Abstract We introduce a novel galaxy classification methodology based on the visible spectra of a sample of over 68,000 nearby (z ≤ 0.1) Sloan Digital Sky Survey lenticular (S0) galaxies. Unlike traditional diagnostic diagrams, which rely on a limited set of emission lines and class dividers to identify ionizing sources, our approach provides a comprehensive framework for characterizing galaxies regardless of their activity level. By projecting galaxies into the 2D latent space defined by the first three principal components (PCs) of their entire visible spectra, our method remains robust even when data from individual emission lines are missing. We employ Gaussian kernel density estimates of the classical Baldwin-Phillips-Terlevich (BPT) activity classes in the new classification subspace, adjusted according to their relative abundance in our S0 sample, to generate probability maps for star-forming, Seyfert, composite, and LINER galaxies. These maps closely mirror the canonical distribution of BPT classes shown by the entire galaxy population, demonstrating that our PC-based taxonomy effectively predicts the dominant ionizing mechanisms through a probabilistic approach that provides a realistic reflection of galaxy activity and allows for refined class membership. Our analysis further reveals that flux-limited BPT-like diagrams are inherently biased against composite and star-forming galaxies due to their weaker [O iii] emission. Besides, it suggests that although most low-activity galaxies excluded from these diagnostics exhibit visual spectra with LINER-like characteristics, their remaining activity is likely driven by mechanisms unrelated to either star formation or supermassive black hole accretion. A machine-readable catalogue listing BPT-class probabilities for the galaxies analysed is available online at CDS website.
Read full abstract