Inference of Star Formation and Metallicity Histories from Galaxy Spectra with Score-Based Models

Star formation histories (SFHs) describe when stars in individual galaxies were formed, and are thus a key quantity in understanding galactic evolution. As SFHs themselves are unobservable, information about a galaxy’s SFH must be inferred from the galaxy’s spectrum, which is an ill-posed inverse problem. In this work, we train a score-based diffusion model to act as a prior over the SFHs. We collect our training data, SFHs and metallicity histories (MHs), from TNG50 simulations, ensuring that the samples include realistic features. We test our model by doing Bayesian posterior inference on mock observations. We apply our model to spectra from the Sloan Digital Sky Survey and find good agreement with previous results. Our work shows it is possible to infer detailed and realistic SFHs and MHs, opening a window to the study of galactic histories at a level of detail never before possible.