[View Abstracts by Session]

Session: Extragalactic Astronomy and Astrophysics

Name: Mr. Charalampos Daoutis (IA-FORTH & Univ. of Crete)
Coauthors: Kyritsis Elias (University of Crete - I.A FORTH)
Zezas Andreas (University of Crete - I.A FORTH)
Kouroumpatzakis Konstantinos (Astronomical Institute, Czech Academy of Sciences)
Type: Oral
Title: A versatile classification tool for galactic activity using optical and infrared colors
Abstract:

The activity classification of galaxies is usually performed using fluxes of optical emission lines (e.g., emission-line ratios). So far, these diagnostics have been applied to a wide range of galaxy samples with great success. However, they suffer from two major disadvantages. The first is that the measurement of the flux of an emission line can often be challenging. The second is that the use of emission lines prohibits the applicability of these diagnostics to dormant systems, i.e., the passive galaxies. In this work, we define a new diagnostic tool by training a machine learning algorithm to characterize galactic activity (or the lack thereof) using as discriminating features two infrared colors, (W1-W2) and (W2-W3; WISE survey), and one optical color, g-r (SDSS or PanSTARRS). This new diagnostic classifies galaxies into five activity classes: star-forming, active galactic nucleus (AGN), LINER, composite, and passive. The overall accuracy achieved is ~81% while the achieved completeness is ~81% for the star-forming and ~85% for the passive galaxies. Besides the excellent scores achieved for the star-forming and passive galaxies, we also find that it offers significantly improved results for the AGNs in the local Universe compared to other commonly used mid-IR diagnostics which tend to miss the lower-luminosity AGN. Furthermore, another advantage of this new classification tool is its ability to identify very blue, metal poor, star-forming galaxies that are often misclassified as AGNs (AGN imposters) by the majority of the available mid-IR diagnostics.