pith. sign in

arxiv: 2604.25543 · v1 · submitted 2026-04-28 · 🌌 astro-ph.GA · astro-ph.HE

Compton-thick AGN Characterisation in a Multi-wavelength Context: Insights from the 70-Month textit{SWIFT}/BAT Catalogue

Muhammad Luqman Hakeem Musa , Zamri Zainal Abidin , Masatoshi Imanishi , Yoshiaki Hagiwara , Adlyka Ainul Annuar This is my paper

Pith reviewed 2026-05-07 15:40 UTC · model grok-4.3

classification 🌌 astro-ph.GA astro-ph.HE
keywords Compton-thick AGNEddington ratioSWIFT/BATmulti-wavelengthAGN unificationobscurationaccretion ratemid-IR diagnostics
0
0 comments X

The pith

Compton-thick AGNs show higher Eddington ratios than less obscured ones, pointing to accretion and obscuration as linked drivers rather than orientation alone.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper compares 26 Compton-thick active galactic nuclei to 217 others drawn from the 70-month SWIFT/BAT catalogue and tracks their behaviour across radio, infrared, optical and X-ray bands. It reports that the heavily obscured objects tend to accrete at higher fractions of their Eddington limit, display redder mid-infrared colours consistent with cooler dust, and sit outside standard WISE colour wedges. These patterns lead the authors to argue that intense accretion helps sustain the high column densities rather than the obscuration arising merely from our line of sight. A reader would care because the result affects how we select and count the most hidden black-hole growth episodes in future large surveys where X-ray data may be sparse. Machine-learning tests show that optical line luminosities plus a mid-infrared colour can recover most of the Compton-thick candidates even without direct X-ray penetration.

Core claim

Using the 70-month SWIFT/BAT sample, the study finds that Compton-thick AGNs exhibit significantly higher Eddington ratios than non-Compton-thick AGNs, accompanied by modestly elevated 2-3 GHz radio luminosities and redder W3-W4 colours. Optical BPT diagrams place most Compton-thick sources in the Seyfert region, indicating that narrow-line emission remains largely isotropic. Principal-component analysis isolates ionizing power and the accretion-obscuration connection as the dominant sources of variance, while a machine-learning classifier trained on intrinsic X-ray luminosity, [OI] and Hα luminosities, and W2-W3 colour achieves 0.80 recall for Compton-thick identification.

What carries the argument

The direct comparison of Eddington ratios between the Compton-thick and non-Compton-thick populations, which underpins the claim that radiation-driven accretion maintains high column densities.

Load-bearing premise

The observed differences in accretion rates and other properties between the 26 Compton-thick sources and the rest of the sample are not mainly produced by selection effects in the BAT catalogue or by uncertainties in the bolometric corrections used to derive those rates.

What would settle it

A larger, less biased sample of Compton-thick AGNs that shows no average difference in Eddington ratio relative to non-Compton-thick AGNs would falsify the central claim.

Figures

Figures reproduced from arXiv: 2604.25543 by Adlyka Ainul Annuar, Masatoshi Imanishi, Muhammad Luqman Hakeem Musa, Yoshiaki Hagiwara, Zamri Zainal Abidin.

Figure 1
Figure 1. Figure 1: Comparison of the normalized distributions of redshift (z), black hole mass (log 𝑀BH), X-ray luminosity (log 𝐿14−195 keV), and Eddington ratio (log 𝜆Edd) for the CT AGN (red solid lines) and non-CT AGN (black dashed lines) samples view at source ↗
Figure 2
Figure 2. Figure 2: Distribution of 𝐿2−3 GHz for different AGN classes. The histogram on the top panel shows the luminosity distribution for non-CT AGN (black dashed line) and CT AGN (red solid line). The bottom panel show the fitted Gaussian for the AGN population with the vertical lines represent the fitted mean radio luminosity, 𝜇. the null hypothesis, showing no significance difference between the distributions. The regre… view at source ↗
Figure 3
Figure 3. Figure 3: Correlation between intrinsic radio luminosity, log 𝐿2–3 GHz, and X-ray luminosities across three energy bands: 2–10 keV (left), 20–50 keV (middle), and 14–195 keV (right). Non-CT AGN are shown in grey and CT AGN in red, with downward arrows indicating upper limits for non-detections. The dashed lines represent the censored regression fits, with grey for non-CT AGN and red for CT AGN. The faint lines aroun… view at source ↗
Figure 4
Figure 4. Figure 4: WISE band luminosity distributions across different AGN populations. Non-CT AGNs are shown in blue, CT AGNs in red, and blazars in green. The top panel shows the histogram of the luminosity distributions, and the fitted Gaussian with the mean luminosity, 𝜇, are shown in the bottom panel view at source ↗
Figure 5
Figure 5. Figure 5: W1-W2 (left panel) and W2-W3 (right panel) colour diagrams for different AGN samples. The W1-W2 colour cut AGN selection criteria from Stern et al. (2012) is shown in the yellow region view at source ↗
Figure 6
Figure 6. Figure 6: WISE colour-colour diagrams using three (left panel) and four (right panel) WISE bands for our sample, comprising of CT AGNs (red triangles) and non-CT AGNs (grey circles). The AGN selection wedges/criteria from Stern et al. (2012) (black), Jarrett et al. (2011) (green), and Mateos et al. (2012) (blue) are indicated by the dashed lines. The contours represent the 1𝜎, 2𝜎, and 3𝜎 of the fiducial density, wit… view at source ↗
Figure 7
Figure 7. Figure 7: Bootstrap scatter analysis in theWISE colour-colour diagram for our sample. The histogram illustrates the mean scatter distance from the centroid, comparing CT AGNs (red dashed line) with non-CT AGNs (black dash-dotted line). The upper panel shows the three-band WISE diagram distribution, while the bottom panel shows the four-band WISE diagram. principal components is shown in view at source ↗
Figure 8
Figure 8. Figure 8: BPT diagrams for our AGN sample based on emission line ratios. The left panel shows the BPT-NII diagram, while the middle and right panels show the BPT-SII and BPT-OI, respectively. Demarcation lines from Kewley et al. (2001) (solid), Kauffmann et al. (2003)(dashed-dotted), and Schawinski et al. (2007)(dashed) are adopted for BPT-NII, while in BPT-SII and BPT-OI, Kewley et al. (2006)(dashed) and Kewley et … view at source ↗
Figure 10
Figure 10. Figure 10: The relationship between Eddington ratio (𝜆Edd) and [NII]𝜆6584/H𝛼 emission-line ratio is displayed in log scale for CT AGNs (red triangles) and non-CT AGNs (grey dots). The solid red line denotes the best-fit linear regression for CT AGNs, with a shaded region indicating the 3𝜎 uncertainty. The dashed grey line represent regression fits for non-CT AGNs. models also provide significant insight from optical… view at source ↗
Figure 9
Figure 9. Figure 9: Bootstrap scatter analysis for all three BPT diagrams. The grey his￾togram illustrates the mean scatter distribution from the centroid, comparing non-CT AGNs (black dash-dotted line) with CT AGNs (red dashed line) importance is defined by the absolute magnitude of the learned coef￾ficients (|𝑤𝑖 |). A higher importance value indicates a stronger influ￾ence on the classification of CT candidates. Interesting… view at source ↗
Figure 11
Figure 11. Figure 11: Scree plot of the explained variance ratio for each principal com￾ponent. The first three components account for the majority of the variance in the dataset, capturing more almost 72% of the total variance. An elbow trend is observed around the third to fourth component, suggesting that these components retain the most informative features of the data view at source ↗
Figure 12
Figure 12. Figure 12: Principal component coefficients for the multi-wavelength features of the AGN sample. Principal Component 1 (blue) is dominated by optical emission-line ratios and X-ray luminosity. Principal Component 2 (orange) shows strong positive contribution from X-ray luminosities, log 𝜆Edd, and W1-W2 colours, in contrast with negative contributions from optical emission lines and radio luminosities view at source ↗
Figure 13
Figure 13. Figure 13: PCA of multi-wavelength properties for CT AGNs (red triangles) and non-CT AGNs (grey circles). The main panel shows the score plot of the first two principal components, while the histograms illustrate the normalized one-dimensional distributions of each principal component axis. servations reported by Chiaraluce et al. (2020), showing diverse radio emissions in hard X-ray-selected AGNs. The observed weak… view at source ↗
read the original abstract

We analyse Compton-thick active galactic nuclei (CT AGNs), a heavily obscured subclass that challenges traditional X-ray diagnostics. Using 243 sources from the 70-Month \textit{SWIFT}/BAT catalogue (26 CT, 217 non-CT), we investigate their properties across radio, infrared (IR), optical, and X-ray bands. VLASS data reveals slightly higher 2--3~GHz mean luminosities in CT AGNs, suggesting active cores attenuated by circumnuclear absorption. Mid-IR diagnostics show redder $W3-W4$ colours in CT AGNs, tracing cooler dust, with significant scatter likely driven by host-galaxy dilution. Most CT AGNs fall outside standard WISE selection wedges, highlighting mid-IR selection limitations. BPT diagnostics show that CT AGNs primarily occupy Seyfert regions, indicating isotropic narrow-line properties. CT AGNs favour significantly higher Eddington ratios ($\lambda_{\text{Edd}}$), supporting radiation-driven unification where intense accretion maintains high-column density. We also observe a moderate anti-correlation between [NII]/H$\alpha$ and $\lambda_{\text{Edd}}$. Principal component analysis identifies ionizing power and the accretion-obscuration link as primary variance drivers, though both populations overlap significantly in the PC1--PC2 plane. Machine learning achieved high recall (0.80) using intrinsic X-ray luminosity, [OI]$\lambda$6300 and H$\alpha$ luminosities, and $W2-W3$ colour. This demonstrates the potential for multi-wavelength signatures to verify CT candidates in future deep surveys where X-ray data is limited. Overall, our findings suggest CT AGNs are driven by high obscuration and accretion rates rather than a simple orientation effect.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript examines the multi-wavelength properties of 26 Compton-thick (CT) AGNs compared to 217 non-CT AGNs selected from the 70-month SWIFT/BAT catalog. It reports slightly higher 2-3 GHz radio luminosities, redder W3-W4 mid-IR colors, primary occupation of Seyfert regions in BPT diagrams, significantly higher Eddington ratios (λ_Edd) for CT sources, a moderate anti-correlation between [NII]/Hα and λ_Edd, PCA results highlighting ionizing power and accretion-obscuration as main variance drivers, and an ML classifier achieving 0.80 recall using intrinsic X-ray luminosity, [OI]λ6300, Hα luminosities, and W2-W3 color. The authors conclude that CT AGNs are driven by high obscuration and accretion rates rather than a simple orientation effect, with implications for radiation-driven unification.

Significance. If the reported elevation in λ_Edd for CT AGNs proves robust after accounting for selection effects, the result would lend observational support to radiation-driven unification scenarios in which intense accretion helps sustain high column densities. The multi-wavelength diagnostics, PCA decomposition, and ML demonstration for identifying CT candidates in X-ray-limited regimes constitute a practical contribution to AGN population studies and future survey strategies.

major comments (3)
  1. [Abstract and §4] Abstract and §4 (Eddington ratio and unification discussion): The central claim that CT AGNs 'favour significantly higher Eddington ratios (λ_Edd)' is presented without a statistical test (e.g., KS or Mann-Whitney p-value), uncertainties on the reported means or distributions, or explicit bolometric correction details. Because λ_Edd is derived from absorption-corrected L_X, this omission directly affects the load-bearing interpretation that intense accretion maintains high N_H rather than orientation.
  2. [§2] §2 (Sample selection and catalogue description): In the flux-limited 14-195 keV BAT sample, CT sources (N_H > 10^24 cm^-2) experience stronger attenuation, so only those with higher intrinsic luminosities enter the catalog. The manuscript does not quantify this selection function (via completeness simulations or comparison to a volume-limited subsample), which could artificially inflate the observed λ_Edd difference even if the underlying populations are identical.
  3. [§5] §5 (PCA and machine-learning results): The reported ML recall of 0.80 is given without cross-validation details, feature-importance rankings, confusion-matrix breakdown, or performance uncertainties. Since the classifier relies on the same intrinsic X-ray luminosities and line luminosities used for the λ_Edd comparison, any uncorrected selection bias propagates into the claimed multi-wavelength verification potential.
minor comments (2)
  1. [Abstract] Abstract: The phrase 'significantly higher' should be replaced or supplemented with the actual quantitative offset (e.g., factor or dex difference) and the associated statistical measure.
  2. Throughout: Ensure all mid-IR color notations (W2-W3, W3-W4) and line ratios ([NII]/Hα, [OI]λ6300) are explicitly defined on first use and that host-galaxy dilution effects are discussed with reference to specific figures or tables.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which highlight important areas for improving the statistical robustness, discussion of selection effects, and transparency of our machine-learning analysis. We address each major comment point by point below, indicating the revisions planned for the next version of the manuscript.

read point-by-point responses

  1. Referee: [Abstract and §4] Abstract and §4 (Eddington ratio and unification discussion): The central claim that CT AGNs 'favour significantly higher Eddington ratios (λ_Edd)' is presented without a statistical test (e.g., KS or Mann-Whitney p-value), uncertainties on the reported means or distributions, or explicit bolometric correction details. Because λ_Edd is derived from absorption-corrected L_X, this omission directly affects the load-bearing interpretation that intense accretion maintains high N_H rather than orientation.

    Authors: We agree that the presentation of the λ_Edd result requires additional statistical support and methodological transparency to strengthen the interpretation. In the revised manuscript we will add a two-sample Kolmogorov-Smirnov test (with p-value) comparing the λ_Edd distributions of the CT and non-CT populations, report mean values together with their uncertainties (standard error of the mean), and explicitly state the bolometric correction adopted to convert intrinsic 2–10 keV luminosity to bolometric luminosity, including the reference used. These additions will be incorporated into both the abstract and §4. revision: yes

  2. Referee: [§2] §2 (Sample selection and catalogue description): In the flux-limited 14-195 keV BAT sample, CT sources (N_H > 10^24 cm^-2) experience stronger attenuation, so only those with higher intrinsic luminosities enter the catalog. The manuscript does not quantify this selection function (via completeness simulations or comparison to a volume-limited subsample), which could artificially inflate the observed λ_Edd difference even if the underlying populations are identical.

    Authors: We acknowledge this as a genuine limitation of the flux-limited BAT sample. In the revised §2 we will expand the discussion of the selection function, explicitly noting that CT sources must possess higher intrinsic luminosities to exceed the survey threshold and that this could contribute to the observed λ_Edd offset. We will also add a comparison of λ_Edd distributions in luminosity-matched subsamples to test whether the difference persists at fixed luminosity. A full set of completeness simulations or construction of a strictly volume-limited subsample lies beyond the scope of the present work and would require dedicated Monte-Carlo modeling of the BAT sensitivity curve; we therefore treat this as a partial revision while flagging the issue as a caveat for the interpretation. revision: partial

  3. Referee: [§5] §5 (PCA and machine-learning results): The reported ML recall of 0.80 is given without cross-validation details, feature-importance rankings, confusion-matrix breakdown, or performance uncertainties. Since the classifier relies on the same intrinsic X-ray luminosities and line luminosities used for the λ_Edd comparison, any uncorrected selection bias propagates into the claimed multi-wavelength verification potential.

    Authors: We agree that the machine-learning section needs more complete reporting. In the revised §5 we will specify the cross-validation scheme (e.g., stratified k-fold), provide feature-importance rankings, include the full confusion-matrix breakdown, and report performance uncertainties (standard deviation across folds or bootstrap estimates). We will also add an explicit caveat discussing how the use of intrinsic X-ray luminosity may carry forward sample-selection effects, while emphasizing that the optical-line and mid-IR colour features supply independent information. These changes will clarify the practical utility of the classifier for future surveys. revision: yes

Circularity Check

0 steps flagged

No significant circularity; observational comparisons are self-contained

full rationale

The paper conducts direct statistical comparisons of multi-band properties (radio luminosities, WISE colours, BPT line ratios, Eddington ratios) between the 26 CT and 217 non-CT sources drawn from the public 70-month BAT catalogue. Eddington ratios are computed from absorption-corrected intrinsic X-ray luminosities via standard bolometric corrections; no parameter is fitted to the CT/non-CT difference and then re-used as a 'prediction'. PCA and ML are applied post hoc to the observed quantities but do not define or force the reported trends. No self-citations, uniqueness theorems, or ansatzes are invoked as load-bearing justification for the central claims. The derivation chain relies on independent public data and standard diagnostics and does not reduce to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an observational study using established catalogues and standard astrophysical diagnostics; no new free parameters, axioms, or invented entities are introduced in the abstract.

pith-pipeline@v0.9.0 · 5655 in / 1208 out tokens · 67047 ms · 2026-05-07T15:40:43.808230+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

  1. [1]

    M., Greene J

    Akylas A., Georgakakis A., Georgantopoulos I., Brightman M., Nandra K., 2012, Astronomy & Astrophysics, 546, A98 Albán M., Wylezalek D., Comerford J. M., Greene J. E., Riffel R. A., 2024, Astronomy & Astrophysics, 691, A124 Alonso-Herrero A., et al., 2014, Monthly Notices of the Royal Astronomical Society, 443, 2766 AnannaT.T.,etal.,2022,TheAstrophysicalJ...

    work page 2012

  2. [2]

    K., Silk J., 2007, Monthly Notices of the Royal Astronomical Society, 382, 1415 SecrestN.,DudikR.,DorlandB.,ZachariasN.,MakarovV.,FeyA.,Frouard MNRAS000, 1–16 (2026) 16M

    pp 459–464 Schawinski K., Thomas D., Sarzi M., Maraston C., Kaviraj S., Joo S.-J., Yi S. K., Silk J., 2007, Monthly Notices of the Royal Astronomical Society, 382, 1415 SecrestN.,DudikR.,DorlandB.,ZachariasN.,MakarovV.,FeyA.,Frouard MNRAS000, 1–16 (2026) 16M. L. H. Musa et al. 2026 J., Finch C., 2015, The Astrophysical Journal Supplement Series, 221, 12 S...

    work page 2007