Ryan Rubenzahl (Flatiron Institute)

The discovery of extrasolar Earth-analogs with radial velocities requires 10 cm/s sensitivity over multi-year-to-decades long surveys. Instrumental and stellar variability across all relevant timescales and amplitudes complicate this goal; these are best studied with high signal-to-noise observations of the Sun, made hundreds of times every clear day by EPRV solar feeds. By comparing between multiple instruments, stellar variability (common) can be disentangled from instrumental drift (uncommon). We analyzed one year of KPF, NEID, EXPRES, and HARPS-N solar data at the line-by-line and pixel-by-pixel level to isolate the sources of each type of spectral variability. We developed a novel state space Gaussian process method to efficiently handle the large volume of solar data (~100k spectra) and account for the effects of exposure-integration when instruments are simultaneously exposing. We map instrumental changes across the detector to diagnose sources of drift, and map solar variability to layers of the solar atmosphere by modeling the contribution function for each spectral segment. The GP method naturally enables us to separate in the time domain how oscillations, granulation, supergranulation, and rotationally modulated effects themselves change throughout the solar atmosphere. We also correlate the pixel-level variability with SDO images to track which parts of the spectrum are especially sensitive to specific surface features, such as spots and plage. Ultimately, our goal is to assess across the spectrum which lines (or parts of lines) covary or not for each solar process, so that we may leverage such “families” of lines to better extract pristine Doppler shifts in stellar observations.