A Closer Look at the Origin of LINER Emission in Later-type Galaxies and Its Connection to Evolved Stars with a Machine Learning Classification Scheme

Identifying the dominant ionizing sources in galaxies is essential for understanding their formation and evolution. Traditionally, spectra are classified based on their dominant ionizing source using strong emission lines and Baldwin, Phillips, & Terlevich (BPT) diagrams. The ionizing source is traditionally determined by the emission line ratios using the BPT diagrams. Low-ionization nuclear emission-line regions (LINERs) are a class of ionizing mechanisms that is observationally identified but with a poorly understood origin, unlike the case of star-forming regions and active galactic nuclei (AGN). LINERs, typically found in early-type galaxies, are often associated with low-luminosity AGN activity but may also be powered by aging stellar populations, particularly post-asymptotic giant branch (p-AGB) stars. In this study, we employ a machine-learning-based encoder, spender, to analyze the full MaNGA integral field unit spectra and identify key spectral features of LINERs. By examining the continuum and line emission of these spaxels, our approach aims to uncover hidden patterns and better understand the dominant ionizing sources. We show in this work that the neural-network-based encoder was able identify LINER sources from the stellar continuum alone. The characteristics of the stellar population underlying LINER regions are consistent with evolved low-mass stars, implying that the source driving LINER emission is probably p-AGB stars rather than AGN activity.