A versatile classification tool for galactic activity using optical and infrared colors

Charalampos Daoutis,Elias Kyritsis,Konstantinos Kouroumpatzakis,Andreas Zezas

DOI: https://doi.org/10.1051/0004-6361/202347016

2023-10-04

Abstract:(abridged) The overwhelming majority of diagnostic tools for galactic activity are focused on active galaxies. Passive or dormant galaxies are often excluded from these diagnostics which usually employ emission line features. In this work, we use infrared and optical colors in order to build an all-inclusive galactic activity diagnostic tool that can discriminate between star-forming, AGN, LINER, composite, and passive galaxies, and which can be used in local and low-redshift galaxies. We explore classification criteria based on infrared colors from the 3 WISE bands supplemented with optical colors from the u, g, and r SDSS bands. From these we aim to find the minimal combination of colors for optimal results. Furthermore, to mitigate biases related to aperture effects, we introduce a new WISE photometric scheme combing different sized apertures. We develop a diagnostic tool using machine learning methods that includes both active and passive galaxies under one unified scheme using 3 colors. We find that the combination of W1-W2, W2-W3, and g-r colors offers good performance while the broad availability of these colors for a large number of galaxies ensures wide applicability on large galaxy samples. The overall accuracy is $\sim$81% while the achieved completeness for each class is $\sim$81% for star-forming, $\sim$56% for AGN, $\sim$68% for LINER, $\sim$65% for composite, and $\sim$85% for passive galaxies. Our diagnostic provides a significant improvement over existing IR diagnostics by including all types of active, as well as passive galaxies, and extending them to the local Universe. The inclusion of the optical colors improves their performance in identifying low-luminosity AGN which are generally confused with star-forming galaxies, and helps to identify cases of starbursts with extreme mid-IR colors which mimic obscured AGN galaxies, a well-known problem for most IR diagnostics.

Astrophysics of Galaxies

What problem does this paper attempt to address?

The aim of this paper is to develop a new diagnostic tool for galaxy activity classification that can simultaneously distinguish between active galaxies (including star-forming galaxies, active galactic nuclei AGN, low-ionization nuclear emission-line region LINER galaxies, and composite galaxies) and passive galaxies. Traditional diagnostic methods mainly rely on optical spectral features to identify active galaxies, but these methods often exclude passive or dormant galaxies and have limitations in identifying low-redshift or low-luminosity AGNs.

The goal of the paper is to construct a comprehensive galaxy activity classification tool by combining infrared and optical colors, which can not only identify various types of active galaxies but also cover passive galaxies and be applicable to local and low-redshift galaxy samples. Specifically, the researchers utilized a random forest algorithm to define a new diagnostic tool to overcome the limitations of traditional optical spectral line-based classification methods, especially for galaxies whose optical spectral lines are difficult to detect due to interstellar material absorption.

To achieve this goal, the research team adopted a method that combines infrared data from the WISE satellite (Wide-field Infrared Survey Explorer) with optical data from the SDSS (Sloan Digital Sky Survey). By using machine learning techniques, they determined the optimal color combinations to improve classification performance. This approach not only handles large-scale datasets but also captures complex relationships in multi-dimensional parameter space, thereby enhancing the accuracy of identifying different types of galaxies.