[Paper Review] Lower-Luminosity Obscured AGN Host Galaxies are Not Predominantly in Major-Merging Systems at Cosmic Noon

Erini Lambrides, M. Chiaberge+16 more|arXiv (Cornell University)|Jul 15, 2021

Galaxies: Formation, Evolution, Phenomena112 references11 citations

TL;DR

This study tests whether low-luminosity obscured AGN at cosmic noon (0.5 < z < 3.1) are predominantly hosted in major-merging systems, as predicted by the AGN-merger paradigm. Using deep Hubble Space Telescope near-IR imaging and a novel Bayesian framework incorporating expert human classifier accuracy, the authors analyze 40 obscured AGN and a matched control sample of 40 inactive galaxies. They find no statistically significant evidence that obscured AGN host galaxies are preferentially in major mergers, challenging the idea that major mergers are the dominant trigger for obscured AGN growth at this epoch.

ABSTRACT

For over 60 years, the scientific community has studied actively growing central super-massive black holes (active galactic nuclei -- AGN) but fundamental questions on their genesis remain unanswered. Numerical simulations and theoretical arguments show that black hole growth occurs during short-lived periods ($\sim$ 10$^{7}$ -10$^{8}$ yr) of powerful accretion. Major mergers are commonly invoked as the most likely dissipative process to trigger the rapid fueling of AGN. If the AGN-merger paradigm is true, we expect galaxy mergers to coincide with black hole accretion during a heavily obscured AGN phase (N$_H$ $ > 10^{23}$ cm$^{-2}$). Starting from one of the largest samples of obscured AGN at 0.5 $<$ $z$ $<$ 3.1, we select 40 non-starbursting lower-luminosity obscured AGN. We then construct a one-to-one matched redshift- and near-IR magnitude-matched non-starbursting inactive galaxy control sample. Combining deep color extit{Hubble Space Telescope} imaging and a novel method of human classification, we test the merger-AGN paradigm prediction that heavily obscured AGN are strongly associated with galaxies undergoing a major merger. On the total sample of 80 galaxies, we estimate each individual classifier's accuracy at identifying merging galaxies/post-merging systems and isolated galaxies. We calculate the probability of each galaxy being in either a major merger or isolated system, given the accuracy of the human classifiers and the individual classifications of each galaxy. We do not find statistically significant evidence that obscured AGN at cosmic noon are predominately found in systems with evidence of significant merging/post-merging features.

Motivation & Objective

• To test the AGN-merger paradigm, which posits that major mergers trigger obscured AGN growth.
• To determine whether low-luminosity, heavily obscured AGN at cosmic noon (z ≈ 1–3) are predominantly hosted in major-merging systems.
• To address potential biases in AGN selection by focusing on a sample of obscured AGN from the deepest X-ray survey to date (7MS CDFS).
• To develop and apply a Bayesian statistical framework that accounts for human classifier accuracy in morphological classification of high-redshift galaxies.
• To compare the merger fraction of obscured AGN hosts with a redshift- and magnitude-matched sample of inactive, non-starbursting galaxies.

Proposed method

• Selecting 40 non-starbursting, low-luminosity obscured AGN from the 7MS Chandra Deep Field-South survey, using X-ray and multi-wavelength data.
• Constructing a control sample of 40 inactive, non-starbursting galaxies matched in redshift and near-IR magnitude to the AGN sample.
• Using deep Hubble Space Telescope near-IR imaging to assess galaxy morphology.
• Employing 14 expert human classifiers to visually assess each galaxy’s merger status, with individual accuracy estimated via a Bayesian model.
• Applying a novel Bayesian probabilistic framework that combines individual classifier accuracy with their classifications to compute the probability of each galaxy being in a merging or isolated system.
• Using the k-sample Anderson-Darling mid-rank test to compare the merger fraction distributions of the AGN and control samples.

Experimental results

Research questions

• RQ1 Is there a statistically significant excess of major-merging systems among low-luminosity, obscured AGN hosts at cosmic noon (0.5 < z < 3.1)?
• RQ2 Do obscured AGN host galaxies show a higher merger fraction than non-starbursting inactive galaxies with similar redshift and near-IR magnitude?
• RQ3 Does the inclusion of human classifier accuracy in the analysis alter the interpretation of morphological merger fractions compared to standard classification methods?
• RQ4 Is the AGN-merger paradigm valid for obscured AGN at cosmic noon, particularly for lower-luminosity systems?
• RQ5 Are minor mergers or other dynamical processes more relevant for triggering obscured AGN than major mergers?

Key findings

• The study finds no statistically significant evidence that non-starbursting, low-luminosity obscured AGN at cosmic noon are predominantly hosted in major-merging or post-merging systems.
• The merger fraction among obscured AGN hosts is not significantly higher than in the matched control sample of inactive galaxies, as determined by the k-sample Anderson-Darling test.
• The Bayesian framework estimates individual classifier accuracy and computes a probability of merger status per galaxy, reducing bias from subjective classification.
• The results challenge the long-standing AGN-merger paradigm, suggesting that major mergers are not the dominant trigger for obscured AGN at cosmic noon.
• The analysis confirms that the merger fraction does not vary significantly across different bins of galaxy stellar mass, size, or concentration, indicating no enhanced merger signature in any host galaxy property subgroup.
• The study highlights the importance of human classification and statistical modeling in high-redshift morphology studies, especially where automated methods fail due to resolution and redshift limitations.