A SPECTACULAR OUTFLOW IN AN OBSCURED QUASAR

We first introduce the basic observational components of SDSS J1356+1026 and its immediate environment, as seen in Figure 2. We observe two nuclei, the northern (N) nucleus and the southern (S) nucleus, separated by ∼2.5 kpc (11) on the sky. We find distinct stellar continuum and ionized gas emission from each nucleus in SDSS J1356+1026. Although the continua are spatially resolved, the southern source is faint enough that we are unable to determine the velocity separation between the nuclei from the stellar spectra. To the north is a companion galaxy located 62 kpc (in projection on the plane of the sky) from the N nucleus. In velocity space, the stellar features of the companion galaxy are blueshifted by ∼250 km s⁻¹ with respect to the stellar features of the N nucleus.

Zoom In Zoom Out Reset image size

In addition to stellar continua, there is strong, spatially resolved [O iii]λλ4959, 5007 line emission, as well as Hβ and other high-ionization forbidden lines. We focus here predominantly on the [O iii]. A ∼51 kpc long [O iii] emission-line feature (denoted "TT" for "tidal tail"; Figure 2) extends between SDSS J1356+1026 and the northern companion galaxy and somewhat beyond it. No associated stellar continuum is detected for this feature, nor is it seen in the broadband image. Additional likely tidal features are seen in the broadband image, such as the low surface brightness feature ∼10'' (22 kpc) SE of the nuclei. This feature is tentatively detected in [O iii] emission in the two-dimensional spectrum (see feature Bv2 in Figure 3).

Zoom In Zoom Out Reset image size

The focus of this paper is on the well-organized kinematic feature to the south of SDSS J1356+1026, "the bubble" (see Figure 3). The feature appears as a ring in the two-dimensional spectrum, with a spatial extent of ∼12 kpc and apparent velocity extent of ∼460 km s⁻¹. The central velocity of the bubble at its base is close to that of the N nucleus, ∼8 kpc (projected) from its center. Equidistant to the north of the N nucleus, we find three clumps spanning a comparable velocity range. The bubble and the clumps are pure emission-line features, with no stellar continuum detected in the spectra. Emerging at a 45° angle to the bubble are two "wisps" of emission associated with the same outflow.

We subtract the stellar continuum and co-add the spectrum in velocity space to evaluate the [O iii]/Hβ line ratio as a function of position along the slit (Figure 4). The ratio is constant along the slit, with [O iii]/Hβ ∼10, suggesting that the gas is photoionized by an accreting BH. This line ratio is constant for a remarkable number of components, including both the S and N nuclei, bubble, and the northern clumps. We have not corrected these observations for extinction, but note that assuming R_V = 3.1 and the reddening law of Cardelli et al. (1989), there is only an 8% difference in the extinction of Hβ and [O iii]. Thus, if we invoke (for instance) higher levels of extinction in the galaxy center, we would still only expect the [O iii]/Hβ ratio to decrease by <10% in the central region.

Zoom In Zoom Out Reset image size

As argued in Paper II, the most plausible explanation for the constant line ratios is that both nuclei host obscured quasars with similar luminosities. Another possible scenario is that only one quasar is present and it photoionizes the other nucleus; however, we find that the fine-tuning required to satisfy the available observables makes this hypothesis unlikely (Paper II). Future observations (in X-rays with Chandra or in the radio with ∼1'' resolution) will help distinguish between these two possibilities.

Recently, Fu et al. (2012) presented integral field observations of the 6'' × 6'' field around SDSS J1356+1026 (e.g., largely excluding the bubble itself). It is not obvious from their [O iii] map that they see emission-line sources associated with both nucleus N and nucleus S. They conclude that SDSS J1356+1026 harbors only a single active BH in nucleus N, and that the multiple kinematic components are due to large-scale motions of the gas. It is possible that our slit did not intercept the peak emission from nucleus N, thus explaining why the continuum strengths of nucleus S and nucleus N are roughly comparable in our spectrum, even though source N is considerably more luminous (Paper I). We still believe we are seeing a distinct [O iii] peak associated with nucleus S, but cannot perform more detailed comparisons at this time.

Taking the spectrum of the bubble alone and subtracting the continuum, we use the velocity profile of the [O iii] line to fit all other emission features in the spectrum and thus calculate line ratios without modeling the complex velocity structure of the lines. In particular, we find a ratio of [O iii]λλ4959,5007/[O iii]λ4363 = 90.5 ± 7.2 and therefore derive an electron temperature T_e = 13, 500 ± 500 K (Osterbrock & Ferland 2006). Furthermore, we detect the [Ar iv]λλ4711,4730 doublet with a ratio of 1.32 ± 0.37, consistent with its low-density limit of 1.4 (Krueger et al. 1970). Because the critical density for this transition is fairly high, even if we take our measurement as a lower limit on the ratio, it places only a weak limit on the electron density (n_e ≲ 850 cm⁻³). The ratios of lines of the Balmer series are consistent with no reddening.

The Magellan spectrum does not cover the density diagnostics [O ii]λλ3726, 3729 and [S ii]λλ6716, 6731, but both of these features are covered by the SDSS spectrum taken in the 3'' circular aperture centered on the nuclei. The components of the [O ii] doublet are separated only by 240 km s⁻¹ and are far too blended to determine their ratio, given that $\mathrm{FWHM}_\mathrm{[O\,\mathsc{iii}]}=652$ km s⁻¹ in this aperture. The [S ii] doublet has an acceptable velocity separation, but the velocity structure of the [O iii] emission line does not provide a good fit to this feature, so we cannot determine the line ratio in the same model-independent way as for the Magellan spectrum. We measure the peak ratio of the two components of the doublet to be 1.20, which is likely to be a lower limit on the actual ratio because of blending. Taking [S ii]6716/6731 >1.2 yields n_e < 250 cm⁻³.

The object is strongly linearly polarized, with the polarization fraction reaching 8% at shortest observed wavelengths (Figure 5). These observations are consistent with the hypothesis that SDSS J1356+1026 is powered by a luminous quasar whose emission is largely blocked along the line of sight to the observer, but escapes along some other directions, and then scatters off the interstellar material within the host galaxy (Smith et al. 2003; Zakamska et al. 2005) and reaches the observer. In particular, the emission from the quasar escapes predominantly in the N–S direction in the plane of the sky, as indicated by the ∼90° P.A. of the plane of polarization. Furthermore, the wavelength dependence of the polarization fraction can be attributed to the dilution of the scattered quasar light by the unpolarized emission from the extended narrow-line region and starlight from the host galaxy. Indeed, the dilution of the polarized flux by starlight increases to the red, as expected from a red stellar population, and an abrupt minimum in P is seen around [O iii] (Figure 5). The generally higher values of P for the N–S slit orientation of the 2011 May 27 observation are likely caused by the fact that the spectral extraction aperture runs along the axis of the presumed scattering cone, thereby including a larger fraction of the scattering regions than the earlier measurement made with the Kuiper telescope. Other small discrepancies between the two spectropolarimetric observations are likely caused by differing positions of the object within the slit and by the difference in the slit direction, suggesting that polarized structures within the source are partially resolved.

Zoom In Zoom Out Reset image size

We propose that SDSS J1356+1026 is powering a large-scale outflow, which is entraining the warm ionized gas that we see radiating in [O iii]. Here we present the observed kinematics along each slit position that we will use to estimate the energetics of the outflow.

Starting with the bubble along the first (N–S) slit position PA1, we take two-pixel-wide extractions along the spatial direction starting from nucleus S and working our way to the bottom of the bubble (all within the bubble as demarcated in Figure 3). We use Gaussian fits to identify the primary velocity components at each spatial position. These velocities are plotted in Figure 6, ignoring some weak components. This velocity pattern, increasing symmetrically to a projected velocity of ∼250 km s⁻¹, is the signature of an expanding sphere of gas with limb brightening.

Zoom In Zoom Out Reset image size

In addition to the bubble, there are three clumps to the north of nucleus N that we see in slit position PA1. We note that (1) their central velocity is similar to that of the bubble and (2) the maximum velocity extent across the clumps is comparable to that seen in the bubble (see Figure 3). Therefore, we suggest that these clumps are the northern side of the same outflow. The bubble is systematically blueshifted by ∼215 km s⁻¹ relative to the clumps; therefore, the axis of symmetry of the outflow is likely not quite in the plane of the sky but is tipped so that the bubble is expanding slightly toward us and the clumps are from the component moving slightly away from us (see Figure 6). The northern and southern components of the outflow are discontinuous; we suggest three possible corresponding geometries. One possibility is that we are seeing two unrelated components, but given the similarity of the systemic velocities and the velocity spread we find this unlikely. Another possibility is that our slit did not pass through the center of the outflow, but instead was 1''–2'' (2.3–4.6 kpc) offset (see arguments above suggesting that the PA1 slit was not centered on nucleus N). In this case, an outflow that forms a figure "8," with the center taken out, could appear as a pair of disconnected features. Yet another possibility (that we favor on the basis of kinematic model presented in the next section) is that the merger provides plentiful gas in the inner several kpc of the system, pinching the flow, which finally breaks free of this dense gas on either side, as shown in Figure 5.

The signal-to-noise ratio along PA2 is not very high, and thus we do not measure the velocities as a function of position in this case. The two linear structures to the south of nucleus N, Bv1 and Bv2 (Figure 3), show a clear trend of increasing line-of-sight velocity with projected distance. These wisps represent gas that is escaping at a 45° angle to the gas in the bubble.

We see outflowing gas to the north, south, and southeast of the nuclei of SDSS J1356+1026. The emission-line gas is clearly expanding away from the nuclei. Is it being actively driven from the nuclei and accelerating away, or is it decelerating as it slams into the interstellar or intergalactic medium? The linearly expanding structures along PA2 (wisps Bv1 and Bv2) are the telltale signs of an accelerating outflow (Veilleux et al. 1994). Both a conical outflow with v_3D∝r_3D and a spherical outflow blown from one side with v_3D∝r²_3D will produce v∝r when projected onto the line of sight (v) and on the plane of the sky (r). The cap of the bubble along the southernmost edge is produced by a shell where the outflow interacts with gas outside the galaxy, either tidal debris or the intergalactic medium. Because of projection effects, the cap is brightest in the plane of the sky, where its line-of-sight velocity is zero, so we do not know whether the shell is still accelerating or whether it has already started to decelerate.

If we assume that there is a single accelerating agent causing all of the outflowing components, then a credible picture of the geometry can be built. Due to interactions with the ambient medium of varying density, the outflow is complicated enough that we cannot uniquely determine its geometry from two slit positions, but we are able to bracket the outflow velocities using a reasonable range of models. Our primary goal is to explain the symmetric velocity structure along the bubble in a way that is also consistent with the clumps to the north and the linearly increasing velocity structures ("wisps") to the southeast along PA2.

We first consider an accelerating (v_3D∝r_3D) biconical outflow, which is a natural geometry for an obscured quasar and easily explains the wisps. This model fails to reproduce the concave velocity structure (the southern "cap") observed in the bubble (Figure 6). Another straightforward model is a spherical outflow centered on nucleus N. If only the shell is emitting, it is then difficult to reproduce v∝r at small radii in the bubble and the wisps. Both the cap and the accelerating features are naturally reproduced in the framework of a bubble blown from its side, not from its center, perhaps due to the high density of matter near the energy injection point (Veilleux et al. 1994). This model is reminiscent of outflows in starburst galaxies in which propagation in the plane of the galaxy is strongly suppressed and the outflows escape predominantly perpendicular to the plane (Veilleux et al. 2005). Note that if the outflow axis is inclined toward the plane of the sky, then the velocity offset between the clumps and the bubble (relative to the maximum span in the velocity of the bubble) leads to an estimate of the inclination of the axis of ∼20°, which could in principle boost the space velocities relative to those assumed here.

A simple version of this model is shown in Figure 6. We assume that the flow is spherical, with the origin on the side of the sphere and the axis in the plane of the sky, that only the shell is emitting, and that v_3D∝rⁿ_3D where n = 1, 1.5, 2. Then we project the system onto the sky and the velocities along the line of sight and plot the model curves along with the observations of the bubble in SDSS J1356+1026. We perform a χ² fit to the observed position–velocity diagram taking 1 < n < 2.5. We exclude all data points with velocities greater than ±400 km s⁻¹, since they are not treated by our model. We further use a minimum velocity error of 20 km s⁻¹ (a fraction of a pixel). Models with 1.2 < n < 1.5 all yield a reduced χ² differing from the minimum by less than one. The corresponding range of projected velocities is 250 < v_3D < 1000 km s⁻¹. At the lowest velocity there is no projection effect: the maximum observed velocity is also the maximum space velocity. This is a strict lower limit to the energetics.

We postulate that all of the line emission is part of a quasi-spherical outflow blown by the AGN that extends from the southernmost edge of the bubble to the clumps in the north (illustrated schematically in Figure 6). However, due to interference from gas and dust in the central pair of galaxies, the outflow cannot escape freely in all directions. Along PA2, the wisps represent gas that has broken free and is in free expansion. In contrast, along the N–S direction (PA1) the outflow runs into the two galaxies. The ring shape in the PA1 two-dimensional spectrum results from gas breaking out of the high-density material associated with galaxy S and accelerating away from the merger with v_3D∝r^1.5_3D. In addition, the structure is disconnected from its origin (nucleus N) because of the intervening matter. The clumps in PA1 are the northern counterpart of the same flow, produced closer to nucleus N than the bubble since there is less material to penetrate. In this picture, the similar velocity spread seen along the bubble, clumps, and wisps (Figure 3) are naturally explained as arising from the same outflow. The opening angle of the outflow is no less than the 45° angle between PA1 and PA2, since both slit orientations captured it.

With our crude outflow model, and thus deprojected velocities, we can now estimate the energy injection $\dot{E}$ required to power the observed bubble, following Nesvadba et al. (2006). One method is to estimate the kinetic energy in the shell, based on an estimate for the gas mass and the observed velocity, and then divide by the dynamical time to estimate an energy injection rate (e.g., Martin 1998). The other approach is to treat the bubble like a supernova remnant and estimate the energy needed to expand the bubble into a low-density ambient medium. In both cases we are forced to make assumptions, but with the two methods we can at least bracket the kinetic luminosity of the outflow.

First, we derive a lower limit on the energy in the outflow based on its apparent kinetic energy. The dominant uncertainty here is the mass of gas in the bubble. In principle, with an electron density, we can turn the observed Hβ luminosity (L_Hβ = 7.5 × 10⁴¹ erg s⁻¹) into a hydrogen gas mass. As explained in detail in Paper II, we have two methods to estimate gas density in the extended nebulae of obscured quasars. One is the standard diagnostic line ratios, which give density estimates n_e ≲ 1000 cm⁻³. Unfortunately, this method is biased toward regions of the brightest line emission, and thus regions of the highest density. Furthermore, in clumpy media, density diagnostics often yield values close to the critical density in a wide range of conditions (Paper II). The other method is to use the observations of scattered light at large distances from the nucleus to infer a gas density, but this has only been possible for a couple of objects observed with the Hubble Space Telescope (HST) and with particularly well-measured scattered light nebulae (Zakamska et al. 2006). This method gives a mean electron density of n_e ≲ 1 cm⁻³. Clearly, the two estimates are discrepant, suggesting that gas clumping is significant. Defining the clumping factor as κ = 〈n²_e〉/〈n_e〉², we can write the mass in gas as

$$\begin{equation} M_g = 1.7 \times 10^9 \,M_{\odot } \cdot L^{41}_{\rm H\beta } \langle n_e\rangle ^{-1} \kappa ^{-1}. \end{equation} \tag{ 1 }$$

In Section 3.1, we used diagnostic line ratios to find that in the nuclear region n_e < 250 cm⁻³ and that [O iii]/Hβ is constant throughout the nuclei and the bubble. Since at the location of the bubble the ionizing source is 5–10 kpc away, the density must drop with radius by at least a factor of 20–100 to maintain the same ionization parameter (the ratio of the number of ionizing photons to electrons, U). This decrease can plausibly be somewhat offset by a clumping factor of similar magnitude. Therefore, our final mass estimate is M_g ≳ 5 × 10⁷ M_☉. This mass is a lower limit, because it takes into account only the ionized gas that produces emission lines. Any lower density gas at higher temperature filling the bubble would not be included in this estimate. On the other hand, without a measurement of the clumping factor κ, we cannot say for sure what fraction of the observed volume is filled by high-density gas. In the case of the observed bubble in NGC 3079 (e.g., Filippenko & Sargent 1992; Veilleux et al. 1994), the energetics estimates were revised downward based on high-resolution HST imaging of the ionized gas filaments (Cecil et al. 2001). We plan to obtain integral-field observations of SDSS J1356+1026 in the Spring, which will hopefully help settle the outstanding ambiguities.

Based on the extent of the bubble, R = 12 kpc, and the maximum deprojected velocity of v = 250–1000 km s⁻¹, we find the dynamical time, t_d ≈ R/v ≈ 10–40 Myr. It is interesting to note that this timescale is similar to typical AGN lifetimes (Martini & Weinberg 2001; Martini 2004). With our previous estimate of the mass of the emitting gas, this corresponds to a mass outflow rate of $\dot{M} \sim M_g/t_d > 2\hbox{--}5$ M_☉ yr⁻¹. The corresponding kinetic energy is E_kin ∼ M_gv²/2 > 3–50 × 10⁵⁵ erg. For a constant energy injection rate over the lifetime of the bubble, we find $\dot{E} > 3\hbox{--}10 \times 10^{41}$ erg s⁻¹. This is the kinetic energy of the emission-line gas alone and, therefore, a strict lower limit on the total energy of the outflow.

The energy injection rate required to expand the bubble into a low-density medium, n₀, in the galaxy halo can be calculated in a manner similar to the Sedov–Taylor solution for supernovae. Following Nesvadba et al. (2006), we assume that the radiative losses are minimal and that energy injection rate is constant rather than a single explosion. We then find

$$\begin{equation} \dot{E} \approx 1.5 \times (\Omega /4\pi) 10^{46} R_{10}^2 v_{1000}^3 n_0 \,{\rm erg \,s^{-1}}, \end{equation} \tag{ 2 }$$

where the extent of the bubble R₁₀ is in units of 10 kpc, v₁₀₀₀ is the expansion velocity in units of 1000 km s⁻¹, and Ω/4π is the covering factor of the outflow in steradians. Taking R₁₀ = 1.2, n₀ = 0.5 cm⁻³, and v₁₀₀₀ = 1, $\dot{E} \approx (\Omega /4\pi) \times 1.1 \times 10^{46}$ erg s⁻¹. If we instead adopt the lower limit on velocity (i.e., no deprojection) of v₁₀₀₀ = 0.25, we find $\dot{E} \approx (\Omega /4\pi) \times 2 \times 10^{44}$ erg s⁻¹.

In addition to the deprojected velocity, the major uncertainty in this value comes from the lack of good measurements of n₀ and Ω. Both slit positions, separated by 45°, captured the signature of the flow and Fu et al. (2011) find a strong component of [O iii] emission to the northeast of N, which is not covered by our observations. We therefore conclude that Ω is not too small. It is much harder to estimate the proper value for n₀. The HST observations of Zakamska et al. (2008) alluded to above find clear evidence for electron scattering off of a medium with n_e ≲ 1 cm⁻³, even at these large projected distances of ∼10 kpc from the nucleus. We clearly see evidence for denser gas from the observed line emission. Thus, we view our choice of n₀ to be a conservative lower limit.

The kinetic energy injection rate of the outflow is bracketed by the values 10⁴² erg s⁻¹ (where we likely underestimated the amount of mass by a large factor) and 10⁴⁶ erg s⁻¹ (where we likely overestimated the outflow covering factor and the density of the ambient medium by smaller factors). We will therefore adopt a fiducial range of 10^44–45 erg s⁻¹ for the kinetic energy injection rate in the discussion that follows. It is interesting to note that our upper limit on the kinetic energy injection rate is roughly consistent with our estimate for the radiative bolometric luminosity of L_bol ∼ 10⁴⁶ erg s⁻¹.

Can the energy of the outflow be plausibly supplied by any source other than accretion onto the supermassive black hole? Vigorous star formation can lead to large-scale outflows (Veilleux et al. 2005) that are observable both in ionized gas in emission (e.g., Chevalier & Clegg 1985; Heckman et al. 1990) and in neutral gas in absorption (e.g., Rupke et al. 2005b; Martin 2005). In an active merger like SDSS J1356+1026, with plenty of gas available (as evidenced by our emission-line observations), it is natural to expect that some ongoing star formation is present. While from the high-ionization line ratios (such as the uniformly high [O iii]/Hβ ≃ 10) it is clear that the photoionization of the gas is completely dominated by the quasar, star formation may still be the primary driver of the kinetic outflow. Unfortunately, despite the availability of abundant multi-wavelength data and deep spectra of the host galaxy, the relative contributions of star formation and obscured quasar activity to the bolometric output of SDSS J1356+1026 are very hard to determine. This is a generic difficulty arising due to the fact that most of the flux from the object is coming out at mid- and far-infrared wavelengths, where the emission is due to thermal radiation of dust particles heated by young stars or the quasar in a manner notoriously insensitive to the spectrum of the underlying power source.

The bolometric luminosity of the quasar, as expected from its [O iii] luminosity (Liu et al. 2009), is 1.5 × 10⁴⁶ erg s⁻¹, although such scalings have dispersion of ≳ 0.5 dex. The bolometric luminosity estimated using the 25 and 60 μm IRAS fluxes and quasar bolometric corrections (Richards et al. 2006) ranges between 7.2 × 10⁴⁵ and 2.4 × 10⁴⁶ erg s⁻¹. The reasonable agreement between these values indicates that most, if not all, of the mid- and far-infrared luminosity from SDSS J1356+1026 (2.7 × 10⁴⁵ erg s⁻¹) can plausibly be produced by the quasar activity rather than star formation. The high degree of polarization likewise suggests that the AGN is dominant.

If, however, all of the infrared luminosities of this source were powered by star formation, it could just barely power the observed outflow. We use the calculation of the mechanical energy injection rate from a constant rate of star formation by Leitherer & Heckman (1995) and Veilleux et al. (2005). For every solar mass of stars formed, the combination of stellar winds and supernovae can produce at most L_mech ≈ 7 × 10⁴¹ erg s⁻¹ at time t ≈ 10⁷ yr. Thus, to produce a 10^44–45 erg s⁻¹ outflow would require >100–1000 M_☉ yr⁻¹ of star formation and predict a star formation luminosity of L_FIR = 3 × 10^45–46 erg s⁻¹. Thus, star formation is marginally consistent with powering the outflow at the lower limit of the energy range. On the other hand, a starburst of >100 M_☉ yr⁻¹ is hard to reconcile with our optical spectroscopic and spectropolarimetric observations. We see no evidence in the spectrum for an intermediate age population (i.e., the Balmer absorption lines are consistent with an old stellar population), and the blue continuum that we do see is consistent with scattered light (e.g., Liu et al. 2009). Therefore, we consider power from star formation as unlikely given the currently available data.

Radio flux densities at 1.4 GHz can be obtained from the Faint Images of the Radio Sky at Twenty cm (FIRST) survey (Becker et al. 1995) and the NRAO VLA Sky Survey (NVSS; Condon et al. 1998). The FIRST detection is unresolved—the nominal deconvolved size of 12 × 06 is smaller than half of the beam size of 54 (Ho & Ulvestad 2001). The FIRST peak flux is 57.90 mJy beam⁻¹ and the integrated flux is 59.58 mJy; the ratio of these two measures is consistent with that seen in point sources (Ivezić et al. 2002; Kimball & Ivezić 2008). The high-resolution FIRST observations tend to resolve out the low surface-brightness-extended emission, so the comparison with lower-resolution NVSS observations (45'' beam) provides yet another measure of extended emission. The NVSS catalog flux is 62.9 mJy, and again the difference between NVSS- and FIRST-measured fluxes is consistent with that seen in point sources (Kimball & Ivezić 2008). From these observations, it appears that SDSS J1356+1026 lacks extended radio emission on >4 kpc (2'') scales at the few mJy level.

Using the radio–infrared correlation of star-forming galaxies (Helou et al. 1985) and its convenient normalization presented by Morić et al. (2010), we find that no more than 20% of the observed radio emission can be due to star formation. Thus, most of the radio emission is attributable to one or both quasar nuclei. Nevertheless, the position of the SDSS J1356+1026 in the [O iii]–radio diagram (Xu et al. 1999; Zakamska et al. 2004; Lal & Ho 2010) puts it squarely into the radio-quiet regime, 1.5–3 dex below radio-loud sources in Nesvadba et al. (2008) and Fu & Stockton (2009). For direct comparison with Xu et al. (1999), we use a radio spectral index of α = −0.47 (defined as F_ν∝ν^α, see below) to convert a rest-frame-corrected value of log  (P_1.4 GHz, erg s⁻¹ Hz⁻¹) = 31.32 into log  (P_5 GHz, erg s⁻¹ Hz⁻¹) = 31.06 at $\log \,(L_{\mathrm{[O\,\mathsc{iii}]}},\,{\rm erg \,s^{-1}})=42.78$ taken from Reyes et al. (2008). Obscured quasars with such radio luminosities have only a ∼10% chance to have radio extents >3 kpc (Lal & Ho 2010), consistent with the lack of any sign of extended emission in SDSS J1356+1026. In principle, we cannot rule out a <4 kpc scale jet as the accelerating agent, as has been proposed for NGC 3079 (Cecil et al. 2001). On the other hand, the outflow we observe is considerably more extended and more energetic than that in NGC 3079, so it is not clear that energy injection on 1 kpc scales can power the observed 20 kpc outflow. Deeper radio interferometry at 1'' resolution would settle the issue.

The source is undetected in the Texas survey at 365 MHz (Douglas et al. 1996) and in the VLA Low-Frequency Survey at 74 MHz (Cohen et al. 2007), with the 74 MHz observation providing a stronger upper limit, so we focus on this observation. This implies a limit on the spectral index α > −0.47 (hence the use of this value in the calculation above). Such a flat spectrum effectively excludes a "relic" radio source (Komissarov & Gubanov 1994).

High-resolution, high-sensitivity radio observations would be useful in establishing the presence of two quasars and of any extended radio emission, especially on 5''–10'' scales. Since the separation of the nuclei is ∼1'', observations with Very Large Array would be well suited for these purposes. The archival radio observations available to us at the moment indicate that the outflow in SDSS J1356+1026 is unlikely to be jet-driven.