First Results from the JWST Early Release Science Program Q3D: The Warm Ionized Gas Outflow in z ∼ 1.6 Quasar XID 2028 and Its Impact on the Host Galaxy

XID 2028 was observed on 2022 November 20 by JWST using the NIRSpec Instrument in IFU mode (Böker et al. 2022; Jakobsen et al. 2022). These data are publicly available on the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute. The specific observations analyzed can be accessed via doi:10.17909/04tb-mn90. The NIRSpec field of view (FOV) in IFU mode is ∼3'' × 3'' or ∼25 × 25 kpc for this object. We used the filter/grating combination F100LP/G140H, with corresponding wavelength coverage of 0.97–1.89 μm or ∼0.37–0.73 μm at the redshift of XID 2028. The grating has a near-constant dispersion Δλ = (2.30 − 2.40) × 10⁻⁴ μm per pixel, corresponding to a velocity resolution ∼80−180 km s⁻¹. This allows us to easily spectrally resolve the profiles of the emission lines in XID 2028, which have typical velocity widths of several hundred km s⁻¹. As in the case of the J1652 observations (Wylezalek et al. 2022), we used a nine-point small cycling dither pattern with 25 groups and one integration per position to improve the spatial sampling, and help us more accurately measure and characterize the point-spread function (PSF). We also took one leakage exposure at the first dither position to account for light leaking through the closed microshutter array (MSA), as well as light from failed open shutters. Following the recommendation of the STScI staff, we used the NRSIRS2 readout mode, which improves the signal-to-nose ratio and reduces data volume compared to the NRSIR2RAPID mode. No pointing verification image was taken. The total integration time was 177 minutes on target and 20 minutes for the leakage exposure.

The NIRSpec data of XID 2028 were reduced following largely the same method used by Vayner et al. (2023, submitted) to reduce the NIRSpec data cube on J1652. We refer the interested reader to this paper for more details. Here, we describe the main steps with an emphasis on aspects that are specific to the NIRSpec data on XID 2028.

Data reduction was done with the JWST Calibration pipeline version 1.8.4 (Bushouse et al. 2022) using CRDS version 11.16.16 and context file jwst 1019.pmap. The first stage of the pipeline, Detector1Pipeline, performs standard IR detector reduction steps, such as dark current subtraction, fitting ramps of non-destructive group readouts, combining groups and integrations, data quality flagging, cosmic ray removal, bias subtraction, linearity, and persistence correction.

Afterward, we ran Spec2Pipeline, which assigns a world coordinate system to each frame, applies flat field correction, flux calibration, and extracts the 2D spectra into a 3D data cube using the cube build routine. Here, we adopt the emsm weighting method instead of the standard drizzle method to build the 3D data cube from the 2D data. ²⁴ The emsm weighting reduced the oscillating spectral pattern in the point-source spectrum compared to the drizzle method at the cost of minor degradation in the spatial and spectral resolution. Additional steps were taken to flag pixels affected by open MSA shutters. At this point, we skipped the imprint subtraction step due to increased spatial variation in the background across many spectral channels. Due to known issues with the outlier detection step in the Spec3Pipeline (JWST Help Desk, priv. communication), we opted to use the reproject package ²⁵ to combine the different dither positions into a single data cube using their drizzle algorithm. The dither positions were combined onto a common grid with a spatial pixel size of 005.

The astrometry of the cube obtained by the JWST Calibration pipeline features a minor offset with respect to archival Hubble Space Telescope (HST) data. We use the position of the quasar in the HST WFC3 F814W image, previously discussed in Brusa et al. (2010, 2015, 2018) and Scholtz et al. (2020), to align the astrometry of the JWST data with respect to the HST image. We find a small offset of ΔR.A. = −007 and Δdecl. = 003.

The data cube produced by the pipeline was put on an absolute flux scale by using the data on flux standard star TYC 4433-1800-1 (PID 1128, o009) reduced in exactly the same way as the data cube of XID 2028.

[O iii] 5007 Å is the strongest emission line in the NIRSpec data cube, as expected from the published ground-based data, and therefore the best tracer of the warm ionized gas in XID 2028; which is the focus of the present section. The q3dfit analysis is restricted to 3860–5400 Å in the quasar rest frame to allow us to fit Hβ, [O iii] 4959, 5007 simultaneously, as well as the underlying continuum emission from the quasar and galaxy host.

Figure 3 shows the quasar-subtracted narrowband images of the [O iii] 5007 line emission produced by extracting the [O iii] 5007 flux from the q3dfit-processed data cube across the velocity ranges of [−1200, − 300] km s⁻¹, [−300, + 300] km s⁻¹, and [+300, + 500] km s⁻¹ relative to the quasar rest frame (the systemic velocity of XID 2028, z = 1.5933, is derived from our kinematic analysis of the q3dfit-processed data cube, discussed in Section 4.2). Figure 3 shows that nearly all the [O iii] line-emitting gas outside of the unresolved quasar lies either in a bright western plume of gas with negative radial velocities down to −1000 km s⁻¹ (90-percentile) or in fainter blueshifted gas clouds that surround the bright plume and trace a wide-angle cone with an opening angle ϕ ∼90°. The line emission extends to the western edge of the field of view of NIRSpec, i.e., at least 17 kpc from the quasar.

Zoom In Zoom Out Reset image size

As we will discuss later in Section 4.2, the measured negative velocities of the gas are well in excess of those expected for rotational motion in the host galaxy. Therefore, the NIRSpec data nicely confirm the presence of a one-sided fast [O iii] outflow in XID 2028, which was first reported by Cresci et al. (2015) and Perna et al. (2015), and was more recently reported by Scholtz et al. (2020), based on independent analyses of ground-based Xshooter and SINFONI IFS data. The improved angular resolution, PSF characterization, and sensitivity of the JWST data reveal intricate bubble-like substructures in the bright plume that were not evident in the ground-based data and resolves outflowing gas clouds, which trace a cone with a wider opening angle than previously suspected. The absence of a redshifted counterpart to the outflow, now confirmed by the NIRSpec data down to a 3σ f_λ limit of 1.4 × 10⁻²⁰ erg s⁻¹ cm⁻² Å⁻¹ arcsec⁻², combined with the small extent (R ≈ 05 = 4 kpc) of the near-systemic (±300 km s⁻¹) line emission from the host ISM, puts strong constraints on the outflow geometry. We return to this issue in Section 5.

The kinematics of the [O iii]-emitting gas derived from our q3dfit analysis of the data cube centered around Hβ–[O iii] 5007 are shown in Figure 4. The 50-percentile (median, v₅₀) and 90-percentile velocities (v₉₀) are, respectively, the velocities at 50% and 90% of the total [O iii] flux, calculated starting from the red side of the line profile. The v₅₀ and v₉₀ velocity fields in Figure 4 highlight the fact that most of the gas outside of the inner arcsecond region (projected distances R ≳ 4 kpc from the quasar) is significantly blueshifted with respect to the systemic velocity of the system (z = 1.5933), derived from the median velocity, v₅₀, of the line emission in the central arcsecond region (after removal of the quasar light with q3dfit). This redshift is consistent within the errors with published values derived from other optical and ALMA data, z = 1.5930 (Cresci et al. 2015; Brusa et al. 2018).

Zoom In Zoom Out Reset image size

A velocity gradient spanning the range ∼[−250, +250] km s⁻¹ along PA = −10° is seen in the inner arcsecond (R ≲ 4 kpc) of the nebula centered on the quasar position (Figure 4(a)). This gradient is largely consistent in direction, amplitude, and location with the gradient of the molecular disk detected in the ALMA data (Brusa et al. 2018), which is attributed to galactic rotation in the potential of the host galaxy. The implied circular velocity (∼400 km s⁻¹, assuming that the inclination of the host galaxy disk is the same as that of the molecular disk, i = 30°; Brusa et al. 2018) is consistent with that estimated from the stellar mass of this galaxy: v_circ = $\sqrt{2}S\approx 400$ km s⁻¹, where log S = 0.29 log M_* − 0.93 (Weiner et al. 2006; Kassin et al. 2007) and a stellar mass log M_*/M_⊙ = ${11.65}_{-0.35}^{+0.35}$ is used for XID 2028 (Brusa et al. 2018). Therefore, the observed blueshifted velocities in the outer nebula are well in excess of the rotation velocity and are indicative of a large-scale outflow along our line of sight.

Figure 4 also shows velocity gradients on larger scales, along and perpendicular to the bright plume. The gas along the plume is systematically more blueshifted with increasing distance from the quasar: v₅₀ (v₉₀) range from ∼−200 (−500) km s⁻¹ within ∼1−2 kpc of the quasar to −800 (−1000) km s⁻¹ at a distance of 17 kpc. Meanwhile, the gas near the central axis of the plume is systematically more blueshifted by 200−300 km s⁻¹ than the off-axis gas.

The 80-percentile line widths (defined as w₈₀ = v₁₀−v₉₀ using the same convention as above to calculate v₁₀) are shown in Figure 4(c). The measured values of w₈₀ near the quasar are on average broader by ∼200−300 km s⁻¹ than those further out (recall that the quasar light has been scaled and removed here, so it does not contribute to the observed line broadening). Clear line splitting is seen on the northern and southern edges of the bright outflow plume, where both blueshifted and systemic-velocity gas is detected, which is apparent in Figures 2(d), and 3(a) and (b). On average, the [O iii] line profiles outside of the plume are slightly narrower by ∼100−200 km s⁻¹ than those in the plume.

We ran a separate q3dfit analysis of the NIRSpec data cube covering 3860–7200 Å in the quasar rest frame to capture the extranuclear Hβ, [O iii] 4959, 5007, Hα, [N ii] 6548, 6583, and [S ii] 6716, 6731 line emission. The results are shown in Figure 5. This analysis allows us to derive the [O iii] 5007/Hβ, [O iii] 5007/Hα, [N ii] 6583/Hα, and [S ii] 6716, 6731/Hα line ratio maps shown in the top panels of Figure 6. We also derive the Hα/Hβ and [S ii] 6716/6731 line ratio maps shown in Figure 7. For this analysis, we fit all the above listed emission lines simultaneously to get a consistent fit across all lines, and thus allow inter-line comparisons. The velocity centroids and widths of the individual Gaussian components are once again kinematically locked but allowed to vary in intensity relative to each other. While this approach is necessary to create self-consistent line ratio maps, it does limit the extent of these line ratio maps to only those spaxels where line emission is detected in both lines of the ratios rather than just [O iii] 4959, 5007, as in Section 4.2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Figures 6(e)–(f), the line ratios measured in the extranuclear nebula of XID 2028 are displayed in the diagnostics line ratio diagrams of (Baldwin et al. 1981, BPT; Figure 6(e)) and (Veilleux & Osterbrock 1987, VO87; Figure 6(f)) to assess the primary source of ionization of the line-emitting gas. The theoretical and empirical curves of Kewley et al. (2001) and Kauffmann et al. (2003), which separate AGNs from star-forming galaxies in these line ratio diagrams, are shown for comparison. We find that virtually all of the line ratios are consistent with photoionization by the central quasar, regardless of the location in the nebula. A more quantitative statement can be made by measuring the distance to the peak of the AGN branch, D_AGN, in the [N ii] diagram, following Yuan et al. (2010). We get D_AGN = 0.7−1 for nearly all data points in Figure 6(e), confirming that the AGN is the dominant source of ionization of the nebula. The lack of a significant gradient in the line ratios with distance from the quasar also indicates that the nebula is matter-bounded, i.e., it has no ionization edges. We discuss the implications of these results in Section 5.2.

Note that shock models with photoionized precursors (e.g., Allen et al. 2008) are able to reproduce some of the AGN-like line ratios of the XID 2028 nebula, but the lack of obvious trends between the line ratios and gas kinematics (v₅₀, v₉₀, w₈₀) outside of the inner arcsecond region of the host galaxy suggests that shock ionization and heating are not important in the nebula of XID 2028, contrary to J1652 (Vayner et al. 2023) and shock-dominated systems (e.g., Veilleux et al. 1995; Allen et al. 1999; Sharp & Bland-Hawthorn 2010; Rich et al. 2011, 2014, 2015; Hinkle et al. 2019, and references therein).

Figure 7(a) displays spatial variations in the log(Hα/Hβ ratios) that are suggestive of an extinction gradient across the host and outflow regions, with the inner host galaxy region generally displaying larger Hα/Hβ ratios consistent with A_V ≈ 1.1 mag, while the median value of A_V across the outer outflow region is 0.52 mag. Note that the Hα/Hβ ratios are available over only a limited area of the [O iii] nebula. The measurements are systematically more uncertain in the fainter portion of the nebula dominated by the outflowing gas. To avoid introducing noise bias between the host and outflow regions, we did not apply reddening corrections to the line ratios presented in Figure 6. However, we note that our interpretation of the line ratio maps and diagnostic diagrams is robust to these reddening corrections. This is not surprising because, by design (VO87), these line ratio diagrams involve emission lines that are close in wavelengths, and are therefore insensitive to extinction.

The [O iii] λ5007 emission line is by far the brightest line emitted by the outflow within the rest-frame visible range covered by the NIRSpec data cube. It is also well separated in wavelength from neighboring emission lines in contrast to Hα, which is blended with [N ii] 6548, 6583. Finally, this forbidden line is unaffected by line emission from the central high-density (n_e > 10⁹ cm⁻³) broad-line region (BLR), while broad-line Hβ and Hα are prominent in this type 1 quasar (Figure 1). Therefore, we use the strength of [O iii] 5007 rather than that of Hα or Hβ to estimate the mass of the outflowing gas (Cano-Díaz et al. 2012; Veilleux et al. 2020):

$$\begin{eqnarray}&&{M}_{\mathrm{ionized}}=5.3\times {10}^{8}\,\displaystyle \frac{{C}_{e}{L}_{44}([{\rm{O}}\,{\rm\small{III}}\,5007])}{{n}_{e,2}{10}^{[{\rm{O}}/{\rm{H}}]}}{M}_{\odot },\end{eqnarray} \tag{ 1 }$$

where L₄₄([O iii] 5007) is the luminosity of [O iii] λ5007, normalized to 10⁴⁴ erg s⁻¹, n_e,2 is the average electron density, normalized to 10² cm⁻³, ${C}_{e}\equiv \langle {n}_{e}^{2}\rangle /\langle {n}_{e}{\rangle }^{2}$ is the electron density clumping factor, which can be assumed to be of order unity on a cloud-by-cloud basis (i.e., each cloud has uniform density), and 10^[O/H] is the oxygen-to-hydrogen abundance ratio relative to the solar value. The ionized mass derived from this expression assumes an electron temperature T ∼ 10⁴ K and electron density n_e ≲ 7 × 10⁵ cm,⁻³ the critical density associated with the [O iii] 5007 transition above which collisional de-excitation becomes significant. The electron density in the extended [O iii] nebula is well below this value. We measure log ([S ii] 6716/6731) ≈ 0.0−0.1 in the outflowing portion of the nebula (Figure 7(b)), corresponding to a median electron density of ∼410 cm⁻³. We take the [O iii] 5007 line flux, 2.4 × 10⁻¹⁶ erg s⁻¹ cm⁻², in the portion of the extended nebula where the individual Gaussian component(s) have v₅₀ ≤−300 km s⁻¹, which is well in excess of observed velocities due to rotation, to be representative of the outflowing gas in this object. We assume a solar oxygen abundance in the nebula ([O/H] = 0) and use the median extinction in the outflow of A_V = 0.52 mag to derive an extinction-corrected ionized mass M_ionized = 1 × 10⁷ M_⊙ in the outflow. Note that the oxygen abundance may be higher than solar near the center of this massive galaxy. Meanwhile, up to 40% of the oxygen atoms may be depleted onto dust grains (Baron & Netzer 2019). These two effects may partly cancel each other. The readers should thus be cautious when interpreting the results.

This value of M_ionized is nearly two orders of magnitude smaller than the value derived from the observed (not corrected for extinction) Hβ line emission in the ground-based IFS data analyzed by Cresci et al. (2015). A factor of ∼4 is accounted for by the higher n_e used in our calculations (410 versus 100 cm⁻³). We have compared our absolute flux measurements with those in the literature and found an excellent agreement to within ±20% of the values published in Scholtz et al. (2020). The outflow mass derived in our data is not sensitive to our adopted definition for the outflow, v₅₀ ≤−300 km s⁻¹. The [O iii] flux is boosted by only ∼33% if we instead use v₅₀ ≤−200 km s⁻¹ for the outflow threshold. The significant gain in resolution and sensitivity of the NIRSpec IFU over ground-based IFS, combined with our careful removal of the quasar light with q3dfit, provides an unprecedented view of the outflow in XID 2028 and likely accounts for most of this discrepancy. This is compounded by the fact that the Hβ line used as a mass tracer by Cresci et al. (2015) is more sensitive to errors in the quasar light removal than [O iii] 5007 because it is typically an order of magnitude fainter than [O iii] 5007 (Figures 6(a), (e)) and lies on top of the quasar broad-line Hβ emission.

To estimate the outflow mass rate and energetics, we need to know the dynamical timescale of each parcel i of outflowing gas, τ_dyn,i ≈(R_deproj,i/v_deproj,i) = (R_i /sin θ_i )(v_i /cos θ_i )⁻¹ = (R_i /v_i ) cot(θ_i ), where R_i is the measured distance from the center of the gas parcel on the sky, v_i is the measured outflow radial velocity of that same gas parcel, and θ_i is the angle between the outflow velocity of the gas parcel and our line of sight. The integrated mass outflow rate is the sum of m_i /τ_dyn,i over all gas parcels, namely

$$\begin{eqnarray}&&\dot{M}={\rm{\Sigma }}\,\dot{{m}_{i}}={\rm{\Sigma }}\,{m}_{i}\,({v}_{i}/{R}_{i})\,\tan ({\theta }_{i}).\end{eqnarray} \tag{ 2 }$$

The corresponding momenta and kinetic energies and their outflow rates are

$$\begin{eqnarray}&&p={\rm{\Sigma }}\,{m}_{i}\,{v}_{i}\,\sec \,{\theta }_{i},\end{eqnarray} \tag{ 3 }$$

$$\begin{eqnarray}&&\dot{p}={\rm{\Sigma }}\,\dot{{m}_{i}}\,{v}_{i}\,\sec \,{\theta }_{i},\end{eqnarray} \tag{ 4 }$$

$$\begin{eqnarray}&&E=\displaystyle \frac{1}{2}\,{\rm{\Sigma }}\,\{{m}_{i}\,[{\left({v}_{i}\,\sec \,{\theta }_{i}\right)}^{2}+3\,{\sigma }_{i}^{2}]\},\end{eqnarray} \tag{ 5 }$$

$$\begin{eqnarray}&&\dot{E}=\displaystyle \frac{1}{2}\,{\rm{\Sigma }}\,\{\dot{{m}_{i}}[\,{\left({v}_{i}\,\sec \,{\theta }_{i}\right)}^{2}+3\,{\sigma }_{i}^{2}]\},\end{eqnarray} \tag{ 6 }$$

where the energy includes both the bulk kinetic energy due to the outflowing gas and the turbulent kinetic energy (where we assume the same velocity dispersion σ in each dimension).

In the following discussion, we set v_i = v₅₀ of the individual Gaussian component(s) in the [O iii] line profile at each spaxel where the outflow is detected (following our definition of the outflow, v₅₀ must be ≤−300 km s⁻¹). A derivation of the values of θ_i requires detailed kinematic modeling of the outflow, which is beyond the scope of this first-look paper (W. L. Liu et al. 2023, in preparation). Here, we make a number of simplifying assumptions to estimate the energetics of the outflow. First, we note that most (∼97%) of the outflowing gas is in the bright [O iii] plume and shows only moderate velocity variations across, and perpendicular to, the plume. This is consistent to first order with a simple kinematic model where all gas parcels in the plume move coherently along the same direction, i.e., θ_i = θ. The value of θ may be constrained if the absence of the redshifted counterpart to the outflow in our data is due to obscuration by the host galaxy, assuming biconical symmetry with the observed blueshifted outflow. Recall that the outflow extends beyond the bright plume to trace a cone with an opening angle ϕ ∼90° (Figures 3(a) and 4(b)). The main axis of the cone must therefore be tilted by ≲45° from our line of sight, i.e., θ ≲ 45°, assuming biconical symmetry and a favorable face-on galaxy orientation. Note, however, that this assumption breaks down if we are dealing with a one-sided outflow. A sketch of the outflow geometry, as seen on the sky, is shown in Figure 8.

Zoom In Zoom Out Reset image size

Assuming θ_i = θ = 45° in Equations (2)–(6), we get $\dot{M}$ = 1.9 M_⊙ yr⁻¹, $\dot{p}$ = 2.4 × 10³⁴ dynes, and $\dot{E}$ = 3.6 × 10⁴² erg s⁻¹. Not surprisingly, these outflow mass rate and energetics are two orders of magnitude smaller than the values derived from the ground-based IFS data (Cresci et al. 2015). This difference is due to the ∼100× smaller outflowing ionized gas mass in the NIRSpec data rather than to our simplifying assumption that the outflow is largely along θ = 45°. Under this simple assumption, the dynamical timescale of the outflowing gas ranges from ≲0.1 Myr nearest the quasar to ≳10 Myr at 17 kpc from the quasar. In an independent analysis of the exact same Q3D NIRSpec data cube on XID 2028, Cresci et al. (2023) recently reported [O iii]-based outflow mass rate and energetics that agree within the errors (factor of ∼3) with the values reported here.

The radiative pressure due to the AGN, L_AGN/c ≈7 × 10³⁵ dynes (where we used L_AGN ∼ 2 × 10⁴⁶ erg s⁻¹, derived from the spectral energy distribution decomposition of Lusso et al. 2012), is 35× larger than the measured momentum rate $\dot{p}$ in the outflow. The quasar can thus in principle easily drive this outflow via radiation pressure. In the single-scattering optically thin limit, the momentum flux imparted onto the gas is simply ${L}_{\mathrm{absorbed}}/c\simeq (1-{e}^{-{\tau }_{\mathrm{UV}}}){L}_{\mathrm{AGN}}/c$, where τ_UV is the optical depth to the UV radiation. Dust in the outflowing gas (Figure 7(a)) will boost this term significantly.

The ratio of the kinetic energy outflow rate to the AGN luminosity in this object, $\dot{E}/{L}_{\mathrm{AGN}}$ = 1.8 × 10⁻⁴, is low but not unusually so for a type 1 quasar with this luminosity (e.g., Fiore et al. 2017; Rupke et al. 2017; Harrison et al. 2018). Recently, a weak, loosely collimated, radio jet has been purported to exist in XID 2028 (source #10964 in Figure 6 of Vardoulaki et al. 2019). The radio emission is roughly aligned along the same southwestward direction, and extends on a similar scale as the brightest blueshifted [O iii] line-emitting cloud in Figure 3(a). This loose jet may contribute to driving the observed outflow or may be the by-product of shocks produced where the fast outflowing material collides with the ambient host ISM (e.g., Whittle et al. 1988; Whittle 1992; Zakamska & Greene 2014).

Note that the mass outflow rate is much smaller than the dust-obscured star formation rate in XID 2028, SFR = ${134}_{-70}^{+132}$ M_⊙ yr⁻¹, inferred from the ALMA observations in the rest-frame far-IR by Scholtz et al. (2020), adjusted to our cosmology, assuming that the AGN does not contribute to the far-IR fluxes. Since the implied mass-loading factor, $\dot{M}$/SFR, is much smaller than unity, we cannot formally rule out the possibility that stellar processes, rather than the quasar itself, drive this outflow. Indeed, the energy rate from a starburst with SFR = 134 M_⊙ yr⁻¹ is $\dot{{E}_{* }}$ ≈7 × 10⁴¹ SFR erg s⁻¹ = 1 × 10⁴⁴ erg s⁻¹ (e.g., Veilleux et al. 2005), or ∼30× larger than the measured kinetic power $\dot{E}$ in the outflow. While the maximum outflow velocity in XID 2028, ∼1000 km s⁻¹, is large in comparison to those of local starburst-driven winds, it is not exceptionally high for more distant compact starburst galaxies such as XID 2028 (e.g., Tremonti et al. 2007; Rupke et al. 2019).

As discussed in Section 4.3, the optical line ratios in the extended nebula of XID 2028 are consistent with those expected for AGN photoionization (D_AGN = 0.70−1). Thus, hot and young stars, if present in the host galaxy, do not contribute significantly to the ionization of the warm ionized gas probed by the optical lines. This statement seems to be inconsistent with the large star formation rate (SFR = 134 M_⊙ yr⁻¹) derived by Brusa et al. (2018) and Scholtz et al. (2020) using the rest-frame far-IR flux from ALMA. Barring contamination of the far-IR flux by AGN continuum emission, most of the star formation activity in XID 2028 must therefore be shrouded in dust to make it undetectable in the rest-frame optical band. ²⁷ This seems plausible given that XID 2028 is the prototypical obscured quasar selected from the Cosmic Origins Survey (COSMOS) survey based on its observed red color (r − K = 4.81) and high X-ray to optical flux ratio (Hasinger et al. 2007; Brusa et al. 2018). Indeed, the host of XID 2028 has a large (dusty) molecular gas mass of ∼1 × 10¹⁰ M_⊙ (Brusa et al.2018). The obscured star formation distribution in this object, traced by the ALMA high-resolution 1.3-mm (rest-frame 500 μm) flux map presented in Brusa et al. (2018) (see also Scholtz et al. 2020), is remarkably symmetric around the quasar except for a narrow plume of emission that juts out ∼11 (∼9 kpc) in the northeast direction (PA ≈ 45°). Faint [O iii] 5007 line emission at near-systemic velocity is detected in this general direction (Figure 3(b)) but no other emission lines lie above the detection limit in our data (Figures 4(a)–(d)), and so the dominant source of ionization of the gas responsible for this faint [O iii] emission cannot be constrained from the optical line ratios.

Given the crucial role played by the dust in shaping our views of XID 2028, it may be surprising to realize that this object is a relatively bright source in the rest-frame near-UV (NUV; ∼3000 Å). Following Brusa et al. (2010, their Figure 13), Cresci et al. (2015, their Figure 4), and Scholtz et al. (2020, their Figure 11), we compare the extent of the outflow traced by the blueshifted [O iii] line emission shown in Figure 3(a) with the archival HST/ACS F814W image that was obtained as part of COSMOS (Scoville et al. 2007). Figure 9 shows a clear asymmetry of the NUV emission in the general direction of the outflow. Some of the extended NUV emission roughly coincides with the brighter [O iii] clouds in the outflowing plume of gas, while other NUV emission loosely follow the fainter wide-angle cone of outflowing [O iii]-emitting material.

Zoom In Zoom Out Reset image size

The NUV emission is often used as a star formation indicator in unobscured regions of galaxies (e.g., Kennicutt & Evans 2012). Taken at face, the flux of the extended NUV emission beyond r = 005, ∼7 × 10⁻¹⁶ erg s⁻¹ cm⁻², translates into a star formation rate SFR_NUV ∼1 M_⊙ yr⁻¹, which may be considered an upper limit on the rate of (unobscured) star formation triggered by the outflow as it propagates through the host ISM (positive feedback). However, the AGN-like line ratios in this region are inconsistent with this interpretation of the NUV emission. Moreover, this interpretation is fraught with errors when dealing with dusty systems such as XID 2028, especially around intense sources of UV radiation such as quasars. Extended NUV emission has been detected in several luminous obscured quasars, tracing giant scattering cones where the quasar radiation field is scattered off dust in the surrounding material, confirmed by spectropolarimetric data (Zakamska et al. 2006; Obied et al. 2016; Wylezalek et al. 2016). About 75% of the observed flux at ∼3000 Å in luminous obscured quasars may be due to scattered light, which if left unaccounted for may strongly bias estimates of the star formation rates of quasar hosts.

An analysis of the rest-frame NUV emission in the first target of the Q3D JWST ERS program, J1652, revealed a giant scattering cone along the direction of the outflow in this quasar (Wylezalek et al. 2022). Given the presence of dust in the outflowing gas of XID 2028 (Figure 7(a)) and the loose connection between this gas and the rest-frame NUV emission (Figure 9), we argue that the same scattering process is taking place in this object. The loose spatial correlation of the extended NUV emission with the wind-angle conical outflow of XID 2028 is reminiscent of other scattering cones in starburst-driven winds (e.g., M82; Hoopes et al. 2005) and quasar-driven winds (Zakamska et al. 2006; Obied et al. 2016; Wylezalek et al. 2016). Contrary to the [O iii] emission that traces n_e ², the spatial distribution of the scattered NUV light is a complex function of several variables (e.g., geometry and strength of the NUV radiation field, location, and dust column density distribution of the scattering material, and our viewing angle with respect to this material). Consequently, this loose correlation between NUV emission and the [O iii] clouds is actually expected.

The absence of a tight one-to-one correlation between the extended NUV emission and brightest [O iii] 5007 clouds also rules out the possibility that the extended NUV emission is entirely due to [Ne v] 3426 and Mg II 2800 line emission within the F814W filter bandpass (∼7000–9500 Å or ∼2700–3650 Å in the rest frame; note that [O ii] 3727 lies outside of the bandpass). This is confirmed quantitatively: the AGN photoionization models of Groves et al. (2004) predict fluxes for ([Ne v] 3426 + Mg ii 2800) that are ≲25% of the [O iii] 5007 fluxes in cases where [O iii] 5007/Hβ ≈10 as seen in the outflow (Figure 6). Dust extinction in the outflowing material (A_V = 0.5 mag) will reduce the strength of [Ne v] 3426 + Mg II 2800 relative to [O iii] 5007 by a factor of ∼2. In the end, we find that the [Ne v] 3426 + Mg II 2800 line emission contributes at most ∼10% of the observed extended NUV emission.

Overall, the JWST and HST data paint a picture of XID 2028 in a blow-out phase where the combined action of radiative and mechanical modes of quasar feedback have accelerated warm ionized gas up to high velocities, breaking through the thick dusty shroud of the obscured quasar. While the bulk (∼97%) of the outflowing material lies in a loosely collimated plume west of the quasar, fainter outflowing clouds trace a wider cone centered on the plume spanning an angle ϕ ∼90°. The intense radiation field from the quasar is ionizing and possibly also driving the warm ionized outflow out to at least 17 kpc. The lack of ionization edges in the outflow line-emitting nebula indicates that it forms a matter-bounded structure where some of the ionizing radiation manages to escape the host galaxy and may potentially ionize the surrounding CGM, keeping it warm and perhaps preventing cold gas from accreting back onto the galaxy and forming new stars.

The outflow rate of 1.9 M_⊙ yr⁻¹ derived from our analysis of the [O iii] nebula is small for such a powerful quasar (e.g., Cano-Díaz et al. 2012; Fiore et al. 2017; Rupke et al. 2017). It is also much smaller than the cold-gas mass outflow rate reported by Brusa et al. (2018), 50–350 M_⊙ yr⁻¹, based on a tentative (<5σ) detection of high-velocity CO gas with ALMA and a CO-to-H₂ conversion factor α_CO = 0.13–0.80 M_⊙/(K km s⁻¹ pc²). The neutral-gas outflow traced by the broad Na i D absorption feature in the spectrum of XID 2028 (Figure 1) is spatially unresolved in the JWST data (r ≲ 1 kpc), so the neutral-gas mass outflow rate remains unconstrained (as is the case of the neutral/low-ionization Mg II 2800 outflow detected by Perna et al. 2015). Given the median outflow velocity in the [O iii]-emitting material (v₅₀ ≈550 km s⁻¹ for the entire line profile) and estimated stellar mass of log M_*/M_⊙ = ${11.65}_{-0.35}^{+0.35}$ from Brusa et al. (2018), the escape fraction ²⁸ of the warm ionized gas is ∼3%, so only 0.06 M_⊙ yr⁻¹ is able to escape the galaxy potential and make it to the IGM. This warm ionized outflow event is thus a negligible contributor to the enrichment of the CGM, let alone the IGM. The neutral- and cold-gas outflows of XID 2028 have lower velocities and smaller sizes than the [O iii] outflow, so they will also not contribute to the enrichment of the CGM and IGM.

Under these circumstances, the unusually small molecular gas fraction in XID 2028 that was reported by Brusa et al. (2018; a ratio of molecular gas to stellar mass ≲5%, significantly smaller than those typically measured in high-z galaxies with similar specific star formation rates) is hard to explain as due purely to ejective quasar mechanical feedback. We instead favor joint action of radiative and mechanical feedback where a significant fraction of the quasar hard ionizing radiation is able to escape through the cone created by the wide-angle outflow, dissociating/ionizing the molecular gas on its path.

In this scenario, the quasar radiation field may only affect the gas within the outflow (bi)cone. ²⁹ Assuming a simple (bi)conical symmetry for the outflow, the solid angle subtended by the outflow, normalized to 4π steradians, is (1 − cos ϕ/2) × 50% (100%) = 15% (30%). If the galaxy ISM and CGM are, to first order, distributed spherically symmetrically around the quasar, then only about 15% (30%) of this gas may be affected by the quasar ionizing radiation. This fraction may be less in the case of the molecular disk in the inner arcsecond (Brusa et al. 2018). This may explain why this object lies on the main sequence of star-forming galaxies at z ≈ 2 (Brusa et al. 2018). In the end, we find no convincing evidence for star formation quenching by quasar mechanical or radiative feedback in this object.