THE JET-DRIVEN OUTFLOW IN THE RADIO GALAXY SDSS J1517+3353: IMPLICATIONS FOR DOUBLE-PEAKED NARROW-LINE ACTIVE GALACTIC NUCLEUS

The most current version of the SDSS J1517+3353 survey data set was downloaded from the SDSS DR7 Data Access Server (http://das.sdss.org), comprised of images of the galaxy and its surrounding field in the five survey filter bands (ugriz), as well as a spectrum from the SDSS spectroscopic survey taken through a 3'' fiber aperture centered on the galaxy. All data were processed and calibrated by the standard SDSS data pipeline through the SDSS DAS (Adelman-McCarthy et al. 2008).

A major limitation of the SDSS spectroscopic data is its lack of resolved spatial information, which, we will show, is necessary to disentangle the various components in this system. To gain this added dimension, follow-up spectroscopic observations of the galaxy were made on the Keck I telescope, using the High-Resolution Echelle Spectrometer in red-optimized mode (HIRES-r). The spectra were taken on the night of 2008 April 21, in 15° twilight as part of a bright object program planned for the end of the night. The median seeing over the period of the observations was about 08. Sky conditions were clear and transparent.

A 70 long slit with decker C5 (slit width of 115) was used. The slit was placed over the centroid of the galaxy in the telescope guide camera and aligned along a position angle (P.A.) of 122° in order to cover both central peaks in the galaxy. This P.A. placed the slit at an angle of ∼35° from parallactic, implying that atmospheric dispersion effects, while small, are not negligible (see Section 6). Image grabs of the slit relative to a half arcminute field of sky around the galaxy were taken a few times during the observing sequence with the HIRES guider camera. These images were subsequently compared to the SDSS images to verify the alignment of the HIRES slit on the sky.

The echelle and cross-disperser grating angles were chosen to place as many emission lines as possible on the green and red CCD chips of the HIRES science camera. Two exposures of 600 s each were taken with the same slit alignment. Quartz Halogen and ThAr arc-lamp calibration spectra with the same instrument configuration as the science frames were taken at the end of the night after sunrise.

Spectroscopic calibration and data reduction of the raw HIRES data were performed using the XIDL HIRES reduction package developed by Jason Prochaska.⁵ Following bias subtraction and the application of pixel-to-pixel superflats to the science and calibration spectra, the quartz halogen lamp exposures were used to trace the echelle orders. Archived XIDL/HIRES ThAr arc templates were tweaked to match our wavelength calibration arc-lamp exposures. These were then used to generate a wavelength solution for each order. Finally, each curved order was extracted and rectified, giving a set of two-dimensional long slit spectra, one per exposure for each echelle order, spanning a wavelength range from 5289 Å to 8379 Å, with small gaps in coverage redward of 6000 Å. Cosmic rays were flagged using a sigma-clipping algorithm and the resultant cosmic-ray masks were examined visually to verify that compact substructure in the two-dimensional spectra were not flagged in error.

Following the observations, it was discovered that the HIRES slit had shifted slightly between the two exposures, by less than 1'' perpendicular to the long axis of the slit. Since the observations were made in growing twilight, the guiding drift was most likely a result of the strong change in the level of sky background between the two exposures. The most immediate effect is a difference in the shape of the emission lines from the spectra taken during the two exposures. Therefore, in our subsequent analysis, we treated each exposure independently. Interestingly, the line profile differences between exposures largely affect the blue part of the lines, which provides some valuable insight into the spatial structure of the emission-line region. Therefore, when appropriate, a comparative approach between the two spectral exposures was employed in the analysis of our HIRES data set.

The moderately bright radio source (B2 1515+34; Colla, et al. 1970) corresponding to J1517+3353 was detected in the FIRST survey (Becker et al. 1995), with a 1.4 GHz flux density of 106 ± 5 mJy, making it an ideal target for high-resolution radio imaging. Consequently, observations were taken with VLA in A-configuration on 2008 October 30 in four bands: 1.4, 5, 8, and 22 GHz. Between five and ten minutes of data were obtained at each frequency. The observations were processed and reduced using the Astronomical Image Processing System in the standard fashion with flux densities tied to 3C286. CLEANed images were produced with the Difmap package using natural weighting. Integrated flux densities at each frequency band were estimated from the sum of the clean components. Flux densities for the core components were derived in Difmap using Gaussian fits to the visibility data. The images and the relation between radio and optical structures are discussed in Section 8. Flux density measurements are tabulated in Table 3. At the redshift of the galaxy (z = 0.135) the FIRST measurement corresponds to a monochromatic luminosity at 1.2 GHz of 5.1 × 10³¹ erg s⁻¹ Hz⁻¹. Comparing to the SDSS optical photometry of the host galaxy, this yields a Kellerman R radio-loudness index of 440 (defined in Kellermann et al. 1989), which is strictly a lower limit, since the host luminosity is dominated by stellar light. The high value of R places J1517+3353 in the category of narrow-line radio galaxies.

VLA images for a second radio bright AGN with double-peaked narrow lines were also taken as part of this program. These data are presented in the Appendix.

The [O iii]λ5007 emission line from the HIRES spectrum of SDSS J1517+3353 is shown in Figure 5. The spectrograph dispersion is miniscule compared to the kinematic width of the line and the its shape tracks the kinematics of the line emitting gas within the active region. In general, three relatively distinct structures can be distinguished: two bright features separated by 14.2 Å or about 750 km s⁻¹ with blue- and red-shifted kinematics with respect to the systemic velocity of the galaxy, and a third component at intermediate velocities, which spans the location of the nucleus. By comparing the position of the features along the slit to the position of the knots in the SDSS images of the galaxy, it is clear that the bluer feature corresponds to knot B, while the redder feature corresponds to knot A. We will call the third region of emission, cospatial with the nucleus, as feature C. There are tentative clues that this may be a distinct structure: its kinematics appear to differ from knot B, though both overlap in the sky and may possibly be contiguous in space. Henceforth, we will use the term "feature" to label a kinematic component and the term 'knot" to label substructure seen in the images. Therefore, we associated feature A with knot A and feature B with knot B. A close examination of the color image in Figure 2 indicates a possible faint blue knot coincident with the nucleus, which may be associated with feature C.

Zoom In Zoom Out Reset image size

Features A and B display fairly asymmetric line profiles. Feature A has a sharp blue edge and a high velocity, extended red wing. In contrast, feature B exhibits a somewhat less pronounced extended blue wing. Broadly, the kinematic structure of feature B is roughly antisymmetric to that of feature A, and a rough mirror symmetry appears to exist between the profiles of these two features. However, there are some differences in detail. Feature B shows definite kinematic substructure, in the form of a narrow ridge of emission which extends over the length of the feature and defines its red edge. Feature A, on the other hand, has a much smoother, less peaky profile. Also, feature A shows a definite gradient in both line asymmetry and peak velocity as a function of distance from the nucleus. If this gradient exists along feature B, it is far less pronounced and not evident under the ridge-like substructure which dominates the peak of the line.

Visually, the kinematics of feature C appear more symmetric, but its emission has to be disentangled from that of the much brighter feature B before a systematic study of its kinematics can be attempted. To do this, we adopt the technique of Gaussian decomposition of line profiles.

A parameterization of the kinematics of the line-emitting gas may be obtained by fitting Gaussian subcomponents to the profile of the [O iii]λ5007, keeping in mind that these components may not necessarily be identified with physically independent structures. Not surprisingly, a single Gaussian is insufficient to model the profile of the [O iii] line, either integrated or even at any given location along the slit. However, an inspection of the [O iii] line highlights some simplifications that serve to guide our fits. Visually, the kinematics of features A and B are relatively continuous along the slit, so we adopt the following approach. First, we independently fit one to two Gaussians to the line profile for the parts of the slit that do not sample feature C (nuclear distances of 15–25 for feature A and 08–2''for feature B). Then, we adopt these same Gaussians at inner nuclear radii, maintaining their peak wavelength and width, and scale them while fitting additional Gaussians to feature C. In this way, we can determine physically motivated Gaussian fits to all three features and model the [O iii] line reasonably well (Figure 6(a)).

Zoom In Zoom Out Reset image size

Because of its strong blue wing, feature B requires at least two Gaussians to describe its structure—a narrow component centered at 5676.16 Å with a line width of 5.58 Å(294 km s⁻¹), and a broad component centered at 5672.69 Å with a line width of 8.00 Å (423 km s⁻¹). Feature A can be adequately modeled by a single Gaussian centered at 5690.57 Å with a line width of 9.30 Å (489 km s⁻¹), which slightly underestimates the red wing, though within the tolerance of this fitting exercise. After accounting for these three components, we are left with the emission from feature C, which we fit by a single Gaussian centered at 5680.97 Å with a line width of 8.57 Å (452 km s⁻¹). The ratio of the fluxes in the three components is measured to be Flux(A:B:C) = 2.1:1.5:1.0. Since the various features are larger than the HIRES slit width, we do not have a good handle on slit losses and, therefore, these flux ratios are probably not representative of the actual fluxes in the features.

The other set of bright emission lines in the HIRES spectrum is the Hα, [N ii]λλ6583, 6548 complex. These three lines are strongly blended in the HIRES spectra due to their large line widths. Therefore, we used the Gaussian subcomponents from the [O iii]λ5007 line fits to attempt a constrained fit to the Hα+[N ii] complex, by shifting and broadening each [O iii] subcomponent appropriately, and scaling all nine shifted Gaussians in strength to fit the integrated profile of these lines, as shown in Figure 6(b). We require that the ratio of the [N ii]λλ6583, 6549 lines be 3:1, dictated by atomic physics. It may be argued that this process is preferable as a way to fit Hα+[N ii], since, in the absence of a strong dependence between gas kinematics and ionization, the average line profiles of all emission lines from a spatially resolved emission-line region should be very similar. From Figure 6(b), it is clear that this approach yields very acceptable fits to these lines. The quality of the fit was equally good for the line profiles extracted from both HIRES exposures, strengthening the validity and robustness of our approach.

Valuable insight into the ionization and extinction properties of the gas in the features may be obtained by combining the results of the HIRES subcomponent fitting with the wider suite of emission lines in the SDSS spectrum, in particular the intermediate strength lines of Hβ and the [S ii]λ6717, 6731 doublet.

We start with a three-component Gaussian fit to the [O iii]λ5007 line in the SDSS spectrum. In the case of the HIRES data, four Gaussian components were needed to completely describe the [O iii] line. However, the much lower spectral resolution of the SDSS data prevents us from consistently deblending the two components that constitute the profile of feature B. Therefore, we treat these two components as equivalent to a single one at lower resolution and only used three Gaussians to fit the [O iii] profile in the SDSS spectrum. The fit was guided by the results of the HIRES analysis. Given the different sky coverage of the SDSS aperture compared to the HIRES slit, coupled with the highly complex and nonuniform structure of the galaxy's active region, the subcomponents that were used to fit the [O iii] line in the HIRES spectrum do not exactly match the [O iii] line the SDSS spectrum. However, slightly altered versions of the HIRES subcomponents, carefully broadened to match the resolution and dispersion of the SDSS spectrum, provide a very reasonable fit. The properties of these subcomponents are listed in Table 2. Note that we did not attempt to fit the wings of the [O iii] line in any detail. The wings of emission lines from Seyfert NLRs frequently display ionization properties that differ from the bulk of the line (Whittle 1985; Veilleux 1991).

While Hβ is detected in the HIRES data set, it is much too weak for a profile analysis study. Instead, we use the fit to the SDSS [O iii] line, described above, as a template to fit Hβ, as well as few other strong lines in the SDSS spectrum. The [O iii] subcomponents were shifted to the observed wavelength of each line to be fitted, corrected for the wavelength-dependent differences in SDSS spectral resolution between [O iii] and the line, and then scaled in strength to reproduce the profile of line. Lines fitted in this manner were Hβ, the Hα+[N ii] complex, and the [S ii] doublet at 6720 Å.

To conclude, using a combination of constrained Gaussian subcomponent fits and scaled template matching, we estimate, for each spectral feature, kinematic parameters such as a bulk velocity and line width, as well as the strengths of six strong emission lines in a fairly self-consistent and controlled manner (Table 2). We are now poised to explore the physical properties of the ionized gas in some detail.

We start by evaluating some basic average properties of the line-emitting gas: its density, temperature, and dust content. The electron density of the gas can be estimated approximately from the ratio of the [S ii]λλ6717, 6731 doublet. For both the high velocity knots A and B, we estimate the ratio R_6717/6731 to be 0.71 ± 0.10, which translates to an electron density log n_e = 3.40 ± 0.24 in cm⁻³. For the nuclear knot C, R_6717/6731 = 0.99 ± 0.14, giving log n_e = 2.87 ± 0.25 cm⁻³. The densities of all three knots are comparable, though the nuclear gas may be slightly denser, at least in the parts of the clouds which emit most of the [S ii] line.

For typical AGN NLRs, the Balmer decrement Hα/Hβ has a value of 3.1. Deviations from this standard value are a signature of dust extinction in, or along the line of sight (LOS) to, the line-emitting gas. We have measured Hα and Hβ for each individual subcomponent in Section 6, from which, for the three features A/C/B, we get the ratio of the two lines to be 5.3/4.6/3.3. This implies that feature A is significantly more dust reddened than feature B, with feature C in between. Assuming that the dust is distributed as a foreground screen and follows a Galactic extinction law (Cardelli et al. 1989), we derive the visual extinction A_V toward features A/C/B to be 1.35/0.98/0.1, after correcting by A_V = 0.062 for the dust in our own galaxy toward the direction of SDSS J1517+3353. This may be compared to the average value of A_V = 0.59 implied by the spectral synthesis analysis in Section 4.

The [O iii]λ4363 auroral emission line is sensitive to the electron temperature of the ionized gas. Though weak, this line is clearly detected in the SDSS spectrum. We measure the ratio R_5007/4363 to be 87 ± 36, after applying a reddening correction for an average extinction of A_V = 0.9. Using the standard calibration from Osterbrock (1989, p. 422) and taking a mean density of log n_e = 3.0, this yields an electron temperature estimate of T_e = 12000⁺²³⁰⁰₋₁₅₀₀ K, somewhat lower than a typical AGN NLR, though within the range measured in the normal population.

Finally, we explore whether the emission-line region may be composed of an extended complex of H ii regions from recent star formation, rather then AGN activity. For this, we look at line ratio combinations which discriminate between AGN and star-forming systems (e.g., Kewley et al. 2006, and references therein). We measure a ratio of [N iiλ6584]/Hα to be 0.65/0.42/0.59 for knots A/B/C. At these values, AGN/star-forming composite galaxies have [O iii]/Hβ less than 1.83/2.95/2.08, whereas we measure this ratio to be 7.37/7.81/6.09, placing all three knots significantly above the transition between AGN and the star-forming sequence. While [N ii]/Hα in SDSS J1517+3353 is somewhat lower than the median value for Seyfert-like AGN, the high [O iii]/Hβ ratio clearly places the galaxy in this general class, separate from LINERs. Ionization by the UV radiation from young stars is clearly not very important in exciting the line emission. This is consistent with the continuum spectral analysis of Section 4, in which we found that the light of the host galaxy is dominated by an old stellar population, with little or no evidence for blue light from young stars, even blueward of the Balmer break. Hence, whatever mechanism is responsible for ionizing the gas does not contribute greatly to the visible continuum.

The richness of the emission-line spectrum allows a detailed look into the processes that ionize the emission-line region of the AGN. If the kinematics of the gas are driven primarily by a radio jet, then shocks should be widely present throughout the NLR and serve as a good source of ionizing photons. If, on the other hand, the emission-line region is primarily photoionized by merging accreting black holes, we may expect normal nuclear photoionization to dominate the excitation of the NLR.

The dynamical and physical modeling of shocks is a complex process, involving, among other variables, important dependences on geometry and magnetic fields. Shocks compress and collisionally ionize gas, leading to post-shock emission which produces a profuse ionizing spectrum. These ionizing photons then propagate outward from the post-shock region and produce an extended "precursor" region. The emergent spectrum is therefore a combination of shock and precursor emission. The canonical models of Dopita & Sutherland (1995), which we adopt here, successfully reproduce the basic features of emission-line spectra of Seyferts and radio galaxies, as long as both shock and precursor emission are taken into account. They predict emission-line ratios of most strong emission lines that are very similar to the predictions from standard photoionization models. Unfortunately, this means that we are not able to use strong, well-measured lines to verify the presence or absence of shock-ionized gas, since both shocks and AGN photoionization models generally produce the same relative strengths of these lines. Instead, we turn to line ratios involving weaker lines, such as [Ne v]λ3426, He iiλ4681, and [O i]λ6300 which may help discriminate between the two types of models, as shown in Figure 7. The shock models are shown as solid lines in the figure, spanning shock velocities from 200 km s⁻¹ to 500 km s⁻¹.

Zoom In Zoom Out Reset image size

For comparison to the shock models, we employ a set of modern nuclear photoionization models from Groves et al. (2004), which include the effects of dust and a broad spectrum of cloud properties, and are quite successful at matching the strengths of weak AGN lines, while generally maintaining the fit to strong lines. For our purposes, we fix the gas metallicity at twice solar (appropriate for the nuclear regions of massive early-type galaxies) and choose a gas density of 1000 cm⁻³ to match those measured from the [S ii] lines. The remaining degrees of freedom in the models are the ionization parameter U, defined as the ratio of the number densities of ionizing photons to gas at the face of a cloud, and the shape of the ionizing continuum, assumed to be a power law in frequency, with a spectral index α. The two models tracks plotted in Figure 7 as dotted lines are for α of 1.2 and 2.0: each track is a sequence in ionization parameter over the range 3.0 ⩽ U ⩽ 0.0. These tracks encapsulate most of the ionization variation in the dusty models.

Note that Figure 7 employs only the integrated line strengths measured from the SDSS spectrum, not measurements from the Gaussian fitting treatment. The weak lines in the spectrum do not have the necessary S/N for accurate subcomponent analysis. In order to place J1517+3353 in the context of the broader AGN population, the figure also contains data points from a sample of nearby Seyferts and narrow-line radio galaxies drawn from the literature.

In general, both sets of models match the line ratios of J1517+3353 well. The photoionization models fail badly in one case (Figure 7(a)), where they severely overpredict the strength of [Ne iii]λ3869, considerably more than is observed in almost all the narrow-line AGNs on the plot. This failure has been noted in previous studies of Seyfert ionization (e.g., Whittle et al. 2005). All in all, the shock models seem to perform consistently the best in matching the measurements. Shock velocities around 400 km s⁻¹ are needed to explain the line ratios—at the low end, but of the same order as the measured FWHM of the kinematic subcomponents. Based on this, we tentatively conclude that shocks are the most likely source of ionization in most of the emission-line gas in the galaxy. The limitations of current NLR ionization models prevent any conclusive statements to be made, as photoionization models can also explain most of the line ratios quite adequately.

In summary, the profiles of a number of emission lines from the AGN in SDSS J1517+3353 can be characterized by a minimal set of four Gaussian subcomponents which appear to be associated with three spatially and kinematically distinct features seen in the central 5 kpc of the galaxy. These features are bluer than the mean color of the host galaxy and do not trace the galaxy's light. From our line profile analysis, we see that the kinematics of these features are strongly disturbed, far in excess of any motion that may be driven by virial processes, even in a massive early-type galaxy such as this AGN host. Two of the features, A and B, are in overall bulk motion with respect to the nucleus of the galaxy. Their profiles display some symmetries, but the relative distance of these features from the nucleus are quite different, with feature B being significantly closer to the center of the galaxy than feature A. Their gas excitation properties are similar and produce a highly ionized Seyfert-like spectrum, though feature A is associated with much greater dust extinction, either intrinsic or along the LOS, than feature B. The third feature, C, is coincident with the galaxy's center of light and probably surrounds the true nucleus of the galaxy. Unlike features A and B, feature C has a kinematic profile which is quite symmetric. Its excitation properties are a bit different from the two off-nuclear features, with stronger lines from low ionization species.

Despite its initial selection as a candidate for a multiple merging black hole system, several lines of evidence confirm that the double-peaked lines of J1517+3353 are due to a strong emission-line outflow produced by a jet–interstellar medium (ISM) interaction. The radio jet and emission-line gas are co-aligned and appear to trace each other quite closely. This is characteristic of radio galaxies and Seyferts that show jet-ISM interaction signatures (Whittle & Wilson 2004, and references therein). The profiles of the kinematically distinct components appear to show a particular mirror symmetry (Section 6.1), consistent with gas accelerated by a bipolar flow or jet. In addition, the emission-line spectrum is consistent with ionization by strong shocks with a mean velocity of 400 km s⁻¹, consistent with the line widths of the emission-line sub-components. Finally, the host galaxy does not appear to display any obvious signatures of recent merger activity. The isophotes are quite regular and the galaxy colors outside the emission-line region are quite red. The stellar continuum appears to be that of an early-type galaxy with some extinction in the central regions, presumably due to dust near the nucleus. There is no sign of the substantial recent star formation which may indicate a gas-rich merger of the sort that can host merging accreting SMBHs.

Through a combination of spectral line decomposition and high-resolution radio imaging, it is possible to now associate the eastern and western radio jets with specific emission-line knots. In this way, a comparison can be made between the dynamics of the emission-line gas and the energy content of the radio source.

We begin by estimating the mass and kinetic energies of the ionized gas. The extinction-corrected Balmer line luminosity is a direct measure of the total amount of ionized gas:

$$\begin{equation} M_{{\rm em}} = 171\, F_{{\rm H}\beta }\,d^2 \, n_{{\rm em}}^{-1} \, {\rm g}, \end{equation} \tag{ 1 }$$

where d is the luminosity distance of the galaxy in cm and F_Hβ is the Hβ flux. Using the measurements in Table 2, the masses of feature A/B are both estimated to be around 10^5.9 solar masses. Taking the peak velocity and FWHM as a measure of the motion in the gas, we calculate bulk and turbulent kinetic energies. These are combined to yield total kinetic energies in features A and B of 10^54.4 and 10^54.1 erg, respectively.

The internal energy of the ionized gas is related to its density and temperature, and can be parameterized by the mass of ionized gas:

$$\begin{equation} E_{{\rm th}} = 1.25\times 10^{8}\; M_{{\rm em}}\; \mu ^{-1} \; T_{e} \, {\rm erg}, \end{equation} \tag{ 2 }$$

where μ ∼ 0.6 is the mean molecular weight of the gas. For the features A/B, we get E_th ∼ 10^51.5 for both. This is 2.5–3 orders of magnitude lower than the kinetic energy of the gas. If we assume that the emission-line luminosity of the gas is L_em ≈ 10 × L_[O iii] ∼ 10^43.5 erg s⁻¹, we derive a timescale at which this internal energy is radiated away τ_th ≈ E_th/L_em = 3 years. It is clear that the thermal energy of the ionized gas has to replenished, either through jet-driven shocks or photoionization from the central AGN.

The ionized gas energies may be compared to that of the radio source. A listing of the measured flux densities in the four VLA bands is given in Table 3. The mean radio spectral index of the extended emission between 5 and 8 GHz is α_R = −1.5. Assuming that this synchrotron spectrum extends between 0.01 GHz and 100 GHz, the total radio luminosity (L_R) of the jets can be estimated by normalizing the spectrum to the flux of the jets at 5 GHz, which gives L_R = 10⁴² erg s⁻¹. From the 5 GHz image, we estimate the size of the eastern jet to be about 12 × 11, which, if we assume a cylindrical jet geometry, gives a radio-emitting volume of V_jet ≈ 15.2 kpc³. Combining this with equipartition energy arguments of Miley (1980), and assuming that the radio emitting plasma has a filling factor ∼1, mean relativistic pressures, magnetic fields, and energies are estimated to be P_me = 10^−9.4 dyne cm⁻², B_me = 125 μG, and E_me = 10^56.7 erg. The synchrotron lifetime of the radio source τ_R ≈ E_rel/L_R ∼ 16 million years, comparable or longer than a typical lifetime for Seyfert-like activity, and much smaller than the timescale on which the radio-source imparts energy to the emission-line gas through shocks and pressure-driven acceleration (see below). Synchrotron emission is probably not a major loss mechanism for the energy in the radio plasma. In addition, the energy content of the radio source, assuming equipartition, is more than 2 orders of magnitude higher than the kinetic energy content of the ionized gas. Only a small fraction of the energy in the lobes needs to go into accelerating the ISM of the galaxy, through radio source expansion or ram pressure-driven shocks, to fully account for the dynamics of the emission-line material. We envision a scenario where the jet propagates rapidly to kpc scales, at velocities much larger than the ionized gas, sweeping up the ISM of the host galaxy and shocking it as it expands. The denser component of the shocked ISM in direct contact with the jet is visible as the NLR. There may be a potentially large fraction of shocked gas which cannot cool efficiently by line emission and may remain invisible (e.g., Mellema et al. 2002). Therefore, the amount of energy from the jet that is lost to line emission is highly unconstrained.

The pressure of the relativistic plasma, acting over the area of the jet, is the principal source of momentum for the ionized gas. A rough interaction timescale can be derived by equating P_me to the measured bulk momentum in the gas:

$$\begin{equation} \tau = \frac{M_{{\rm em}}\; v_{{\rm em}}}{P_{{\rm me}}\; A_{{\rm jet}}} \, {\rm s}, \end{equation} \tag{ 3 }$$

which yields τ ∼ 10⁵ years, for A_jet = 5.3 kpc² and an average bulk flow velocity of v_em ≈ 400 km s⁻¹. The timescale for the acceleration of the gas could be significantly smaller if the major part of the acceleration happened at sub-kpc radii in the galaxy, where P_me is expected to be higher. However, the gas would then have to coast out to where it is currently observed on kpc scales at or below its measured bulk velocity of v_em, adding more than 10⁶ years to the overall timescale for the interaction. The timescale would also be longer if the accelerating force was smaller than determined by P_me, or if the accelerated gas contained a substantial neutral component not detected in emission lines, as implied by some studies (e.g., Morganti et al. 2005). Therefore, τ is probably a lower limit to the actual time that the jet interaction has been happening in this system.

An upper limit to the energy flux F_E of the jet then follows

$$\begin{equation} F_{E} \, = \, E_{{\rm me}}/\tau \, =\, 1.65 \times 10^{44} \: {\rm erg\; s}^{-1}. \end{equation} \tag{ 4 }$$

This flux is about an order of magnitude larger than the emission-line luminosity and 2 orders of magnitude greater than the radio luminosity. If the timescale of the interaction is indeed as short as τ, we infer that only a small fraction of the energy output of the jet is lost to shock-excited radiation or other radiative mechanisms.

How does the energy in the lobes compare to the binding energy of the host galaxy? We estimate a velocity dispersion of the galaxy from the NaD absorption line, which is strong and well-defined in the SDSS spectrum. A Gaussian fit to the line yields an FWHM of 14.76 Å  which gives a velocity dispersion σ_* = 274 km s⁻¹, after a small correction for the SDSS spectrograph resolution. The amount of mass that can be unbound by the energy content of the radio source (i.e,, E_me) is given by

$$\begin{equation} M_{{\rm bind}} \, \approx \, \frac{E_{{\rm me}}}{\sigma _{*}^{2}}, \end{equation} \tag{ 5 }$$

where we have taken the gravitational binding energy of a unit mass of gas in the central regions of the galaxy to be approximately σ²_* erg gm⁻¹, assuming that the underlying mass density traces a singular isothermal sphere.

Combining our estimates of E_me and σ_* gives M_bind ∼ 10^8.5 M_☉ which should be accurate to 0.5 dex, given our uncertainties in the mass distribution of the host galaxy and the velocity dispersion. The energy in the lobes can easily unbind the total mass in ionized gas (∼10⁶M_☉). If we assume a gas fraction in the galaxy of ∼10⁻² and taking its stellar mass to be 10¹¹M_☉, the gas content of the galaxy, in both hot and cold ISM phases, is estimated to be 10⁹M_☉. The energy in the radio lobes can unbind a good fraction of all the gas in the galaxy and is certainly capable of heating most of this gas to its virial temperature, as long as it is effectively coupled with the ISM. The synchrotron lifetime of the lobes sets a simple upper limit to the efficiency of thermalization of the lobe energy. If the radio source is to serve as a source of feedback to the hot atmosphere of the host galaxy, then the energy in the non-thermal population of the lobes has to be exchanged with the surrounding gas within about 16 million years, or else most of it will be lost to radiation, which couples inefficiently with the hot gas.

J1517+3353 serves as a case study of a powerful AGN outflow that effectively masquerades as a candidate for merging accreting SMBHs. Only through spatially resolved spectroscopy and radio imaging is its true nature revealed. In the presence of a jet or strong nuclear outflow, it is much more likely that complex line kinematics on kpc scales are the result of gas interacting with the outflow rather than from inspiraling black holes. Smith et al. (2009) find that their double-peaked AGN sample includes more FIRST detected radio bright AGN than the parent QSO population, implying that a non-negligible fraction of the split lines probably come from jet interactions. Whittle (1992) points out that clear non-gravitational kinematics are seen even in radio-quiet Seyferts with 20 cm luminosities >10^22.5 W Hz⁻¹. We suggest that a radio prior be applied to double-peaked samples to identify such outflows. A radio galaxy like J1517+3353 can be detected to z ≈ 1 in wide-area 20 cm mJy-level surveys like FIRST and NVSS, while radio-quiet sources above the threshold from Whittle (1992) are detectable with 100 μJy pointed radio observations out to z ≈ 0.3 and with wide-area surveys to z ≈ 0.1. Modestly deep follow-up of double-peaked narrow-line samples with existing radio facilities to separate outflows from true black hole mergers.

About 10% of local optically selected Seyfert galaxies have total radio luminosities greater than 10²² W Hz⁻¹ (de Bruyn & Wilson 1978; Rush et al. 1996; Ho & Ulvestad 2001), significantly higher than the fraction of SDSS AGNs which show double-peaked profiles. If only one in ten AGNs with bright nuclear radio sources have extended jets, outflows can conceivably account for most of, if not the entire, double-peaked population. Clearly, stronger constraints on the radio content of these samples is needed before they can be used to study merger rates and SMBH evolution.

On the other hand, we have demonstrated that double-peaked emission can indicate the existence of powerful, galaxy-wide outflows, where massive AGN feedback is caught in the act. These samples may identify the best laboratories with which to directly tackle the details of feedback physics, such as the importance of momentum-driven winds, the effects of anisotropic obscuration, and the efficiency of coupling between the output of the AGN and material in the host galaxy.