A machine learning algorithm for reliably predicting active galactic nucleus absorbing column densities^⋆

¹ Department of Physics and Astronomy, Clemson University, Kinard Lab of Physics, Delta Epsilon Ct., Clemson, SC, 29634 USA
e-mail: rmsilve@clemson.edu
² Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA, 02138 USA
³ Dipartimento di Fisica e Astronomia (DIFA), Universià di Bologna, Via Gobetti 93/2, 40129 Bologna, Italy
⁴ INAF – Osservatorio di Astrofisica e Scienza dello Spazio di Bologna, Via Piero Gobetti, 93/3, 40129 Bologna, Italy

Received: 23 January 2023
Accepted: 16 May 2023

Abstract

We present a new method for predicting the line-of-sight column density (N_H) values of active galactic nuclei (AGN) based on mid-infrared (MIR), soft X-ray, and hard X-ray data. We developed a multiple linear regression machine learning algorithm trained with WISE colors, Swift-BAT count rates, soft X-ray hardness ratios, and an MIR–soft X-ray flux ratio. Our algorithm was trained off 451 AGN from the Swift-BAT sample with known N_H and has the ability to accurately predict N_H values for AGN of all levels of obscuration, as evidenced by its Spearman correlation coefficient value of 0.86 and its 75% classification accuracy. This is significant as few other methods can be reliably applied to AGN with Log(N_H < 22.5). It was determined that the two soft X-ray hardness ratios and the MIR–soft X-ray flux ratio were the largest contributors toward accurate N_H determinations. We applied the algorithm to 487 AGN from the BAT 150 Month catalog with no previously measured N_H values. This algorithm will continue to contribute significantly to finding Compton-thick (CT) AGN (N_H ≥ 10²⁴ cm⁻²), thus enabling us to determine the true intrinsic fraction of CT-AGN in the local Universe and their contribution to the cosmic X-ray background.