An Active Galactic Nucleus Caught in the Act of Turning Off and On

The spectrum of SDSS J1354+1327 is classified as a Type 2 AGN and it has no evidence of broad emission lines (Oh et al. 2015). We obtain the fluxes of the emission lines in the SDSS spectrum of the galaxy using the OSSY catalog (Oh et al. 2011), which simultaneously fits an entire spectrum using stellar templates for the stellar kinematics and Gaussian templates for the emission components. The host galaxy redshift, based on the template fit to the stellar absorption features, is $z=0.063316\pm 0.000070$, which we use to convert the [O iii] λ5007 flux to a luminosity L_{[O iii] λ5007} = 4.56 × 10⁴¹ erg s⁻¹. The [O iii] λ5007 emission line is redshifted by 69.1 ± 15.0 km s⁻¹ relative to the systemic velocity of the galaxy's stars (Comerford & Greene 2014). We find that the line flux ratios are log ([O iii] λ5007/Hβ) = 1.03 ± 0.05 and log ([N ii]/Hα) = −0.50 ± 0.05, which place SDSS J1354+1327 in the Seyfert region of the Baldwin–Phillips–Terlevich (BPT) diagram (Baldwin et al. 1981; Kewley et al. 2006).

The companion galaxy, SDSS J1354+1328, has a redshift of $z=0.063586\pm 0.000048$ and an [O iii] λ5007 luminosity of L_{[O iii] λ5007} = 2.4 × 10⁴⁰ erg s⁻¹. The line flux ratios are $\mathrm{log}([{\rm{O}}\,{\rm{III}}]\lambda 5007/{\rm{H}}\beta )=0.44\pm 0.09$ and $\mathrm{log}([{\rm{N}}\,{\rm{II}}]/{\rm{H}}\alpha )\,=-0.26\pm 0.08$, which are in the Seyfert region of the BPT diagram.

SDSS J1354+1327 was observed with Chandra/ACIS for 9443 s on UT 2014 June 25 as part of the program GO4-15113X (PI: Comerford). The data were taken with the telescope aimpoint on the ACIS S3 chip in "timed exposure" mode and telemetered to the ground in "faint" mode. We reduced the data with the latest Chandra software (CIAO 4.6.1) and the most recent set of calibration files (CALDB 4.6.2).

We used dmcopy to make sky images of the field in the soft (S, restframe 0.5–2 keV), hard (H, restframe 2–10 keV), and total (T, restframe 0.5–10 keV) energy ranges with pixels binned to 1/10 the native pixel size. Then, we used Sherpa to model the X-ray source as a 2D Lorenztian function in beta2d. Our model also included a background component of fixed count rate, which we determined from an adjacent circular region of 30'' radius. We confirmed that the background region does not contain any sources detected by wavdetect with a threshold of ${\mathtt{sigthresh}}={10}^{-8}$. We set the initial position of the beta2d component to the location of the SDSS galaxy coordinates. After using psfSize to estimate that the radius of the point spread function (PSF) is $1\buildrel{\prime\prime}\over{.} 07$, we ran the fit in Sherpa and allowed the model to fit a region around the galaxy coordinates of three times the PSF size at that location on the chip. We determined the best-fit model parameters by minimizing the Cash statistic using Sherpa's implementation of the "Simplex" minimization algorithm (Lagarias et al. 1998). To test for additional sources, we also attempted fitting a two-component beta2d model. However, the amplitude of any second component is detected at $\lt 1\sigma$ significance, and therefore we adopt the single beta2d model.

To determine the relative positions of the Chandra sources and the HST sources, we registered the Chandra and HST/F160W images and estimated the relative astrometric uncertainty between the two images. Due to the small number of Chandra sources and the relatively small HST/F160W field of view, we could not directly register the two images. Instead, we registered both images with external images: SDSS (u, g, r, i, and z) and the 2MASS point source catalog (Cutri et al. 2003). This provided us with six independent estimates of the transformations and relative astrometric uncertainty between the Chandra and HST images. Finally, to minimize the relative astrometric uncertainty, we computed an error-weighted average of the six independent transformations. In the Chandra physical (X,Y) coordinate system, the error-weighted averages of these six transformations yield the final astrometric errors $\overline{{\rm{\Delta }}X}=0\buildrel{\prime\prime}\over{.} 0583$ and $\overline{{\rm{\Delta }}Y}=0\buildrel{\prime\prime}\over{.} 1104$.

The best-fit positions of the soft, hard, and total X-ray sources, with astrometric shifts applied to align the positions in the HST coordinate frame, are (13:54:29.052, +13:27:56.70), (13:54:29.050, +13:27:56.81), and (13:54:29.047, +13:27:56.84), respectively. The errors on the positions are ($0\buildrel{\prime\prime}\over{.} 19$, $0\buildrel{\prime\prime}\over{.} 50$), ($0\buildrel{\prime\prime}\over{.} 07$, $0\buildrel{\prime\prime}\over{.} 27$), and ($0\buildrel{\prime\prime}\over{.} 06$, $0\buildrel{\prime\prime}\over{.} 20$), respectively. The X-ray positions are shown in Figure 2.

Zoom In Zoom Out Reset image size

Next, we measured the numbers of soft, hard, and total counts attributed to the X-ray source using the Bayesian Estimation of Hardness Ratios (BEHR) code described in (Park et al. 2006). BEHR takes as input the observed soft and hard counts from both the source region and a background region (which we measured using calc $\_$data $\_$sum). The code then uses a Bayesian approach to estimate the expected values and uncertainties of the soft counts, hard counts, total counts, and hardness ratio. Using this approach, we found $S={13.7}_{-4.2}^{+3.1}$ counts, $H={238}_{-17.0}^{+13.9}$ counts, and $T={252}_{-17.0}^{+14.8}$ counts.

We then used Sherpa to model the unbinned energy spectrum of the extracted region over the observed energy range of 2–8 keV. For our purposes, we are interested in the observed and intrinsic fluxes integrated over restframe soft, hard, and total energy ranges. Therefore, we fit the spectrum with a redshifted power law, $F={E}^{-{\rm{\Gamma }}}$, (intended to represent the intrinsic AGN X-ray emission at the SDSS spectroscopic redshift, ${z}_{\mathrm{SDSS}}$). We also included two multiplicative model components of absorbing neutral hydrogen column densities. The first absorbing component is fixed to the Galactic value, ${n}_{H,\mathrm{gal}}$, which we determined using an all-sky interpolation of the neutral Hydrogen in the Galaxy (Dickey & Lockman 1990). The second absorbing component is allowed to vary and assumed to be intrinsic to the source, ${n}_{H,\mathrm{exgal}}$, at the redshift ${z}_{\mathrm{SDSS}}$. All fluxes are k-corrected, with the observed values calculated from the model sum (including the absorbing components), while the intrinsic values are calculated from the unabsorbed power-law component. Because we have used Cash statistics, we did not subtract the background before modeling, but we have confirmed that the counts over the 2–8 keV range are dominated by the source region, with a negligible contribution from the background region.

For our first fit to the spectrum, we allowed Γ and ${n}_{H,\mathrm{exgal}}$ to vary freely. We found a best-fit value of ${\rm{\Gamma }}=-1.81$ (and a corresponding upper limit on the column density of $3.4\,\times {10}^{21}$ cm⁻²), which is not within the typical range of observed power-law indices for AGNs, i.e., $1\leqslant {\rm{\Gamma }}\leqslant 3$ (Nandra & Pounds 1994; Reeves & Turner 2000; Piconcelli et al. 2005; Ishibashi & Courvoisier 2010). Consequently, we fixed Γ at a value of 1.70 and ran the fit again. We obtained the best-fit model parameters for each combination of parameters by minimizing the Cash statistic using Sherpa's implementation of the Levenberg–Marquardt optimization method (Bevington 1969). The best-fit column density is ${1.96}_{-0.23}^{+0.29}\,\times {10}^{23}$ cm⁻², where the errors do not account for the uncertainty in Γ, and the reduced Cash statistic is 0.94. Figure 3 shows the best fit to the spectrum.

Zoom In Zoom Out Reset image size

The restframe hard absorbed (unabsorbed) X-ray fluxes are ${1.1}_{-0.1}^{+0.2}\times {10}^{-12}$ erg s⁻¹ cm⁻² (${2.8}_{-0.3}^{+0.3}\times {10}^{-12}$ erg s⁻¹ cm⁻²) for 2–10 keV. We then used the redshift ${z}_{\mathrm{SDSS}}$ to convert the X-ray fluxes to X-ray luminosities. The restframe hard absorbed (unabsorbed) luminosities are ${L}_{{\rm{X}},2-10\mathrm{keV}}={1.1}_{-0.1}^{+0.2}\,\times {10}^{43}$ erg ${{\rm{s}}}^{-1}({2.7}_{-0.3}^{+0.3}\times {10}^{43}$ erg ${{\rm{s}}}^{-1})$.

Finally, we searched for an X-ray source at the center of the companion galaxy SDSS J1354+1328, but found none (detection of ${0.001}_{-0.001}^{+0.002}$ counts in the restframe 0.5–10 keV energy range).

SDSS J1354+1327 was also observed with HST/WFC3 on UT 2014 April 28 (GO 13513, PI:Comerford), and the observations covered three bands: UVIS/F438W (B band, 954 s), UVIS/F606W (V band, 900 s), and IR/F160W (H band, 147 s). The HST field of view included SDSS J1354+1328, which is the companion galaxy to the northeast of SDSS J1354+1327. We used GALFIT V3.0 (Peng et al. 2010) to model the light profiles of these two galaxies. GALFIT is capable of decomposing images of galaxies into multiple components, including a Sérsic profile (e.g., a galactic stellar bulge), exponential disk (e.g., a galactic disk), and an image PSF for bright unresolved sources (e.g., AGNs).

We fit a Sérsic profile (plus a fixed, uniform sky component) to locate the position of each central stellar bulge, as Sérsic profiles have been empirically shown to be good approximations of the light profiles of stellar bulges (Graham & Driver 2005). To avoid modeling complex and irregular gas kinematics, we only ran GALFIT models on the F160W image because it does not contain significant line emission from ionized gas and it is sensitive to the central stellar bulge, which is the primary component of interest. As shown in Comerford et al. (2015), an additional unresolved point source is not necessary in this model because the AGN contributes negligibly to the F160W flux. The fit was run on a square region of projected physical size 40 kpc on each side to explore within 20 kpc of the AGN position.

We find that the position of SDSS J1354+1327's central stellar bulge is (13:54:29.045, +13:27:56.88), with errors of ($0\buildrel{\prime\prime}\over{.} 06$, $0\buildrel{\prime\prime}\over{.} 11$). The separation between SDSS J1354+1327 and the companion galaxy is 12.5 kpc ($10\buildrel{\prime\prime}\over{.} 2$). Using the results of the best-fit GALFIT model to each galaxy, we also find the H-band luminosity for SDSS J1354+1327 (${L}_{H}=4.0\,\times {10}^{10}\,{L}_{\odot }$) and for the companion galaxy (${L}_{H}=4.1\times {10}^{10}\,{L}_{\odot }$).

To explore the distribution of dust in the galaxy, we made a V − H color map (Martini et al. 2003). We convolved the V and H images with a Gaussian kernel to put them at the same angular resolution, and then computed the magnitude difference V − H. The resultant color map is shown in Figure 1 (right). We find that the center of SDSS J1354+1327 is the most obscured part of the galaxy ($V-H=2$).

Next, we turn our attention to the F438W and F606W observations, which cover a wavelength range that includes $H\delta$, $H\gamma$, and [O iii] λ4363 (F438W) and Hβ, [O iii] $\lambda \lambda$ 4959,5007, [O i] $\lambda \lambda$ 6300,6363, [N ii] $\lambda \lambda$ 6548,6583, and Hα (F606W). The F438W and F606W observations are dominated by ionized gas emission and not stellar emission, as the optical longslit observations show this to be the case for these wavelength ranges and since the morphologies of the F438W and F606W emission are different from the morphology of the stellar continuum (traced in the F160W observations). We identify two main structures in the F438W and F606W data:

Spatially extended ionized gas to the south. South of the galaxy center is a cone of clumpy, ionized gas (Figure 1, left panel). We map the gas detected in F606W at $\gt 10\sigma$ above the background, and we find that it is distributed in a cone-like structure with an opening angle of 67°. The gas detected at $\gt 3\sigma$ then extends ∼10 kpc radially from the galaxy center. The range of colors in the cone is $0.2\lt V-H\lt 1.2$, and the patchiness of the color indicates that in some regions the far side of the cone may be obscured. We explore this in more detail in Section 3.2.

Northern bubble. North of the galaxy center is a loop of emission that extends to a maximum distance of $0\buildrel{\prime\prime}\over{.} 6$ (0.7 kpc) from the galaxy center. The bubble's color is V−H = 0.7. By fitting a Sérsic component, we find that the brightest component of this bubble is located northeast of the galaxy center. We call this feature the NE source, and in F606W its apparent magnitude is 21.4 (compared with an apparent magnitude of 18.7 for the galaxy). The positions of the NE source in F438W (13:54:29.064, +13:27:57.10) and F606W (13:54:29.062, +13:27:57.10) agree to within $0\buildrel{\prime\prime}\over{.} 04$, so we take the NE source position to be the average of the positions measured in F438W and F606W. The NE source is located $0\buildrel{\prime\prime}\over{.} 33$ (0.40 kpc) from the galaxy center.

We observed SDSS J1354+1327 with APO/DIS on UT 2015 December 18 and UT 2016 June 30. We used a 1200 lines mm⁻¹ grating, centered so that the wavelength range covered Hβ and [O iii] on the blue channel and [N ii] and Hα on the red channel. We used a $1\buildrel{\prime\prime}\over{.} 5$ slit width, and the DIS pixel scales are $0\buildrel{\prime\prime}\over{.} 42$ on the blue channel and $0\buildrel{\prime\prime}\over{.} 40$ on the red channel. Following the approach used to study active galaxies with double-peaked narrow emission lines (Comerford et al. 2012; Nevin et al. 2016), we observed SDSS J1354+1327 at multiple different slit position angles to understand the kinematics of the features of the galaxy that are seen in the HST images. We observed the galaxy at three different slit position angles (given in degrees East of North): 0° to cover the emission seen in the north-south direction and the companion galaxy to the north, 153° to cover the bright clumps of emission seen to the south of the galaxy center, and 45° to cover the NE emission source. We observed for 1200 s at each position angle, and the data were reduced using standard IRAF procedures.

The longslit spectra are shown in Figure 4. We fit Gaussians to the [O iii] λ5007 and ${\rm{H}}\alpha$ flux at each position along the slit where the emission line is detected with a signal-to-noise ratio (S/N) $\gt 10$, and we extracted the line-of-sight velocities. These are shown in Figure 5.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Next, we identify where the peak of [O iii] λ5007 flux is located. The longslit orientation of position angle 45 has the strongest [O iii] λ5007 flux (integrated flux of $2.6\times {10}^{-13}$ erg s⁻¹ cm⁻², which is twice the integrated flux at position angle 0 and 2.4 times the integrated flux at position angle 153), and we fit a Gaussian to the spatial distribution of [O iii] λ5007 flux along this slit position angle. We find that the peak flux of [O iii] λ5007 occurs $0\buildrel{\prime\prime}\over{.} 41\pm 0\buildrel{\prime\prime}\over{.} 08$ to the north of the galaxy along position angle 45, which is spatially coincident with the NE source seen in the HST images. The velocity of the [O iii] λ5007 peak flux, relative to systemic, is 66.8 ± 11.5 km s⁻¹. This explains the velocity offset of 69.1 ± 15.0 km s⁻¹ seen in the SDSS spectrum of SDSS J1354+1327.

We then use emission line flux ratios in the BPT diagram (Baldwin et al. 1981; Veilleux & Osterbrock 1987; Kewley et al. 2006) to identify the sources of the ionized emission in SDSS J1354+1327. We measure these line flux ratios at each position along the slit where the S/N is $\gt 3$ for each of the four emission lines, and the results are shown in Figure 6. At all observed position angles, the ionized gas has Seyfert-like or composite line ratios.

Zoom In Zoom Out Reset image size

To probe the relative roles of photoionization and shocks in the galaxy, we measure the line flux ratios [O i] λ6300/Hα and He ii λ4686/Hβ at each position along the slit where the S/N is $\gt 3$ for each of the four emission lines. These line flux ratios discriminate between AGN photoionization and shock models (e.g., Clark et al. 1997; Moy & Rocca-Volmerange 2002), and we have illustrated these models in Figure 7. The photoionization models are computed with CLOUDY (Ferland 1996) and the shock models are computed with MAPPINGSIII (Dopita & Sutherland 1996).

Zoom In Zoom Out Reset image size

To explore the densities of the material within the galaxy, we also measure the line flux ratio [S ii] $\lambda \lambda$ 6717/6731 at each position along the slit where the S/N is $\gt 3$ for both emission lines. Then, we convert this to an electron density (Osterbrock & Ferland 2006; Schnorr-Müller et al. 2016) as shown in Figure 8. We find that the electron densities are consistent with those of a typical narrow-line region (${10}^{2}\lesssim {n}_{{\rm{e}}}$ (cm${}^{-3}$) $\lesssim {10}^{3};$ Osterbrock & Ferland 2006), and that the electron density is higher south of the galaxy center and lower north of the galaxy center.

Zoom In Zoom Out Reset image size

From the longslit spectra, we identify three main kinematic components in the galaxy:

Rotating disk. The rotating disk is most apparent at position angle 45 (Figure 5, middle), where the velocities progress from redshifts to the south to blueshifts to the north. We find that the Hα emission is best fit by a rotating disk that has a position angle on the sky of 53° East of North, is inclined 50° into the plane of the sky, has an intrinsic velocity of 182 km s⁻¹, has a radius of 2.5 kpc, and whose kinematic center is located at the galaxy's stellar centroid. This rotating disk is also seen in Paα, with the same disk parameters.

Spatially extended ionized gas to the south. South of the galaxy center, we find ionized gas that extends to distances $\gtrsim 4\,\mathrm{kpc}$ from the galaxy center. We focus our analysis on [O iii] λ5007 because it is the strongest emission line and because Hα is contaminated by the disk rotation. We find that the [O iii] λ5007 velocities south of the galaxy center range from 24 km s⁻¹ redshifted relative to systemic (along position angle 153) to 20 km s⁻¹ blueshifted relative to systemic (along position angle 153; Figure 5). The line flux ratios indicate a photoionized origin for emission south of the galaxy center (Figure 7).

Northern bubble. North of the galaxy center is ionized gas that extends ∼1 kpc from the galaxy center. The [O iii] λ5007 velocities north of the galaxy center range from 72 km s⁻¹ redshifted relative to systemic (along position angle 45) to 20 km s⁻¹ blueshifted relative to systemic (along position angle 0; Figure 5). The [O iii] λ5007 emission is strongest along position angle 45, which is the position angle aligned with the brightest component of the northern bubble seen in the HST observations. The line flux ratios at the position of the NE source are inconsistent with pure photoionization models, and the high [He ii] λ4686/Hβratio can instead be explained by a shock with a precursor H ii region (e.g., Sutherland et al. 1993); Moy & Rocca-Volmerange (2002) find regions of AGN emission with similar line flux ratios and interpret them also as produced by a shock with a precursor.

We observed SDSS J1354+1327 on UT 2013 April 3 using Laser Guide Star Adaptive Optics (LGS-AO) with Keck/OSIRIS IFS. We used the Kn1 filter with the $0\buildrel{\prime\prime}\over{.} 1$ pixel scale and the associated $3\buildrel{\prime\prime}\over{.} 6\times 6\buildrel{\prime\prime}\over{.} 4$ field of view to ensure that the observations would cover the entire area of a 3'' diameter SDSS fiber. The observations covered an observed wavelength range of 1.955–2.055 μm and detected the Paα 1.87 μm emission line, which traces the spatial distribution and kinematics of the ionized gas in the galaxy. We used the galaxy nucleus as the tip-tilt star, observed at a position angle of 0 degrees East of North, and integrated for 30 minutes. The data were reduced with the OSIRIS data reduction pipeline, and we used the IDL code LINEFIT (Davies et al. 2007) to create 2D kinematic and flux distribution maps (Figure 9).

Zoom In Zoom Out Reset image size

From the OSIRIS observations, we identify three main kinematic components in the galaxy:

Rotating disk. To model the rotating disk, we first mask the pixels in the Paα velocity map that have a flux density lower than 10% of the peak of the continuum emission. We then use the KINEMETRY code (Krajnović et al. 2006) to model the rotating disk structure seen extending from the northeast (blueshifted) to the southwest (redshifted) in Paα. The parameters of the best-fit model are a position angle of 55 degrees east of north, an inclination of 53 degrees, and a kinematic center that is spatially offset by $0\buildrel{\prime\prime}\over{.} 2$ northeast of the peak of the stellar continuum. The parameters of the rotating disk observed in Paα are consistent with those of the rotating disk observed in the optical longslit spectra. After the disk model is subtracted from the data, we examine the velocity residuals, which have ±20 km s⁻¹ errors (Figure 9, bottom right panel). The velocity residuals are not significant enough to fit a secondary component. We find that the kinematic center of the rotating disk is offset by $0\buildrel{\prime\prime}\over{.} 2$ (0.3 kpc) from the stellar continuum center, and we attribute this perturbation to a recent pericenter passage and interaction with the companion galaxy.

Spatially extended ionized gas to the south. While the OSIRIS field of view is dominated by the rotating disk, there is also extraplanar Paα emission visible to the south of the galaxy center, and this emission is detected at $5\sigma$ significance out to the edge of the field of view (Figure 9, top middle). The line-of-sight velocities of the gas are ±30 km s⁻¹. The Paα emission also follows the conical structure seen in the HST observations.

Northern bubble. There is spatially extended Paα emission associated with the loop seen in the HST data.

The X-ray source in SDSS J1354+1327 has an observed luminosity of ${L}_{{\rm{X}},2-10\mathrm{keV}}=1.1\times {10}^{43}$ erg s⁻¹, which is brighter than the typical AGN luminosity threshold of ${L}_{{\rm{X}},2-10\mathrm{keV}}\gt {10}^{42}$ erg s⁻¹ and shows that the X-ray source is associated with an AGN. Mid-infrared colors from the Wide-field Infrared Survey Explorer (WISE) also show that an AGN is present (SDSS J1354+1327 has $W1-W2=1.14\pm 0.03$, where $W1-W2\geqslant 0.8$ indicates that the mid-infrared flux is dominated by an AGN; Stern et al. 2012).

The fit to the X-ray spectrum reveals a relatively high column density ${n}_{H,\mathrm{exgal}}=2\times {10}^{23}$ cm⁻², which indicates that the source is an X-ray absorbed AGN. This is consistent with the V − H dust map, which shows that the very center of the galaxy is the most obscured (Figure 1, right panel).

Next, we consider the position of the X-ray source within the host galaxy. We have identified the position of the X-ray source, the positions of the central stellar nucleus of the galaxy and the NE emission source visible in the HST F438W and F606W observations, and performed astrometric corrections. We compare the positions to determine if the X-ray AGN position aligns with any of these galaxy features (Figure 2, right panel). If the X-ray AGN is associated with the NE source, then the AGN is located 0.40 kpc from the center of SDSS J1354+1327 and this spatial offset could be explained as an offset AGN, a gravitational recoil AGN, or a gravitational slingshot AGN. We find that the position of the X-ray AGN is more consistent with the position of the stellar centroid of the galaxy (consistent to within $\lesssim 1\sigma$) than with the position of the NE source (consistent to within $\lesssim 2.5\sigma$). Consequently, we conclude that the X-ray AGN is most likely located at the galaxy center.

There is extraplanar ionized gas extending south of the galaxy center, as seen in the HST F606W and F438W observations, the APO/DIS optical longslit observations, and the Keck/OSIRIS near-infrared IFS observations. The gas is clumpy, as seen morphologically in the HST data (Figure 1, left panel) and in the mixture of redshifted and blueshifted line-of-sight velocities of the emission lines observed with APO/DIS (Figure 5).

Here, we aim to determine whether the gas is inflowing, outflowing, or passively photoionized by the central AGN. If the AGN is passively photoionizing the gas in the galaxy, then there should be symmetric photoionized cones on either side of the AGN due to the collimating torus (e.g., Schmitt et al. 2003), but this is not the case for SDSS J1354+1327. In the passive photoionization scenario, the gas also should be rotating (with possible deviations due to the interaction with the companion galaxy) because it is following the galaxy potential. We fit a rotating disk to the optical and near-infrared spectra, and find that the rotating disk is oriented with a position angle 54 degrees east of north, but no rotation is seen in the gas extending to the south of that disk. Instead, the southern gas has blueshifts and redshifts observed along directions not coincident with the axis of rotation, as can be seen in the velocities in the longslit data. This is evidence of an inflow or outflow, rather than passively photoionized gas.

The morphology of the southern gas distinguishes between the inflow and outflow scenarios. As shown in the HST images (Figure 1, left panel), the gas has a conical morphology with the cone beginning at the galaxy center and extending to the south, outside the plane of rotation of the galaxy. This conical morphology is typical of AGN outflows (e.g., Mulchaey et al. 1996; Schmitt et al. 2003), where the torus provides the collimation (e.g., Antonucci & Miller 1985; Malkan et al. 1998). In contrast, conical morphologies are not expected for inflows, which are typically radial streamers of gas (e.g., Iono et al. 2004; Müller-Sánchez et al. 2009).

We conclude that the spatially extended ionized gas to the south is an outflow. We apply an analytic Markov Chain Monte Carlo model to the longslit observations to model the data as a biconical outflow, as done for similar longslit observations of galaxies in Nevin et al. (2017). We find that the data are best fit (with a reduced ${\chi }^{2}=1.5$) by a bicone extending south of the galaxy center at an inclination of ${27}_{-22}^{+14}$ degrees out of the plane of the sky, a position angle of ${190}_{-25}^{+26}$ degrees east of north on the plane of the sky, a half opening angle of ${37}_{-15}^{+10}$ degrees, a turnover radius of ${1.9}_{-1.7}^{+2.4}$ kpc, a maximum velocity at the turnover radius ${V}_{\max }={40}_{-30}^{+40}$ km s⁻¹, and a lateral surface area $A={4}_{-4}^{+14}$ kpc². The observed blueshifted and redshifted velocities of the gas imply that we are seeing parts of the front and rear facing walls of the outflow (e.g., Bae & Woo 2016), which fits with the clumpiness of the gas and dust partially obscuring the view (Figure 1, right). When we project the cone onto the plane of the sky, we find an observed opening angle of ${66}_{-27}^{+18}$ degrees, which is consistent with the opening angle of 67 degrees measured from the HST data.

Then, we use the best-fit parameters of the outflow model, our measurement of the electron density n_e, and assume a filling factor f = 0.01 to be conservative in our energy calculation to estimate the mass outflow rate $\dot{M}={m}_{{\rm{p}}}{n}_{{\rm{e}}}{V}_{\max }{fA}$, where m_p is the proton mass. We find $\dot{M}={6.5}_{-6.3}^{+52.3}\ {M}_{\odot }\ {\mathrm{yr}}^{-1}$.

This outflow could be driven by either star formation or by an AGN. First, we consider the scenario of a star formation driven outflow. To set an upper limit on the star formation rate (SFR) in SDSS J1354+1327, we assume that all of the ${\rm{H}}\alpha$ emission is associated with star formation. Using the luminosity measured from the SDSS spectrum, ${L}_{{\rm{H}}\alpha }\,=1.4\times {10}^{41}$ erg s⁻¹, the SFR from Kennicutt & Evans (2012) is then $\mathrm{SFR}=5.37\times {10}^{-42}\ {L}_{{\rm{H}}\alpha }=0.75\ {M}_{\odot }\ {\mathrm{yr}}^{-1}$. When we convert this SFR to a mass outflow rate, we find an upper limit of ${\dot{M}}_{\mathrm{SF}}=0.2\ {M}_{\odot }\ {\mathrm{yr}}^{-1}$ (Veilleux et al. 2005), which is too weak to drive the observed outflow. We also note that a star formation driven outflow that is powered by supernovae in the disk of the galaxy typically has a wide, chimney-shaped morphology of the gas being driven from the disk into the halo (e.g., Norman & Ikeuchi 1989), which is inconsistent with the conical morphology of the ionized gas observed in SDSS J1354+1327.

Consequently, we conclude that the outflow is AGN driven. This is consistent with the [O iii] λ5007/Hβ and [N ii]/Hα line flux ratios at the location of the outflow, which show that the emission is Seyfert driven. Further, the [O i] λ6300/Hα and [He ii] λ4686/Hβ line flux ratios and the conical morphology each indicate that the emission is photoionized (e.g., Wilson & Tsvetanov 1994). The AGN luminosity is also sufficient for photoionizing the cone of gas, based on photoionization models computed using the spectral synthesis code CLOUDY (Ferland et al. 2013; Richardson et al. 2014). We conclude that the spatially extended ionized gas to the south is a photoionized AGN outflow.

Although we do not know the velocity of the outflow at the time it was launched, we can use the current observed velocity and spatial extent of the outflow to find an upper limit of 10⁸ years on the timescale for the outflow to produce the cone of gas. The light-travel timescale plus the recombination timescale (${t}_{\mathrm{rec}}\approx {(\alpha {n}_{{\rm{e}}})}^{-1}$, where the recombination coefficient $\alpha =1.72\times {10}^{-11}$ cm³ s⁻¹ for [O iii] and n_e is from Section 2.4; Osterbrock & Ferland 2006) to illuminate the gas is then ∼50 years, which is negligible. Given that the emission can be observed for up to 10⁵ years (e.g., Schawinski et al. 2015), we conclude that the outflow was likely launched $\lesssim {10}^{5}\,\mathrm{years}$ ago.

The HST F438W and F606W observations reveal a loop of gas extending 0.7 kpc north of the galaxy center, as well as a NE source that is located 0.40 kpc northeast of the galaxy center and is embedded in the loop of emission (Figure 1, left panel). The peak of [O iii] λ5007 emission is located at the NE source as well.

As we did with the spatially extended ionized gas to the south, we consider whether the northern bubble is caused by gas that is inflowing, outflowing, or passively photoionized by the AGN. First, the off-nuclear peak of the [O iii] λ5007 emission indicates a shock at the position of the NE source and explains the velocity-offset emission line in the SDSS spectrum. We also find that the observed line flux ratios cannot be explained by pure photoionization, and instead are explained by shocks. Further evidence for shocks comes from the looped morphology of the northern bubble, which indicates that shocks are being driven into the ambient gas (e.g., Faucher-Giguère & Quataert 2012; Gabor & Bournaud 2014). Where we see redshifted velocities, we are seeing the back side of the bubble.

The presence of shocks rules out the scenario of an AGN passively photoionizing the northern gas in SDSS J1354+1327, and leaves inflowing or outflowing gas as the remaining scenarios. Inflowing gas typically streams radially toward the galaxy center (e.g., Iono et al. 2004; Müller-Sánchez et al. 2009), and the bubble morphology of the observed gas is difficult to explain for an inflow. In contrast, many AGN outflows are seen blowing bubbles of gas (e.g., Fabian 2012; Greene et al. 2014). We conclude that the northern bubble is an AGN outflow driving shocks into the ambient material.

The original velocity of the outflow is unknown, but we measured the current observed velocity and spatial extent of the bubble. Using these values, we estimate that the outflow was launched $\lesssim {10}^{7}\,\mathrm{years}$ ago. Taking into account the length of time that the emission signatures can linger, the outflow must have been launched $\lesssim {10}^{5}\,\mathrm{years}$ ago.

SDSS J1354+1327 has a companion galaxy, SDSS J135429.17+132807.3, which is located 12.5 kpc to the northeast and is redshifted by 76 km s⁻¹ relative to SDSS J1354+1327. Emission line diagnostics of the companion's SDSS spectrum suggest that it hosts an AGN, and Liu et al. (2011) also noted that SDSS J1354+1327 and SDSS J1354+1328 are a pair of active galaxies. However, there is no WISE source detected at the location of the companion galaxy. Further, we do not detect an AGN in our Chandra observations of the companion galaxy.

The lack of an X-ray detection suggests that the AGN is obscured, the AGN's X-ray luminosity is low on the predicted [O iii] λ5007 to X-ray luminosity scaling relation, some of the gas covered by the SDSS fiber is ionized by SDSS J1354+1327's AGN, or that the [O iii] λ5007 emission is a light echo of past AGN activity (e.g., Lintott et al. 2009; Schawinski et al. 2015; the AGN may be flickering on and off, or the SMBH may have been ejected from the host galaxy center). The other possibility is that the [O iii] λ5007 luminosity is not associated with an AGN. In this scenario, the observed line flux ratio (log ([O iii]/Hβ) = 0.44) could be produced by a starburst instead of an AGN, given that models predict that starbursts can produce line ratios that high. Given that the dust map indicates obscuration at the galaxy center ($V-H=2.4$), we conclude that the galaxy most likely hosts an obscured AGN.

The HST observations show tidal tails of stars connecting the primary and companion galaxies, indicating that they are already interacting (Figure 2, bottom left panel). Taking the mass ratio to be the luminosity ratio of the two stellar bulges, the merger's mass ratio is 0.98 (SDSS J1354+1327 is slightly less massive).