JWST meets Chandra: a large population of Compton thick, feedback-free, and intrinsically X-ray weak AGN, with a sprinkle of SNe

TL;DR

This work uses the deepest Chandra data in GOODS to characterize the X-ray emission of a large JWST-selected AGN sample at 2<z<11, revealing widespread X-ray weakness relative to standard AGN SEDs. The authors argue that a combination of Compton-thick, dust-poor absorption (potentially from BLR clouds) and intrinsic X-ray weakness (e.g., high accretion rates, reduced corona) explains the observations, with minimal ejective feedback evidenced by weak [OIII] outflows. The findings imply a large unseen high-z AGN population that does not overproduce the X-ray background, necessitating revisions to black hole mass estimates and impacting models of early galaxy evolution and AGN feedback. Local analogues and multiwavelength diagnostics support the proposed scenarios and highlight the importance of X-ray–infrared joint analyses for understanding black hole growth in the early universe.

Abstract

We investigate the X-ray properties of a sample of 71 broad line and narrow line AGN at 2$<$z$<$11 discovered by JWST in the GOODS fields, which have the deepest Chandra observations ever obtained. Despite the widespread presence of AGN signatures in their rest-optical and -UV spectra, the vast majority of them is X-ray undetected. The stacked X-ray data of the non-detected sources also results in a non-detection. The upper limit on the X-ray emission for many of these AGN is one or even two orders of magnitude lower than expected from a standard AGN SED. X-ray absorption by clouds with large (Compton-thick) column density and low dust content, such as the Broad Line Region (BLR) clouds, can explain the X-ray weakness. In this scenario the BLR covering factor should be much larger than in low-z AGN or luminous quasars; this is supported by the larger equivalent width of the broad component of H$α$ in JWST-selected AGN. We also find that the JWST-discovered AGN lack prominent, fast outflows, suggesting that, in JWST-selected AGN, dense gas lingers in the nuclear region, resulting in large covering factors. We also note that a large fraction of JWST-selected AGN matches the definition of NLSy1, typically accreting at high rates and characterized by a steep X-ray spectrum -- this can further contribute to their observed weakness at high-z. Finally, we discuss that the broad Balmer lines used to identify type 1 AGN cannot be ascribed to Very Massive Stars or Supernovae, although we show that some of the faintest broad lines could potentially be associated with superluminous SNe.

Figures (15)

• Figure 1: Chandra X-ray image of the type 1.8 AGN GS_3073, which is one the most luminous z$>$4 targets in our sample, thoroughly studied Ubler2023Ji2024, and whose AGN nature is unambiguously confirmed by numerous rest-optical and rest-UV diagnostics. The image does not reveal a detection, while a $\sim$20$\sigma$ detection would have been expected based on a standard AGN SED. The radius of the circular region is 3 arcsec.
• Figure 2: Stacked X-ray spectrum for the whole sample of JWST-identified type 1 AGN. The upper grey spectrum shows the background level.
• Figure 3: Stacked Chandra X-ray images of the whole type 1 AGN sample (left panel) and for the high-luminosity and low-luminosity subsamples (middle and right panels, respectively). The radius of the circular region is 3 arcsec. The individual images are from different positions in the CDF-N and CDF-S fields, so they are shown for illustrative purposes only. All the fluxes have been derived from a spectral analysis of each individual source, as described in the text.
• Figure 4: Stacked Chandra X-ray images of the whole type 2 AGN sample. The radius of the circular region is 3 arcsec.
• Figure 5: Ratio $k_{\rm bol,X}$ between the AGN bolometric luminosity and the X-ray (2-10 keV) luminosity as a function of the bolometric luminosity L$_{BOL}$ for our sample of type 1 AGN. Light blue small points are from the low-redshift sample of Lusso2020, and are representative of "normal", optically/UV selected blue quasars. The light black lines show the best-fit (continuous line) and dispersion (dashed lines) of the $k_{\rm bol,X}$-L$_{BOL}$ relation for quasars from Duras2020. Purple diamonds show local NLSy1 while the light green diamond shows the prototypical broad line Seyfert 1 NGC 4151. Orange solid circles show all lower limits in $k_{\rm bol,X}$ for our sample of type 1 AGN at z$>$4. The star shows the lower limit for the stacked spectra from the whole sample at z$>$4. We also identify a few relevant sources, as shown in the figure legend (see text). The black squares show a group of lower redshift (2$<$z$<$4) type 1 AGN in JADES, two of which are detected. Sources with indication of absorption features in their broad H$\alpha$ profile are highlighted with blue empty circles. The bottom and top points relative to the source XID 403 refer to the observed and intrinsic X-ray luminosity, respectively. The bottom and top points relative to GN-z11 indicate two different ways of estimating its X-ray upper limit (see text).