Tip of the iceberg: overmassive black holes at 4<z<7 found by JWST are not inconsistent with the local $\mathcal{M}_{\rm BH}$-$\mathcal{M}_\star$ relation

TL;DR

The paper interrogates whether JWST-detected overmassive black holes at 4<z<7 imply evolution of the M_BH-M_* relation or arise from selection biases and measurement uncertainties. It develops a forward-modeling, Bayesian framework that incorporates the Eddington ratio distribution, the galaxy stellar mass function, and JWST selection effects to infer a bias-corrected intrinsic relation. The results show that the observed offsets can be reproduced by biases, yielding an intrinsic relation at z~5 that is steeper and more scattered than the naive, observed trend, with a potentially sizable hidden population of low-mass BHs. These findings refine constraints on BH seed scenarios and highlight the essential role of bias-aware modeling in JWST-era studies of BH-galaxy coevolution.

Abstract

JWST is revealing a new remarkable population of high-redshift ($z\gtrsim4$), low-luminosity Active Galactic Nuclei (AGNs) in deep surveys and detecting the host galaxy stellar light in the most luminous and massive quasars at $z\sim 6$ for the first time. Latest results claim supermassive black holes (SMBHs) in these systems to be significantly more massive than expected from the local BH mass - stellar mass ($\mathcal{M}_{\rm BH} - \mathcal{M}_\star$) relation and that this is not due to sample selection effects. Through detailed statistical modeling, we demonstrate that the coupled effects of selection biases (i.e., finite detection limit and requirements on detecting broad lines) and measurement uncertainties in $\mathcal{M}_{\rm BH}$ and $\mathcal{M}_\star$ can in fact largely account for the reported offset and flattening in the observed $\mathcal{M}_{\rm BH} - \mathcal{M}_\star$ relation toward the upper envelope of the local relation, even for those at $\mathcal{M}_{\rm BH} < 10^8\,M_{\odot}$. We further investigate the possible evolution of the $\mathcal{M}_{\rm BH} - \mathcal{M}_\star$ relation at $z\gtrsim 4$ with careful treatment of observational biases and consideration of the degeneracy between intrinsic evolution and dispersion in this relation. The bias-corrected intrinsic $\mathcal{M}_{\rm BH} - \mathcal{M}_\star$ relation in the low-mass regime suggests that there might be a large population of low-mass BHs (${\rm log}\,\mathcal{M}_{\rm BH} \lesssim 5$), possibly originating from lighter seeds, remaining undetected or unidentified even in the deepest JWST surveys. These results have important consequences for JWST studies of BH seeding and the coevolution between SMBHs and their host galaxies at the earliest cosmic times.

Figures (7)

• Figure 1: The $\textit{$\mathcal{M}_{\rm BH}$}-\textit{$\mathcal{M}_{\star}$}$ relation of high-redshift AGNs used in this study Harikane2023Maiolino2023Stone2023Yue2023Ding2023. For comparison, we display the SDSS quasar sample ($\textit{${\rm log}\, L_{\rm bol}$}/\textit{$\rm erg\ s^{-1}$}\lesssim46$) at $0.2<z<1$ in Li2021sizeLi2021mass as green contours ($1\,\sigma-3\,\sigma$ level), along with the local $\textit{$\mathcal{M}_{\rm BH}$}-\textit{$\mathcal{M}_{\rm bulge}$}$ relation in Kormendy2013 and the local $\textit{$\mathcal{M}_{\rm BH}$}-\textit{$\mathcal{M}_{\star}$}$ relation in Greene2020. The gray shaded regions display a schematic view of the inaccessible region of high-redshift observations, where the BHs are either undetectable or under-represented (sparsely sampled) due to the limited sensitivity and survey volume, with the exact limits depending on specific survey designs.
• Figure 2: Impact of selection effects and measurement uncertainties on the $\textit{$\mathcal{M}_{\rm BH}$}-\textit{$\mathcal{M}_{\star}$}$ relation. The square data points, color-coded by the average $L_{\rm bol}$ in each mass bin, represent mock AGNs simulated using Equation \ref{['eq:p_mo_so']} based on the local KH13 relation (red dashed line) with an intrinsic scatter of 0.3 dex. The cyan solid curve represents the mean $\textit{$\mathcal{M}_{\rm BH}$}-\textit{$\mathcal{M}_{\star}$}$ relation averaged over the square data points. Panel (a): The original KH13 relation. Panel (b): no mass uncertainties added but with the selection function ($\textit{${\rm log}\, L_{\rm bol}$}>44.1\ \textit{$\rm erg\ s^{-1}$}$ and $\rm FWHM > 1000\ \textit{$\rm km\ s^{-1}$}$) applied to the input KH13 relation. The effect of applying the $\textit{$L_{\rm bol}$}$ and the FWHM cut separately is displayed for comparison. Panel (c): mass uncertainties added on both $\mathcal{M}_{\rm BH}$ (0.35 dex) and $\mathcal{M}_{\star}$ (0.45 dex), but without the selection function applied. Panel (d): mass uncertainties added and selection function applied.
• Figure 3: The predicted "observed" $\textit{$\mathcal{M}_{\rm BH}$}-\textit{$\mathcal{M}_{\star}$}$ relations (solid curves) of detectable AGNs at $z\sim5$ in the three surveys, assuming that the underlying BH population (i.e., including those below the detection limits) intrinsically follow the local KH13 and G20 relations (dashed curves with the same colors). The intrinsic scatter at $1\,\sigma$, $2\,\sigma$, and $3\,\sigma$ levels of each local relation are shown as shaded regions. Only BHs located in the upper envelope of the assumed intrinsic mass relations can be detected as luminous AGNs. Their positions are further biased upwards by the substantial mass uncertainties (Figure \ref{['fig:sample_simu']}), resulting in a significantly elevated $\mathcal{M}_{\rm BH}$/$\mathcal{M}_{\star}$ compared to the intrinsic ratio.
• Figure 4: Model-predicted permitted region ($15<\textit{$N_{\rm obs}$}<65$ for AGNs with $\textit{${\rm log}\, L_{\rm bol}$}>44.1$, $7.8<\textit{${\rm log}\,\textit{$\mathcal{M}_{\star}$}$}<10.6$, and $4<z<7$) in the $\beta-\sigma$ plane for $\alpha=1.17$ (KH13) and $\alpha=1.61$ (G20), respectively. The values of $\beta$ and $\sigma$ for the KH13 and G20 relations are marked by the red stars. Each bin is color-coded by its average $N_{\rm obs}$. The forbidden region shown in white indicates parameter spaces in $[\alpha, \beta, \sigma]$ that result in either too few ($<15$, left side) or too many ($>65$, right side) observable AGNs compared to observations ($\textit{$N_{\rm obs}$}=22$), assuming $\textit{$p_{\rm ac}$}=10\%$.
• Figure 5: Left: best-fit intrinsic $\textit{$\mathcal{M}_{\rm BH}$}-\textit{$\mathcal{M}_{\star}$}$ relation (black line) and its $1\,\sigma$ uncertainty (green shaded region) inferred from the JDEEP sample with solid host galaxy detections (blue points). The inset panel shows the expanded $1-3\,\sigma$ intrinsic scatter (gray shaded regions). The intrinsic $\textit{$\mathcal{M}_{\rm BH}$}-\textit{$\mathcal{M}_{\star}$}$ relation for the same sample derived by Pacucci2023 and its $1\,\sigma$ uncertainty are shown in magenta. Local relations from KH13, G20, and RV15 are also shown for comparison. Right: posterior parameter distributions and best-fit values of the mass relation and BL AGN fraction for the observed AGN sample show on the left.