excluding magnetic activity. In a line-by-line approach, we measure variations in line depth and shift over a simulated time series of spectra. For both properties, we apply singular value decomposition (SVD) to model line responses to granulation. We use a reduced-rank representation with a small number of principal components, which serve as decorrelation vectors. The principal components reveal distinct line responses to granulation, depending on their formation temperature and atomic parameters. These decorrelation vector patterns vary significantly with spectral type due to differences in spectral line characteristics.
Similar line responses are observed in real data from the Sun and other stars. In particular, years of high-cadence observations with ultra-stable spectrographs provide ideal datasets for studying granulation. By applying the decorrelation vectors derived from synthetic spectra to a low-activity subset of HARPS-N solar data, we achieve a significant reduction in scatter in full-spectrum RV measurements. This approach to mitigating granulation-induced RV variations can be applied to other stars, improving RV precision and advancing the search for low-mass exoplanets.
