zELDA II: Reconstruction of galactic Lyman-alpha spectra attenuated by the intergalactic medium using neural networks

Abstract

The observed Lyman-alpha line profile is a convolution of complex radiative transfer through the interstellar, circumgalactic, and intergalactic media. Discerning these components is crucial for using Lyα as a probe of galaxy formation and intergalactic-medium evolution. We present the second version of zELDA, an open-source Python module for modeling and fitting observed Lyα line profiles, with a focus on disentangling galactic and intergalactic effects. We construct realistic profiles by combining Monte Carlo shell-model simulations with intergalactic transmission curves generated from TNG100, and use them to train artificial neural networks. The networks take an observed spectrum as input and infer the best-fitting shell-model outflow parameters, source redshift, and intergalactic Lyα escape fraction. On mock line profiles, zELDA reconstructs the emerging interstellar-medium Lyα profile with high accuracy for 95% of HST/COS-like observations and 80% of MUSE-WIDE-like observations. It measures intergalactic transmission with typical uncertainties below 10% for both datasets. zELDA therefore enables high-precision reconstruction of intergalactic-medium-attenuated Lyα profiles and separates the galactic and intergalactic contributions shaping the line.