HSTbanner

AR 16626 (Archival Research)

Mon Nov 11 04:48:02 GMT 2024

Principal Investigator: Desika Narayanan
PI Institution: University of Florida
Investigators (xml)

Title: SED Fitting in the Modern Era: Fast and Accurate Machine-Learning Assisted Software
Cycle: 29

Abstract
Recovering physical properties from galaxies via spectral energy distribution (SED) modeling is dependent on a number of assumptions, including galaxy star foramtion histories and dust attenuation curves. As we show in this proposal, uncertainties in these assumptions canpropagate to significant errors in the derived stellar masses and star formation rates of galaxies. The fundamental goal of this proposal is to minimize the uncertainties incurred by these assumptions. To accomplish this, we propose to develop a new public-facing machine-learning assisted software that maps observational photometry to the derived physical properties of galaxies. Our proposed software utilizes synthetic SEDs from cosmological hydrodynamic galaxy formation simulations as a training set, and as we demonstrate, shows significant promise in outperforming modern state-of-the-art traditional SED fitting software. We will develop our new machine learning-assisted SED fitting software to (a) enable more accurate physical property determinations from both archival and future HST observations of galaxies near and far, and (b) couple with the ETC to aid observers in planning future observations that are motivated by determining particular physical properties to below a certain error threshold.