Classification of Galactic Planetary Nebulae Among HASH Candidates

Abstract

We present a multi-diagnostic classification of 149 new planetary nebula (PN) candidates drawn from the HASH database for which optical spectra are available. Emission-line fluxes were extracted using the automated fitting code alfa. Of the 149 candidates, 26 exhibit emission lines above a 3σ detection threshold and form the working sample. We apply layered diagnostic criteria comprising classical emission-line ratio criteria, BPT and SMB diagnostic diagrams, a symbiotic star (SySt) discriminator based on [O III] auroral-to-nebular ratios, and a group-aware supervised machine-learning classifier (CatBoost; ROC–AUC ∼0.96). Two objects are classified as PN candidates (PN G022.5+14.4 and PN G023.2+13.6) and one as a tentative PN candidate (PN G021.3−05.0), one as a possible symbiotic-star candidate (PN G025.6-05.9), nine as SNRs, three as objects with mixed HII Region/SNR characteristics, and ten as HII regions. Under a strict high-purity operating threshold ($\hat{p}_{\rm thr} \approx 0.75$), the CatBoost classifier flags three high-confidence candidates. We find that the machine-learning classification is consistent with the diagnostics derived from the emission-line analysis.