JADES -- The Rosetta Stone of JWST-discovered AGN: deciphering the intriguing nature of early AGN

TL;DR

This study analyzes GN-28074, a z=2.26 broad-line AGN discovered by JWST, using a deep NIRSpec spectrum to reveal blueshifted Balmer and He I absorption from a very dense, dust-free neutral absorber consistent with BLR clouds. Through detailed line fitting, SED modeling, and Cloudy photoionization simulations, it demonstrates that the extreme X-ray weakness is driven at least in part by Compton-thick absorption by BLR-like gas located between the BLR and the dusty torus, with outflowing neutral gas inferred at a few hundred to ~1000 km s^-1 and substantial mass flow. The work rules out alternative broad-line scenarios such as hyperdense ultra-metal-poor outflows or Raman scattering, and it places GN-28074 in the broader context of JWST-discovered, X-ray–weak AGN, including local analogues like NGC 4151 and SBS 0335-052E. Together, these findings provide a practical framework for interpreting the new population of high-z AGN and illuminate the connection between BLR structure, accretion physics, and multiwavelength obscuration.

Abstract

JWST has discovered a large population of Active Galactic Nuclei (AGN) at high redshift. Many of these newly discovered AGN have broad permitted lines (typically H$α$), but are extremely weak in the X-rays. Here we present the NIRSpec spectrum of the most extreme of these objects, GN-28074, an AGN at $z=2.26$ with prominent Balmer, Paschen and \HeI broad lines, and with the highest limit on the bolometric to X-ray luminosity ratio among all spectroscopically confirmed AGN in GOODS. This source is also characterized by a mid-IR excess, most likely associated with the AGN torus' hot dust. The high bolometric luminosity and moderate redshift of this AGN allow us to explore its properties more in depth relative to other JWST-discovered AGN. The NIRSpec spectrum reveals prominent, slightly blueshifted absorption of H$α$, H$β$ and \HeI$λ$10830. The Balmer absorption lines require gas with densities of $n_{\rm H}> 10^8~{\rm cm}^{-3}$, inconsistent with an ISM origin, but fully consistent with clouds in the Broad Line Region (BLR). This finding suggests that at least part of the X-ray weakness is due to high (Compton thick) X-ray absorption by (dust-free) clouds in the BLR, or in its outer, slowly outflowing regions. GN-28074 is also extremely radio-weak. The radio weakness can also be explained in terms of absorption, as the inferred density of the clouds responsible for H$α$ absorption makes them optically thick to radio emission through free-free absorption. Alternatively, in this and other JWST-discovered AGN, the nuclear magnetic field may have not developed properly yet, resulting both in intrinsically weak radio emission and also lack of hot corona, hence intrinsic X-ray weakness. Finally, we show that recently proposed scenarios, invoking hyper-dense and ultra-metal-poor outflows or Raman scattering to explain the broad H$α$, are completely ruled out.

Figures (13)

• Figure 1: R1000 spectrum of GN-28074 with emission lines marked. The fluxes on the y axis are presented in logarithmic scale to better showcase weaker lines and the shape of the continuum. All bright permitted lines appear to have broad components while the continuum showcases the characteristic 'v' shape found in some objects of the LRD population. The false-colour NIRCam image (inset) highlights the point-source nature of the object, and the position of the MSA slitlets (grey rectangles). We note a pink region to the south-east, denoting a deficiency in F182M flux (see text for a discussion). The rest of the NIRCam single-band cutouts highlight the extended emission (only discernible at SW wavelengths); the angular size of these cutouts is the same as the RGB image.
• Figure 2: R1000 spectrum expanded around some of the permitted lines of GN-28074 and showing the associated spectral fits. The broad line region (BLR) and the (extended) outflow components are plotted as solid blue and red lines respectively. The narrow components are shown with dashed green lines. The combined fit, including the absorption absorption components, is shown in magenta. Top left: H$\beta$ and [O iii]$\lambda\lambda$5007,4959 complex, showcasing the deep absorption in H$\beta$ and broadening of [O iii]$\lambda\lambda$5007,4959 lines due to the presence an outflow. Top right: H$\alpha$ , [N ii]$\lambda\lambda$6548,6583 and [S ii]$\lambda\lambda$6716,6731 . Notable is the relative weakness of the [N ii]$\lambda\lambda$6549,6585 and [S ii]$\lambda\lambda$6716,6731 doublets. Bottom left: Combined He i$\lambda$10830 and Pa$\gamma$ line fit. The ionized outflow component is not required here by the data. Bottom right: Fit to the Pa$\beta$ line. Due to wavelength calibration issues, this line is slightly offset to the rest and thus requires separate kinematics to properly fit. The ionized outflow is absent just as in the Pa$\gamma$ + He i$\lambda$10830 fit.
• Figure 3: The best-fitting CIGALE model. It can be seen that the mid infrared emission is well explained by hot dust emission from an attenuated accretion disk emission (magenta) with comparatively little contribution from stellar dust (in red). The excess flux around 2.0$\mu$m that is not well reproduced can be attributed to the strong H$\alpha$ emission.
• Figure 4: Same as \ref{['fig:AGN_SED']}, except for models without AGN emission. It can readily be seen that stellar-only templates leave significant residuals from 5$\mu$m onwards resulting in a poorer reduced $\chi^2$ value.
• Figure 5: Narrow emission-line diagnostic diagrams, showing our target (green circles with errorbars; the measurement uncertainties are often smaller than the size of the marker), and the distribution of galaxies at $z=0.1$ from SDSS Data Release 7 (contour lines and scatter points). The demarcation lines in panels a--c are from kewley+2001kauffmann+2003ckewley+2006schawinski+2007. Our target lies in a scarcely populated region of the diagram, formally consistent with star-formation photoionisation (panels a and b) or AGN (panel c). Shock-driven emission in the NLR is also unlikely, because shock models (red and yellow lines) do not cover the location of our target in these diagnostic plots (see text for a description).