Black Holes in the Shadow: The Missing High-Ionization Lines in the Earliest JWST AGNs

TL;DR

This paper investigates the ionizing continua and broad-line region (BLR) conditions in a sample of 34 JWST-selected Type 1 AGNs spanning $1.7 < z < 9$, with a focus on the $z>5$ regime where high-ionization lines are anomalously weak. By combining prism and grating JWST/NIRSpec data, stacking in three redshift bins, and applying detailed spectral fitting, the authors measure line fluxes and equivalent widths that reveal a suppression of high-ionization lines (e.g., CIV, HeII) relative to Balmer and [OIII] emission. They compare the observations to Cloudy photoionization models built from sub-Eddington SEDs ( Jin2012a and Pezzulli2017) across a grid of ionization parameter $\log U$ and BLR covering factors, finding that only soft continua with $\log U \sim -2.5$ to $-3$ and modest covering factors ($\mathrm{CF} \sim 0.1$–$0.5$) can reproduce the data, while harder SEDs overpredict high-ionization lines. The results support scenarios in which the ionizing continuum is intrinsically soft, filtered by inner disk structures, or anisotropically emitted in super-Eddington accretion flows, with potential contributions from intervening absorbing gas; these findings have important implications for early MBH growth and BLR physics in the young universe.

Abstract

Observations with the James Webb Space Telescope (JWST) have uncovered a substantial population of high-redshift, broad-line active galactic nuclei (AGNs), whose properties challenge standard models of black hole growth and AGN emission. We analyze a spectroscopic sample of 34 Type 1 AGNs from the JWST Advanced Deep Survey (JADES) survey, spanning redshifts 1.7 < z < 9, to constrain the physical nature of the accretion flows powering these sources with broad-line diagnostics statistically for the first time. At z > 5, we find a marked suppression of high-ionization emission lines (HeII, CIV, NV) relative to prominent broad Halpha and narrow [OIII] features. This contrast places strong constraints on the shape of the ionizing spectral energy distribution (SED) and on the physical conditions in the broad-line region (BLR). By comparing the observations to photoionization models based on SEDs of black holes accreting at sub-Eddington ratios, we show that standard AGN continua struggle to reproduce the observed broad line ratios and equivalent widths across a wide ionization parameter range. These results suggest the need for modified SEDs -- either intrinsically softened due to super-Eddington accretion or radiative inefficiencies in the innermost disk, or externally filtered by intervening optically thick gas that absorbs or scatters the highest-energy photons before they reach the BLR.

Figures (15)

• Figure 1: Spectral wavelength coverage across dispersers for each object. Invalid or missing values are masked and shown as empty. The dashed vertical line at 3000 Å indicates a conventional division between the UV and optical regimes, though it is not used in the kinematic classification of emission lines. Treatment of lines near this boundary (e.g., [NeV]$\lambda$3426) is discussed in Section \ref{['sec:spectralfitting']}.
• Figure 2: Stack of the JWST NIRSpec (G140M+G235M+G395M) grating spectra of JADES Type 1 AGNs in the high-redshift bin ($z>5$). For visualization, the normalized stack was multiplied by the mean [OIII]$\lambda$5007 flux of the 15 contributing objects; this rescaling is used only for the figure and not for any measurements.
• Figure 3: Comparison of emission-line EWs in the JWST grating stack for the $z > 5$ sample (purple) with median values from the SDSS quasar catalog of Wu2022 (green). Circles and squares represent measured lines with $1\sigma$ uncertainties, while left-pointing arrows denote $3\sigma$ upper limits for undetected lines in the JWST stack. Only the broad components are considered for C IV, He II, and H$\beta$. In most cases, error bars are smaller than the marker symbols and are not visually apparent.
• Figure 4: $\nu F_{\nu}$ SEDs of the Jin et al. (2012) and Pezzulli et al. (2017) families, normalized at 13.6 eV. The normalization energy is marked by the dashed vertical line; amplitudes are in arbitrary units and are used solely to compare spectral shapes. The Pezzulli et al. (2017) SED is harder near the hydrogen ionization edge, leading to stronger excitation of high-ionization lines such as He II, whereas the Jin et al. (2012) SED exhibits a more prominent soft X-ray tail, contributing extra ionizing photons above the He II edge (54.4 eV).
• Figure 5: Broad H$\beta$ EWs as a function of ionization parameter $\log U$, derived from sub-Eddington Cloudy models (parameters listed in Table \ref{['tab:cloudy_par']}). Different line styles indicate varying broad-line region CFs, ranging from 1.0 to 0.1; models with the same SED share the same color. The shaded region for each SED represents the envelope spanned by CF values from 1.0 down to 0.1. The dark violet horizontal line marks the observed EW of broad H$\beta$ measured in the $z > 5$ grating stack. The vertical axis uses a symlog scale (linear below 200 Å, logarithmic above).