Identifying Compton-thick active galactic nuclei in the COSMOS. II. Searching among mid-infrared selected AGNs

This study identifies Compton-thick active galactic nuclei within a sample of 1,104 mid-infrared-selected, X-ray-undetected AGNs in the COSMOS field by combining MIR diagnostics and X-ray stacking analysis, confirming 23 candidates but finding that they constitute only 2.1% of the sample, suggesting that a significant population of Compton-thick AGNs remains undetected by current selection methods.

The Gist

Here is an explanation of the paper, translated into everyday language with some creative analogies.

The Big Picture: The "Ghost" Black Holes

Imagine the universe is a giant, noisy party. Most of the guests are Active Galactic Nuclei (AGNs)—supermassive black holes at the centers of galaxies that are actively eating matter and shooting out huge beams of light and energy.

Astronomers have a theory about how many of these black holes should be "heavy eaters" (Compton-thick AGNs). They think about 30% of all these black holes are so heavily wrapped in thick blankets of gas and dust that they are almost invisible to our X-ray telescopes. It's like trying to see a campfire through a thick brick wall; the light is there, but it's blocked.

However, when astronomers look at the sky, they only find a tiny fraction of these "blanketed" black holes. It's a mystery: Where are the missing 90%?

This paper is the second part of a detective story trying to find these missing black holes in a specific region of the sky called COSMOS.

---

The Detective's Strategy: Looking for the "Heat" Instead of the "Light"

Usually, astronomers hunt for black holes by looking for X-rays (the "light" from the fire). But if the black hole is wrapped in a thick blanket, the X-rays can't get out.

The New Trick:
Instead of looking for the light, the team decided to look for the heat.

• The Analogy: Imagine a person wrapped in a thick winter coat. You can't see their face (the X-rays), but if you stand close enough, you can feel the heat radiating off their body (the Mid-Infrared light).
• The Method: The team used a special list of galaxies that glow brightly in Mid-Infrared (MIR) light. They reasoned that if a black hole is wrapped in a thick dust blanket, that dust gets hot and glows in infrared. So, they picked 1,104 galaxies that looked like they had these "hot blankets" but didn't show up in X-ray surveys.

The Investigation: Stacking the Evidence

The team had a problem: Even though these 1,104 galaxies were glowing in infrared, they were still too faint to see individually in X-rays. It was like trying to hear a whisper in a noisy room; you can't hear one person, but if 20 people whisper at the same time, you might hear a hum.

The "Stacking" Technique:

1. The Filter: They used a mathematical formula to guess which of these 1,104 galaxies were the most likely "blanketed" black holes. They found about 23 candidates.
2. The Stack: They took the X-ray data for these 23 specific galaxies and "stacked" them on top of each other.
3. The Result: When they combined the data, a signal appeared! It was a clear "hum" in the soft X-ray band. This proved that these 23 galaxies do have black holes, but they are indeed heavily hidden.

The Verdict: We Found Some, But Not Enough

The team confirmed that these 23 sources are indeed Compton-thick AGNs (the heavily wrapped ones).

However, there is a twist.

• The Expectation: Theory says 30% of the black holes should be hidden.
• The Reality: In their sample of 1,104 galaxies, they only found 23 hidden ones. That's only 2.1%.

The Conclusion:
The team realized that even their clever "heat-seeking" method missed a huge number of these black holes. It's like using a metal detector to find gold, but the detector is only sensitive enough to find big nuggets, missing all the tiny flakes.

They estimate that there are still hundreds of these "ghost" black holes hiding in the data that their current tools can't find. To find them, we will need even deeper, more sensitive X-ray telescopes in the future.

The Side Note: Do They Live in Different Neighborhoods?

The team also asked: Do these hidden black holes live in different types of galaxies compared to the visible ones?

• The Question: Maybe the hidden ones live in galaxies that are super busy making new stars (like a construction zone), while the visible ones live in quiet neighborhoods.
• The Answer: Surprisingly, no. They found that the "hidden" black holes and the "visible" ones live in galaxies that look very similar in terms of size and star-making activity. They are just neighbors who happen to be wearing different coats.

Summary in a Nutshell

1. The Mystery: We think there are many black holes hidden behind thick dust, but we can't find them.
2. The Plan: Look for galaxies that glow in "heat" (infrared) but are silent in X-rays.
3. The Action: We found 1,104 candidates, stacked their weak X-ray signals, and confirmed 23 of them are indeed hidden black holes.
4. The Problem: We only found 2.1% of what we expected. The "missing" black holes are still out there, hiding deeper than our current tools can see.
5. The Future: We need better telescopes to peel back the layers of the universe and find the rest of the ghosts.

Technical Summary

Here is a detailed technical summary of the paper "Identifying Compton-thick active galactic nuclei in the COSMOS II. Searching among mid-infrared selected AGNs."

1. Problem Statement

Compton-thick Active Galactic Nuclei (CT-AGNs), defined by a hydrogen column density $N_H \geq 1.5 \times 10^{24} \text{ cm}^{-2}$, are heavily obscured by gas and dust. This obscuration renders their X-ray emission extremely faint or undetectable by current instruments, despite their significant contribution to the Cosmic X-ray Background (CXB).

• The Discrepancy: Theoretical population synthesis models predict that CT-AGNs should comprise 20–30% (or even higher at high redshifts) of the total AGN population to reproduce the observed CXB spectrum. However, X-ray surveys typically identify only <15% of CT-AGNs.
• The Gap: A substantial population of "missing" CT-AGNs is believed to be hidden in samples of AGNs that are either undetected in X-rays or have very low photon counts. Previous work in the COSMOS field (Part I) identified some missing CT-AGNs, but a significant fraction remains elusive.
• Objective: This study aims to identify CT-AGNs hidden specifically among mid-infrared (MIR) selected AGNs that are individually undetected in X-rays within the COSMOS field.

2. Methodology

Sample Selection

• Source Catalog: The authors utilized the COSMOS2020 FARMER catalog.
• MIR Selection: They applied the stringent color-color criteria defined by Donley et al. (2012) using Spitzer/IRAC bands (3.6, 4.5, 5.8, and 8.0 $\mu$m) to select a pure sample of AGNs.
• Constraints:
  • Redshift range: $0.1 < z < 2$.
  • Exclusion of known X-ray sources (to focus on the "missing" population).
  • Exclusion of sources within $6''$ of known X-ray sources to avoid flux contamination.
  • Final Sample: 1,104 MIR-selected AGNs individually undetected in X-rays.

Data Reduction & Analysis

• X-ray Data: Used Chandra X-ray observations covering the COSMOS field (totaling ~4.6 Ms exposure). Data were reprocessed using CIAO 4.17 and CALDB 4.12.0.
• Flux Upper Limits: Since individual sources were undetected, the authors calculated 3$\sigma$ X-ray flux upper limits (2–10 keV) based on local background counts and effective exposure times.
• Spectral Energy Distribution (SED) Fitting:
  • Multiwavelength data (FUV to FIR, 27+ bands) were compiled.
  • The CIGALE (Code Investigating GALaxy Emission) code was used to fit SEDs using combined galaxy and AGN templates (Fritz et al. 2006 for AGN).
  • Derived parameters included AGN 6 $\mu$m luminosity ($L_{6\mu m}$), stellar mass ($M_*$), and Star Formation Rate (SFR).

CT-AGN Diagnostics

• MIR-X-ray Correlation: The study utilized the known correlation between intrinsic X-ray and MIR luminosities. Since CT-AGNs have absorbed X-rays but unabsorbed MIR emission, they appear "X-ray faint" relative to their MIR luminosity.
• Identification Criteria: Sources falling below the theoretical $L_X$–$L_{6\mu m}$ relations (adjusted for $N_H = 1.5 \times 10^{24} \text{ cm}^{-2}$ absorption) were flagged as candidates.
• Confidence Filtering: Only candidates with a high confidence coefficient for the fitted AGN component ($\alpha_{AGN} > 0.95$) were retained.

Stacking Analysis

• To confirm the heavy absorption, the authors performed an X-ray stacking analysis on the identified candidates using the CSTACK tool.
• They analyzed the stacked signal in soft (0.5–2 keV) and hard (2–8 keV) bands.
• Model Fitting: The stacked fluxes were fitted with a physical model (zphabs*zpowerlw+constant*zpowerlw+pexmon) to constrain the column density ($N_H$).

3. Key Results

• Candidate Identification: Using MIR diagnostics, the study identified 23 CT-AGN candidates (ranging from 7 to 23 depending on the specific relation used, but 23 met the high-confidence criteria).
• Stacking Detection:
  • Soft Band: The stacked sample was detected with >3$\sigma$ significance.
  • Hard Band: Detection was marginal (>1$\sigma$ significance).
  • Fluxes: Soft band flux $\approx (7.57 \pm 2.25) \times 10^{-17} \text{ erg cm}^{-2} \text{ s}^{-1}$; Hard band flux $\approx (2.34 \pm 1.55) \times 10^{-16} \text{ erg cm}^{-2} \text{ s}^{-1}$.
• Confirmation of CT Nature:
  • Model fitting of the stacked fluxes yielded two solutions: Compton-thin and Compton-thick.
  • The Compton-thin solution was ruled out as it would imply the sources are intrinsically X-ray weak, which is statistically improbable.
  • The Compton-thick solution was confirmed with $95%$confidence, requiring$\log N_H > 24.38$. This provides direct evidence that these 23 sources are indeed CT-AGNs.
• Fraction of CT-AGNs:
  • The identified CT-AGNs constitute only 2.1% (23/1104) of the MIR-selected sample.
  • This is significantly lower than the theoretical prediction of ~30% required to explain the CXB.
  • Conclusion: The MIR selection method, while effective, still misses a vast population of CT-AGNs (estimated at least 300 in this sample).

4. Host Galaxy Properties

• The authors compared the stellar mass ($M_*$) and Star Formation Rate (SFR) of the identified CT-AGNs against non-CT-AGNs.
• Kolmogorov-Smirnov (KS) Tests:
  • Stellar Mass distribution: $p = 0.33$.
  • SFR distribution: $p = 0.20$.
• Finding: No statistically significant difference was found between CT-AGN and non-CT-AGN host galaxies in this sample. This contrasts with some previous studies suggesting CT-AGNs reside in highly star-forming galaxies, though the authors note their sample size may be too small for robust conclusions.

5. Significance and Implications

• Validation of Method: The study successfully demonstrates that combining MIR selection with X-ray stacking can confirm the presence of heavily obscured AGNs that are invisible to standard X-ray surveys.
• The "Missing" Population: The primary finding is that even with MIR selection, the current census of CT-AGNs is incomplete. The low fraction (2.1%) suggests that many CT-AGNs are either:
  1. Not selected by current MIR color criteria.
  2. Too faint in X-rays even for stacking analysis to distinguish them from background noise without deeper exposures.
• Future Strategy: The authors propose that significantly deeper X-ray exposures of the COSMOS field are necessary to:
  • Directly detect individual CT-AGNs.
  • Lower the X-ray flux upper limits, thereby increasing the discriminatory power of MIR diagnostics.
• Cosmological Context: The results reinforce the need for multi-wavelength approaches to solve the "missing CT-AGN" problem, which is crucial for accurately modeling the evolution of supermassive black holes and the Cosmic X-ray Background.