Narrow line AGN selection in CEERS: spectroscopic selection, physical properties, X-ray and radio analysis

TL;DR

This paper presents a spectroscopic census of narrow-line AGNs (NLAGNs) in JWST CEERS data, identifying 52 NLAGNs across $2<z<9$ using a broad set of optical and UV emission-line diagnostics, including [OIII]$\lambda4363$ auroral and high-ionization lines. Through spectral fitting, stacking, and comprehensive SED analysis, the authors show that NLAGNs at high redshift typically have modest bolometric luminosities, do not strongly perturb their host ISM kinematics or star formation, and often appear X-ray weak. Multiwavelength stacking reveals significant X-ray deficits relative to their inferred bolometric output, while radio stacking remains inconclusive in distinguishing obscured from intrinsically X-ray weak sources, though in at least one low-$z$ case radio data suggest a Compton-thick scenario. The work highlights the need for complementary, deep radio and X-ray follow-up to fully resolve the nature of X-ray weakness in the early AGN population and demonstrates the value of JWST-era, multi-diagnostic NLAGN selection for building a complete census of AGN demography across cosmic time.

Abstract

In this work, we spectroscopically select narrow-line AGNs (NLAGNs) among the ~300 publicly available medium-resolution spectra of the Cosmic Evolution Early Release Science Survey (CEERS). Using both traditional and newly identified emission line NLAGN diagnostics diagrams, we identified 52 NLAGNs at 2<z<9, on which we performed a detailed multiwavelength analysis. We also identified four new z<2 broad-line AGNs (BLAGNs), in addition to the eight previously reported z>4.5 BLAGNs. We found that most of the high-z NLAGN were selected using the recently proposed AGN diagnostic diagrams based on the [Oiii] $λ$4363 auroral line or high-ionization emission lines. We compared the emission line velocity dispersion and the obscuration levels of the sample of NLAGNs with those of the parent sample without finding significant differences between the two distributions, suggesting a population of AGNs heavily buried and not significantly impacting the host galaxies physical properties, as was further confirmed by spectral energy distribution fitting. The bolometric luminosities of the high-z NLAGNs selected in this work are ~1.5dex below the ones sampled by surveys before JWST, potentially explaining the weak impact of these AGNs. Finally, we investigated the X-ray properties of the selected NLAGNs and of the sample of high-z BLAGNs. We found that all but four NLAGNs are undetected in the deep X-ray image of the field, as well as all the high-z BLAGNs. We did not obtain a detection even by stacking the undetected sources, resulting in an X-ray weakness of ~1-2 dex from what was expected based on their bolometric luminosities. To discriminate between a heavily obscured AGN scenario or an intrinsic X-ray weakness of these sources, we performed a radio (1.4GHz) stacking analysis, which did not reveal any detection and left open the questions about the origin of the X-ray weakness.

Figures (19)

• Figure 1: Stacked spectra of the final sample of NLAGNs (red) and non-AGNs (blue) derived in the manner described in Sec. \ref{['sec:spec_stack']}. We also marked the positions of some relevant lines.
• Figure 2: Left: R3N2 diagnostic diagram (also called BPT). The gray points represent the parent sample of galaxies analyzed in this work. Gold, green and red colors are used, respectively, for X-ray detected sources, radio-detected sources, and the high-$z$ BLAGNs reported in Harikane23. The NLAGNs selected using this diagram are marked with magenta edges. In magenta and cyan, we show the line ratios derived from the stacked spectra of all the selected NLAGN (52 sources) and non-AGN sources. The dashed lines refer to three different demarcation lines, as labeled, the one in green is the more conservative demarcation line derived in Scholtz23b. The shaded blue and red area represents the regions covered by the SFG and AGN photoionization models computed in Gutkin16 and Feltre16. The gray contours mark the distribution of SDSS sources Abazajian09. In the lower-right corner are reported the median errors of the sample. Right: R3S2 diagnostic diagram (also called VO87) with the demarcation line originally presented in Kewley01 (in black) and the new (more conservative) demarcation derived in Scholtz23b. Bottom: R3O1 diagnostic diagram with the demarcation line presented in Kewley01 (dashed black line) and the new one derived in this work (solid green line).
• Figure 3: He2N2 diagnostic diagram. The colors follow the same scheme as in Fig. \ref{['fig:BPT']}
• Figure 4: M1 and M2 diagnostic diagrams, firstly presented and discussed in Mazzolari24b. The colors follow the same scheme as in Fig. \ref{['fig:BPT']}.
• Figure 5: C3He2 versus C43 diagnostic diagrams, selecting only one NLAGN due to the difficulties in detecting rest-UV lines in high-$z$ galaxies with the short exposure times of the CEERS survey. The stacking line ratios and the average errors are not reported because of the poor statistics. The colors follow the same scheme as in Fig. \ref{['fig:BPT']}. The demarcation lines are those defined in Hirschmann22.