GA-NIFS: Powerful and frequent outflows in moderate-luminosity AGN at $z\sim3-6$

TL;DR

This work uses JWST NIRSpec IFU data from the GA-NIFS survey to spatially resolve ionised AGN outflows in a sample spanning $z\sim3-6$ and $L_{\rm bol}\sim10^{45-46}$ erg s$^{-1}$. By decomposing rest-frame optical emission lines into narrow and broad components, the authors derive outflow velocities ($v_{\rm out} \sim 600-2000$ km s$^{-1}$), maximum extents ($R_{\rm max} \lesssim 1-4$ kpc), and mass outflow rates ($\dot{M}_{\rm out} \sim 0.1-100\ M_\odot\ \mathrm{yr}^{-1}$), with mass loading factors near unity in many cases. The incidence of outflows is high ($>75\%$), consistent with strong feedback activity in the early Universe, and after correcting for luminosity bias, the data hint at a cosmic evolution toward stronger, more impactful outflows at higher redshift. The study emphasizes the role of outflows in regulating star formation and the necessity of homogeneous measurements of AGN properties and host-galaxy SFR to robustly quantify feedback across cosmic time.

Abstract

The period between z ~ 3-6, a key transformational phase in galaxy evolution preceding `cosmic noon' (z ~ 1-3), is very poorly explored in terms of feedback from AGN acting through gas outflows. In this work, we study the properties of outflows in AGN (mostly X-ray-selected) from the GOODS-S field, exploiting JWST NIRSpec IFU observations as part of the GA-NIFS GTO survey. Together with its twin sub-sample from COSMOS reported in a previous GA-NIFS work, this constitutes the largest spatially resolved sample of AGN outflows at these redshifts to date, comprising 16 targets with outflows (out of a total of 19 AGN), and probes the unexplored regime of AGN at z ~ 3-6 with bolometric luminosities ~$10^{45-46}$ erg/s. We mapped the rest-optical ionised gas emission lines at sub-kpc scales and spectrally isolated the broad wings tracing fast outflows from the gas at rest in the host galaxies. The incidence of ionised outflows in the GOODS-S + COSMOS GA-NIFS sample is high (>75\%), among the highest at any redshift. We inferred outflow velocities between ~600-2000 km/s, maximum radii of <1-4 kpc, and ionised gas mass outflow rates of ~0.1-100 $M_\odot$/yr, which in some cases can exceed the host galaxy star formation rate (SFR). We find that the outflow properties inferred for the GOODS-S + COSMOS GA-NIFS AGN sample and their relations with $L_{\rm bol}$ and SFR generally align with those observed for other spatially resolved literature samples of AGN outflows across different redshifts and AGN luminosities. Nonetheless, after accounting for any luminosity bias, our analysis suggests a cosmic evolution of the outflow properties, with higher median mass outflow rates (and possibly also mass loading factors) at higher redshifts, especially at z>3, indicating that AGN outflows were stronger in the early Universe than at later times, and thus potentially more capable of affecting their host galaxy.

Figures (14)

• Figure 1: AGN bolometric luminosity ($L_\textrm{bol}$) versus redshift ($z$). The sources observed as part of the GA-NIFS survey are reported with golden stars (GS-AGN; this work) and salmon diamonds (COS-AGN; Bertola2025); out of the sample of 19 GS- and COS-AGN, only the 16 sources with detected outflows are shown. We also plot the targets included in other spatially resolved studies of AGN outflows at different $z$ and $L_\textrm{bol}$, comprising other GA-NIFS sources not part of the two samples above (the $z \sim 3$ QSO from Perna2023b, the $z \sim 7$ QSOs from Marshall2023Marshall2024, the $z \sim 4$ source in GOODS-N from Ubler2024b, and the QSO + AGN-host SMG from Zamora2024). The plot shows that the parameter space at $z \gtrsim 3$ and $L_\textrm{bol}$$\lesssim$$10^{46}$ erg s$^{-1}$ had remained uncharted until the advent of GA-NIFS.
• Figure 2: Rest-optical emission maps of GS 133 from JWST NIRSpec $R$2700 observations. Top row, from left to right: flux (moment 0) and velocity (moment 1) of total [O iii]$\lambda$5007 line profile and centroid velocity of 1st (narrower) and 2nd (broader) Gaussian components. Bottom row: continuum emission in the rest-frame range 4400--4800 Å, velocity dispersion (moment 2) of total line profile and of components 1 and 2. The black contour on the [O iii] velocity dispersion map encompass the extraction region of the integrated spectrum used for the outflow analysis (see Sect. \ref{['sec:outf_integr_anal']} and Fig. \ref{['fig:spectra']}). The white contours on the [O iii] and continuum flux maps mark the continuum emission, while the black ones on the [O iii] velocity and velocity dispersion maps trace the [O iii] emission. In the maps for all the other targets in Figs. \ref{['fig:gs539']}-\ref{['fig:gs20936']}, different colours may be used for the contours to simply enhance the contrast, but have the same meaning as in this figure (when a different emission line instead of [O iii] is reported, the [O iii] contours are substituted with those of that line). The lime cross marks the position of the nucleus, as traced by the line emission peak.
• Figure 3: Integrated spectra for all of the sources (black). The spectrum extraction region is shown by the black contour superimposed on the velocity dispersion map of each source in Figs. \ref{['fig:gs133']} and \ref{['fig:gs539']}-\ref{['fig:gs20936']}. The coloured curves mark the narrow host-galaxy (blue; blue and orange for GS 10578), broad outflowing (purple), and BLR (green; for type-1 AGN) components, while the red curve marks the overall best-fit model. Vertical, dashed labelled lines mark the wavelength of the peak of the fitted narrow component of each fitted emission line, adopted as the systemic velocity of each system (except for GS 10578, for which we adopted the systemic redshift based on stellar absorption lines from DEugenio2024). Wavelengths are in the observed frame. Uncertainties on the data are reported as shaded areas. The displayed wavelength range is smaller than the entire spectral range covered by the observations and than the fitted one; the displayed range is the same for all the objects, being [--250, +50] Å around [O iii]$\lambda$5008 and [--100, +200] Å around H$\alpha$ as calculated in rest-frame wavelengths.
• Figure 4: Bolometric luminosity ($L_\textrm{bol}$) versus outflow properties for AGN from GA-NIFS and from the literature. From left to right: AGN $L_\textrm{bol}$ versus outflow velocity ($v_\textrm{out}$), mass outflow rate ($\dot{M}_\textrm{out}$), kinetic energy rate ($\dot{E}_\textrm{out}$), and mass loading factor ($\eta$ = $\dot{M}_\textrm{out}$/SFR). Symbols and colours are the same as in Fig. \ref{['fig:Lbolvz']}. The green dashed, purple dotted, and blue dash-dot lines represent the best-fit relations from Fiore2017, Musiimenta2023, and Bischetti2019, respectively, re-scaled to $n_\mathrm{e}$ = 1000 cm$^{-3}$. The diagonal black dashed lines in the lower-left plot mark $\dot{E}_\textrm{out}$ = 100%, 10%, 1%, 0.1%, and 0.01% $L_\textrm{bol}$. The horizontal black dashed line in the lower-right plot indicates $\eta$ = 1. The grey arrow quantifies the linear scaling factor that would be applied to all of the points when assuming a different value of the electron density, specifically when using 100 cm$^{-3}$ instead of the value of 1000 cm$^{-3}$ assumed.
• Figure 5: Redshift evolution of the outflow properties for sources with AGN $L_\textrm{bol}$ < 10$^{46}$ erg s$^{-1}$, matching the bolometric luminosities of our GA-NIFS GS- and COS-AGN targets. From left to right: Outflow velocity ($v_\textrm{out}$), mass outflow rate ($\dot{M}_\textrm{out}$), and mass loading factor ($\eta$ = $\dot{M}_\textrm{out}$/SFR) versus redshift. Measurements are reported for both single sources (top row) and in redshift bins (bottom row). For the latter, each point is the median (50th percentile) and the first and third quartiles (25th and 75th percentiles, respectively) of the distribution of values at each $z$ bin are marked by the dotted horizontal segments within each violin. The bins are $z$ = 0--1, 1--2, 2--3, 3--4, and 4--8. The area of each violin is proportional to the fraction of sources in the given redshift bin out of the total. The upper panels report the number of sources in each redshift bin.