Unveiling Obscured Accretion in the Local Universe

Indrani Pal, Stefano Marchesi, Ross Silver, Marco Ajello, Vittoria Gianolli, Núria Torres-Albà, Isaiah Cox, Xiurui Zhao, Dhrubojyoti Sengupta, Anuvab Banerjee, Kouser Imam, Andrealuna Pizzetti

First listed 2025-09-25 | Last updated 2025-09-25

Abstract

Heavily obscured Active Galactic Nuclei (AGN), especially Compton-thick sources with line-of-sight column density ($N_{\rm H,los}$) $>$ 10$^{24}$ cm$^{-2}$, are critical to understanding supermassive black hole (SMBH) growth and the origin of the Cosmic X-ray Background (CXB). However, their observed fraction remains significantly below model predictions, due to strong absorption bias, even in the hard X-ray (i.e., above 10 keV) band. We analyze a sample of 26 nearby ($z < 0.1$) AGN from the Swift-BAT 150-month catalog, selected via mid-IR to X-ray diagnostics and observed with NuSTAR and soft X-ray telescopes (Xmm-Newton, Chandra, or Swift-xrt). Using self-consistent torus models (MyTorus, Borus02, and UXCLUMPY), we aim to constrain $N_{\rm H,los}$, the average torus column density, and other geometrical parameters of the obscuring medium. A comparative analysis among the three torus models showed that while estimates of $N_{\rm{H,los}}$ were generally in agreement, Borus02 tended to classify a slightly larger number of sources as Compton-thick AGN (CT-AGN). Building on this comparison, we benchmark two prediction schemes -- a mid-IR/X-ray relation and a machine-learning model -- against our broadband best-fit $N_{\rm H,los}$ measurements to assess which approach more effectively bridges the gap between predicted and measured obscuration, finding that while the former works effectively in the heavily obscured region (log$\rm{N_H} \gtrsim$ 23.5 $\rm{cm^{-2}}$), the latter provides improved accuracy, particularly for Compton-thin to moderately thick regimes (log$\rm{N_H} \lesssim$ 23.5 $\rm{cm^{-2}}$).

Short digest

A 26-object, z<0.1 Swift-BAT sample selected by 12 μm–to–X-ray ratios is modeled with broadband (0.5–50 keV) NuSTAR+soft-X spectra using MyTorus, Borus02, and UXCLUMPY to recover N_H,los and torus geometry. N_H,los values are broadly consistent across models, but Borus02 yields a slightly higher Compton-thick fraction. Benchmarking predictors shows the MIR/X relation performs best at heavy obscuration (log N_H ≳ 23.5), while a machine-learning model is more accurate for Compton-thin to moderately thick regimes (log N_H ≲ 23.5). Together these results reduce obscuration bias in the local CT census, improving links to CXB synthesis and SMBH growth.

Key figures to inspect

• Figure 1: Check how the WISE 12 μm to 2–10 keV flux ratio maps onto the Asmus-predicted N_H,los; the regression line and color-coded types show how the MIR/X selection targets obscured systems and where outliers sit.
• Figure 2: For IC 2227, compare the unfolded spectra and residuals across MyTorus, Borus02, and UXCLUMPY to see differences in Compton hump strength and Fe Kα behavior that drive model-dependent N_H,los and CT classification.
• Figure 3: Inspect the CT-AGN counts per model; note how Borus02 increases the CT bin and where gray “uncertain N_H” bars cluster near the CT threshold.
• Figure 4: Cross-compare N_H,los from all three models; look for scatter and systematic offsets around the 1:1 line, especially for sources straddling the Compton-thick boundary.