High confidence AGN candidates among unidentified Fermi-LAT sources via statistical classification

TL;DR

This study develops a statistical classification method to identify high-confidence AGN candidates among unassociated gamma-ray sources in the Fermi-LAT catalog, significantly aiding future multi-wavelength identification efforts.

Contribution

The paper introduces a dual-algorithm statistical approach that improves confidence and reduces false positives in classifying AGN candidates among unassociated gamma-ray sources.

Findings

Identified 231 high-confidence AGN candidates among unassociated sources.

Reduced false-association rate to 11%.

Enhanced candidate selection for follow-up observations.

Abstract

The second Fermi-LAT source catalog (2FGL) is the deepest survey of the gamma-ray sky ever compiled, containing 1873 sources that constitute a very complete sample down to an energy flux of about 10^(-11) erg cm^(-2) s^(-1). While counterparts at lower frequencies have been found for a large fraction of 2FGL sources, active galactic nuclei (AGN) being the most numerous class, 576 gamma-ray sources remain unassociated. In these proceedings, we describe a statistical algorithm that finds candidate AGNs in the sample of unassociated 2FGL sources by identifying targets whose gamma-ray properties resemble those of known AGNs. Using two complementary learning algorithms and intersecting the high-probability classifications from both methods, we increase the confidence of the method and reduce the false-association rate to 11%. Our study finds a high-confidence sample of 231 AGN candidates among the population of 2FGL unassociated sources. Selecting sources out of this sample for follow-up observations or studies of archival data will substantially increase the probability to identify possible counterparts at other wavelengths.