Title:
The faults in our stars
Abstract:
Stellar spectroscopy is the backbone of astrophysics — and it’s broken. Imperfect models lead to systematic errors that cascade to nearly every astronomical measurement, limiting our understanding of how planets, stars, and galaxies form and evolve. Even our best models do not agree on the chemical composition of the Sun, the yardstick we use to compare all light in the universe. Unfortunately, stellar models will probably remain imperfect for the rest of our lives: time-varying 3D models remain prohibitively expensive for survey-scales despite compute availability, and critical labs that measure electronic transitions have recently closed after over 120 years of service. Without a new approach to tame systematics, stellar astrophysics may stagnate amidst the 300 million spectra expected by 2030. In this talk I will describe concerted efforts we have made in the Sloan Digital Sky Survey to address systematic effects at every stage: theoretically by including additional physics in spectral synthesis codes and quantifying systematics; by modernizing data reduction pipelines; and using hierarchical Bayesian models to learn pixel-level systematic effects that would otherwise bias our inferences. In doing so we are delivering fundamental stellar properties that are — for the first time — statistically consistent with non-spectroscopic techniques, and precise chemical abundances. These measurements will enable science that is otherwise impossible, and define a legacy standard for the next decade.