2025ApJS..277...38Z

Atmospheric retrieval is a modeling technique used to determine a planet's atmosphere's temperature and composition from spectral data. The retrieved atmospheric composition can provide understanding into the surface fluxes of gaseous species necessary to maintain the stability of that atmosphere, leading to insights into the geological as well as biological processes active on the planet. Among exoplanets, rocky terrestrial ones are of particular interest because of their theoretical habitability. Atmospheric retrieval is both time consuming and computationally intensive. Traditional retrieval methods involve the use of complex algorithms that generate numerous atmospheric models. These models are then compared to observational data, and a posterior distribution is constructed to determine the most likely value and associated uncertainty for each model parameter. Runtimes scale with the number of model parameters, and when many molecular species are considered, become prohibitively long. The issue will become especially prohibitive as the number of detected exoplanets will grow tremendously in the near future. Machine learning (ML) offers a way to reduce the time to perform a retrieval by orders of magnitude, given a sufficient data set to train with. Here we present a large data set of 3,112,620 synthetic planetary systems generated with our Intelligent Exoplanet Atmospheric Retrieval framework based on the NASA Planetary Spectrum Generator. The data set contains the parameters defining each planetary system and the simulated spectra of stellar, planetary and noise components. The data set was designed to enable the first ML retrieval model for rocky terrestrial exoplanets, and it is publicly available through the NASA Exoplanet Archive.