Is Morphology All You Need? Quantifying Imprints of Past Star Formation in the Present-Day Morphologies of Galaxies
The next generation of galaxy surveys will bring a plethora of new observations, a vast majority of which will be observed with a few photometric broadbands, limiting the information we can infer from traditional SED fitting methods. To maximize the scientific value of these surveys, techniques that can go beyond traditional summary statistics in extracting information from galaxy observations are needed. Moving beyond integrated photometry, we introduce the Katachi-PRIME method to do spectral energy distribution (SED) based inference that jointly incorporates morphological and SED information. Katachi-PRIME uses implicit likelihood inference (ILI) to incorporate realistic morphological priors from gri broadband images, to make predictions of: stellar mass (M$_*$), specific star formation rate (sSFR), redshift (z), age (t$_{50}$), dust attenuation (Av), and gas-phase metallicity (12+log(O/H)) of the galaxy. We find that this improves constraints compared to traditional 5-band photometry. We then use the learned morphological priors to examine the extent to which morphology encodes information about past star formation, finding a stark dichotomy in how morphology encodes star formation history. While the stable morphologies of main-sequence and quiescent galaxies can predict past star formation out to $\sim 4-7$ Gyr and $\sim3-5$ Gyr, respectively, starburst and green-valley populations above and below the main sequence tend to experience structural disruptions that obscure all but the most recent star formation activity. Methods like Katachi-PRIME maximize the scientific value of future telescopes by providing broad-baseline SED-level property estimates by incorporating morphological information.