betaFeedback

WorksPhysics and Astronomy

Recently launched AGN-driven outflow in a lensed quiescent galaxy at Cosmic Noon

Abstract

We analyze JWST/NIRSpec IFU spectroscopy of MRG-M0150, a gravitationally lensed, massive quiescent galaxy at z=2.6z=2.6 from the INQUEST-JWST survey. Thanks to the combination of lensing magnification with JWST's angular resolution, we are able to build spatially resolved maps of the ionized gas spatial distribution, kinematics, and ionization properties in this system. We detect ionized gas emission throughout the galaxy: the lack of residual star formation inferred from line-ratio diagnostic diagrams, together with the decoupling between the gas and stellar kinematics, suggest that the ionized gas is in a wind. We use the 3D gas kinematics modeling code MOKA3D to constrain the geometry and physical properties of the ionized gas wind. We find that the observations are consistent with a biconical wind driven by the low-luminosity AGN at the center of the galaxy, which is detected in X-rays. The wind reaches high velocities of 1000−14001000-1400 km/s, implying that part of the gas may escape the halo. The ionized gas mass outflow rate, (0.6±0.2)(0.6 \pm 0.2)~M⊙{\rm M}_{\odot}~yr−1{}^{-1}, is modest but exceeds the upper limit on the residual star formation rate, and it is likely that the wind is multiphase and carries more mass. We show that typical unresolved observations would not detect the wind, suggesting that the incidence of such outflows may be higher than previously recognized. Simple geometric arguments show that the observed outflow must have been launched ≈0.45\approx 0.45~Myr ago from the time of observation. This evidence suggests that low-level AGN activity is an important mechanism for maintaining quiescence in massive galaxies at cosmic noon, potentially over cosmological timescales.

Introduction

The physical processes responsible for the quenching of star-formation in massive galaxies remain one of the key open problems in galaxy formation and evolution. Numerous mechanisms have been proposed to explain the quenching of massive galaxies, including active galactic nuclei (AGN), stellar feedback, and virial shock heating [see 90 for a recent review], but no single mechanism seems to be universally applicable. Semi-analytical models and cosmological hydrodynamical simulations typically invoke AGN feedback as a mechanism capable of quenching massive galaxies. Feedback from the AGN can drive galaxy-scale outflows in multiple gas phases, quenching the galaxy both by expelling gas from the system and by heating the circumgalactic medium, thereby preventing the cooling and accretion of gas that could otherwise fuel subsequent star formation [e.g. 35, 51, 11, 75, 28]. However, direct observational evidence linking AGN feedback to the quenching of star formation remains elusive. At cosmic noon (z∼2−3z \sim2-3), a significant fraction of massive galaxies stopped forming new stars (e.g. Muzzin et al. 2013; Straatman et al. 2014; Tomczak et al. 2014; Weaver et al. 2023). Observing recently quenched systems at this epoch is crucial for investigating the main mechanisms responsible for quenching star formation in the most massive galaxies.

Galaxy-scale gas outflows have been extensively studied at z∼1−3z \sim1-3 through large galaxy surveys. It has been found that outflows are ubiquitous in massive star-forming galaxies at cosmic noon [76, 89, 44] and in those hosting the most luminous AGN (Cano-Díaz et al. 2012; Cresci et al. 2015; Kakkad et al. 2016; Davies et al. 2020), where they are expected to be powerful enough to be able to suppress star formation. However, the study of gas outflows and nuclear activity in the first quiescent galaxies at high redshift is challenging from the ground, relying either on deep exposures of individual galaxies (e.g. Kriek et al. [57]; Belli et al. [5]) or stacked spectra (e.g. Belli et al. [6]).

The launch of the James Webb Space Telescope (JWST) has opened a new window on the high-redshift Universe, revealing an abundant population of massive quiescent galaxies already in place at z>2z > 2 (Carnall et al. [19]; Weibel et al. [88]; Nanayakkara et al. [63]; de Graaff et al. [33]). Early JWST spectra of individual galaxies have revealed the presence of AGN activity in these systems (Carnall et al. [20]; Belli et al. [7]; D’Eugenio et al. [34]; Wu [92]) and subsequent studies with larger samples have confirmed a high AGN incidence in massive quiescent galaxies at z>2z > 2 (Bugiani et al. [13]; Skarbinski et al. [74]). These AGN, for the most part, have been classified as low-luminosity AGN (LL-AGN), based on their faint inferred bolometric luminosities and the lack of more overt tracers of powerful AGN activity such as X-rays detection or IR excess (Almaini et al. [1]; Hewitt et al. [47]).

At the same time, JWST observations have revealed a widespread presence of gas outflows in massive quiescent galaxies at high redshift, detected both in the neutral and ionized phases (Davies et al. [30]; Taylor et al. [79]; Valentino et al. [81]; Sapori et al. [73]; Sun et al. [77]). These outflows can be remarkably powerful, often carrying large amounts of cold gas. In particular, observations of neutral-phase outflows in z>2z > 2 galaxies through the Na D absorption line have shown that powerful winds are prevalent among massive galaxies at cosmic noon and may represent the dominant pathway for quenching at this epoch (Davies et al. [30]; Sapori et al. [73]). In some systems, the inferred neutral-phase mass outflow rates exceed the current star-formation rates by more than an order of magnitude, suggesting that efficient removal of cold gas may contribute directly to the rapid quenching observed in these galaxies (Belli et al. [7]; D’Eugenio et al. [34]).

However, the origin of these galactic winds and their connection to star-formation quenching remain unclear, as the low-level AGN activity observed in many massive systems struggles to explain the high energy and momentum rates characterizing the observed outflows. While some studies attribute the origin of the observed winds to ongoing LL-AGN activity (e.g. Belli et al. [7]; Wu [92]; Bugiani et al. [13]), others interpret them as relics of past starburst episodes or previous periods of more powerful black hole activity, with the central AGN having since faded into a low-accretion state (Valentino et al. [81]; Sun et al. [77]; Taylor et al. [79]). Distinguishing between these scenarios is therefore crucial for establishing whether the outflows observed in massive quiescent galaxies represent ongoing AGN feedback or the remnants of the events that originally quenched the galaxy.

With only a few exceptions (Gangula et al. [45]; D’Eugenio et al. [34]), most studies rely on spatially unresolved data and therefore are not able to directly constrain the geometry, kinematics, and timescales of the outflows. While the ionized phase may represent a small contribution to the total outflowing mass, as observed in high-zz quiescent galaxies (Belli et al. [7]) and predicted by simulations of galactic winds (Kim et al. [55]), it is the best suited to measuring these properties. Ionized gas can be detected directly in emission without requiring a bright background continuum. Its emission traces the full spatial extent of the wind, and, provided that it can be spatially resolved, yields detailed information about the kinematics and 3D structure of the outflow. This can reduce some of the uncertainties associated with assumptions about the geometry and kinematics when deriving mass outflow rates and energetics. Moreover, reconstructing the extent, geometry, and velocity of the winds provides a way to constrain their characteristic timescales, which can be used to distinguish between winds powered by ongoing AGN activity and fossil outflows launched by a past quasar or starburst event. Resolved observations of the ionized phase can therefore offer a unique view on the connection between AGN activity, galactic outflows, and the maintenance of quiescence in massive galaxies at high redshift.

In this work, we investigate ionized gas emission in MRG-M0150, a strongly-lensed, massive quiescent galaxy at z=2.64z = 2.64 (Newman et al. [64]) using spatially resolved observations obtained by JWST/NIRSpec in IFU mode. We make use of the combined sensitivity of JWST and gravitational lensing magnification to characterize the geometry, velocity, and energetics of the ionized gas outflow observed in this system, and use it to investigate the connection between the outflow and the quenching of star-formation in this galaxy. In § 2, we present the observations used in this paper. In § 3, we describe the emission-line fitting and discuss the resulting maps and resolved emission-line diagnostics, as well as the AGN activity and residual star formation. In § 4, we construct a 3D kinematic model of the observed ionized gas outflow and derive its main properties. In § 5, we perform SED fitting of the integrated light from the galaxy to reconstruct its star-formation history (SFH). We discuss our results in § 6 and summarize our main conclusions in § 7. Throughout this paper, we assume the WMAP9 cosmology (Hinshaw et al. [48]) and a Chabrier [22] initial mass function.

Target and observations

MRG-M0150 is a massive quiescent galaxy at z=2.64z = 2.64, multiply imaged by the galaxy cluster MACS J0150.3−1005 at z=0.365z = 0.365 (Newman et al. [64]). A JWST/NIRCam image of the cluster, with the multiple images of our target labeled, can be found in Fig. A.1 in Appendix § 8. The galaxy has a stellar mass of log⁡(M/M⊙)=11.30±0.04\log(M/M_{\odot}) = 11.30 \pm0.04 – derived by the integrated SED fitting, see Sect. 5 – and is classified as a “fast rotator,” exhibiting a rotationally supported stellar disk (Newman et al. [66]). It has a circularized half-light radius of Re∼0.2R_{\mathrm{e}} \sim0.2 arcsec (corresponding to 1.6 kpc on the source plane), as found by the lens modeling (Appendix § 8). Dynamical models indicate that the inclination of the stellar disk is ∼30∘\sim30^{\circ} (Newman et al. [66]).2

The target belongs to the INQUEST-JWST galaxy sample, which includes 5 gravitationally lensed, massive and quiescent galaxies at redshift z=2.1−2.6z = 2.1 - 2.6. Data used in this paper comes from the Cycle 3 JWST GO-4903 observational program. The program observed one image of MRG-M0150, labeled Image 1 in the cluster image visible in Fig. A.1, using the JWST/NIRSpec IFU with the G140M/F100LP and G235M/F170LP grating/filter pairs. Exposure times were 4.4 hours for each setting. Additionally, the same program acquired NIRCam imaging with total exposure times of 20-30 minutes in the F115W, F200W, F277W and F444W filters. We also make use of archival imaging data from the Hubble Space Telescope (HST) Advanced Camera for Surveys (ACS) through the F814W filter (Newman et al. [64]) for the full SED fitting (Sect. 6.4).

The NIRSpec data cubes were reduced using the JWST Science Data Processing system version 2025_3a and the JWST Science Calibration Pipeline version 2.0.0. Spectral “wiggles” – artifacts affecting NIRSpec data due to undersampling of the PSF – were mitigated using the adaptive trace model step in the pipeline. We detected some residual wiggles, which we modeled and removed using the procedure described in Newman et al. [67]. We created output data cubes with a 0.1″ spaxel scale for the G140M and G235M observations. The fluxes and variances are logarithmically rebinned onto a wavelength grid with 132 km s−1^{-1} per pixel and shifted to the systemic rest frame (zsys=2.6355z_{\mathrm{sys}} = 2.6355). Then they are combined by splicing them together at λrest=0.505\lambda_{\mathrm{rest}} = 0.505 μm. Noise spectra coming out of the pipeline are rescaled based on residuals to the stellar continuum model fits (Sect. 3.1). The resulting data cube is spatially binned to a signal-to-noise target of 40 using the vorbin algorithm (Cappellari & Copin [18, 17]). A visualization of the resulting Voronoi bins can be seen in the left panel of Figure 1. We also measure PSF-matched photometry in each spatial bin for every photometric filter.

Median signal-to-noise ratio map and example spectrum

Figure 1. Left: Median signal-to-noise ratio along the wavelength axis of the IFU data cube. Overplotted in white are the boundaries of the Voronoi bins analyzed. Right: One example spectrum extracted from the central spaxel (blue) with the pPXF stellar continuum fit overlaid in red.

The final step is to reconstruct the galaxy onto the source plane by applying the lens model described in Appendix § 8. The gravitational magnification of a factor of μ∼5\mu\sim5 in the central parts of the galaxy delivers a resolution of ∼\sim5 pc for the IFU data of σpsf=29\sigma_{\mathrm{psf}} = 29 mas = 235 pc, calculated by casting a PSF placed at the galaxy’s centroid to the source plane (see Appendix § 9).

Spatially resolved gas emission

Emission lines fit

In order to isolate emission lines due to ionized gas, we first have to model the underlying emission from the stellar continuum. We fit the stellar continuum contribution in each spaxel using the pPXF code. The model used for the stellar population synthesis with pPXF and the E-MILES library (Vazdekis et al. [83]) is analogous to the one used by Newman et al. [67]. The best-fit stellar continuum model is subtracted from the data to isolate the emission lines. An example spectrum and stellar continuum fit is given in Figure 1. To measure the stellar kinematics, we perform an additional ppxf run in which the strongest emission lines are masked, since tests showed that fitting them could bias the stellar velocity dispersions (Newman et al., in prep.).

All rest-frame optical emission lines – consisting of Hα\alpha, Hβ\beta, [OII]λλ3727,3729\lambda\lambda3727, 3729, [OIII]λλ5007,4959\lambda\lambda5007, 4959, [NII]λλ6548,6581\lambda\lambda6548, 6581, [SII]λλ6717,6732\lambda\lambda6717, 6732, [NeIII]λ3968\lambda3968 and [OI]λ6300\lambda6300 – are fitted together for each spatial bin. The fitting is performed using EMCEE, an affine invariant Markov Chain Monte Carlo (MCMC) ensemble sampler (Foreman-Mackey et al. [43]). We employ single Gaussian profiles fitting the same velocity shift and dispersion for all lines, and taking into account the NIRSpec wavelength-dependent spectral broadening. The flux ratios [OIII]λ5007\lambda5007/[OIII]λ4959\lambda4959 and [NII]λ6581\lambda6581/[NII]λ6548\lambda6548 are fixed at the theoretical value of 3:1 (Osterbrock & Ferland [69]). Although our spectral resolution is not sufficiently high to resolve the [OIII]λλ3727,3729\lambda\lambda3727, 3729 doublet, we fit the lines separately and impose the 1.5 << [OII]λ3729\lambda3729/[OII]λ3727\lambda3727 << 2.5 theoretical limits on their flux ratio (Osterbrock & Ferland [69]).

We then repeat the fit using double Gaussian components for each line profile, in which the kinematics of both components are tied together across all lines during the fit. We do not assign any physical meaning to each Gaussian component, but rather use them to better reproduce the observed line profiles. Finally, for each spatial bin, we select which model better reproduces the data based on residuals and the Bayesian Information Criterion (BIC) score of the fits. The line surface brightness values provided throughout this paper correspond to the median values from the EMCEE parameters’ posterior distributions, with uncertainties represented by the ±1σ\pm1\sigma percentiles around the median values. We consider a line above the detection threshold when it is detected with a signal-to-noise ratio >3> 3.

We find that ionized gas emission is detected across the entire galaxy, with many bins requiring double Gaussian profile to reproduce the observed data. The high spatial resolution highlights interesting features in the ionized gas lines, that make it possible to thoroughly map the kinematics and physical properties of the ionized gas. In Figure 2, we show the integrated surface brightness maps for the Hα\alpha and [OIII]λ5007\lambda5007 emission lines, and we compare the kinematics of the gas emission and the stellar component. Hα\alpha is well detected in all bins, while [OIII] is mostly detected in the central regions and along the minor axis (upper left and upper right panels, respectively). Both lines’ surface brightness distribution peak towards the center, with the brightest bin overlapping the centroid of the galaxy.

Figure 2. Source plane maps of MRG-M0150. First row: integrated surface brightness of the fitted HαH\alpha (left) and [OIII]λ5007\lambda5007 emission lines (right); light gray areas mark non-detections. Second row: velocity field of the gas (left) from fitted HαH\alpha line profile, and of the stars (right) from pPXF. Third row: gas (left) and stellar (right) velocity dispersions, obtained in the same way. Maps are rotated to align the x and y axis to the major and minor axis, respectively.

Gas kinematics

In the second and third rows of panels in Figure 2 we compare the velocity and dispersion of the gas (which are taken as the moment 1 and moment 2 of the best-fit profile of the Hα\alpha emission line) to the stellar kinematics. It is immediately visible that the kinematics of stars and ionized gas are not well coupled in this system. The stellar velocity field traces a well defined rotating disk, as already observed by Newman et al. [66] (right panel), whereas the gas does not follow at all the stellar rotational pattern, with the red- and blue-shifted sides of the gas velocity field positioned almost perpendicular to the stellar rotational axis. The velocity dispersion field (last row of panels) is also quite different between stars and gas. The gas map shows a symmetric, high-velocity dispersion feature almost aligned with the minor axis of the galaxy that is absent from the stellar map, as well as exhibiting generally higher values of σ\sigma in the central regions.

The decoupling between stellar and gas velocity field suggests that the gas emission we are detecting is not produced by star-forming regions rotating in the disk, but may instead be coming from a galaxy-wide ionized gas outflow, as commonly observed in AGN host galaxies with the same gas kinematics signature [39, 42]. The outflowing gas scenario could also explain the widespread emission from ionized gas observed in this quenched galaxy, in the absence of the star-formation needed to sustain the observed emission. In this case, an outflow centrally driven by an active nucleus would explain both the observed line fluxes and gas kinematics.

Resolved line ratios

We investigate the possibility that MRG-M0150 hosts an AGN by using resolved rest-frame optical line diagnostics, which are widely used to discern between the contributions of star formation (SF) and AGN emission to the ionization of the interstellar medium (ISM). In the top panel of Figure 3 we show the BPT diagram [3], i.e., [OIII]/Hβ\beta as a function of [NII]/Hα\alpha, for each spatial bin in which all lines are detected. The resolved BPT diagram of MRG-M0150 indicates the presence of ionizing radiation that is not due to star formation: as noticeable in the upper panel of Figure 3, all spatial bins for which all 4 emission lines are detected – colored in the map on the right – are clustered in the same region of the BPT diagram. Line ratios typical of Low Ionization Nuclear Emission-line Regions (LINERs) are detected from the center to the edges of the galaxy. LINER-like ratios can arise from AGN photoionization [e.g. 49], shocked gas [e.g. 37] or ionization by hot evolved stellar populations [e.g. 4] and the BPT diagram alone cannot definitively differentiate between excitation mechanisms.

Resolved emission line ratios in MRG-M0150

Figure 3. Resolved emission line ratios in MRG-M0150. Top: resolved BPT [3] diagram (left), in which bins have been colored according to their median distance from the center of the galaxy in kpc, mapped onto the right panel. Bottom: wHaN diagnostic diagram [25] (left) colored by EW of Hα (right): spaxels below the “retired galaxies” purple line show emission that might be produced by an evolved stellar population; all the rest show definite evidence of AGN or shock contribution.

We construct the resolved WHaN diagram [25] by showing the equivalent width (EW) of Hα\alpha as a function of [NII]/Hα\alpha, in Figure 3, lower left panel. The WHaN diagram is useful to identify regions of gas ionized by hot evolved stellar populations, such as post-AGB stars, which can heavily contribute to the total ionizing flux budget particularly in quiescent galaxies [24]. We find that most of the spatial bins in MRG-M0150 are located above the “retired galaxy” line at 3 Å in Hα\alpha EW [25]: this implies that most of the observed emission is not excited by ionization due to old stars, but instead a contribution from AGN radiation or shocks is needed to explain the observations. Furthermore, Byler et al. (2019) found that the EW of Hα\alpha line emission due to post-AGB stars is weaker for younger stellar populations (stellar ages <1< 1 Gyr), shifting down the retired galaxies threshold for high-z quiescent galaxies. Estimating the contribution of shocks from AGN or star-formation feedback to the total ionizing budget of MRG-M0150 is a more difficult endeavor. The negligible level of current star formation in this galaxy, as inferred from Hα\alpha luminosity (Sect. 3.4), makes it unlikely that shocks from star formation feedback are the main source of ionization. Instead, the observed LINER-like line ratios are more likely to be due to AGN photoionization or shocks from a central AGN-driven outflow.

We map the Hα\alpha EW distribution across the galaxy in the lower-right panel of Figure 3: unlike the Hα\alpha surface brightness map in Figure 2, the Hα\alpha EW is higher in the outskirts of the galaxy than in the center. This radial trend cannot be explained by gradients in star formation or dust attenuation, given that the line ratios show little indication of either (Sect. 3.4). It is also inconsistent with hot, evolved stars powering the line emission, since in that case the line emission should more closely follow the stellar light. Instead, we propose that the Hα\alpha EW pattern is analogous to that seen in ‘red geysers’, a category that represents ∼10%\sim10\% of all massive quiescent galaxies at z∼0z \sim0. Red geyser galaxies are characterized by the presence of a low-luminosity AGN (LL-AGN) in their center and centrally-driven ionized gas outflows (Chung et al. 2016; Roy et al. 2021). In these galaxies the same Hα\alpha EW pattern is due to the fact that the Hα\alpha emission is tracing the outflowing gas, while the continuum is tracing the stellar population: going from the center to the outskirts of the galaxy, the stellar surface brightness profile decreases more rapidly then the Hα\alpha emissivity in the outflow, generating the observed inside-out EW distribution. The distribution of Hα\alpha EW in MRG-M0150 is thus further evidence in support of the presence of a centrally-driven, galaxy-wide ionized gas outflow.

Residual star formation and AGN activity in MRG-M0150

Although line ratio diagnostics and gas kinematics indicate that the line emission is not powered by star formation, we can still use the integrated Hα\alpha line luminosity to place an upper limit on the residual star formation rate. We obtain the total integrated Hα\alpha flux by first correcting the flux measured in each spaxel for magnification, and then correcting for dust attenuation. In a small number of bins (13 out of 61) where both Hα\alpha and Hβ\beta lines are above the detection limit of 3σ\sigma, we are able to use the Balmer decrement Hα\alpha/Hβ\beta to estimate the dust attenuation AVA_V; for all other bins, we use the value estimated by the pPXF fits. We find generally low dust attenuation throughout the galaxy using both methods, with a median value of AV=0.3−0.3+0.4A_V = 0.3^{+0.4}_{-0.3}. We then apply the Calzetti et al. [15] dust attenuation curve and obtain the unabsorbed Hα\alpha luminosity. Assuming the conversion provided by Kennicutt [54], we derive a current SFR for MRG-M0150 of 0.30 ±\pm 0.08 M⊙_\odot yr−1^{-1}, which has to be considered an upper limit on the true SFR of the galaxy, since most of the Hα\alpha emission is excited by processes other than star formation, according to the diagnostics of Figure 3. For a galaxy of stellar mass log⁡(M/M⊙)=11.30\log(M/M_\odot) = 11.30, this implies a specific star-formation rate of log⁡(sSFR/yr−1)=−11.8\log(\mathrm{sSFR}/\mathrm{yr}^{-1}) = -11.8. This is very strong evidence of the currently quenched nature of the galaxy, and confirms the presence of an AGN in the center of the galaxy as the main engine powering the observed ionized gas emission.

We investigate the AGN properties by estimating its bolometric luminosity using the [O III] 5007 line luminosity in the central spaxel, corrected for aperture losses assuming a point-like emission. We correct for dust attenuation using the Balmer decrement and compute a central [O III] luminosity of L[O III]=(9±2)⋅1039L_{\mathrm{[O\ III]}} = (9 \pm2) \cdot10^{39} erg s−1^{-1}. Assuming the bolometric correction factor from Heckman et al. [46] we derive the AGN bolometric luminosity Lbol=(3.2±0.9)⋅1043L_{\mathrm{bol}} = (3.2 \pm0.9) \cdot10^{43} erg s−1^{-1}. We estimate the Eddington luminosity of the SMBH from the black hole mass, which we infer from the MBH−σ_{\mathrm{BH}}-\sigma relation of McConnell et al. [61], adopting the light-weighted second stellar velocity moment within the effective radius to define σ=282\sigma= 282 km s−1^{-1}. We find an Eddington ratio of Lbol/LEdd≈10−3.8L_{\mathrm{bol}}/L_{\mathrm{Edd}} \approx10^{-3.8}. Using the [O III] line luminosity integrated over the whole galaxy – as might be done for unresolved studies – yields a bolometric AGN luminosity almost 6×\times higher.

We further constrained LbolL_{\mathrm{bol}} using a 27 ks observation with the Advanced CCD Imaging Spectrometer (ACIS) onboard the Chandra X-ray Observatory (program 18000090, PI H. Ebeling). We used the same methodology as Newman et al. [67] to measure LXL_X, which we define from 2–10 keV in the rest frame, and to estimate Lbol=CXLXL_{\mathrm{bol}} = C_X L_X using CX=15.8C_X = 15.8 [50]. For the multiple images 1, 2, and 3, we find Lbol/1044L_{\mathrm{bol}}/10^{44} erg s−1^{-1} = 1.3 ±\pm 1.2, 2.9 ±\pm 2.3, and 5.1 ±\pm 1.9, respectively. We thus marginally detect X-ray emission in each image. We combined the Poisson posteriors for LbolL_{\mathrm{bol}} from the three multiple images to estimate Lbol=3.3−1.0+1.2⋅1044L_{\mathrm{bol}} = 3.3^{+1.2}_{-1.0} \cdot10^{44} erg s−1^{-1}. This is an order of magnitude higher than [O III]-based estimate, which could be an underestimate due to the LINER-type spectrum. Nonetheless, the implied Eddington ratio is still low, Lbol/LEdd≈10−2.8L_{\mathrm{bol}}/L_{\mathrm{Edd}} \approx10^{-2.8}.

The low bolometric luminosity and Eddington ratio classifies the SMBH as an LL-AGN, a class of AGN occupying the faintest end of the luminosity function and characterized by very low accretion rates, Lbol/LEdd≤10−3L_{\mathrm{bol}}/L_{\mathrm{Edd}} \leq10^{-3} [see 49 for a review]. Recent studies of similar quenched systems at cosmic noon and even higher redshift have found a high incidence of LL-AGN activity [13, 74] based on rest-frame optical line diagnostics.

Ionized gas outflows

Wind model

All the evidence so far points to the presence of an AGN-driven galaxy-wide ionized gas outflow originating from the central part of the galaxy. To test whether our observations are compatible with such model we employ the kinematics fitting code “Modeling Outflows and Kinematics of AGN in 3D” (MOKA3D; Marconcini et al. [60]) and apply it to the de-lensed maps of observed emission line profiles. MOKA3D reconstructs the three-dimensional gas distribution in the outflow by weighting model emitting clouds according to the observed line emission profile and velocity in each spaxel (Sect. 3.1). It then generates a model data cube with the same spatial and spectral resolution as the observations, including PSF convolution. This tomographic reconstruction enables the geometry and kinematics of the outflow to be constrained, and provides estimates of its mass outflow rate and energetics.

The model adopted is shown in Figure 4a. We assume a bi-conical geometry for the wind, with a semi-aperture angle of 45 degrees consistent with other AGN-driven outflows studied at lower or equal redshift. The origin is fixed in the brightest spaxel in the [OIII] emission line map, co-spatial with the geometric center of the galaxy, and the position angle of the bi-cone’s axis is fixed to −5 degrees with respect to the minor axis after visual inspection of the observed gas velocity distribution. We adopt a constant radial velocity profile in both the receding and the approaching cones, which are fitted independently from each other. For each cone, we find best-fit values of radial velocity voutv_{\mathrm{out}} and inclination angle β\beta, defined as the angle between the cone’s axis and the observer’s direction, of the wind in 5 shells of incremental radius, as shown in Figure 4a. Global values for the inclination and intrinsic velocity of the wind in the cone are given by averaging the values obtained in the shells. We repeat the fit twice – once fitting the emission line profiles of the [OIII] line and once with Hα\alpha – and find consistent results. Further details on the 3D kinematics modeling of the gas emission in MRG-M0150 are provided in Appendix § 9.

Best-fit kinematic fit with MOKA3D, showing the bi-conical wind model mask and observed, model, and residual gas velocity and dispersion maps

Figure 4. Best-fit results of the kinematic fit with MOKA3D. Panel (a): bi-conical wind model mask overlayed on the observed Hα\alpha surface brightness distribution; the model is made up of two cones delimited by the pink diagonal lines; each cone is further divided into 5 concentric shells. Panel (b): comparison between the observed and model gas velocity (moment 1) and dispersion (moment 2) maps from Hα\alpha line profiles.

Outflow properties and energetics

The results of the MOKA3D modeling using the Hα\alpha line profiles and a bi-conical wind prescription are summarized in Figure 4. Analogous results and plots for the [OIII] line fitting results can be found in Appendix § 9. The gas kinematics are well reproduced by the model, as visible from the residuals map. The observations match a bi-conical wind whose axis is nearly within the plane of the sky: a schematic representation of the geometry of the outflow with respect to the stellar disk and our line-of-sight is shown in Figure 5. For the approaching (+) side, we find an inclination angle β=80\beta= 80 degrees with respect to the line-of-sight and an intrinsic de-projected velocity of v+=1200±57v_{+} = 1200 \pm57 km s−1^{-1}; for the receding (−) side an inclination angle β=100\beta= 100 degrees and similar velocity v+=1150±49v_{+} = 1150 \pm49 km s−1^{-1}. Consistent results are obtained from the model fitted to the [OIII] line emission, although we measure slightly lower values of wind velocity (see Table 1). The two cones extend out to the edge of the fitted spaxels, corresponding to an extension of at least 3.3 kpc (2.3 times the effective radius ReR_{\rm e} in F444W).

Emission lineFlux 10−1810^{-18} erg s−1^{-1} cm−2^{-2}voutv_{\mathrm{out}} km s−1^{-1} (+) sidevoutv_{\mathrm{out}} km s−1^{-1} (−) sideβ\beta degrees (+) sideβ\beta degrees (−) sideM˙out\dot{M}_{\mathrm{out}} M⊙_\odot yr−1^{-1}E˙kin\dot{E}_{\mathrm{kin}} 104110^{41} erg s−1^{-1}p˙out\dot{p}_{\mathrm{out}} 103310^{33} dyn
Hα3.0±0.83.0 \pm0.81200±571200 \pm571150±731150 \pm7380±580 \pm5100±3100 \pm30.6±0.20.6 \pm0.22.8±0.82.8 \pm0.85±25 \pm2
[OIII]λ50073±13 \pm11000±661000 \pm661050±491050 \pm4980±380 \pm3100±5100 \pm50.08±0.020.08 \pm0.020.3±0.10.3 \pm0.10.5±0.20.5 \pm0.2

Table 1. Results from the 3D kinematics fitting with a bi-conical wind model, for both the Hα and [OIII] line fits.

3D representation of the geometry of the bi-conical outflow

Figure 5. 3D representation of the geometry of the bi-conical outflow fitted with MOKA3D with respect to the stellar disk and our line-of-sight. The galaxy is seen at an inclination of i=30∘i = 30^\circ (gray disk), while the outflow is inclined at β=80∘\beta= 80^\circ and β=100∘\beta= 100^\circ for the approaching and receding cones, indicated with a blue and red arrow respectively. Each cone has an aperture angle of α=45∘\alpha= 45^\circ by construction. We have color-coded the cones according to their apparent velocity along the line-of-sight, with blue and red corresponding to approaching and receding velocities, respectively.

As visible in Figure 5, the bi-conical outflow’s inclination implies strong interactions between the galaxy’s disk and the outflow. This is not often seen in classical QSOs and starburst systems hosting such large-scale ionized phase winds [84], as usually the outflow chooses the path of least resistance out of the galaxy and is seen positioned perpendicular to the disk plane. The quiescent nature of MRG-M0150 may be a contributor to the relative orientation of the outflow and the disk: indeed, a centrally-driven wind expanding inside a quiescent system with a negligible amount of gas in its disk does not encounter much resistance from the ISM along its path and is free to propagate in any direction. This is the case for MRG-M0150, for which ALMA observations of the dust continuum (REQUIEM survey, Whitaker et al. [91]) give an upper limit of 1% on its molecular gas fraction. Interestingly, recent semi-analytical models of dust evolution in quiescent galaxies find that AGN feedback can rapidly destroy dust in after quenching (Donevski et al. [36]), in line with both the ALMA results and the AGN activity detected in MRG-M0150. On the other hand, the close alignment between the outflow axis and the stellar disk could favor strong interactions between the wind and the ISM, contributing to a negative feedback mechanism. The higher residuals between model and observations in the moment 2 (Figure 4, bottom right panel) reflects closely the feature of enhanced gas velocity dispersion of Figure 2 and may be tracing injected turbulence by the wind in the densest ISM regions as a consequence of wind-disk interactions.

We use the model to constrain the wind’s mass outflow rate and energetics, following Marconcini et al. [60]. The tomographical reconstruction of emission clouds in the outflow by the MOKA3D code enables us to accurately estimate the mass outflow rate of the ionized phase outflow and to better constrain wind energetics with respect to standard approaches used when the outflow physical properties are unknown ([21, 16]). But first, we need to estimate an electron density nen_{\mathrm{e}}. This is usually derived using either the [O II]λλ\lambda\lambda3727, 3729 or the [S II]λλ\lambda\lambda6717, 6732 doublet ratio (Osterbrock & Ferland [69]), or is assumed from a broad range of values – from 1000 cm−3^{-3}({citep}Venturi et al. 2025; Bertola et al. 2025) to more widely adopted choices of $n_{\mathrm{e}} \sim 100 - 200$ cm$^{-3}$([53, 12, 42 e.g.] (Cresci et al. 2015)). We infer the average electron density of the gas in the red- and blue-shifted sides of the outflow from the observed [S II]λλ\lambda\lambda6717, 6732 doublet ratio, since the [O II] doublet is not resolved at NIRSpec’s medium spectral resolution. The most rigorous approach involves deriving a resolved map of the [S II] doublet ratio to infer nen_{\mathrm{e}} in each spaxel: however, the [S II] doublet is below detection limit for many bins of the cube, on top of being difficult to de-blend in most of the rest. Thus, we choose to adopt a constant electron density value ne=600n_{\mathrm{e}} = 600 cm−3^{-3} for both sides of the wind, corresponding to the average value found in all spaxels with an accurate estimate of the [S II] doublet ratio, assuming a temperature of 10410^{4} K. This choice constitutes one of the most important sources of uncertainties in the outflow’s energetics estimation – indeed, assuming a constant electron density can significantly impact the outflow energetics ([85] (Davies et al. 2020)) – thus, we attach a 50% systematic uncertainty to nen_{\mathrm{e}} to account both for the averaging procedure and the uniform density assumptions. Moreover, different values adopted for the constant electron density only affect the (inverse) scaling of the mass outflow rate M˙out\dot{M}_{\mathrm{out}} and related quantities by a numerical factor; thus our inferred ne=600n_{\mathrm{e}} = 600 cm−3^{-3} results in more conservative M˙out\dot{M}_{\mathrm{out}} values than the typical assumptions of ne∼100−200n_{\mathrm{e}} \sim100 - 200 cm−3^{-3} would yield.

We finally obtain a total mass outflow rate from the ionized gas of 0.08±\pm0.02 M⊙_{\odot} yr−1^{-1} from the [O III] line flux, and 0.6±\pm0.2 M⊙_{\odot} yr−1^{-1} from the Hα\alpha line flux. This discrepancy in mass outflow rate values between [O III] and Balmer lines tracers is a consequence of the different volumes in which [O III] and Hα\alpha are emitted: [O III] is a higher-ionization tracer that usually misses a substantial amount of lower-ionized gas ([21]), especially in LINERs such as MRG-M0150. Furthermore, no dust correction was applied to the line luminosities before calculations, thus [O III] may be more affected by attenuation than the redder Hα\alpha line, as well as being under the detection limit in many spaxels embedded in the outflow (see Fig. B.3 in Appendix § 9). Thus, we adopt the mass outflow rates and derived properties given by Hα\alpha as our fiducial values. Consequently, we can calculate the energy and momentum rates of the ionized wind as:

E˙kin=12M˙outvout2=(2.8±0.2)⋅1041 erg s−1(1)\dot{E}_{\mathrm{kin}} = \frac{1}{2}\dot{M}_{\mathrm{out}}v_{\mathrm{out}}^{2} = (2.8 \pm0.2) \cdot10^{41}\ \mathrm{erg\ s}^{-1} \tag*{(1)}
p˙out=M˙outvout=(5±2)⋅1033 dyne(2)\dot{p}_{\mathrm{out}} = \dot{M}_{\mathrm{out}}v_{\mathrm{out}} = (5 \pm2) \cdot10^{33}\ \mathrm{dyne} \tag*{(2)}

All the derived outflow energetics are summed up in Table 1.

Alternative models

We also consider a tilted gaseous disk geometry to explain the observed emission. To test this scenario, we use MOKA3D to fit a rotating disk model to the observed Hα\alpha line emission, allowing it to be freely inclined with respect to the stellar disk. We find that the rotating disk model requires extreme rotational velocities of vrot∼600v_{\rm rot} \sim600 km s−1^{-1} and an inclination of the gas disk of 80 degrees. Thus, we can exclude this model from the discussion. We also experiment with an alternative model in which we constrain the inclination angle of the bi-cone to be 30°, in order to simulate a (more commonly observed) nearly-polar bi-cone with a 3D outflow velocity more comparable to the observed projected one. We find, however, that such model does a poor job in fitting the observed data (see Appendix § 9).

SED fitting and star-formation history

We carry out full SED fitting on the integrated photometry and spectroscopy of MRG-M0150 using Prospector (Johnson et al. [52]), to reconstruct its star-formation history (SFH) and verify the quenched state of the galaxy. The integrated photometry and spectrum are extracted from all spatial bins included in the analysis, corresponding to an aperture of 2.3 ReR_{\rm e}. Flux values extracted in each spaxel are corrected for magnification according to the lens model before being added together, and an additional 10% error is added in quadrature to all photometric points to account for systematic uncertainties.

We employ a model akin to the one used in Park et al. [70], adopting MIST isochrones (Dotter [38]), the MILES stellar library (Falcón-Barroso et al. [40]), and a non-parametric star-formation history (SFH) defined on 14 logarithmically-spaced age bins. A “continuity prior” on the SFH is adopted, which discourages drastic variations of SFR between adjacent time bins (Leja et al. [58]). We fit photometry from the ACS/WFC F814W filter and JWST/NIRCam’s F115W, F200W, F277W, F444W filters, together with the integrated spectrum from NIRSpec/IFU in the 4000 − 6700 Å wavelength range. Given that nebular emission lines are tracing the AGN activity and/or the outflowing gas and not star formation, we choose to marginalize over them during the fit. We do not include any AGN emission in the model.

The quality of the photometry + spectrum fitting is shown in Figure 6. In Figure 7, we plot MRG-M0150’s non-parametric star formation history derived by Prospector (green line). This is compared to our upper limit on the current SFR measured from the integrated Hα\alpha luminosity (red diamond) and to the mass outflow rate of the ionized gas outflow (blue squares). The reconstructed SFH of MRG-M0150 is similar to other recently quenched systems observed at z>2z > 2 ([70, 20, 7]), showing a sharp decrease of the SFR after the peak at ∼500\sim500 Myr ago in lookback time. We further discuss the SFH of MRG-M0150 and the impact of outflows in Sect. 6.4.

Results of the SED fitting procedure on the integrated photometry and spectrum

Figure 6. Results of the SED fitting procedure on the integrated photometry and spectrum. Left: observed and predicted photometric points, and galaxy SED from the best-fit Prospector model (green plot). Right: the observed integrated spectrum (blue) and the best-fit model (red) in the 4000 – 6700 Å rest-frame wavelength range, with residuals in the lower panel.

Reconstructed star-formation history of MRG-M0150 according to the Prospector fit

Figure 7. Reconstructed star-formation history of MRG-M0150 according to the Prospector fit. The solid green line indicates the median SFR per time bin, with the light green shaded areas indicating the 16th – 84th confidence intervals from the posterior distributions. The current SFR and mass outflow rates are shown at a lookback time of approximately zero.

Discussion

We have detected widespread line emission throughout MRG-M0150. Given the upper limit on the residual star formation from Hα\alpha and the observed emission line ratios, we concluded that the ionized gas emission is not powered by star formation but instead requires an AGN contribution. Moreover, the bi-conical, centrally-driven outflow morphology and its inclination with respect to the stellar disk (Figure 5) further argues against a stellar origin of the observed winds, which would produce a much different wind geometry (see Veilleux et al. [84], for examples). Finally, the velocities we measure are substantially higher than what observed in the most powerful star-formation driven outflows at z∼2z \sim2 (Davies et al. [31]). Thus, we conclude an AGN in MRG-M0150 is driving the observed outflow, for which we measure a mass outflow rate of 0.6 ± 0.2 M⊙_\odot yr−1^{-1}.

Recently launched ionized gas outflows from a LL-AGN

Thanks to the high spatial resolution achieved by NIRSpec, the lensing magnification in MRG-M0150, and our ionized wind kinematics fitting, we can use this system to directly measure the timescales of an AGN-driven outflow in a quiescent galaxy at z>2z > 2. In Figure 8, we plot the radial wind velocity along both the blue-shifted and red-shifted sides of the bi-conical outflow as fitted by MOKA3D using the [OIII] emission line and the Hα\alpha line. We find ionized gas velocities vout∼1200v_{\rm out} \sim1200 km s−1^{-1}, similar to those of prototypical Seyfert galaxies (Fiore et al. [42]) and much higher than the vout∼300v_{\rm out} \sim300–400 km s−1^{-1} usually observed in the local “red geyser” galaxies, but consistent with ionized-phase outflow velocities found in similar massive, quenched galaxies at cosmic noon, such as Belli et al. [7] and D’Eugenio et al. [34]. These results suggest a difference in the nature of feedback mechanisms and impact of ionized gas outflows in massive quiescent systems between z>2z > 2 and z=0z = 0.

Radial velocity profiles of the ionized gas outflows in MRG-M0150

Figure 8. Radial velocity profile of the ionized gas outflows in MRG-M0150, according to the 3D kinematics fit. The left panel shows the profile for the approaching (blue) and receding (red) sides of the bi-conical outflow derived from [OIII]; the right one shows the values derived from Hα\alpha.

Using the radial velocity profile of the wind in MRG-M0150, we can constrain the launching timescales of the observed wind. Considering the full extent of the wind, we infer that the ionized outflow in MRG-M0150 has been launched over, at least, the past 2.6 Myr. Narrowing our focus on the innermost shell of the bicone and using the lowest measured velocity from the [OIII] line fit, we infer that the innermost resolved portion of the wind must have launched no more than ∼0.45\sim0.45 Myr ago from the nuclear region: this shows that the central engine has been recently launching material. Moreover, the SFH inferred by the SED fitting (Figure 7) further rules out a fossil origin for the observed outflow due to a past starburst, given that the travel times that we measure for the bi-conical winds are much smaller than the time elapsed since the last burst in star formation (∼500\sim500 Myr). The biconical and centrally-driven outflow geometry in MRG-M0150 also differs from that predicted for fossil AGN outflows, which hydrodynamical simulations show to become detached from the nucleus, lopsided and slow once the driving AGN has switched off ([94]).

Following Davies et al. [30] and Taylor et al. [79], in Figure 9 we compare the wind energetics to the AGN bolometric luminosity for MRG-M0150. We also show the ionized outflows reported in the literature for Cosmic-Noon AGN – as compiled by Venturi et al. [86], converted to ne∼600n_e \sim600 cm−3^{-3} (green triangles) – as well as those observed in two massive quiescent galaxies at z>2z > 2 by Belli et al. [7] and D’Eugenio et al. [34], and in the prototype red geyser “Akira” ([23]). Energy-conserving AGN-driven outflows carry kinetic energy rates to up to 5% of the AGN bolometric luminosity ([56]), while their momentum rate p˙out=Lbol/c\dot{p}_{\mathrm{out}} = L_{\mathrm{bol}}/c can be boosted by a factor of 5-20 due to interactions with the ISM ([41]): we mark these theoretical limits with black dashed lines in the plots. As expected, we find substantially different energy and momentum rates when using Hα rather than [OIII] (see Sect. 4.2), however in both cases we find that the system is broadly consistent with an extrapolation of the Venturi et al. [86] outflows to lower LbolL_{\mathrm{bol}}, and within the theoretical limits expected given the power of the central engine. MRG-M0150’s outflow energy and momentum rates (from Hα) are consistent with those found in the other two quiescent galaxies at high-zz, even though they host AGN ∼2\sim2 dex more luminous. Since outflow properties for both of those objects have been estimated using [OIII] emission, a substantial fraction of mass outflow in the ionized phase may be missed by relying only on higher-ionization lines. Thus, especially in the case of LL-AGN hosted in quiescent galaxies, mass outflow rates and related properties estimated from [OIII] should be treated as lower limits on the true values. We also note that the X-ray-based estimate of LbolL_{\mathrm{bol}} is about an order of magnitude higher than the [OIII]-based one (see Sect. 3.4), so that the ratio between the outflow energetics and the AGN luminosity decreases by about one dex when the former is adopted. We conclude that the current LL-AGN in MRG-M0150 is powerful enough to sustain the observed winds, and that a previous, more luminous AGN episode is not required by the energetics.

Energy and momentum rate of the ionized winds in MRG-M0150 compared to other ionized-phase outflows

Figure 9. Energy (left) and momentum rate (right) of the ionized winds in MRG-M0150 estimated from the Hα line (star-shaped points) and from the [OIII] line (pentagons) compared to other ionized-phase outflows in the literature, as a function of the AGN bolometric luminosity. Yellow scatter points have been plotted using LbolL_{\mathrm{bol}} computed from [OIII] line luminosity, while orange points use LbolL_{\mathrm{bol}} computed from the X-rays. Ionized-phase outflows observed in AGN at 1.5<z<3.51.5 < z < 3.5 are marked with empty green triangles ([86]). The blue and magenta filled scatter points are taken from Belli et al. [7] and D’Eugenio et al. [34], respectively; the purple X represents the prototypical red geyser galaxy “Akira” at z∼0z \sim0 ([23]).

Incidence of ionized gas outflows

The incidence of AGN-driven ionized gas outflows in high-zz massive quiescent galaxies appears to be relatively high. In particular, Bugiani et al. [13] report an incidence of ∼20%\sim20\%, and comparable values are found by studies targeting outflows in the neutral phase ([73]; [30]). This is higher than what is found in the local Universe – in which the red geyser galaxies constitute around 10% of the entire quiescent population – indicating that ionized gas outflows in high-zz quiescent galaxies are not only more powerful but also more frequent than in their local counterparts. Interestingly, our detection of the ionized outflows in MRG-M0150 is highly dependent on the availability of both lensed and spatially resolved data. To illustrate this, we simulate the unlensed and unresolved NIRSpec/MSA spectrum of the target with a slit loss of ∼50%\sim50\% by extracting the integrated spectrum within 1 Re1\,R_e of the galaxy, corrected for magnification: when fitting the emission lines in this spectrum, the broad component is no longer statistically preferred over a single-component model (see the BIC values for both fits in Figure 10), implying that such outflows may remain undetected in unresolved observations of high-zz galaxies.

Spectra showing the difficult detection of outflows in unresolved spectra

Figure 10. The difficult detection of outflows in unresolved spectra: [O III] and [N II]/Hα\alpha emission lines observed (blue) and fitted (purple) in the simulated unresolved spectrum of MRG-M0150. Top: fit to the emission lines using single Gaussian components; bottom: fit using broad (red) + narrow (green) components.

The same applies to the X-ray emission of the central LL-AGN, which is only marginally detected in each of the multiple images (Sect. 3.4). This suggests that the relatively high incidence of outflows previously reported is a lower limit, supporting the idea that winds driven by a LL-AGN are a common phase in the evolution of massive quiescent galaxies at cosmic noon, while they become rarer in the present-day Universe.

The fossil outflow scenario

The relationship between the observed multiphase outflows in massive galaxies at z>2z > 2 and the low levels of AGN activity observed in the same sample remains unclear, as these AGN are rarely detected in X-rays and are primarily classified as active through rest-frame optical diagnostic diagrams ([13, 74]). Studies of neutral-phase outflows in massive galaxies at high redshift have raised the idea that these constitute mostly relics of a past starburst or fossils of past quasar events – that is, short-lived bursts of black-hole activity that occurred at earlier epochs than the time of observation ([79, 1, 77]). The remnant nature of outflows observed in high-zz quenched galaxies hosting inactive or low-luminosity AGN is supported by hydrodynamical simulations (e.g. Zubovas et al. [93]) and energetics arguments ([1, 79]), which for many of the observed galaxies struggle to reconcile the lack of powerful AGN signatures with the energetics of the observed neutral-phase winds (while ionized-phase winds generally do not pose a problem, as shown in Figure 9). In this scenario, the low incidence of X-ray AGN among recently quenched galaxies at cosmic noon (∼6%\sim6\%; [1]) is reconciled with the high incidence of fast outflows if the luminous AGN phase is short-lived, while the outflow it launches remains visible for up to ∼10\sim10 Myr after the AGN has faded ([1]). For example, Taylor et al. [79] assume that the detection rate of neutral outflows/inflows in quenched galaxies from the EXCELS survey is set by the duty cycle timescales of the AGN hosted at their center, and conclude that the visibility window of both an X-ray active AGN and the outflow launched by the same event is extremely short, of the order of 1 Myr.

It is useful to distinguish between two versions of the fossil outflow scenario. In its weaker form, it postulates that the outflows observed today were often launched by an AGN in a substantially different state from the one we observe, so that the present-day LbolL_{\mathrm{bol}} is poorly correlated with the wind energetics, even if a lower-luminosity AGN is still active. A single galaxy whose outflow does not exceed the energy budget of its current AGN, such as MRG-M0150, cannot test this version. In its stronger form, the AGN switches between “on” and “off” states, or at least undergoes drastic changes in LbolL_{\mathrm{bol}}, and only the “on” phases are able to drive outflows. MRG-M0150 can be used as a test case for this second version.

Our kinematic modelling shows that the innermost portion of the wind was launched no more than ∼0.45\sim0.45 Myr ago, well within the 1 Myr time window in which an outflow and the X-ray bright AGN that launched it should be observable at the same time ([79]). We do detect the AGN, but it is a LL-AGN with λEdd≈10−3\lambda_{\mathrm{Edd}} \approx10^{-3} (Sect. 3.4) almost an order of magnitude below the Eddington ratio adopted by Taylor et al. [79] as the threshold for X-ray detection and for AGN activity capable of driving outflows (λ≳0.027\lambda\gtrsim0.027). This suggests that AGN in a lower accretion state can also launch and sustain galaxy-scale winds. Such systems are, however, difficult to identify: neither the AGN nor the ionized outflow in MRG-M0150 would have been detected

without lensing and spatially resolved spectroscopy (Sect. 6.2), and in unlensed, unresolved samples, galaxies like MRG-M0150 would likely be counted as hosting neither. Accounting for this population may therefore lengthen the inferred overlap between the AGN duty cycle and outflow phases.

The energetics argument in favor of fossil outflows is also sensitive to how LbolL_{\mathrm{bol}} is estimated. In most quiescent galaxies at these redshifts, LbolL_{\mathrm{bol}} can only be derived from the [O III] line, an estimate subject to significant systematic uncertainties related to dust obscuration and to the physical state of the AGN (see the discussion in Taylor et al. [79]). In MRG-M0150, the X-ray-based LbolL_{\mathrm{bol}} is an order of magnitude higher than our [O III]-based estimate (Sect. 3.4). If similar offsets apply to other LL-AGN in quiescent galaxies, their current AGN may account for a larger fraction of the observed outflow energetics than [O III]-based estimates would suggest.

These considerations suggest an interpretation in which the AGN luminosity in quenched galaxies follows a continuous distribution that declines after quenching, rather than alternating between two distinct states. The AGN may flicker on short timescales, but its overall luminosity fades as the galaxy ages, as also illustrated by the scenario proposed by Almaini et al. [1], in which AGN activity broadly traces the decline of the SFR after the starburst. Consistently, Valentino et al. [82] find that the most luminous AGN in quenched galaxies at z∼1.5z \sim1.5–4.54.5 are preferentially found in the youngest post-starburst systems. Observed shortly after quenching, galaxies such as COSMOS-11142 ([7]) host massive outflows and luminous AGN. MRG-M0150 would represent a later stage, in which both the AGN and the outflow have faded below the sensitivity of typical unlensed observations. Finally, older quiescent galaxies show no outflows and, in some cases, evidence of gas accretion, as in the lensed galaxy MRG-M0138 ([45]). Within this framework, MRG-M0150 shows that AGN just below the detection limits of current samples can still drive galaxy-scale outflows. Such weak activity may be common: stacking the X-ray emission of individually undetected post-starburst galaxies at cosmic noon reveals, on average, low-level AGN accretion (L/LEdd≲10−3L/L_{\mathrm{Edd}} \lesssim10^{-3}; [1]).

Quenching by AGN-driven outflows

A major open question regarding observations of gas outflows in high-redshift quiescent galaxies is whether the presence of these winds can be directly linked to the quenching of the galaxy itself. While it is not trivial to link the rapid quenching of MRG-M0150 in the last 0.5 Gyr with the observed ionized-phase outflows, the mass outflow rate of 0.6±0.20.6 \pm0.2 M⊙_\odot yr−1^{-1} is higher than the current levels of SF, for which we have an upper limit of 0.300.30 M⊙_\odot yr−1^{-1} from Hα\alpha, consistent with the SFR estimated by Prospector in the most recent time bin of [0–30] Myr (SFR = 0.2−0.8+0.20.2^{+0.2}_{-0.8} M⊙_\odot yr−1^{-1}, see Figure 7). Furthermore, the detection of outflowing neutral gas through the Na D absorption line in the central parts of the galaxy (Gangula et al., in prep.) implies a multi-phase nature of the observed winds. It has been observed that neutral-phase gas can reach mass outflow rates more than one order of magnitude higher than in the ionized phase, both at high redshift (Belli et al. [7]) and in the local Universe (Roberts-Borsani [71]; Avery et al. [2]). Thus, the ionized-phase mass outflow rate that we measure is a strictly lower limit on the true rate of the multiphase wind in MRG-M0150.

The high velocities reached by the ionized outflow means that some of the gas may be capable of escaping the gravitational potential of the galaxy. The halo escape velocity is expected to be 2.5–3 times the rotational velocity of the galaxy (e.g. Weiner et al. [89]; Swinbank et al. [78]). Using the rotational velocity of MRG-M0150 inferred from stellar kinematic modelling of vrot≈450v_{\mathrm{rot}} \approx450 km s−1^{-1} (Newman et al., in prep.), we estimate an escape velocity between ∼1100−1300\sim1100 - 1300 km s−1^{-1}. The highest velocities measured in the ionized outflow are comparable to these estimates, suggesting that a substantial amount of ionized outflowing gas may be able to escape the halo. This type of “ejective” feedback from AGN-driven outflows has been invoked also in the post-starburst high-zz galaxies of Belli et al. [7] and D’Eugenio et al. [34], in which the ejection of cold gas at a faster rate than it can be converted into stars has been invoked as the main mechanism through which the galaxy has been quenched. While the fast-moving ionized phase does not transport as much mass as the cold phase, what we observe in MRG-M0150 and similar high-zz systems is qualitatively different than the “maintenance mode” feedback of local red geysers, in which the ionized-phase gas is unable to escape the galaxy and instead keeps the galaxy quenched through the heating of gas in the halo (Cheung et al. [23]).

Although the short launching timescales associated with the current wind prevents us from directly attributing the main quenching event to the observed outflow, the ongoing wind provides evidence that AGN feedback remains active well after star formation has been quenched. Outflows such as the one observed here can be intermittent, and it is possible that similar, possibly more powerful episodes have occurred repeatedly since the last peak of star formation. The cumulative effect of recurrent AGN-driven winds, rather than a single event, may therefore have been ultimately responsible for quenching MRG-M0150. The mass outflow rate we measure exceeds the residual SFR, suggesting that winds driven by a low-luminosity AGN can substantially affect the remaining gas reservoir and suppress subsequent star formation well after the initial “blowout” phase. Given that most of an active galaxy’s lifetime is spent in a low-accretion state (Ho [49]), this type of long-term feedback may represent the key to maintenance of quiescence in massive systems over much longer timescales, potentially linking the population of recently quenched galaxies at cosmic noon to the massive quiescent population observed in the local Universe.

Summary and conclusions

We have used JWST/NIRSpec IFU observations of the gravitationally lensed massive quiescent galaxy MRG-M0150 at z=2.64z = 2.64, from the INQUE-JWST sample. Owing to the spatial resolution and sensitivity of NIRSpec’s IFU and the magnification offered by the strong lensing, we are able for the first time to reconstruct in 3D the bi-conical ionized outflow of a massive quiescent galaxy at z>2z > 2, making it possible investigate the impact of AGN-driven outflows on star-formation quenching at high redshift. The main findings of this study are the following:

  • MRG-M0150 hosts a previously undetected low-luminosity AGN, identified through LINER-like line ratios and marginal X-ray detections, with Lbol≈3×1043−3×1044L_{\mathrm{bol}} \approx3 \times10^{43} - 3 \times10^{44} erg s−1^{-1} and L/LEdd∼10−3L/L_{\mathrm{Edd}} \sim10^{-3}.

  • The ionized gas traces a galaxy-wide, centrally driven bi-conical outflow reaching ∼1200\sim1200 km s−1^{-1}, whose mass outflow rate is substantially underestimated when relying only on high-ionization lines such as [OIII] as opposed to low-ionization lines such as Hα\alpha.

  • The outflow has been recently launched by the central LL-AGN: its innermost portion was launched no more than 0.45 Myr ago, and the current AGN is powerful enough to sustain it without requiring a previous, more luminous episode.

  • Neither the AGN nor the outflow would be detected without lensing and spatially resolved spectroscopy, suggesting that similar systems are missed in unresolved samples, affecting incidence-based estimates of the visibility timescales of AGN and outflows.

  • The ionized mass outflow rate (0.6±0.20.6 \pm0.2 M⊙_\odot yr−1^{-1}), a lower limit on the multiphase wind, exceeds the residual SFR (<0.30< 0.30 M⊙_\odot yr−1^{-1}), and part of the gas may escape the halo, implying ejective feedback is still active in the galaxy after the main quenching event. Combined with the relatively high incidence of such outflows, our results suggest that low-luminosity AGN activity may provide an important long-term mechanism for maintaining quiescence in massive galaxies at cosmic noon, potentially over cosmological timescales.

Overall, our work shows that AGN-driven galaxy-wide outflows can be sustained by low-luminosity AGN in massive quiescent galaxies at cosmic noon. These outflows may play a key role in the maintenance of quiescence over long timescales, and are sufficiently common to likely represent a major mechanism in preventing new star-formation in massive galaxies at this epoch. At the same time, our results highlight the importance of spatially resolved spectroscopy, combined with the magnification provided by gravitational lensing, for uncovering and constraining such outflows in the high-redshift Universe.

Acknowledgements. We acknowledge support from the ERC grant 101076080 “Red Cardinal”. This work is based on observations made with the NASA/ESA/CSA James Webb Space Telescope. The data were obtained from the Mikulski Archive for Space Telescopes at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-03127 for JWST. These observations are associated with program GO-4903. Support for US investigators in program GO-4903 was provided by NASA through a grant from the Space Telescope Science Institute.

Lens Model

The modeling was performed using imaging data of the foreground lensing cluster MACSJ0150.3-1005 at z=0.365z = 0.365 in the NIRCam F200W bands obtained as part of the same Cycle 3 JWST program that provided the spectroscopic data discussed in the text. Additionally, we identified the cluster members and measured their redshifts making use of the IFU spectroscopic data taken with the Multi Unit Spectroscopic Explorer (MUSE) at the Very Large Telescope (VLT), which covers a field of view of 1′×1′1' \times1'. MACSJ0150.3-1005 was observed in September 2019 for one OB of 48 min (program 0103.A-0777, PI: A. Edge). We reduced the MUSE data and measured the redshifts following the procedure described in Bolamperti et al. [10].

We construct the strong lensing model using the software Gravity.jl (Lombardi [59]), which allows for the reconstruction of the total mass distribution of the cluster using a multi-plane approach. The F200W image of the cluster and multiple images of MRG-M0150, together with the final lens model, can be seen in Figure A.1. The following mass components were included into the model. The main cluster-scale DM halo of the cluster, which is parametrized with a nearly isothermal ellipsoid (NIE) mass distribution characterized by six free parameters: the on-sky coordinates (two parameters), ellipticity, position angle, velocity dispersion, and core radius. An elliptical coreless Jaffe profile is used to describe one of the cluster galaxies lying close to Image 2 of MRG-M0150 along the line of sight. Contrary to the NIE profile, its core radius is substituted with a truncation radius, and the galaxy on-sky coordinates are fixed, for a total of 4 free parameters. A second elliptical coreless Jaffe profile with 4 free parameters is used to parameterize a galaxy located in the background at z=0.4907z = 0.4907 (as determined from the NIRSpec IFU data) that is close in projection to Image 1 of MRG-M0150. This galaxy is included in the model employing the multi-plane approach possible within Gravity.jl. Finally, 18 additional spherical Jaffe profiles are used to describe 18 spectroscopically confirmed cluster galaxies identified from the MUSE data, including the BCG. For these galaxies, the central velocity dispersion σ\sigma and truncation radius rcutr_{\mathrm{cut}} are scaled according to the galaxy luminosity in the HST/F160W band assuming the following scaling relations:

σgal=σref(LgalLref)α,(3)\sigma^{\mathrm{gal}} = \sigma^{\mathrm{ref}}\left(\frac{L_{\mathrm{gal}}}{L_{\mathrm{ref}}}\right)^\alpha, \tag*{(3)}
rcutgal=rcutref(LgalLref)β;(4)r_{\mathrm{cut}}^{\mathrm{gal}} = r_{\mathrm{cut}}^{\mathrm{ref}}\left(\frac{L_{\mathrm{gal}}}{L_{\mathrm{ref}}}\right)^\beta; \tag*{(4)}

where LgalL_{\mathrm{gal}} is the F200W luminosity of the cluster member. In the model we assume α=0.25\alpha= 0.25 and β=0.50\beta= 0.50, which is equivalent to assuming a constant mass-to-light ratio across the cluster population (see [8, 68]), and thus our only free parameters for the cluster members’ modeling are σref\sigma^{\mathrm{ref}} and rcutrefr_{\mathrm{cut}}^{\mathrm{ref}}.

To constrain the lens model with the data we adopt a parametric extended modeling approach. In addition to modeling the lens, we also model the brightness distribution of the multiply-imaged source (that is, MRG-M0150) using a Sérsic profile: this adds as free parameters the coordinates, the effective radius, total flux, the ellipticity and the position angle of the source.

In total the model contains 21 free parameters and is constrained by 22781 pixels. The model is optimized to reproduce the observed multiple images of the source; our goodness-of-fit estimation is given by the residuals between the model-predicted and observed shape of the multiple images. To sample the full parameter space of the models employed in this work, we carry out the Gravity.jl fits using a Markov Chain Monte Carlo (MCMC) algorithm to obtain posterior distributions.

Kinematic modeling with MOKA3D

We perform 3D kinematics fitting of the ionized gas emission in MRG-M0150 using the de-lensed maps of ionized gas surface brightness in the source plane, along with the line profiles in each spaxel, in order to to reconstruct the geometry and velocity structure of the outflow. The MOKA3D code requires the observations’ grid to be regularly sampled and to convolve the model with the PSF of the instrument. Since we have to work on the source plane, we need to deal with the distortion of the NIRSpec/IFU spaxels and PSF caused by the application of the lens model, and to make them suitable to be fed to MOKA3D. We re-sample the source-plane cubes onto a regular grid of pixels with area much smaller than the minimum size of the original pixels in the following way: we first construct the regular grid, and then for each of the original pixels we assign the surface signal and noise values to the closest pixel in the new grid, and put NaN values elsewhere. MOKA3D performs a convolution of the model with the PSF of the instrument; in this way we can ignore any correlated noise deriving from the resampling as long as the PSF convolution is applied, since the dimension of the PSF is much greater than the pixels in the new grid. In order to obtain the PSF on the source plane, we apply the same lens model to the instrumental PSF of NIRSpec/IFU. The 2D PSF of NIRSpec/IFU is first positioned to overlap the centroid of the galaxy in the image plane and then de-lensed onto the source plane.

We verify that the magnification does not vary greatly across the analyzed spaxels (as visible in Fig.B.1), and thus we should be able to apply the “centrally de-lensed” PSF across the entire cube. However, spaxels belonging to the bin that is closer to the critical line are experiencing higher values of magnification than the rest of the galaxy (μ∼12\mu\sim12 instead of μ∼4\mu\sim4): applying the same “centrally de-lensed” PSF to these spaxels may not be accurate, and may affect the resulting 3D reconstruction of the outflow by the fitting code. To verify that this is not the case, we show in Fig. B.2 the shape of NIRSpec/IFU’s PSF overlapped onto the galaxy’s centroid in the image plane (upper left) and how that gets distorted when projecting it onto the source plane (upper right). In the central row of panels we verify that changing the PSF position away from the galaxy centroid in the image plane (left) results in a source-plane PSF that does not differ significantly from the centrally-de-lensed one. Fitting both source-plane PSFs with an anisotropic 2D Gaussian model, we find only a small variation of the FWHM of the dislocated PSF’s of 8% along the minor axis and 0.2% along the major axis with respect to the centrally-located one. However, when positioning the PSF onto spaxels nearing the critical line in the image plane (last row), the magnification jumps to 3x the median value elsewhere and the PSF gets highly distorted along a preferential axis, with a variation in FWHM of 60% along the minor axis and 7% along the major one with respect to the center. Nonetheless, spaxels affected by the higher magnification form a small minority of all the spaxels analyzed, on top of being lower in signal-to-noise with respect to the central parts of the galaxy (see Figure 1). For this reason, the spatial bins are large near the critical line, and convolution by the PSF has only a very small effect. Thus, we decide to apply the centrally-projected de-lensed PSF to the entire resampled and de-lensed cubes to be fitted with MOKA3D, neglecting any change in magnification levels. The fact that we get consistent results from the MOKA3D fitting of the Hα\alpha emission – detected in all the analyzed bins – and the [OIII] line – which is not detected in the external parts of the galaxy, thus the fit does not include any of the greatly magnified spaxels – is evidence that our central PSF approximation is not significantly affecting the 3D kinematics fitting results.

Lens model of the MACS J0150 cluster

Figure A.1. Lens model of the MACS J0150 cluster. We show the JWST/NIRCam F115W/F200W/F277W images with an asinh stretch. The three images of MRG-M0150 are labeled; red curves trace the critical line.

Magnification map near Image 1

Figure B.1. Magnification map near Image 1, with contours of the analyzed Voronoi bins of MRG-M0150 overlayed in white.

The results of the fit using the [OIII] emission line is visible in Fig. B.3. We also show, in Fig. B.4, the results of the [OIII] kinematics fit using the model with forced inclination of 30∘30^\circ for the bi-cone discussed in Sect. 4: clearly, residuals are much higher than what is found in the inclination-free case and the model does not perform as well in reproducing the data.

NIRSpec/IFU PSF positioned in three different points of the galaxy disk and projected onto the source plane

Figure B.2. NIRSpec/IFU PSF positioned in three different points of the galaxy disk (left column) and projected onto the source plane using the lens model (right column). The red contours mark the external perimeter of the analyzed Voronoi bins, both in the image plane and in the source plane.

Figure B.3. Comparison between the observed and model gas velocity (moment 1) and dispersion (moment 2) maps from the [OIII] line profiles.

Figure B.4. Comparison between the observed and model gas velocity (moment 1) and dispersion (moment 2) maps from the [OIII] line profiles, assuming a fixed inclination of 30∘30^\circ.

References

  1. [1]Almaini, O., Wild, V., Maltby, D., et al. 2025, MNRAS, 539, 3568
  2. [2]Avery, C. R., Wuyts, S., Förster Schreiber, N. M., et al. 2022, MNRAS, 511, 4223
  3. [3]Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, \pasp, 93, 5
  4. [4]Belfiore, F., Maiolino, R., Maraston, C., et al. 2016, MNRAS, 461, 3111
  5. [5]Belli, S., Genzel, R., Förster Schreiber, N. M., et al. 2017, ApJ, 841, L6
  6. [6]Belli, S., Newman, A. B., & Ellis, R. S. 2019, ApJ, 874, 17
  7. [7]Belli, S., Park, M., Davies, R. L., et al. 2024, Nature, 630, 54
  8. [8]Bergamini, P., Rosati, P., Mercurio, A., et al. 2019, A&A, 631, A130
  9. [10]Bolamperti, A., Grillo, C., Caminha, G. B., et al. 2024, A&A, 692, A239
  10. [11]Bower, R. G., Benson, A. J., Malbon, R., et al. 2006, MNRAS, 370, 645
  11. [12]Brusa, M., Bongiorno, A., Cresci, G., et al. 2015, MNRAS, 446, 2394
  12. [13]Bugiani, L., Belli, S., Park, M., et al. 2025, ApJ, 981, 25
  13. [15]Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682
  14. [16]Cano-Díaz, M., Maiolino, R., Marconi, A., et al. 2012, A&A, 537, L8
  15. [17]Cappellari, M. & Copin, Y. 2003, MNRAS, 342, 345
  16. [18]Cappellari, M. & Copin, Y. 2012, VorBin: Voronoi binning method, Astrophysics Source Code Library, record ascl:1211.006
  17. [19]Carnall, A. C., McLeod, D. J., McLure, R. J., et al. 2023a, Monthly Notices of the Royal Astronomical Society, 520, 3974, aDS Bibcode: 2023MNRAS.520.3974C
  18. [20]Carnall, A. C., McLeod, D. J., Dunlop, J. S., et al. 2023b, Nature, 619, 716, aDS Bibcode: 2023Natur.619..716C
  19. [21]Carniani, S., Marconi, A., Maiolino, R., et al. 2015, A&A, 580, A102
  20. [22]Chabrier, G. 2003, PASP, 115, 763
  21. [23]Cheung, E., Bundy, K., Cappellari, M., et al. 2016, Nature, 533, 504, aDS Bibcode: 2016Natur.533..504C
  22. [24]Cid Fernandes, R., Stasińska, G., Schlickmann, M. S., et al. 2010, MNRAS, 403, 1036
  23. [25]Cid Fernandes, R., Stasińska, G., Mateus, A., & Vale Asari, N. 2011, \mnras, 413, 1687, _eprint: 1012.4426
  24. [28]Davé, R., Anglés-Alcázar, D., Narayanan, D., et al. 2019, MNRAS, 486, 2827
  25. [30]Davies, R. L., Belli, S., Park, M., et al. 2024, MNRAS, 528, 4976
  26. [31]Davies, R. L., Förster Schreiber, N. M., Übler, H., et al. 2019, ApJ, 873, 122
  27. [33]de Graaff, A., Setton, D. J., Brammer, G., et al. 2025, Nature Astronomy, 9, 280, aDS Bibcode: 2025NatAs...9..280D
  28. [34]D’Eugenio, F., Pérez-González, P. G., Maiolino, R., et al. 2024, Nature Astronomy, 8, 1443, aDS Bibcode: 2024NatAs...8.1443D
  29. [35]Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604
  30. [36]Donevski, D., Nanni, A., Whitaker, K. E., et al. 2026, arXiv e-prints, arXiv:2608.27571arxiv.org/abs/2608.27571
  31. [37]Dopita, M. A., Ho, I.-T., Dressel, L. L., et al. 2015, ApJ, 801, 42
  32. [38]Dotter, A. 2016, ApJS, 222, 8
  33. [39]Fabian, A. C. 2012, ARA&A, 50, 455
  34. [40]Falcón-Barroso, J., Sánchez-Blázquez, P., Vazdekis, A., et al. 2011, A&A, 532, A95
  35. [41]Faucher-Giguère, C.-A. & Quataert, E. 2012, MNRAS, 425, 605
  36. [42]Fiore, F., Feruglio, C., Shankar, F., et al. 2017, A&A, 601, A143
  37. [43]Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 306
  38. [44]Förster Schreiber, N. M., Übler, H., Davies, R. L., et al. 2019, The Astrophysical Journal, 875, 21, aDS Bibcode: 2019ApJ...875...21F
  39. [45]Gangula, S., Newman, A. B., Gu, M., et al. 2026, Resolved Maps of Gas and Dust in a Massive Quiescent Galaxy at z=2 from INQUEST-JWST: Evidence of Accretion and Rejuvenation, aDS Bibcode: 2026arXiv260421695GDOI
  40. [46]Heckman, T. M., Kauffmann, G., Brinchmann, J., et al. 2004, ApJ, 613, 109
  41. [47]Hewitt, G., Almaini, O., Maltby, D., et al. 2026, MNRAS, 550, stag1180
  42. [48]Hinshaw, G., Larson, D., Komatsu, E., et al. 2013, ApJS, 208, 19
  43. [49]Ho, L. C. 2008, Annual Review of Astronomy and Astrophysics, 46, 475
  44. [50]Ho, L. C. 2009, ApJ, 699, 626
  45. [51]Hopkins, P. F., Hernquist, L., Cox, T. J., et al. 2006, ApJS, 163, 1
  46. [52]Johnson, B. D., Leja, J., Conroy, C., & Speagle, J. S. 2021, ApJS, 254, 22
  47. [53]Kakkad, D., Mainieri, V., Padovani, P., et al. 2016, Astronomy and Astrophysics, 592, A148, aDS Bibcode: 2016A&A...592A.148K
  48. [54]Kennicutt, Jr., R. C. 1998, ARA&A, 36, 189
  49. [55]Kim, C.-G., Ostriker, E. C., Somerville, R. S., et al. 2020, ApJ, 900, 61
  50. [56]King, A. & Pounds, K. 2015, ARA&A, 53, 115
  51. [57]Kriek, M., Beverage, A. G., Price, S. H., et al. 2024, ApJ, 966, 36
  52. [58]Leja, J., Carnall, A. C., Johnson, B. D., Conroy, C., & Speagle, J. S. 2019, ApJ, 876, 3
  53. [59]Lombardi, M. 2024, A&A, 690, A346
  54. [60]Marconcini, C., Marconi, A., Cresci, G., et al. 2023, Astronomy and Astrophysics, 677, A58, aDS Bibcode: 2023A&A...677A..58M
  55. [61]McConnell, N. J., Ma, C.-P., Gebhardt, K., et al. 2011, Nature, 480, 215
  56. [63]Nanayakkara, T., Glazebrook, K., Jacobs, C., et al. 2024, Scientific Reports, 14, 3724, aDS Bibcode: 2024NatSR..14.3724N
  57. [64]Newman, A. B., Belli, S., & Ellis, R. S. 2015, ApJ, 813, L7
  58. [66]Newman, A. B., Belli, S., Ellis, R. S., & Patel, S. G. 2018b, ApJ, 862, 126
  59. [67]Newman, A. B., Gu, M., Belli, S., et al. 2026, Science, 392, 1065
  60. [68]Oguri, M. 2010, PASJ, 62, 1017
  61. [69]Osterbrock, D. E. & Ferland, G. J. 2006, Astrophysics of gaseous nebulae and active galactic nuclei
  62. [70]Park, M., Belli, S., Conroy, C., et al. 2024, ApJ, 976, 72
  63. [71]Roberts-Borsani, G. W. 2020, MNRAS, 494, 4266
  64. [73]Sapori, M., Belli, S., Bugiani, L., et al. 2026, arXiv e-prints, arXiv:2607.20606arxiv.org/abs/2607.20606
  65. [74]Skarbinski, M., Rowlands, K., Alatalo, K., et al. 2026, ApJ, 1000, 191
  66. [75]Somerville, R. S., Hopkins, P. F., Cox, T. J., Robertson, B. E., & Hernquist, L. 2008, MNRAS, 391, 481
  67. [76]Steidel, C. C., Erb, D. K., Shapley, A. E., et al. 2010, ApJ, 717, 289 Straatman, C. M. S., Labbé, I., Spitler, L. R., et al. 2014, ApJ, 783, L14
  68. [77]Sun, Y., Ji, Z., Rieke, G. H., et al. 2026, The Astrophysical Journal, 997, 140
  69. [78]Swinbank, A. M., Harrison, C. M., Tiley, A. L., et al. 2019, MNRAS, 487, 381
  70. [79]Taylor, E., Carnall, A. C., Maltby, D., et al. 2026, MNRAS, 549, stag918
  71. [81]Valentino, F., Heintz, K. E., Brammer, G., et al. 2025, A&A, 699, A358
  72. [82]Valentino, F., Ito, K., Farcy, M., et al. 2026, arXiv e-prints, arXiv:2605.30424arxiv.org/abs/2605.30424
  73. [83]Vazdekis, A., Koleva, M., Ricciardelli, E., Röck, B., & Falcón-Barroso, J. 2016, MNRAS, 463, 3409
  74. [84]Veilleux, S., Cecil, G., & Bland-Hawthorn, J. 2005, ARA&A, 43, 769
  75. [85]Veilleux, S., Maiolino, R., Bolatto, A. D., & Aalto, S. 2020, A&A Rev., 28, 2
  76. [86]Venturi, G., Carniani, S., Bertola, E., et al. 2025, arXiv e-prints, arXiv:2512.09996DOI
  77. [88]Weibel, A., de Graaff, A., Setton, D. J., et al. 2025, The Astrophysical Journal, 983, 11, aDS Bibcode: 2025ApJ...983...11W
  78. [89]Weiner, B. J., Coil, A. L., Prochaska, J. X., et al. 2009, ApJ, 692, 187
  79. [90]Whitaker, K. E. & Bezanson, R. 2026, arXiv e-prints, arXiv:2606.12156DOI
  80. [91]Whitaker, K. E., Williams, C. C., Mowla, L., et al. 2021, Nature, 597, 485, aDS Bibcode: 2021Natur.597..485W
  81. [92]Wu, P.-E. 2025, ApJ, 978, 131
  82. [93]Zubovas, K., Bialopetravicius, J., & Kazlauskaitė, M. 2022, MNRAS, 515, 1705
  83. [94]Zubovas, K. & Maskeliūnas, G. 2023, MNRAS, 524, 4819

Paper details

vlnc:W_up7hm8ma
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents