Mapping H$\alpha$-Excess Candidate Point Sources in the Southern Hemisphere Using S-PLUS Data

TL;DR

This study uses S-PLUS photometry and machine learning to identify and classify H-alpha-excess point sources in the Southern Hemisphere, improving the distinction between Galactic and extragalactic objects and identifying variable stars.

Contribution

It introduces a novel combination of multiwavelength data, clustering, and machine learning techniques to classify H-alpha-excess sources with enhanced accuracy.

Findings

Identified 6956 H-alpha-excess sources in the Southern sky.

Improved classification of emission-line stars, galaxies, and quasars.

Enhanced separation of variable stars like RR Lyrae using colour data.

Abstract

Context. We use the Southern Photometric Local Universe Survey (S-PLUS) Fourth Data Release (DR4) to identify and classify H$\alpha$-excess point sources in the Southern Sky, combining photometric data from 12 S-PLUS filters with machine learning to improve classification of H$\alpha$-related phenomena. Aims. Our goal is to classify H$\alpha$-excess point sources by distinguishing Galactic and extragalactic objects, particularly those with redshifted emission lines, and identifying variability phenomena like RR Lyrae stars. Methods. We selected H$\alpha$-excess candidates using the ($r - J0660$) vs. ($r - i$) colour-colour diagram from the S-PLUS main survey (MS) and Galactic Disk Survey (GDS). UMAP for dimensionality reduction and HDBSCAN clustering were used to separate source types. Infrared data was incorporated, and a Random Forest model was trained on clustering results to identify key colour features. New colour-colour diagrams from S-PLUS MS and infrared data offer a preliminary classification. Results. Combining multiwavelength data with machine learning significantly improved H$\alpha$-excess source classification. We identified 6956 sources with excess in the $J0660$ filter. Cross-matching with SIMBAD explored object types, including emission-line stars, young stellar objects, nebulae, stellar binaries, cataclysmic variables, QSOs, AGNs, and galaxies. Using S-PLUS colours and machine learning, we separated RR Lyrae stars from other sources. The separation of Galactic and extragalactic sources was clearer, but distinguishing cataclysmic variables from QSOs at certain redshifts remained challenging. Infrared data refined the classification, and the Random Forest model highlighted key colour features for future follow-up spectroscopy.