New AGN diagnostic diagrams based on the [OIII]$λ4363$ auroral line

TL;DR

This work introduces three rest-frame optical diagnostic diagrams that leverage the auroral line $[OIII]\lambda4363$ to identify AGN in high-redshift galaxies observed with JWST. By combining the temperature-sensitive ratio $[OIII]\lambda4363/H\gamma$ with ionization-sensitive tracers $[OIII]\lambda5007/[OII]\lambda3727$, $[Ne III]\lambda3869/[OII]\lambda3727$, and $[OIII]\lambda5007/[OIII]\lambda4363$, and testing against extensive local and JWST high-$z$ samples plus Cloudy photoionization models, the authors define AGN-dominated regions with robust demarcation lines. Stacking analyses of galaxies in the AGN regions reveal broad H$\alpha$ components, confirming hidden Type I AGN in some sources and supporting the efficacy of the diagnostics. The results highlight the impact of AGN heating on $T_e$-based metallicity measurements and call for AGN-specific strong-line calibrations, while demonstrating JWST’s power to uncover AGN demographics and refine metallicity estimates in the early universe.

Abstract

The James Webb Space Telescope (JWST) is revolutionizing our understanding of black holes formation and growth in the early Universe. However, JWST has also revealed that some of the classical diagnostics, such as the BPT diagrams and X-ray emission, often fail to identify narrow line TypeII active galactic nuclei (AGN) at high redshift. Here we present three new rest-frame optical diagnostic diagrams leveraging the [OIII]$\lambda4363$ auroral line, which has been detected in several JWST spectra. Specifically, we show that high values of the [OIII]$\lambda4363/$H$γ$ ratio provide a sufficient (but not necessary) condition to identify the presence of an AGN, both based on empirical calibrations (using local and high-redshift sources) and a broad range of photoionization models. These diagnostics are able to separate much of the AGN population from Star Forming Galaxies (SFGs). This is because the average energy of AGN's ionizing photons is higher than that of young stars in SFGs, hence AGN can more efficiently heat the gas, therefore boosting the [OIII]$\lambda4363$ line. We also found independent indications of AGN activity in some high-redshift sources that were not previously identified as AGN with the traditional diagnostics diagrams, but that are placed in the AGN region of the diagnostics presented in this work. We note that, conversely, low values of [OIII]$\lambda4363/$H$γ$ can be associated either with SFGs or AGN excitation. We note that the fact that strong auroral lines are often associated with AGN does not imply that they cannot be used for direct metallicity measurements (provided that proper ionization corrections are applied), but it does affect the calibration of strong line metallicity diagnostics.

Figures (7)

• Figure 1: Diagnostic diagram showing the [OIII]$\lambda4363\,$/H$\gamma\ $ vs [OIII]$\lambda5007\,$/[OII]$\lambda3727\,$ lines ratios. The [OII]$\lambda3727\,$ is the sum of the doublet [OII]$\lambda\lambda3726,3728$. Left panel: plot of all the observational samples described in Sec. \ref{['sec:low-z_obs']} and Sec. \ref{['sec:high-z_obs']}, with colours and shapes as labelled. Red colours are used for AGN, blue for low-z SFGs, and grey for high-z sources not classified as AGN. The red and blue contours show the distribution at SDSS AGN and SFGs, respectively, including the 90%, 70%, 30%, and 10% of the populations. Right panel: same plot but showing the AGN and SFG models computed by Gutkin16, Feltre16 and Nakajima22, as described in Sec. \ref{['sec:photmod']}. The tracks of the AGN and SFG models according to $\log U$ and Z are show in Fig. \ref{['fig:diag_mod_only']}. The black dashed line indicates the demarcation defined in Sec. \ref{['sec:demarcation_lines']}. Both the distribution of models and of observational samples suggest that the dominant ionizing source in galaxies above the demarcation line is AGN. The error bars reported in the lower-right corner represent the median errors of the low redshift (left) and high redshift (right) samples. The effect of dust attenuation on this diagnostic moves sources towards the right, without contaminating the AGN-only region with SFGs.
• Figure 2: Same as in Fig. \ref{['fig:diagn_O32']}, but for the lines ratios [OIII]$\lambda4363\,$/H$\gamma\ $ vs [Ne III]$\lambda3869\ $/[OII]$\lambda3727\,$. Based on the distribution of observational samples and models, also this diagnostic diagram identifies a region that can be populated only by AGN.
• Figure 3: Same as in Fig. \ref{['fig:diagn_O32']}, but for the lines ratios [OIII]$\lambda4363\,$ / H$\gamma\ $ vs [OIII]$\lambda5007\,$$\bigl/$[OIII]$\lambda4363\,$. In the left panel, the arrow on the AGN reported in Furtak23 is for visualization purposes, since it would be located at O3O3$\sim 0$. The cut in SFG models described in Sec. \ref{['sec:photmod']} allowed us to identify an AGN-only region of the diagnostic to the right of the black dashed line.
• Figure 4: Different fits of the spectrum derived from the stack of the sources not identified as AGN in the literature but lying in the AGN region of the diagnostic diagrams presented in Sec. \ref{['sec:results']}. The stacked spectrum in these plots is resampled at the best resolution among those of the single spectra involved (i.e. $\sim 100$km/s), but the results do not change considering the worst resolution (i.e. $\sim 150$km/s). Top panels: fit of the H$\alpha\ $ and [N II] complex with only narrow component (left) and adding a broad component to the H$\alpha\ $ line (right). The global fit is presented in red, while the narrow and broad H$\alpha\ $ components are in green and yellow, respectively. In the upper panel are shown the residuals of the fit compared with the distribution of the $1\sigma$ (blue) and $3\sigma$ (red) errors on the fluxes. On the upper right part of the plots are reported the FWHM and the velocity offset of the different components considered in the fits. Lower panels: same as above but for the H$\beta\ $ and [O III] $\lambda\lambda4959,5007\,$ doublet complex. In particular, on the right, we added to the fit of the [OIII]$\lambda5007\,$ a broad component with the same FWHM and velocity offset as the broad component of the H$\alpha\ $, but here it is not required by the fit.
• Figure 5: Distribution of the SDSS SFG sample (circles) and local analogues (triangles) in the $\rm \log U$ vs $\rm \log OH$ plane, according to the likelihood procedure with the Feltre16 models described in Sec. \ref{['sec:photmod']}. In each panel, we also show, with a red dashed line, the region of the parameter space that we decided to exclude, a posteriori, from the same photoionization models. The upper panels show the variation of the distribution of the sources by considering models with different electron densities $\rm n_H$, in the middle panels different dust-to-metal ratios $\xi$, and in the bottom panels different $C/O$ abundance. The excluded region is never significantly populated by any source in all the panels.