Molecular and Ionized Gas Phases of an AGN-driven Outflow in a Typical Massive Galaxy at z ≈ 2

Figure 2 shows the CO(3–2) integrated intensity and velocity maps of zC400528. We measure an integrated CO(3–2) flux of F_CO = 1.11 ± 0.05 Jy km s⁻¹, which corresponds to a luminosity of L_CO = 3.45 × 10¹⁰ L_⊙.¹⁹ The molecular gas mass (M_mol,disk), assuming a conversion factor ${\alpha }_{\mathrm{CO}(1\mbox{--}0),{\rm{T}}18}\,=4.3\,{M}_{\odot }\,{\left({\rm{K}}\mathrm{km}{{\rm{s}}}^{-1}{\mathrm{pc}}^{-2}\right)}^{-1}$ (Tacconi et al. 2018)²⁰ and a velocity-integrated Rayleigh–Jeans brightness temperature line ratio R₁₃ = 1.3 (e.g., Dannerbauer et al. 2009; Bolatto et al. 2015; Daddi et al. 2015), is ${M}_{\mathrm{mol},\mathrm{disk}}=1.1\times {10}^{11}\,{M}_{\odot }$.²¹ This corresponds to a gas-to-stellar ratio of ${\mu }_{\mathrm{mol},\mathrm{disk}}={M}_{\mathrm{mol}}/{M}_{* }\,\approx 1$, which is consistent with the typical gas-mass fractions observed in massive star-forming galaxies at z ∼ 2 (e.g., Tacconi et al. 2010, 2013, 2018).

Zoom In Zoom Out Reset image size

If we add up the stellar and molecular gas masses, we obtain ${M}_{* }+{M}_{\mathrm{mol},\mathrm{disk}}\approx 2\times {10}^{11}\,{M}_{\odot }$, which is comparable within the uncertainties to the dynamical mass derived from the ionized gas kinematics, i.e., ${M}_{\mathrm{dyn}}={1.3}_{-0.6}^{+1.5}\times {10}^{11}\,{M}_{\odot }$ (Förster Schreiber et al. 2018a).

The integrated velocity map in the right panel of Figure 2 shows that the molecular disk is rotating approximately in the east–west direction, consistent with the observed rotation in the ionized gas (Newman et al. 2013; Förster Schreiber et al. 2018a).

Figure 3 (left) compares the spatial distribution of the molecular gas, the ionized gas disk traced by Hα emission (Förster Schreiber et al. 2014), and the stellar mass traced by the Hubble Space Telescope (HST) H band (Tacchella et al. 2015b). Both molecular and ionized gas components share a similar spatial structure, including a more diffuse component that extends toward the southeast region. The peak of the H-band emission is shifted ∼02 northeast with respect to the peak of the ALMA CO(3–2) emission. The astrometric precision of our ALMA data is ∼0025,²² which is at least a factor of ∼10 better than the astrometrical accuracy of the HST data. Therefore, it is not possible to determine whether the spatial offset is due to physical reasons or astrometric errors.

Zoom In Zoom Out Reset image size

The right panel of Figure 3 shows the CO(3–2) and Hα+[N ii] spectrum from within a 06 radius circular aperture centered at the peak of emission (the systemic velocity corresponds to a redshift of z = 2.387; see Section 2). The [N ii]/Hα line peak ratio is 0.8 (see also Genzel et al. 2014), which is characteristic of galaxies where the AGN contributes to the ionizing radiation or shocks affect the ionization balance (e.g., Kewley et al. 2013; Newman et al. 2014). As described in detail in Förster Schreiber et al. (2014) and Genzel et al. (2014), an ionized outflow is detected as a strong nuclear broad component in the Hα spectrum (detected out to velocities ∼±1000 km s⁻¹ relative to the systemic velocity). The broad Hα emission is resolved with a spatial resolution of 015 (∼1.2 kpc). In the case of the CO(3–2) spectrum, we observe a high-velocity wing on the red side ([+300, +500] km s⁻¹) that is likely the molecular component of the outflow. We analyze and interpret this high-velocity feature in the CO(3–2) spectrum in the next section.

As a first step to study the high-velocity components in the ionized and molecular spectrum of zC400528, we make position–velocity (P–V) diagrams extracted along the vertical (south–north), horizontal (west–east), and −45° (southeast–northwest) directions through the emission peak. The top panel in Figure 4 shows the results for the CO(3–2) data. We identify the rotation pattern of the galaxy in the [−250, +250] km s⁻¹ velocity range and a high-velocity (v ≳ +300 km s⁻¹) component along all of the examined directions that preferentially extends along the northwest direction for about ∼1'' (which corresponds to a projected distance of ≈8.4 kpc). The contours in the bottom panels of Figure 4 show the P–V curves for the Hα and [N ii] lines (the systemic velocity is set to match the Hα line central velocity) plotted on top of the CO(3–2) P–V diagram. On the receding side, the outflow is detected in both ionized and molecular gas, with the latter showing a more extended spatial distribution. On the other hand, there is no molecular counterpart to the high-velocity ionized gas observed at velocities v ≲ −300 km s⁻¹.

Zoom In Zoom Out Reset image size

4.2.1. Disentangling the Molecular Emission from Disk and Outflow

Quantifying the total amount of molecular gas in the outflow requires identifying and removing the CO(3–2) line emission associated with the rotating disk. To achieve this, here we use two different methods.

• (1)  
  Two-component Gaussian fit to the spectrum.  The most common approach to disentangling emission from the disk and outflow in galaxies is to simultaneously fit two Gaussian profiles to the spectrum. The first Gaussian component centered at the systemic velocity represents the rotating disk, while the second component characterizes the outflow, with the caveat that it might include line emission at low velocities that is associated with the disk. We applied this method to the CO(3–2) spectrum extracted within a 05 radius circular aperture centered at the peak of CO emission. The resulting two-Gaussian fit is shown in Figure 5 (bottom left panel). The component centered at the systemic velocity (${v}_{0,\mathrm{disk}}=-15$ km s⁻¹) reaches a peak flux density of ${S}_{\mathrm{peak},\mathrm{disk}}^{\mathrm{CO}}=2.4$ mJy and has a velocity width of ${\mathrm{FWHM}}_{\mathrm{disk}}=277.3$ km s⁻¹. The second component, which we associate with the outflow, has a central velocity of ${v}_{0,\mathrm{out}}=415$ km  s⁻¹, a peak flux density of ${S}_{\mathrm{peak},\mathrm{out}}^{\mathrm{CO}}=0.38$ mJy ($\sim 0.16\times {S}_{\mathrm{peak},\mathrm{disk}}^{\mathrm{CO}}$), and a velocity width ${\mathrm{FWHM}}_{\mathrm{out}}=254.3$ km s⁻¹. The integrated CO(3–2) flux in the outflow component is ${F}_{\mathrm{CO},\mathrm{out}}=0.10$ Jy km s⁻¹.
• (2)  
  Subtraction of a smooth model of the disk.  Given that we have sufficient spatial resolution and signal-to-noise in our ALMA CO(3–2) data, here we try a different strategy to attempt to remove the disk material. The method consists of creating and subtracting from the original cube a smooth model of the rotating disk. This approach is similar to that used by Veilleux et al. (2017) to disentangle the CO(1–0) line emission associated with the molecular outflow in a luminous infrared galaxy.

Zoom In Zoom Out Reset image size

We use the CASA task imfit to fit in each 50 km s⁻¹ channel of the CO(3–2) cube a two-dimensional Gaussian to the image in the region where the source is detected. We then use the resulting fits in the [−250, +250] km s⁻¹ range to make a source model where the Gaussian position, orientation, and intensity changes as a function of velocity. The velocity range was chosen to encompass the bulk of the CO(3–2) line emission associated with the rotating disk as suggested by the two-component Gaussian fit, the observed line width of the Hα line (Δv/2 ≈ 150 km s⁻¹; Förster Schreiber et al. 2018a), and the P–V diagram (Figure 4). In the last step, we removed the source model from each 50 km s⁻¹ slice, resulting in a residual cube where the outflowing gas is now disentangled from the emission associated with the rotating disk.

The top panels in Figure 5 show the integrated intensity map calculated on the [−250, +250] km s⁻¹ range of the original cube (left) and the cube resulting after we subtract the rotating disk model (i.e., the residual cube; right). The model does a good job of removing the CO(3–2) line emission associated with the rotating disk, and we only observe a relatively high residual at the level of ∼0.06 Jy km s⁻¹ in the more diffuse and extended southeast component. The bottom left panel in Figure 5 shows the CO(3–2) spectrum integrated over a circular aperture with a radius of 05. The black and green lines correspond to the spectrum extracted in the original cube and the residual cube, respectively. The integrated signal of the high-velocity red-wing emission in the residual cube is clear, and there is no evidence for a molecular outflow component on the approaching side of the spectrum. The bottom right panel of Figure 5 shows the integrated intensity map of the residual cube in the velocity range where we identified the red-wing emission ([+300, +700] km s⁻¹). The outflow emission peaks around 02 (∼2 kpc) west of the molecular disk center, and there is a tail of outflowing gas line emission that extends beyond the molecular disk toward the northwest for about 1'', which corresponds to a projected distance of ∼8 kpc.

4.2.2. The Structure of the Molecular Outflow

A more detailed view of the molecular outflow components as a function of velocity is shown in the velocity channel map of the residual cube in Figure 6. For reference, overplotted on each panel are the contours from the velocity-integrated CO(3–2) emission of the rotating disk. The first significant outflow structure appears at v = +300 km s⁻¹ and is spatially coincident with the central region of the molecular gas disk. Between v = +350 and +400 km s⁻¹, the main component of the outflow moves toward the west about 02, which corresponds to a projected distance of ∼2 kpc. At v =+450 km s⁻¹, the main outflow component starts to extend toward the north and continues until v = +500 km s⁻¹, which is the last channel at which we detect CO(3–2) line emission.

Zoom In Zoom Out Reset image size

The last two panels in Figure 6 show the integrated intensity and velocity maps of the molecular outflow detected in the [+250, +500] km s⁻¹ range. The bulk of the line emission associated with the outflow is shifted about 02 (projected distance of ∼2 kpc) west of the center of the molecular disk and extends to the north for approximately 07 (projected distance of ∼6 kpc), where it shifts to the west in what seems to be a second outflow component—still connected to the main component—at about 12 (projected distance of ∼10 kpc) from the center. One could suspect that this spatially extended feature of the outflow is in reality a tidal tail product of the interaction with a lower-mass companion, but visual inspection of a deep HST F160W image of the field reveals no stellar emission associated with a close neighbor in the northeast quadrant.

We observe a velocity gradient in the outflow: the gas velocity increases from ∼+300 to ∼+500 km s⁻¹ as the outflow extends north of the nucleus. This could be indicative of an accelerating continuous outflow. Another possibility could be that the ejection of molecular gas triggered during the last outflow burst had a distribution of velocities, so the faster ejecta components traveled further away than the slower components. The most distant outflow component in the northwest has a velocity of ∼+350 km s⁻¹, a lower velocity that could be the result of gravitational pull.

In summary, perhaps the most striking characteristic of the molecular outflow in zC400528 is its asymmetry: we only detect the receding component of the wind. In the local universe, it is not uncommon to observe asymmetric outflows. For example, Pereira-Santaella et al. (2018) found that in four out of five spatially resolved molecular gas outflows in ultra-luminous infrared galaxies (ULIRGs), the receding component of the wind is stronger than the approaching one. This is also true for the molecular outflow in Mrk 231 (Cicone et al. 2012; Feruglio et al. 2015). High-resolution simulations of z ∼ 2 isolated disks also find that AGN-driven outflows are typically unipolar as a result of dense cloud structures in the vicinity of the black hole blocking the expansion in one direction (Gabor & Bournaud 2014). Another possibility is that the molecular outflow on the approaching side experienced a change of phase, similar to what is observed in M82, where the dominant phase of the wind transitions from molecular to atomic at about 1 kpc from the disk (Leroy et al. 2015).

4.2.3. The Molecular Gas Mass in the Outflow

For each outflow component, we measure CO(3–2) flux densities with Gaussian fits to the spectra in the residual cube extracted from the circular and elliptical apertures shown in Figure 7. The flux densities and velocity dispersions (σ_CO) are listed in Table 1.

Zoom In Zoom Out Reset image size

Table 1.  Molecular Outflow Mass in zC400528

Component	σ	Integrated Flux	${M}_{\mathrm{out},\mathrm{mol}}$ a
No.	(km s−1)	(Jy km s−1)	$({10}^{9}\,{M}_{\odot }$)
1	158.0	0.098 ± 0.016	1.81 (0.77–9.72)
2	60.0	0.029 ± 0.007	0.50 (0.21–2.68)
3	173.5	0.032 ± 0.007	0.59 (0.25–3.17)
4	167.0	0.184 ± 0.029	3.36 (1.42–18.06)

Note.

^aMolecular gas masses calculated using an α_CO,ULIRG conversion factor. The values in parentheses correspond to the molecular gas masses we would obtain if we applied an α_CO,thin or α_CO,MW conversion factor, respectively.

Download table as:  ASCIITypeset image

The velocity dispersions derived from the Gaussian fit to the spatially resolved outflow components 1, 2, and 3 are in the σ_CO ∼60–170 km s⁻¹ range. These are comparable to the velocity dispersion observed in the spatially resolved molecular outflows of M82 (σ_co ∼ 30–70 km s⁻¹, ${\sigma }_{[\mathrm{oi}]63\mu {\rm{m}}}$ ∼ 80–130 km s⁻¹; Contursi et al. 2013; Leroy et al. 2015), the LIRG ESO 320-G030 (σ_CO ∼ 30–70 km s⁻¹; Pereira-Santaella et al. 2016), and the AGN/LIRGs I12112 and I14348 (σ_CO ∼ 30–120 km s⁻¹; Pereira-Santaella et al. 2018).

How to convert the flux densities associated with the outflow into molecular gas masses is still an open question. The CO-to-H₂ conversion factor depends mainly on the metallicity and surface density of the gas (e.g., Bolatto et al. 2013b). In the case of the CO gas in the outflow, in the literature, there are three CO-to-H₂ conversion factors commonly used: (1) a Galactic conversion factor of ${\alpha }_{\mathrm{CO},\mathrm{MW}}=4.3\,{M}_{\odot }\,{({\rm{K}}\mathrm{km}{{\rm{s}}}^{-1}{\mathrm{pc}}^{-2})}^{-1}$ (where MW stands for Milky Way; Bolatto et al. 2013b), (2) a ULIRG-like conversion factor of ${\alpha }_{\mathrm{CO},\mathrm{ULIRG}}=0.8\,{M}_{\odot }\,{({\rm{K}}\mathrm{km}{{\rm{s}}}^{-1}{\mathrm{pc}}^{-2})}^{-1}$ (e.g., Cicone et al. 2014; Feruglio et al. 2015; Veilleux et al. 2017), and (3) an optically thin conversion factor of ${\alpha }_{\mathrm{CO},\mathrm{thin}}\,=0.34\,{M}_{\odot }\,{({\rm{K}}\mathrm{km}{{\rm{s}}}^{-1}{\mathrm{pc}}^{-2})}^{-1}$ (e.g., Bolatto et al. 2013a; Richings & Faucher-Giguère 2018). Assuming that the molecular gas in the outflows is optically thin—as observed, for example, in the jet-accelerated wind of IC 5063 (Dasyra et al. 2016)—would yield the most conservative (or "minimum") estimate of the molecular gas masses.

In this paper, we adopt the ULIRG-like conversion factor, which seems to be a good compromise, given the range of conversion factors available. In addition, the detection of dense molecular gas entrained in the outflow of starbursts and LIRGs (e.g., Aalto et al. 2012, 2015; Walter et al. 2017) argues in favor of a conversion factor higher than α_CO,thin. In the main streamer of NGC 253, for example, the outflowing gas is not optically thin, as the observed CO(2–1)-to-CO(1–0) brightness temperature line ratio is about unity (Zschaechner et al. 2018). We also expect that the molecular gas in the outflow is likely not cold but warm, because it has been shocked at some level and/or is immersed in hot gas and subjected to a substantial external radiation field. In that case, the physical conditions of the molecular gas in the wind better resemble those found in the interstellar medium (ISM) of (U)LIRGs rather than the Milky Way.

The molecular gas masses for each of the outflow components are listed in Table 1. The total molecular gas mass in the outflow of zC400528 is

$$\begin{eqnarray*}&&{M}_{\mathrm{out},\mathrm{mol}}=3.36\times {10}^{9}\,{M}_{\odot }\times \left(\displaystyle \frac{{\alpha }_{\mathrm{CO}}}{{\alpha }_{\mathrm{CO},\mathrm{ULIRG}}}\right),\end{eqnarray*}$$

which corresponds to ∼3% of the molecular gas mass in the disk. As Figure 8 shows, this fraction is similar to those measured in local (U)LIRG and Seyfert galaxies of similar stellar mass (Fiore et al. 2017 and references therein).²³

Zoom In Zoom Out Reset image size

4.2.4. Molecular Mass Outflow Rate

There are two common approaches to measure outflow mass rates in galaxies. They are often referred to as instantaneous (maximum) and average (minimum), and a detailed description of the assumptions that go into the calculations can be found in Rupke et al. (2005). In the instantaneous approach, the outflow mass rate (dM_out/dt) is calculated as the product of the outflow gas mass and the timescale taken by the gas to cross the thickness of the outflowing shell, i.e.,

$$\begin{eqnarray}&&{\dot{M}}_{\mathrm{out}}^{\mathrm{inst}}={M}_{\mathrm{out}}\times \displaystyle \frac{{v}_{\mathrm{out}}}{{\rm{\Delta }}R},\end{eqnarray} \tag{ 1 }$$

where v_out is the velocity of the outflow and ΔR ($={R}_{\mathrm{out}}-{R}_{\mathrm{int}}$) is the thickness of the outflow shell. This approach has been used in a number of studies, including Sturm et al. (2011) and González-Alfonso et al. (2014).

In the case of the average approach, the assumption is that the outflowing gas extends to r = 0, so the outflow mass rate (${\dot{M}}_{\mathrm{out}}$) is given by the outflow gas mass time-averaged over the flow timescale, i.e.,

$$\begin{eqnarray}&&{\dot{M}}_{\mathrm{out}}^{\mathrm{avg}}={M}_{\mathrm{out}}\times \displaystyle \frac{{v}_{\mathrm{out}}}{{R}_{\mathrm{out}}}.\end{eqnarray} \tag{ 2 }$$

If the emitting volume (spherical or multiconical) is filled with uniform density, then the mass outflow rate in Equation (2) should be a factor of 3 higher (e.g., Feruglio et al. 2010; Maiolino et al. 2012; Rodríguez Zaurín et al. 2013; Fiore et al. 2017).

In this paper, we use the average approach, which represents a more conservative way to characterize the outflow than the instantaneous method, as ${\dot{M}}_{\mathrm{out}}^{\mathrm{avg}}$ is smaller than ${\dot{M}}_{\mathrm{out}}^{\mathrm{inst}}$ by a factor $({R}_{\mathrm{out}}/{\rm{\Delta }}R)$. Following this, the first step is to calculate the flow timescale, t_out = R_out/v_out. As described in Section 4.2.2, zC400528 has three main molecular outflow components. If we estimate the flow timescale based on the maximal extension of the outflow, then ${t}_{\mathrm{out},\mathrm{mol}}\approx 8.5\,\mathrm{kpc}/400\,\mathrm{km}\,{{\rm{s}}}^{-1}\approx 2\times {10}^{7}\,\mathrm{yr}$. A more representative or characteristic timescale would result from considering only the main outflow component (which corresponds to regions 1 and 2 in Figure 7). This component encompasses ∼80% of the outflow total mass and has a size of ≈4.2 kpc (${R}_{\mathrm{out},\mathrm{mol}}\approx 0\buildrel{\prime\prime}\over{.} 5$, according to a two-dimensional Gaussian fit and deconvolved from the beam) and a velocity of ∼450 km s⁻¹. From these values, the resulting flow timescale is ${t}_{\mathrm{out},\mathrm{mol}}\approx 4.2\,\mathrm{kpc}/450\,\mathrm{km}\,{{\rm{s}}}^{-1}\approx 9\times {10}^{6}\,\mathrm{yr}$, which yields a mass outflow rate of

$$\begin{eqnarray}\begin{array}{rcl}{\dot{M}}_{\mathrm{out},\mathrm{mol}}^{\mathrm{avg}} & \approx & 256\,{M}_{\odot }\,{\mathrm{yr}}^{-1}\times \left(\displaystyle \frac{{\alpha }_{\mathrm{CO},\mathrm{out}}}{{\alpha }_{\mathrm{CO},\mathrm{ULIRG}}}\right)\\ & & \times \ \left(\displaystyle \frac{{v}_{\mathrm{out}}}{450\,\mathrm{km}\,{{\rm{s}}}^{-1}}\right)\times \left(\displaystyle \frac{4.2\,\mathrm{kpc}}{{R}_{\mathrm{out}}}\right).\end{array}\end{eqnarray} \tag{ 3 }$$

This translates into an outflow depletion time (${\tau }_{\mathrm{dep},\mathrm{out}}\,={M}_{\mathrm{mol},\mathrm{disk}}/{\dot{M}}_{\mathrm{out},\mathrm{mol}}^{\mathrm{avg}}$) of

$$\begin{eqnarray}\begin{array}{rcl}{t}_{\mathrm{dep},\mathrm{out}} & \approx & 0.47\,\mathrm{Gyr}\times \left(\displaystyle \frac{{\alpha }_{\mathrm{CO},\mathrm{disk}}}{{\alpha }_{\mathrm{CO},{\rm{T}}18}}\right)\times \left(\displaystyle \frac{{\alpha }_{\mathrm{CO},\mathrm{ULIRG}}}{{\alpha }_{\mathrm{CO},\mathrm{out}}}\right)\\ & & \times \ \left(\displaystyle \frac{450\,\mathrm{km}\,{{\rm{s}}}^{-1}}{{v}_{\mathrm{out}}}\right)\times \left(\displaystyle \frac{{R}_{\mathrm{out}}}{4.2\,\mathrm{kpc}}\right).\end{array}\end{eqnarray} \tag{ 4 }$$

Relative to the star formation depletion timescale (${t}_{\mathrm{dep},\mathrm{SF}}\,={M}_{\mathrm{mol},\mathrm{disk}}/\mathrm{SFR}$), ${t}_{\mathrm{dep},\mathrm{out}}$ is a factor of ∼2 shorter or longer, depending on whether we consider the UV-corrected or UV-plus-infrared SFR measurement of zC400528, respectively.

So far, we have ignored the effect that the inclination of the outflow has on the determination of the mass outflow rate and, consequently, the outflow depletion timescale. The reason is that we do not have enough information to constrain the geometry of the outflow. If θ corresponds to the inclination angle of the outflow with respect to the line of sight, then the deprojected outflow size is ${R}_{\mathrm{out}}={R}_{\mathrm{out},\mathrm{proj}}/\sin (\theta )$, and the deprojected outflow velocity is ${v}_{\mathrm{out}}={v}_{\mathrm{out},\mathrm{proj}}/\cos (\theta )$. This implies that the inclination-corrected mass outflow rate is proportional to tan(θ). The inclination of zC400528 is i ≈ 37° (sin(i) = 0.61; Förster Schreiber et al. 2018a). If we assume that the outflow is perpendicular to the disk, then the deprojection effects in velocity and radius almost cancel out, and the correction is small and of the order of ∼0.75.

Figure 9 compares the outflow and star formation depletion timescales of galaxies as a function of their separation from the MS (after removing the dependence with z using the Tacconi et al. 2018 scaling relations). The plot includes zC400528 and a compilation of local (U)LIRG and Seyfert galaxies (Fiore et al. 2017 and reference therein) and the z = 1.59 QSO XID2028 (Brusa et al. 2018). We observe that local systems where the depletion of the molecular gas is dominated by star formation lie at least ∼1.5 dex above the MS. On the other hand, galaxies where the depletion of molecular gas is dominated by outflows are located within ∼1 dex of the MS, with the exception of the above outlier, Mrk 231 (Feruglio et al. 2010).

Zoom In Zoom Out Reset image size

Similar to the local ULIRG/QSO Mrk 231 and the z ≈ 1.5 QSO XID2028, zC400528 is exhausting its nuclear molecular gas reservoir via the outflow at a comparable rate, on average, at which stars are forming (keeping in mind the uncertainties and assumptions involved in the calculation). Note, however, the fundamental differences between these systems. While zC400528 represents a typical massive galaxy at z ∼ 2, Mrk 231 lies a factor of 100× above the MS, and QSO XID2028, although in the MS, has a molecular gas depletion timescale of only ∼40–75 Myr (Brusa et al. 2018). This is at least an order of magnitude lower than the typical depletion time for MS galaxies at a similar redshift (Tacconi et al. 2018). We discuss in more detail the effect that the powerful molecular outflow detected in zC400528 may have on quenching its star formation in Section 6.3.

The molecular outflow in zC400528 could be driven by the AGN, the nuclear star formation activity, or, most likely, a combination of the two. If the dominant power source is the AGN, then the outflow will expand, conserving energy or momentum depending on how efficiently the inner quasi-relativistic wind bubble (v ∼ 0.1c) communicates its energy to the surrounding shocked gas.

In the energy-conserving case, the gas expands adiabatically, and most of the kinetic energy from the AGN wind is transmitted to the outflow. This results in large-scale (∼kpc size) outflows with kinetic energies up to ∼2.5% of the AGN radiative power (assuming that half of the thermal energy goes into the bulk motion of the ISM) and momentum rates boosted to ∼5–20 times L_AGN/c (e.g., Faucher-Giguère & Quataert 2012; Zubovas & King 2012). In the momentum-driven case, the shocked expanding gas rapidly cools, and only the ram pressure of the wind is communicated to the outflow. This results in outflows that are confined to the nuclear region and have kinetic energies only up to ∼0.1% of the AGN luminosity and momentum rates of the order ∼L_AGN/c (e.g., King 2010; King & Pounds 2015).

If the dominant power source for the outflow is star formation activity, then the combined momentum deposition of supernovae and stellar winds can supply a maximum momentum of²⁴ ∼2 × L_SF/c (where L_SF is the luminosity of the starburst) and inject a maximum mechanical energy of ∼0.2% × L_SF in the energy-conserving case (e.g., Murray et al. 2005; Veilleux et al. 2005).

By calculating the momentum flux (${\dot{P}}_{\mathrm{out}}={\dot{M}}_{\mathrm{out}}\times {v}_{\mathrm{out}}$) and energy flux (${\dot{E}}_{\mathrm{out}}=\tfrac{1}{2}{\dot{M}}_{\mathrm{out}}\times {v}_{\mathrm{out}}^{2}$) of the molecular outflow in zC400528, we can explore if the AGN and/or the nuclear star formation activity can provide enough power to blow the gas and determine if the gas flow is energy- or momentum-conserving.

From the mass outflow rate ${\dot{M}}_{\mathrm{out}}^{\mathrm{avg}}$ calculated in Section 4.2.4, the outflow momentum flux in zC400528 is

$$\begin{eqnarray}\begin{array}{rcl}{\dot{P}}_{\mathrm{out},\mathrm{mol}} & \approx & 7.2\times {10}^{35}\,\mathrm{dynes}\times \left(\displaystyle \frac{{\alpha }_{\mathrm{CO},\mathrm{out}}}{{\alpha }_{\mathrm{CO},\mathrm{ULIRG}}}\right)\\ & & \times \ {\left(\displaystyle \frac{{v}_{\mathrm{out}}}{450\mathrm{km}{{\rm{s}}}^{-1}}\right)}^{2}\times \left(\displaystyle \frac{4.2\,\mathrm{kpc}}{{R}_{\mathrm{out}}}\right)\end{array},\end{eqnarray} \tag{ 5 }$$

and the outflow kinetic energy flux is

$$\begin{eqnarray}\begin{array}{rcl}{\dot{E}}_{\mathrm{out},\mathrm{mol}}\, & \approx & \,1.6\times {10}^{43}\,\mathrm{erg}\,{{\rm{s}}}^{-1}\times \left(\displaystyle \frac{{\alpha }_{\mathrm{CO},\mathrm{out}}}{{\alpha }_{\mathrm{CO},\mathrm{ULIRG}}}\right)\\ & & \times \ {\left(\displaystyle \frac{{v}_{\mathrm{out}}}{450\mathrm{km}{{\rm{s}}}^{-1}}\right)}^{3}\times \left(\displaystyle \frac{4.2\,\mathrm{kpc}}{{R}_{\mathrm{out}}}\right).\end{array}\end{eqnarray} \tag{ 6 }$$

We can now compare these quantities with the maximum estimated momentum and energy rates that the AGN and the star formation activity in zC400528 can provide. For this, we calculated the AGN bolometric luminosity (${L}_{\mathrm{AGN},\mathrm{bol}}$) using the rest-frame 7 μm dust continuum emission (see the Appendix for details) and the star formation luminosity based on the SFR values following L_SF (L_⊙) ≈ 10¹⁰ × SFR (M_⊙ yr⁻¹) (e.g., Kennicutt & Evans 2012).

Figure 10 shows the energetics of the molecular outflow in zC400528 compared to results from local (U)LIRGs (González-Alfonso et al. 2017). The left and right panels depict the cases where we consider that the outflows are driven by AGN and star formation activity, respectively. In the AGN-driven case, the momentum boost in zC400528 is ${\dot{P}}_{\mathrm{out},\mathrm{mol}}/({L}_{\mathrm{AGN},\mathrm{bol}}/c)\approx 7$. This value is at the low end of the distribution of momentum boost found in local (U)LIRGs (${\dot{P}}_{\mathrm{out}}\sim 5\mbox{--}30\,{L}_{\mathrm{AGN},\mathrm{bol}}/c;$ González-Alfonso et al. 2017) but is high enough to consider the outflow as energy-conserving. This is consistent with the observed ~kpc extent of the molecular outflow, as momentum-conserving outflows are expected to be confined to a small region (King & Pounds 2015).

Zoom In Zoom Out Reset image size

The outflow energy flux in zC400528 is ${\dot{E}}_{\mathrm{out},\mathrm{mol}}\,\approx 0.5 \% {L}_{\mathrm{AGN},\mathrm{bol}}$. This is a factor of ∼5 below the maximum power that can be supplied by the AGN (shaded region) and could be the result of a low coupling efficiency of the wind with the ISM. We conclude that the AGN activity in zC400528 is sufficient to explain the observed momentum boost and kinetic power of the molecular outflow.

In the case where we consider the star formation–driven scenario (right panels in Figure 10), we observe that star formation activity can only provide sufficient energy to power the outflow if the star formation rate of zC400528 is close to the SFR_UV+IR value. Remember that the latter is most likely an upper limit to the SFR, as it may include contributions from AGN activity and heating of the dust by evolved stars (see Section 2).

In summary, the most likely scenario is that the molecular outflow in zC400528 is powered by a combination of nuclear star formation and AGN activity, with a dominant contribution from the latter. The molecular outflow mass and energetics measured in zC400528 for different assumptions on α_CO can be found in Table 2.

Table 2.  Molecular and Ionized Gas Outflow Masses and Energetics

Phase	Mass	${\dot{M}}_{\mathrm{out}}$	${\dot{P}}_{\mathrm{out}}$	${\dot{E}}_{\mathrm{out}}$
	109 M⊙	M⊙ yr−1	1035 dynes	1043 erg s−1
Moleculara	3.36 (1.42–18.06)	256 (108.8–1376.0)	7.2 (3.1–38.7)	1.6 (0.68–8.6)
Ionizedb	0.22 (0.07–0.66)	53 (15.9–159.0)	2.7 (0.81–8.1)	1.10 (0.33–3.30)

Notes.

^aFor molecular gas quantities, we assume an α_CO,ULIRG conversion factor. The values in parentheses correspond to the values we would obtain if we apply an α_CO,thin or α_CO,MW conversion factor, respectively. ^bFor ionized gas quantities, we assume n_e = 300 cm⁻³ following Förster Schreiber et al. (2018b). The values in parentheses correspond to the values we would obtain if we assume n_e = 10³ or 100 cm⁻³, respectively.

Download table as:  ASCIITypeset image

Figure 11 shows the spatial distribution of the molecular disk (background), the Hα blue and red broad (outflow) component (green), and the molecular wind (red). The bulk of the projected Hα broad emission extends ∼4 kpc north from the nuclear region. Beyond this point, the only two extranuclear ionized outflow components detected, A and B, are cospatial with the molecular outflow gas, and—at least in the case of region A, as Figure 12 shows—the molecular and ionized gas velocities are similar. The velocity dispersions extracted from the Gaussian fit to the spectra of regions A and B are comparable with those observed in the spatially resolved molecular outflows of local starbursts and (U)LIRGs (see Section 4.2.3; Contursi et al. 2013; Leroy et al. 2015; Pereira-Santaella et al. 2016) and appear to be higher than those measured in the tidal tails of interacting LIRGs (σ_CO ∼ 15 km s⁻¹; Pereira-Santaella et al. 2018) or in the Antennae (σ_{H i} ∼ 5−15 km s⁻¹; Hibbard et al. 2001). With the current data, however, it is difficult to rule out the possibility that regions A and B are part of a tidal tail or a second merger galaxy.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Similar to zC400528, Cresci et al. (2015) and Brusa et al. (2018) found that the ionized and molecular phases of the outflow in the QSO XID2028 (z = 1.59) extend for about ∼8 kpc over the same spatial region. This is consistent with the scenario where AGNs with high-duty cycles on massive galaxies can drive ionized and molecular outflows over scales of tens of kpc (see Section 6.2).

The spatial distribution of the outflow in zC400528 suggests that the ionized gas is outflowing inclined (∼60°) with respect to the major kinematic axis, likely along the path of least resistance outside the galaxy plane. There are two main molecular outflow components in the inner ∼4 kpc region: one that extends alongside the ionized wind (from the center to region A) and one that is aligned with the major kinematic axis and could therefore be equatorial. This picture resembles the structure of the AGN-driven outflow in NGC 1068 (Tecza et al.2001; Cecil et al. 2002). There, the ionized cone is inclined ∼40° with respect to the galaxy disk, and the molecular outflow extends preferentially along the galaxy plane launched as a result of the ionization cone sweeping the molecular gas in the inner disk (García-Burillo et al. 2014). The fact that the outflow in zC400528 appears to be not perpendicular to the disk is a common feature observed in AGN-driven outflows. Star formation–driven outflows, on the other hand, tend to align with the minor kinematic axis of the disk (e.g., Leroy et al. 2015; Pereira-Santaella et al. 2018).

The configuration of the ionized and molecular phases of the outflow in zC400528 is also in qualitative agreement with the modeling work of Zubovas & Nayakshin (2014). They find that AGN-driven outflows that start spherical quickly develop a bipolar morphology that expands faster and further in the polar direction (the direction of least resistance) than in the equatorial direction. After ∼10 Myr (the flow timescale in zC400528), several cold dense clumps have formed embedded in the hot gas and are moving upward, while in the plane of the galaxy, the cold gas, squeezed by the expanding bubbles, is being pushed outward with a mean velocity of ∼400 km s⁻¹.

The ionized mass in the outflow of zC400528 is ${M}_{\mathrm{out},\mathrm{ion}}\,=2.2\times (300\,{\mathrm{cm}}^{-3}/{n}_{{\rm{e}}})\times {10}^{8}\,{M}_{\odot }$ (Genzel et al. 2014), where n_e corresponds to the electron density of the ionized gas in the outflow. Recent results by Förster Schreiber et al. (2018b) find that the mean electron density in the outflow gas of z ∼ 2 galaxies is n_e ∼ 350 cm⁻³.

For an outflow size of ${R}_{\mathrm{out},\mathrm{ion}}=3\,\mathrm{kpc}$ (Förster Schreiber et al. 2014) and an outflow velocity of ${v}_{\mathrm{out},\mathrm{ion}}=802$ km s⁻¹ (Genzel et al. 2014), the ionized mass outflow rate is ${\dot{M}}_{\mathrm{ion},\mathrm{out}}=53$ M_⊙ yr⁻¹. Based on the same assumptions, the ionized momentum and energy outflow rates are ${\dot{P}}_{\mathrm{out},\mathrm{ion}}=2.7\,\times {10}^{35}$ dynes and ${\dot{E}}_{\mathrm{out},\mathrm{ion}}=1.1\times {10}^{43}$ erg s⁻¹, respectively. Figure 13 summarizes the mass and mass-loss rates measured in the molecular and ionized outflow gas in zC400528. The molecular phase dominates the total observed budget of both M_out and ${\dot{M}}_{\mathrm{out}}$. Note, however, that there are gas phases in the outflow that remain unobserved and could change the overall balance, including very hot gas that could contribute to the ionized outflow mass (e.g., Veilleux et al. 2005).

Zoom In Zoom Out Reset image size

A summary of the ionized outflow mass and energetics measured in zC400528 for different assumptions of n_e and the comparison to the quantities measured in the molecular phase can be found in Table 2.

Another quantity of interest in the study of galactic outflows is the mass-loading factor (η_phase), which is defined as the ratio between the mass outflow rate and the SFR. Figure 14 shows the molecular and ionized mass-loading factors as a function of the separation from the MS (removing the z dependence using the Tacconi et al. 2018 scaling relations) for nearby starbursts and AGN galaxies (Heckman et al. 2015; Fiore et al. 2017) and massive (log(M_*/M_⊙) ≥ 10.9) star-forming galaxies at z ∼ 1−3 (this work; Genzel et al. 2014). The ionized mass-loading factors of these high-z galaxies²⁵ are found to be comparable to those observed in local MS outliers. The ionized and mass-loading factors in zC400528 are also similar to those found in the z ∼ 1.5 QSOs XID2028 (η_mol ∼ 1.3 and η_ion ≳ 1; Cresci et al. 2015; Brusa et al. 2018) and 3C 298 (η_mol ∼ 2.5 and η_ion ∼ 1.6; Vayner et al. 2017).

Zoom In Zoom Out Reset image size

The molecular mass outflow rate of zC400528 is comparable to some of the powerful molecular outflows observed in local starbursts and Seyfert galaxies (see also Section 4.2.4). In addition, the molecular mass-loading factor is ∼4 times higher than its ionized counterpart. Note, however, that these results depend on assumptions of the properties of the outflow gas that are not yet fully constrained (e.g., the CO-to-H₂ conversion factor, electron density, geometry, etc.). The same caveat applies to all of the other measurements shown in the figure.

Understanding whether baryons ejected by winds escape the galaxy or rain back onto the disk is key to reproducing the observed galaxy stellar masses and the metal enrichment of the intergalactic medium (IGM; e.g., Oppenheimer et al. 2010; Muratov et al. 2015). Simulations show that about half of the ejected outflow mass across all galaxy masses is later reaccreted (Übler et al. 2014; Christensen et al. 2016), and that in massive z ∼ 2 galaxies, about ∼30% of the stellar mass forms from gas contributed by wind recycling (Anglés-Alcázar et al. 2017).

To determine if the molecular gas in the outflow of zC400528 can permanently escape the galaxy, we first estimate the escape velocity (v_esc) from the host. For this, we follow Rupke et al. (2002) and use a simple gravitational model based on a truncated (r < r_max) isothermal sphere, so ${v}_{\mathrm{esc}}(r)\,=\sqrt{2}{v}_{\mathrm{circ}}{[1+\mathrm{ln}({r}_{\max }/r)]}^{1/2}$.²⁶ Assuming that the dark matter halo extends to r_max ∼ 100 kpc, and using a circular velocity for zC400528 of v_circ = 344 km s⁻¹ (Förster Schreiber et al. 2018a), the escape velocity at r = 10 kpc is ${v}_{\mathrm{esc},10\mathrm{kpc}}\sim 880$ km s⁻¹. The highest-velocity molecular outflow material we detect in zC400528 has a velocity of v_out ≈ 500 km s⁻¹. If we take inclination effects into account (see Section 4.2.4), the deprojected outflow velocity could increase to v_out ∼ 650 km s⁻¹, which is still lower than the escape velocity. This suggests that the bulk of the expelled molecular gas mass will be reaccreted back onto the galaxy on timescales—according to simulations—no shorter than ∼1–3 Gyr (Oppenheimer et al. 2010; Übler et al. 2014; Christensen et al. 2016; Brennan et al. 2018).

In the case of the ionized outflowing gas, the escape velocity at 3 kpc—a distance about the size of the ionized outflow—is ${v}_{\mathrm{esc},3\mathrm{kpc}}\sim 1000$ km s⁻¹. The highest-velocity gas in the ionized wind can reach velocities up to ∼800 km s⁻¹ (Genzel et al. 2014), which, if we correct by inclination, can increase to ∼1000 km s⁻¹ and potentially escape the galaxy.

The very small fraction of the molecular and ionized outflow gas that can escape the gravitational field of zC400528 is consistent with the low escape fractions (≲20%) measured in neutral, ionized, and molecular gas winds in local (U)LIRGs (e.g., Rupke et al. 2005; Arribas et al. 2014; Fluetsch et al. 2019; Pereira-Santaella et al. 2018).

Observational and theoretical arguments suggest that the AGN lifetime of ∼10⁷–10⁹ yr is, in reality, a succession of hundreds or thousands of short (∼10⁵ yr) phases of supermassive black hole growth (e.g., Schawinski et al. 2015)—a process called AGN flickering. These periods of episodic AGN activity inflate large-scale outflows that can continue to expand even after the AGN finally switches off as a result of clearing the gas in the central region.

Zubovas & King (2016) used an analytical model to follow the evolution of the molecular and ionized phases in an outflow driven by an AGN that "flickers" on timescales of 5 × 10⁴ yr. They found that the oscillations caused by the AGN flickering smooth out relatively fast (∼3 Myr), so from an observational standpoint, it would appear that the outflow is driven by an AGN with constant luminosity. For a massive galaxy similar to zC400528 (${M}_{{\rm{h}}}\approx {10}^{12.5-13}\,{M}_{\odot }$, based on the relation between halo and stellar mass in Moster et al. 2013), the model predicts that in the outflow timescale of zC400528 (∼10⁷ yr), the wind should have expanded between ∼4 and 8 kpc with expansion velocities at that distance of ∼300 and 700 km s⁻¹ for AGN duty cycles of f_AGN = 20% and 100%, respectively. The observed extent (∼8 kpc) and velocity (∼500 km s⁻¹) in the molecular outflow of zC400528 is consistent with these model results and favors the scenario with a high AGN duty cycle, in agreement with the results of Genzel et al. (2014) and Förster Schreiber et al. (2018b).

There is growing evidence that suggests that when massive star-forming galaxies at z ≈ 2 reach central stellar surface densities similar to those observed in massive ellipticals at z ≈ 0 (Σ_⋆ ≳ 10¹⁰ M_⊙ kpc⁻² inside the inner ∼kpc region), an effective quenching mechanism must act on relatively short timescales to shut down the star formation activity. Feedback from AGNs, prevalent in massive z ∼ 2 galaxies (e.g., Genzel et al. 2014; Förster Schreiber et al. 2018b), is one of the obvious quenching candidates. In the specific case of zC400528, Tacchella et al. (2015a) suggested that the star formation activity in the inner ∼3 kpc region should be fully quenched on a ≲0.5 Gyr timescale.

To evaluate if zC400528 is on the path of significantly suppressing its star formation activity, we compare the timescales of depletion (t_dep) and replenishment (t_rep) of cold gas. For a galaxy to experience a long-term episode of quenching, we should crudely expect that t_rep/t_dep ≳ 10 (e.g., Tacchella et al. 2016).

From the analysis of the molecular outflow (Section 4.2.4) and star formation activity (Tacchella et al. 2018) in zC400528, we know that the outflow and star formation depletion times are ${t}_{\mathrm{dep},\mathrm{out}}\sim {t}_{\mathrm{dep},\mathrm{SF}}\sim 0.5\,\mathrm{Gyr}$. These timescales can be a factor of ∼3 shorter if we only consider the molecular gas inside the inner ∼4 kpc region. Regarding the replenishment of gas, according to the prescription in Bouché et al. (2010; and ignoring the supply of gas provided by potential merger episodes; see also Lilly et al. 2013), a massive galaxy like zC400528 at z ∼ 2 should accrete ${\dot{M}}_{\mathrm{gas},\mathrm{in}}\sim 240$ M_⊙ yr⁻¹ of fresh gas.²⁷ At this rate, and in the absence of strong preventive feedback, the inflow of fresh and recycled gas into the system should be enough to roughly maintain the level of star formation activity (i.e., ${t}_{\mathrm{rep}}\sim 1\mbox{--}3\times {t}_{\mathrm{dep}}$). This is consistent with the results from the cosmological zoom-in simulations by Brennan et al. (2018) that follow the evolution of massive, gas-rich galaxies (M_⋆ ∼ 10¹¹ M_⊙ and ${M}_{\mathrm{cold}\mbox{--}\mathrm{gas}}/{M}_{\star }\sim 0.5\mbox{--}1$ at z ≈ 2) from z ≈ 4 to z = 0. For the simulated galaxies that only include stellar and supernova feedback, Brennan et al. (2018) found that over the galaxy's lifetime, there is an overall balance between the inflow of gas and the level of star formation activity.

If AGN activity is considered,²⁸ Brennan et al. (2018) showed that this equilibrium drastically changes. As AGN feedback starts to kick in at z ∼ 2, the inflow rate of fresh gas into the simulated galaxies is suppressed by ≳1 dex. In addition, the timescale for reaccretion of gas ejected by the outflow increases from ∼1 to ∼2–3 Gyr. If the baryon cycle in zC400528 is affected by the AGN feedback in a similar way to that described in Brennan et al. (2018), then the gas depletion-to-replenishment timescale ratio in zC400528 should reach the critical level of t_rep/t_dep ∼ 10, where long-term quenching of the star formation activity is expected.

We can further explore the quenching scenario with a simple back-of-the-envelope calculation that yields a rough estimate of the rate at which a galaxy is moving up or down the MS based on its balance term, which is defined as $B=\mathrm{inflow}\,\mathrm{rate}/(\mathrm{SFR}+\mathrm{outflow}\,\mathrm{rate})$. Keeping in mind the many uncertainties associated with the inflow, outflow, and star formation rate measurements, the balance of gas input and drainage in zC400528 is

$$\begin{eqnarray*}&&B=\displaystyle \frac{240\,{M}_{\odot }\,{\mathrm{yr}}^{-1}}{(352+256)\,{M}_{\odot }\,{\mathrm{yr}}^{-1}}\approx 0.4.\end{eqnarray*}$$

For simplicity, for the SFR, we have used the average between the SFR_UV,corr and SFR_UV+IR values. From the scaling relation described in Tacchella et al. (2016) between the balance term B (in the central 5 kpc) and the rate of change of distance from the MS, ${\dot{{\rm{\Delta }}}}_{\mathrm{MS}}$, we speculate that zC400528 is moving down from the MS at a rate of ${\dot{{\rm{\Delta }}}}_{\mathrm{MS}}\sim 2$ dex Gyr⁻¹ (note that the scaling relations in Tacchella et al. 2016 take into account the evolution of the MS of star-forming galaxies with redshift and the reaccretion of gas ejected by outflow episodes). At this rate, it should only take zC400538 ∼1 Gyr to transition from the MS to the regime of passive quenched galaxies ∼2 dex below the MS. This is consistent with the expectations from inside-out quenching that claim that massive star-forming galaxies such as zC400528 should be fully quenched by z ∼ 1 (∼3 Gyr later than the z ∼ 2 epoch).