Resolving X-ray Obscuration Biases with Isotropic AGN Selection

Abstract

Most mass is accreted onto supermassive black holes behind thick columns of gas and dust. An accurate assessment of the fraction of heavily obscured, "Compton-thick" AGN in the local Universe provides important insights into the composition and structure of the circum-nuclear AGN obscurer, as well as its connection with the evolution of supermassive black holes and their surrounding host galaxies across cosmic time. However, current estimates of the Compton-thick fraction vary dramatically between ~20-70%, and it remains unclear whether this large range is driven by selection effects, inadequate sample sizes, luminosity/Eddington rate dependencies, statistical issues associated with fitting low signal-to-noise X-ray spectra or something else entirely. The main handicap of previous works has been the inability to effectively select objects that are *representative* in terms of sampling N(H) parameter space, i.e. are unbiased even by Compton-thick obscuration. To investigate such issues, we present NuLANDS - a large far-infrared legacy survey with the X-ray satellites NuSTAR, XMM-Newton and Swift (~2 Ms of total new time) aimed at constructing an unbiased census of AGN obscuration in the local Universe. The infrared selection using AGN-like colours guarantees that we are not affected by line-of-sight X-ray obscuration biases, even into the log N(H)/cm-2 > 25 regime. In this talk, I will first report on multiple new Compton-thick AGN identified with our novel fitting approach, combining Nested Sampling with a large library of different geometrical models for the AGN obscurer. By fitting from the global multi-dimensional prior parameter space, Nested Sampling robustly estimates parameter uncertainties without requiring parameter tuning. Our results ultimately show that hard X-ray selection alone remains biased against the most heavily obscured AGN, and I will highlight the importance of multi-wavelength selection in completing the local AGN census with future next generation instruments. NuLANDS thus marks a major step in completing this census, and will provide vital boundary conditions for determining the composition of the Cosmic X-ray Background, as well as geometrical insights into the densest regions of the AGN torus.

2. The NuLANDS sample

**To select AGN isotropically of their intrinsic flux**, we used the warm IRAS selection from de Grijp+87 (see also Keel+94) using only detections at 25 & 60 micron in the IRAS PSC v2.1. In the unified model, IR continuum emission arises from dust reprocessing of accretion disc photons in the obscurer, **such that the selection should be largely isotropic of Seyfert type**.

*Left:* To weed contaminants, we classified warm IRAS sources with the spectral index (alpha) defined in de Grijp+85 to **identify flatter (i.e. hotter) IR spectra than cooler inactive galaxies**. *Middle:* Optical spectroscopic classifications were acquired and **non-active/ambiguous sources were removed**. *Right:* A volume cut defines **the NuSTAR Local AGN NH Distribution Survey** - a legacy survey of warm IRAS Seyferts with NuSTAR, Swift, XMM, Chandra & Suzaku.

*Left:* Approximate AGN schematic **highlighting the emission processes & size scales associated with the NuLANDS selection & classification techniques**. *Middle & right:* The distribution of [OIII] to 60 micron flux ratios is fit with a Gaussian model using UltraNest separately for type 1 (S1n, S1, S1.2, S1.5, S1.8) and type 2 (S1h, S1i, S2) AGN. **The istropic nature of the sample is confirmed from the nearly identical best-fit distributions for type 1 and 2 AGN.**

3. Results & contact info

**The obscurer geometries considered in NuLANDS.** To infer the most information possible from potentially low signal-to-noise spectra, **we paired an extensive X-ray spectral model library with nested sampling-based fitting** using the Bayesian X-ray Analysis software (Buchner+14). We fit a total of 24 models to every source, and use a combination of the **Akaike Information Criterion (AIC) and Bayes Factor (logZ) to determine favoured models**.

*Left:* The global line-of-sight logN_H distribution for the sample is determined from each individual source logN_H posterior with a **Bayesian Hierarchical Model** using UltraNest (Buchner+19) with a non-parametric histrogram model & Dirichlet prior. *Middle & right:* The resulting (consistent) N_H distributions generated depending on whether the favoured model is chosen with AIC or logZ.

Clearly a substantial fraction of Compton-thick AGN are directly detected in NuLANDS, highlighting the importance of isotropic multi-wavelength selection paired with the enhanced hard X-ray sensitivity of NuSTAR in probing the local AGN population. Furthermore, nested sampling paired with our large library of spectral models enables us to rule out a number of obscurer geometries per source, indicating current CCD-level spectra are capable of constraining the geometry of the AGN torus.

If you have any questions/comments or want to say hi, please join the Zoom link in the presentation information. If I am not able to join you, please drop me a message on Slack/e-mail to setup a different time to connect.

To see the larger version of my poster, click here!

2. The NuLANDS sample

3. Results & contact info

1. Motivation