NuSTAR Hard X-Ray Data and Gemini 3D Spectra Reveal Powerful AGN and Outflow Histories in Two Low-redshift Lyα Blobs

NuSTAR carries two independent focal plane modules (FPMA and FPMB), sensitive in the 3–80 keV range. We reprocess the event files using the standard nupipeline script as described in the "NuSTAR Analysis Quickstart Guide."⁸ Our targets are faint (∼10⁻³ cts s⁻¹), and periods of high background (such as passages through or near the South Atlantic Anomaly (SAA)) must be excluded. Typical background rates observed with NuSTAR are ≲1 cts s⁻¹, integrated over the focal plane (Forster et al. 2014). Times of high background can be identified by simultaneously increased count rates in the detectors as well as in the shields that surround the focal planes. Using the telemetry reports made by the NuSTAR team, we first check the total event rates during all orbital passages of our observations. In the J1155 observation, the event rate increases around the standard SAA area (above ∼30 cts s⁻¹). We run nupipeline with the option saamode = optimized to reject times with high count rates in both the shields and detectors. High count rates may occasionally occur in the so-called tentacle region (Forster et al. 2014) near the SAA, and we exclude these times for J1155 by setting tentacle = yes. Background rates during the observation of J0113 are stable and low (saamode = none and tentacle = no).

The FWHM of the NuSTAR point-spread function is ∼20''. The source spectra are extracted from circular regions with a radius of 1' centered on the centroid of the counts. The background spectra are extracted from identical apertures, located in an off-source region on the same detector. To increase the signal-to-noise ratios, the FPMA and FPMB spectra are combined without renormalization due to lack of statistical power in the count rates using the addascaspec command in FTOOLS. J0113 and J1155 are significantly detected in the 3–30 keV band with a signal-to-noise ratio of ∼8 and 20, respectively.

The Chandra/ACIS data provide the soft energy band (0.5–10 keV) for the spectral analysis. The raw data are reprocessed by using the chandra_repro pipeline included in the Chandra Interactive Analysis of Observations (CIAO) software (version 4.8) and referring to the latest CALDB at the time (version 4.7.2). The typical angular resolution of Chandra/ACIS is ∼1''. Our targets are barely resolved in the Chandra data, and no offsets are observed between the X-ray and [O iii] peak emission (see Schirmer et al. 2016, for details). The source spectra were extracted from a circular region with 2'' radius, centered on the Chandra detections. The background spectra are taken from similar off-source regions. Background correction is minimal, as the source count rates are ∼100 times higher. Consequently, we detect J0113 and J1155 at the 11σ and 23σ levels, respectively, in the 0.5–10 keV band.

We analyze the Swift/XRT spectra covering the soft energy band (0.3–10 keV). We find low detection significances of 2.3σ and 3.6σ for J0113 and J1155, respectively. This is likely due to the low exposure and effective area. Thus, these spectra are ignored in the detailed broadband X-ray spectral analysis. Here, we only constrain the observed fluxes to confirm the consistency with those estimated from the NuSTAR spectra. The power-law model is fitted based on the C-statistics appropriate to low-counts spectra instead of the chi-squared method. The 2–10 keV fluxes are found to be $\lt 2.0\times {10}^{-13}$ erg cm⁻² s⁻¹ for J0113 and ${4.7}_{-2.9}^{+6.3}\times {10}^{-13}$ erg cm⁻² s⁻¹ for J1155. They agree well with those from the NuSTAR spectra (see Section 3 and Table 2).

Table 2.  Best-fit Parameters

(1)	Name	J0113	J1155
(2)	${N}_{{\rm{H}}}^{\mathrm{Gal}}$ ($\times {10}^{20}$)	3.03	2.04
(3)	${N}_{{\rm{H}}}$ ($\times {10}^{22}$)	${51}_{-33}^{+31}$	${6.0}_{-1.1}^{+1.4}$
(4)	Γ	${2.1}_{-1.2}^{+0.5}$	${1.9}_{-0.4}^{+0.3}$
(5)	${A}_{\mathrm{PL}}^{{Chandra}}$ $(\times {10}^{-4})$	${9.0}_{-8.8}^{+29.0}$	${1.9}_{-0.8}^{+0.9}$
(6)	${A}_{\mathrm{PL}}^{{NuSTAR}}$ $(\times {10}^{-4})$	${1.6}_{-1.0}^{+11.0}$	${3.8}_{-1.9}^{+2.6}$
(7)	${f}_{\mathrm{scat}}$	$\lt 9.9$	${1.2}_{-0.9}^{+1.6}$
(8)	${A}_{{\rm{K}}\alpha }\ (\times {10}^{-6})$	$\lt 2.2$	1.9 ± 1.3
(9)	${R}_{\mathrm{ref}}$	2.7a	$1.0(\lt 2.5)$
(10)	${F}_{2-10}^{{Chandra}}$ ($\times {10}^{-13}$)	3.6	3.6
(11)	${F}_{2-10}^{{NuSTAR}}$ ($\times {10}^{-13}$)	1.3	6.1
(12)	$\mathrm{log}{L}_{2-10}^{{Chandra}}$	${44.5}_{-0.6}^{+0.5}$	${44.0}_{-0.2}^{+0.1}$
(13)	$\mathrm{log}{L}_{2-10}^{{NuSTAR}}$	43.7 ± 0.8	44.3 ± 0.1
(14)	${\chi }^{2}/\mathrm{dof}$	$3.3/6$	$28.9/35$
(15)	NuSTAR cts (src/bg)	212/451	797/717
(16)	Chandra cts (src/bg)	117/1	526/1

Note. Columns: (1) Galaxy name. (2) Galactic absorption in units of cm⁻². (3) Intrinsic absorption in units of cm⁻². (4) Photon index of the power-law component. (5)–(6) Photon fluxes in units of photons keV⁻¹ cm⁻² s⁻¹ at 1 keV of the primary power-law component in the Chandra and NuSTAR spectra. (7) Scattered fraction in units of percent. (8) Total photon flux in the zgauss model in units of total photons cm⁻² s⁻¹. (9) Reflection strength defined as ${R}_{\mathrm{ref}}=2\pi /{\rm{\Omega }}$ in the pexrav model. (10)–(11) Observed fluxes in the 2–10 keV band in units of erg cm⁻² s⁻¹. (12)–(13) Intrinsic (primary power-law component) luminosities at the galaxy rest-frame in the 2–10 keV band. (14) Reduced chi-squared over degrees of freedom. (15)–(16) Source and background counts in the NuSTAR3–30 keV and Chandra 0.5–10 keV bands.

^aThe error at the 90% confidence level cannot be constrained within a range, where we allowed the parameter to vary (${R}_{\mathrm{ref}}=0\mbox{--}10$).

Download table as:  ASCIITypeset image

The 3D spectra were obtained using GMOS at the Gemini-South telescope in Chile, using the IFU-2 configuration. The 50 × 70 field of view is sampled with 02 fibers, and corresponds to $22\times 31\,\mathrm{kpc}$ at z = 0.3.

J0113 was observed on 2016 November 06 UT (Gemini program ID GS-2016B-DD-4) using $4\times 1200\,{\rm{s}}$ integrations in excellent seeing conditions (03–04); the physical resolution corresponds to 1.5 kpc at z = 0.281. The pointing was centered on the nucleus and contains the bulk of the flux, including the two superbubbles (see Figure 2, left panel). The [O iii]λ5008 line was compromised by a bad detector column that had developed around the time of the observations. The [O iii]λ5008 line luminosity and principal velocity structures could be reconstructed from the weaker [O iii]λ4960 line because of the fixed [O iii]λ5008/4960 line ratio of 2.984 (Storey & Zeippen 2000) and the good signal-to-noise ratio (S/N).

Zoom In Zoom Out Reset image size

J1155 was observed on 2015 March 31 (GS-2014B-Q-63). Because of its large physical extent, three adjacent pointings were obtained with the IFU-2 (Figure 2, right panel), exposed for $2\times 900\,{\rm{s}}$ each, in ∼07–08 seeing conditions. The physical resolution is 3.5 kpc.

The spectra were processed using Python/IRAF scripts⁹ developed for GMOS IFU data. A detailed description of the reduction can be found in Davies et al. (2015) for SDSS J2240−0927, a low-redshift LAB in our sample for which we have performed extensive emission line diagnostics.

The ionized gas kinematics were derived using the [O iii] emission lines. For J1155, the [O iii]λ4960 and [O iii]λ5008 lines were both used. For J0113, the brighter [O iii]λ5008 line from half the IFU field of view was projected on a bad detector column and lost, and therefore only the weaker line was used. For each spaxel, a constant continuum value was subtracted, and then each emission line was fit with a single Gaussian. If both lines were fit, then we used their fixed line ratio and required the same velocity and velocity dispersion.

The base-line model fits the spectra with ${\chi }^{2}/\mathrm{dof}=3.3/6$. The AGN in J0113 is heavily obscured with a hydrogen column density of ${N}_{{\rm{H}}}={5.1}_{-3.3}^{+3.1}\times {10}^{23}$ cm⁻², and we find ${\rm{\Gamma }}={2.1}_{-1.2}^{+0.5}$. The Kα line and the scattered emission are insignificant, and their upper limits are constrained to be ${A}_{{\rm{K}}\alpha }\lt 2.2\times {10}^{-6}$ photons cm⁻² s⁻¹ ($\lt 1.3\times {10}^{-14}$ erg cm⁻² s⁻¹) and ${f}_{\mathrm{scat}}\lt 9.9 \%$, respectively. The reflection strength is estimated to be ${R}_{\mathrm{ref}}=2.7$. If we assume that the incident flux into the reflecting and scattering material is based on the primary X-ray emission in the Chandra spectrum (instead of NuSTAR), then ${R}_{\mathrm{ref}}\sim 0.5$ and ${f}_{\mathrm{scat}}\lesssim 4 \%$. Due to the poor statistics, the best-fit parameters in both cases are consistent with each other.

From the confidence contours in Figure 3, the intrinsic luminosities during the Chandra and NuSTAR observations are $\mathrm{log}({L}_{2-10\mathrm{keV}}/\mathrm{erg}\ {{\rm{s}}}^{-1})={44.5}_{-0.6}^{+0.5}$ and 43.7 ± 0.8, respectively (90% confidence errors on the luminosities). This AGN has likely faded by one order of magnitude in two years.

The base-line model reproduces the spectra well with ${\chi }^{2}/\mathrm{dof}\,=28.9/35$. We find ${\rm{\Gamma }}={1.9}_{-0.4}^{+0.3}$ and ${N}_{{\rm{H}}}={6.0}_{-1.1}^{+1.4}\times {10}^{22}$ cm⁻². The Kα line is detected at the 90% confidence level with ${A}_{{\rm{K}}\alpha }=(1.9\pm 1.3)\times {10}^{-6}$ photons cm⁻² s⁻¹, or $(1.9\,\pm 1.4)\times {10}^{-14}$ erg cm⁻² s⁻¹. This supports the existence of a torus. Although the spectra do not require the reflection continuum (pexrav) significantly as ${R}_{\mathrm{ref}}=1.0$ $(\lt 2.5)$, this continuum emission should be taken into consideration to explain the Kα line in a self-consistent way. The scattered emission is significantly detected below 1 keV, and the scattering fraction is ${f}_{\mathrm{scat}}={1.2}_{-0.9}^{+1.6} \%$. This falls within a typical range of obscured Seyfert galaxies (0.1%–10%; e.g., Kawamuro et al. 2016). Note that ${R}_{\mathrm{ref}}$ and ${f}_{\mathrm{scat}}$ with respect to the primary emission in the Chandra spectrum are ${R}_{\mathrm{ref}}\approx 2$ and ${f}_{\mathrm{scat}}\approx 2 \%$, consistent with those derived above within uncertainties. Analogously to J0113, we find $\mathrm{log}({L}_{2-10\mathrm{keV}}/\mathrm{erg}\ {{\rm{s}}}^{-1})={44.0}_{-0.2}^{+0.1}$ and 44.3 ± 0.1 for the epochs of the Chandra and NuSTAR observations, respectively. This AGN has brightened slightly between 2002 and 2016.

J0113 extends over $7^{\prime\prime} \times 13^{\prime\prime}$, featuring two prominent superbubbles at a projected distance of 5 kpc on either side of the nucleus. The bubbles have a diameter of 5–8 kpc each. A further outflow is found 24 kpc (northwest) from the nucleus and measures 14 kpc. Judging from its r-band brightness and the [O iii] equivalent line width of the central region, this outflow contributes 15%–20% to the total [O iii] luminosity, which we disregard in the further analysis. For more details see Figure 2 in this paper and Appendix A4 in Schirmer et al. (2016).

The total [O iii]λ5008 flux in the IFU-2 data is $(4.36\pm 0.44)\times {10}^{-14}$ erg cm⁻² s⁻¹, equivalent to an observed (reddened) luminosity of $1.1\times {10}^{43}$ erg s⁻¹. The archival spectrum from the Sloan Digital Sky Survey (SDSS) has a mean global Hα/Hβ line ratio (Balmer decrement) of 4.4, indicative of significant dust extinction. Using a standard R = 3.1 dust extinction curve (Cardelli et al. 1989; Osterbrock & Ferland 2006, their Table 7.1) and an intrinsic Hα/Hβ line ratio of 3.0, we find $\mathrm{log}({L}_{[{\rm{O}}\,{\rm{III}}]}^{c}/\mathrm{erg}\ {{\rm{s}}}^{-1})=43.6\pm 0.1$ for the dereddened luminosity. Likewise, we measure a corrected Hβ line luminosity of $\mathrm{log}({L}_{{\rm{H}}\beta }^{c}/\mathrm{erg}\ {{\rm{s}}}^{-1})=42.9\pm 0.1$. The 1σ errors include the uncertainty of the flux calibration and allow for a relative uncertainty of 10% in the intrinsic Hα/Hβ line ratio; the latter is subject to metallicity and hardness of the ionizing spectrum (Gaskell & Ferland 1984). Possible variations of the internal dust extinction remained unaccounted for because we do not have resolved Hα maps at hand (for examples, see Schirmer et al. 2013; Davies et al. 2015).

The kinematic maps in Figure 4 reveal a bipolar outflow. Relative radial velocities increase linearly from the nucleus to about 120 km s⁻¹ at the position of both superbubbles at 5 kpc distance. The northern superbubble is blueshifted, and the southern bubble is redshifted. Velocities then continue to increase in the same fashion beyond the bubbles, exceeding 200 km s⁻¹ at 10 kpc from the nucleus, where the S/N becomes too low for our analysis. The outflow does not appear to encounter significant mechanical resistance from the interstellar medium, and its motion is not yet governed by gravitation. The measured radial velocity is fairly low, suggesting that most of the motion happens perpendicular to our line of sight.

Zoom In Zoom Out Reset image size

Another feature is that the gas southwest (northeast) of the nucleus is redshifted (blueshifted). This velocity gradient is perpendicular to the outflow and with a lower amplitude of about 70 km s⁻¹. There is no indication of an outflow in this direction, and we interpret this as rotational motion of the central cloud of ionized gas.

The velocity dispersion is low (Figure 4, upper middle), and it is remarkable that the nucleus is not detectable in this map. We note a band of increased dispersion, 150–180 km s⁻¹, that appears to gird the nucleus and the bipolar outflow near its equatorial plane. In the two superbubbles and beyond, the dispersion drops to $\lesssim 100$ km s⁻¹, which is lower than in other outflow systems. We discuss these properties in Sections 5.2.3 and 5.2.4.

J1155 has the highest X-ray flux in our sample. The optical nebula in the gri broadband images of Schirmer et al. (2016) extend over $9^{\prime\prime} \times 19^{\prime\prime}$ or more than 60 kpc, and fragments into numerous smaller clouds of 3–7 kpc diameter scattered around a very bright center measuring 5 kpc in diameter. For more details see Figure 2 in this paper and Appendix A6 in Schirmer et al. (2016).

We have no spectrum yet of J1155 that includes the Hα line, and therefore we cannot determine a dereddened [O iii] luminosity. The observed total line flux including calibration errors is $(8.0\pm 0.8)\times {10}^{-14}$ erg cm⁻² s⁻¹, corresponding to $\mathrm{log}({L}_{[{\rm{O}}{\rm{III}}]}/\mathrm{erg}\ {{\rm{s}}}^{-1})=43.4\pm 0.05$. If the Balmer decrement were a moderate 3.5, then $\mathrm{log}({L}_{[{\rm{O}}\,{\rm{III}}]}^{c}/\mathrm{erg}\ {{\rm{s}}}^{-1})=43.6\pm 0.1$ and $\mathrm{log}({L}_{{\rm{H}}\beta }^{c}/\mathrm{erg}\ {{\rm{s}}}^{-1})=42.5\pm 0.1$. If the Balmer decrement were as high as for J0113, then $\mathrm{log}({L}_{[{\rm{O}}\,{\rm{III}}]}^{c}/\mathrm{erg}\ {{\rm{s}}}^{-1})=43.9\pm 0.1$. For the rest of this paper we adopt the more conservative lower luminosities associated with less extinction.

J1155 has a very complex kinematic structure and [O iii] morphology. The line emission is strongly concentrated in the center of the galaxy and has a low velocity dispersion. Ionized gas extends out to ∼25 kpc in these shallow IFU data, creating a patchy appearance and complex velocity and velocity dispersion maps. Particularly interesting is a long stream of gas connected to one side of the nucleus, stretching over 20 kpc and receding with a constant radial velocity of ∼120 km s⁻¹. The width of the stream is barely resolved in the data (Figure 4, lower left); the velocity dispersion is increased farther away from the nucleus. The remainder of the surrounding gas appears to be passively illuminated by the AGN.

Like J0113, the velocity dispersion in J1155 is fairly low, between 100–200 km s⁻¹ across the entire analysis area of $30\times 15\,\mathrm{kpc}$. The nucleus is not detectable either (Figure 4, lower middle).

5.1.1. Constraint on the X-Ray Luminosity

5.1.2. Narrow Iron-Kα Line

5.1.3. Scattering Fraction

5.2.1. Comparison Samples

5.2.2. Ionized Gas Mass

We derived Hβ logarithmic line luminosities of 42.9 and 42.5 for J0113 and J1155, respectively (Sections 4.1 and 4.3). Assuming the gas is purely photoionized, the gas mass M associated with the Hβ emission is estimated as

$$\begin{eqnarray}&&M=\displaystyle \frac{{L}_{{\rm{H}}\beta }}{4\pi {j}_{{\rm{H}}\beta }}\,{n}_{{\rm{e}}}\,{m}_{{\rm{p}}},\end{eqnarray} \tag{ 2 }$$

where $4\pi {j}_{{\rm{H}}\beta }$ is the emission per unit volume, ${n}_{{\rm{e}}}$ is the electron number density, and ${m}_{{\rm{p}}}$ is the proton mass. Taking $4\pi \,{j}_{{\rm{H}}\beta }$ from Osterbrock & Ferland (2006) (their Table 4.4; see also Storey & Sochi 2015) for $T={10}^{4}$ K and ${n}_{{\rm{e}}}=100$ cm⁻³, we rewrite

$$\begin{eqnarray}&&M=6.77\times {10}^{8}\,{M}_{\odot }\,\displaystyle \frac{{L}_{{\rm{H}}\beta }}{{10}^{43}\,\mathrm{erg}\,{{\rm{s}}}^{-1}}{\left(\displaystyle \frac{{n}_{{\rm{e}}}}{100{\mathrm{cm}}^{-3}}\right)}^{-1}.\end{eqnarray} \tag{ 3 }$$

Individual measurements of ${n}_{{\rm{e}}}$, derived from the [S ii]λ6719,33 line ratio, yield several 100 cm⁻³ in the inner regions of narrow-line regions (see e.g., Nesvadba et al. 2006; Davies et al. 2015). Using line ratios that are sensitive at higher densities, ${n}_{{\rm{e}}}\sim {10}^{3}$ (10^4–5) cm⁻³ were found for the narrow (broad) lines of another luminous quasar (Holt et al. 2011). In the outskirts, densities drop below 100 cm⁻³, where the outflows have thinned and are difficult to detect (and where line ratio methods break down because the collision effects diminish). Our spectrum of J0113 does not include the [S ii] doublet; however, there is an SDSS spectrum taken with a 3''fiber centered on the nucleus, including the flux from the nucleus and most of the superbubbles. Fitting the [S ii] doublet with a simple double Gaussian with identical line widths, we derive a line ratio of 1.24, i.e., a mean ${n}_{{\rm{e}}}\sim 180$ cm⁻³. We expect the actual electron densities to deviate significantly from this number because outflows are known to be clumpy (Nesvadba et al. 2006; Greene et al. 2011; Davies et al. 2015).

Using Equation (3), we find $5.4\times {10}^{8}\,{M}_{\odot }$ and $2.0\times {10}^{8}\,{M}_{\odot }$ for J0113 and J1155, respectively, for a fixed value of ${n}_{{\rm{e}}}=100$ cm⁻³. This is near the upper end of the gas masses of $(0.2\mbox{--}4)\times {10}^{8}\,{M}_{\odot }$ found by Harrison et al. (2014) using the same density normalization. The outflows described in Liu et al. (2013) entrain higher masses, $(2\mbox{--}10)\times {10}^{8}\,{M}_{\odot }$ (again for ${n}_{{\rm{e}}}=100$ cm⁻³).

We note that Harrison et al. (2014) and Liu et al. (2013) have a numeric prefactor that is erroneously about a factor 4 too high compared to ours in Equation (3). This is caused by an incorrect calculation of the recombination coefficient ${\alpha }_{{\rm{H}}\alpha }^{\mathrm{eff}}$ in Equation (1) of Nesvadba et al. (2006) (N. Nesvadba 2017, private communication). Taking this into account, the ionized gas masses of J0113 and J1155 are among the highest measured in type-2 AGN at $z\lesssim 0.6$.

5.2.3. Gas Velocities

Assuming a typical true outflow velocity of 1000 km s⁻¹ for the [O iii] gas (Harrison et al. 2014), it would take the two superbubbles in J0113 at least ∼10⁷ years to reach their observed positions. However, a sustained high AGN output over such a long time is implausible, as single AGN duty cycles are expected to be comparatively short (10^4–6 years; e.g., Schawinski et al. 2010, 2015; Sun et al. 2017) and followed by longer low states.

What happens with such an outflow during the low states of the AGN? King et al. (2011) calculate the internal properties of outflows and their dynamics. They find that the thermal energy contained in an energy-driven outflow can support its expansion for periods 10 times longer than the actual quasar phase that launched the outflow. However, their numerical results do not show that the velocity in such a stream is increasing over kpc-scales from the nucleus, as we observe in J0113. In a more recent work, Zubovas & King (2016) study the evolution of outflows taking into account recurrent AGN duty cycles lasting 5 × 10⁴ years. In these calculations, the recurrent duty cycles manage to accelerate existing outflows after about ∼10⁶ years and out to radii larger than 10 kpc. Zubovas & King (2016) also find that the AGN should switch off once sufficient energy has been injected to unbind all gas; this is not (yet) the case in J0113 as our X-ray observations still reveal a very powerful AGN (one that has albeit likely faded in the last two years). Other processes could also explain the apparent acceleration. J0113 is weakly detected in the radio VLA FIRST survey (1.18 mJy; Schirmer et al. 2013), suggesting that a former radio jet could have played a role in shaping the current velocity profile. However, the survey's spatial resolution and depth (White et al. 1997) are insufficient for further conclusions. Alternatively, differential velocities in the stream could be at play (faster components reach farther in the same amount of time). However, Liu et al. (2013) argue that in this case, the velocity dispersion should increase as well with increasing distance, yet we observe a decrease in dispersion. We conclude that with the current data at hand, recurrent outflows are the best explanation for the peculiar velocity profile in the outflow of J0113.

Faucher-Giguère & Quataert (2012) have shown that an outflow preferentially breaks out from a denser disk or sphere along paths of least resistance. The hot buoyant gas would then inflate the bubbles, and bipolar systems should be produced eventually by many powerful AGN. J0113 seems prototypical in this respect; we condense our observational results in a simplistic graphical representation in Figure 5.

Zoom In Zoom Out Reset image size

J1155, on the other hand, could not be more different. While we do trace a 20 kpc long linear feature of enhanced recession velocity (Figure 4, lower left), it is not particularly present in the line flux map (Figure 4, lower right). The [O iii] nebula fragments into numerous blobs at all azimuthal angles around the AGN, which do not appear to be connected to any currently active outflow, nor to any of the member galaxies of the small group that hosts J1155 in its center (see also Section 5.2.4 and Figure 2). We think that J1155 represents a more advanced stage, where the ejected materials from various duty cycles have accumulated in the halo, but have not yet fully dispersed. Currently, J1155 experiences another powerful accretion event, as indicated by the very high central [O iii] surface brightness and its high X-ray luminosity.

5.2.4. Velocity Dispersion

What comes to attention with our IFU data are the fairly low velocity dispersions in both systems, in particular for their nuclei (see Figure 4). Near the center of J0113 we observe 150–180 km s⁻¹, and for the two superbubbles and beyond the dispersion drops to 100 km s⁻¹ and below. For J1155, we observe a mean dispersion of 160 km s⁻¹. In Figure 6 we compare our targets against the type-2 sample of Mullaney et al. (2013). We recall that our data of J1155 are shallow, and therefore any broad components in the outskirts of this nebula would remain undetected; consequently, we would underestimate its line FWHM.

Zoom In Zoom Out Reset image size

The low velocity dispersions show that the gas is little turbulent, in particular for J0113. There, it is flowing away from the AGN in a radial fashion, probably still pushed by recurrent duty cycles and its internal thermal energy. If powerful earlier outflows have already "cleared" the escape path, then the current stream would find less resistance and thus be less turbulent. Indeed, our imaging data show the presence of older outflows that must have traveled along the same paths.

Liu et al. (2013) find high (∼500–1500 km s⁻¹) velocity dispersions in the inner 5–10 kpc of their sample, which in about half of the cases drop to 100–200 km s⁻¹ at 10–15 kpc distance. Like us, they argue that recurrent outflows could have cleared the escape path. Additionally, Liu et al. (2013) explain that the narrow opening angle of the outflow (once it managed to break out) would leave less space for projected velocities, naturally decreasing the observed velocity dispersions. A noticeable difference between their and our data is that J0113 and J1155 have low velocity dispersions everywhere, even near the nucleus (our data of J0113 has a physical resolution of 1.5 kpc). Perhaps the more turbulent central components are simply hidden by dust. The enhanced Balmer decrement of 4.4 (integrated over 15 kpc by the SDSS fiber) shows that at least some part of the ionized gas must be severely affected by absorption, which goes in line with the high column density of ${N}_{{\rm{H}}}/{\mathrm{cm}}^{2})=5.1\,\times \,{10}^{23}$ cm⁻² (J0113).

J1155 also reveals a rather low velocity dispersion, about 160 km s⁻¹ averaged over our IFU data. Hardly any of its numerous extended features appear to be connected to actively driven outflows. This material was probably ejected during several AGN bursts in the past, and is now passively ionized by the AGN. A notable environmental difference with respect to J0113 is that J1155 is not isolated, but located at the center of a small galaxy group with ${M}_{200}\sim 1.3\times {10}^{13}\,{M}_{\odot }$. Schirmer et al. (2016) speculated that the unusual morphology of this system could be due to interaction with infalling gas from an intragroup medium (undetected in X-rays). Alternatively, the surrounding inner material could have a lower filling factor, offering more escape paths for any buoyant hot gas. This would also be consistent with the lower absorbing X-ray column density.

One of the best-studied (and most luminous) LABs is SDSS J2143−4423 (also known as "B1," z = 2.38). Its Lyα emission extends over $30\times 100\,\mathrm{kpc}$ (Francis et al. 2001), and Palunas et al. (2004) report a Lyα line flux of $1.35\times {10}^{-15}$ erg cm⁻² s⁻¹, or log $({L}_{\mathrm{Ly}\alpha }/\mathrm{erg}\ {{\rm{s}}}^{-1})=43.8$. Overzier et al. (2013) measure a dereddened [O iii] luminosity of log $({L}_{[{\rm{O}}{\rm{III}}]}/\mathrm{erg}\ {{\rm{s}}}^{-1})\,=43.4-44.1$, subject to uncertain extinction correction. The optical nebula extends over $32\times 40\,\mathrm{kpc}$. The [O iii] line FWHM within 10 kpc of the nucleus is 600–800 km s⁻¹, and drops to 200 km s⁻¹ beyond 10 kpc (similar to many of the targets studied by Liu et al. 2013). The radial gas velocities in B1 are low, mostly within ±50 km s⁻¹, and the [O iii] line flux is strongly concentrated. Any substantial bulk motions must happen perpendicular to the line of sight. Overzier et al. (2013) argue that B1 is powered by an AGN that is undetected in X-rays.

Apart from the high velocity dispersion, B1 seems quite similar to J1155. Both show a high concentration of the line flux, surrounded by a very extended ($\gtrsim 20$ kpc) disk of much lower surface brightness with somewhat patchy structure. The [O iii] luminosities are comparable, and J1155 is about twice as large (judging from our deep imaging data). Its Lyα luminosity estimated from GALEX far-UV data (Schirmer et al. 2016) is the same as that of B1. The diameter of the Lyα emission of J1155, and the accurate Lyα luminosity, are subject of our currently active HST proposal. J0113 has a similar luminosity and extent as J1155, but its estimated Lyα luminosity is three times lower than that of B1.

We conclude that the recently discovered low-z LABs match high-z LABs in terms of luminosity and physical extent. Like B1, our targets host powerful AGN. At this point, it is too early to speculate about the morphological similarities between B1 and J1155. While the formation of J0113 with its characteristic bipolar outflow and other signs of historic accretion events is quite clear cut (see Faucher-Giguère & Quataert 2012; Zubovas & King 2016), the formation and current state of J1155 remain unclear and require further observations (in particular, optical line diagnostics and deeper data).

There is increasing evidence for AGN flickering (duty cycles lasting 10^4–5 years, amplitudes of orders of magnitude), obtained by the identification of ionization echoes (e.g., Schawinski et al. 2010; Ichikawa et al. 2016; Schirmer et al. 2016; Keel et al. 2017), by means of statistical arguments and outflow rates (e.g., Schawinski et al. 2015; Sun et al. 2017), the transverse proximity effect in the Lyα forest (Kirkman & Tytler 2008, albeit on longer scales of 10⁶ years), and also on the theoretical side (e.g., Hopkins & Quataert 2010; Novak et al. 2011; Sijacki et al. 2015; Zubovas & King 2016).

The typical timescales of AGN flickering are on the same order as the light crossing times of galaxy-scale outflows extending over 10 kpc and more. The optical appearance of these extended narrow-line regions is invariable over a human life time given their kinematic timescale of Myr, and the limited physical resolution of current instrumentation. However, because of their large extent, these nebulae reflect the recent history of the AGN, and this affects how we interpret their observed morphologies.

First of all, the effective recombination timescale for the ${{\rm{O}}}^{++}$ ion is very short. Using the calculations of Binette & Robinson (1987),

$$\begin{eqnarray}&&{t}_{\mathrm{rec}}({{\rm{O}}}^{++})=158\,\mathrm{years}{\left(\displaystyle \frac{{n}_{{\rm{e}}}}{100{\mathrm{cm}}^{-3}}\right)}^{-1}.\end{eqnarray} \tag{ 4 }$$

Accordingly, the response of the [O iii] line to a change in ionizing radiation can be considered instantaneous for observations with a physical resolution of ∼1 kpc. An outward-propagating wave of ionizing radiation (or a shutdown thereof) would alter the nebulae surface brightness.

Second, we know from the Changing Look quasars (LaMassa et al. 2015; MacLeod et al. 2016; Runnoe et al. 2016) that AGN may change their line-of-sight obscurations on scales of years, affecting the opening angle of an ionization cone (e.g., due to intervening dust or structural changes in the torus in response to a change in bolometric luminosity). Spin precession of the SMBH accretion disk happens on scales of 10^3–7 years (Lu & Zhou 2005) and may cause illumination (ionization) patterns on a screen of gas extending over 10s of kpc.

Last, absence of line emission in some area does not imply absence of gas. These areas could be shielded from ionization by intervening dust. In turn, absence of line emission does not mean that there is no ionizing radiation; and AGN may well shine into empty volumes.

It is tempting to link the optical morphologies from our IFU data with the structural parameters of the central AGN derived from our X-ray analysis. Indeed, the consistencies are remarkable. From the X-ray data of J0113 we infer an AGN that is buried in a torus with a small opening angle. Accordingly, this AGN is able to photoionize gas within a relatively narrow ionization cone only, matching the collimated bipolar outflow. However, the highest surface brightness of the line emission is located in the central 2–4 kpc (upper right in Figure 4), showing that ionizing radiation can leak into other directions as well. In J1155, we find a weakly obscured AGN capable of ionizing gas over a wide solid angle, consistent with the optical nebula ionized at all azimuthal angles. However, our sample is highly limited (two), and as we cautioned above, the observable nebular morphologies are subject to outflow histories and variabilities in obscuration, luminosity, and illumination direction.

Two possible explanations for the ionization deficits in LABs are deeply buried AGN and time variable (faded) AGN (see Section 1). Our previous analysis of individual sources in Schirmer et al. (2016) was limited by poor knowledge of their X-ray obscurations. We could argue in favor of faded AGN for the sample as a whole, but not for individual sources. This limitation is overcome with hard X-ray (>10 keV) data.

Currently, powerful (${L}_{{\rm{X}}}\sim {10}^{44}$ erg s⁻¹) and Compton-thin AGN exist in both J0113 and J1155 (Section 3). To test for recent long-term variability, we compare their X-ray luminosities against other proxies of AGN power that retain some memory of the past AGN activity. These proxies are mid-infrared (MIR) emission from circum-nuclear dust, and [O iii] emission from the extended ionized gas. The built-in "response" times of these "echo screens" are caused by light travel time from the nucleus, their intrinsic physical properties, and also their orientation along the line of sight (a response from the parts located closer to us would arrive sooner at our telescopes). In the following, we check whether these proxies bear any signs for substantial variability in both AGN over the last 10^2–4 years.

5.5.1. The Mid-infrared X-Ray Relation

Numerous studies have established a tight correlation between the MIR and X-ray luminosities in X-ray selected AGN; radiation from the central engine heats the dusty torus, which subsequently emits thermal radiation. This correlation is independent of AGN classifications (type-1/2, Compton thickness; see e.g., Gandhi et al. 2009; Asmus et al. 2011, 2015; Ichikawa et al. 2012, 2017). Rapid and prolonged fading of AGN, however, would move an individual AGN off the MIR X-ray correlation. Hönig & Kishimoto (2011) simulated the MIR response of a dusty torus to a finite pulse from the AGN as a function of the torus structural parameters. Ichikawa & Tazaki (2017) also estimated the typical cooling timescale of the dusty torus once the AGN is suddenly quenched, and they even considered the dust heating with the energy exchange by the gas. Accordingly, the thermal response of a torus may have decay times of 10–1000 years, depending on its thickness, and in addition to the time lag caused by the light travel time from the nucleus to the sublimation radius. We have presented observational data supporting such thermal echoes in Schirmer et al. (2016), pending confirmation by accurately measuring the X-ray obscurations.

Figure 7 shows scatter plots of the absorption corrected 2–10 keV luminosity versus MIR (12 μm and 22 μm) luminosity for the hard X-ray selected AGN of Ichikawa et al. (2017, gray dots). This is one of the most complete samples to date for the hard X-ray selected AGN. To make the plots, we converted the 14–195 keV luminosities into the 2–10 keV band using a photon index of ${\rm{\Gamma }}=1.9$ (i.e., ${L}_{14-195\mathrm{keV}}/{L}_{2-10\mathrm{keV}}=2.1$). The data for J0113 and J1155 are also shown, using their WISE W3 and W4 band fluxes (Wright et al. 2010) to estimate the 12 μm and 22 μm luminosities, respectively. The data for J1155 place it within the $1\sigma$ scattering of the reference sample. The NuSTAR observations of J0113 taken in 2016, however, deviate by one order of magnitude from the best-fit correlation, whereas its Chandra data taken in 2014 are in much better agreement with the reference sample. This illustrates the effect of rapid X-ray fading on this plot. Because dusty tori typically have parsec scales and delayed intrinsic response times (Hönig & Kishimoto 2011; Ichikawa & Tazaki 2017), the MIR flux of J0113 is not expected to change much just yet (and we used the single WISE measurement from 2010). Sparse long-term MIR monitoring of J0113 is worthwhile, in particular, if the X-ray flux continues to fade.

Zoom In Zoom Out Reset image size

Since both AGN are still within the general scatter, we cannot infer long-term time (∼10–1000 years) variability from the MIR X-ray relation. Note that we initially proposed seven targets to be observed with NuSTAR, covering a large range of Chandra fluxes. Only J0113 and J1155 were approved, however, because they are by far the X-ray brightest, which maximizes the chances for appreciable NuSTAR count rates. Therefore, highly significant deviations from the MIR X-ray scaling relation are not expected for these two targets, but could be present for the X-ray fainter LABs in our sample.

For clarity, we illustrate the simplified time evolution (arrows) after AGN quenching in the left panel of Figure 7. For this, we adopt an instantaneous decline by two orders of magnitude, supported by observational and theoretical data alike (Schawinski et al. 2010; Novak et al. 2011; Schirmer et al. 2016). The X-ray luminosity before the quenching is set to $\mathrm{log}{L}_{2\mbox{--}10\mathrm{keV}}=44.5$, the maximum luminosity we observe for J0113 and J1155. The 12 μm luminosity is estimated using the third equation in Table 3 of Ichikawa et al. (2017) for luminous AGN ($43\lt {L}_{14\mbox{--}195\mathrm{keV}}\lt 46$). Applying Equation (21) by Marconi et al. (2004), we derive the bolometric corrections for the 2–10 keV band, and then use Figure 3 of Ichikawa & Tazaki (2017) to estimate the cooling times. Note that Ichikawa & Tazaki (2017) assume a sudden and complete AGN shutdown, whereas in reality it could take the AGN much longer to move back onto the general MIR X-ray correlation (depending on the radial dependence of the dust density; Hönig & Kishimoto 2011).

5.5.2. Negligible Effect from MIR Emission Lines

[O iii]$\lambda 5007$ emission is another proxy commonly used to infer AGN power (e.g., Mulchaey et al. 1994). A typical [O iii] nebular size observed in local Seyfert galaxies can reach several kpc (e.g., Schmitt et al. 2003; Bennert et al. 2006). In Figure 8 we compare our LABs with local AGN detected in the Swift/BAT 9-month catalog (Ueda et al. 2015). Both LABs have lower X-ray luminosities at a given [O iii] luminosity than predicted from the average properties of the Swift/BAT sample, but not at a level that would distinguish them as fading. The apparently high [O iii] luminosity could simply be a fact of our selection (targeting optically overluminous nebulae), and that our type of targets is not included in the commonly used SDSS AGN catalogs (the latter are incomplete concerning extreme sources, see Schirmer et al. 2013; Baron & Poznanski 2017). Nonetheless, for J0113 we know that it has been fading for the last two years. In addition, Ueda et al. (2015) report a lower [O iii] to X-ray luminosity ratio on average for AGN with low scattering fractions (${f}_{\mathrm{scat}}\lt 0.5$%; green circles in Figure 8), to which J0113 belongs. J0113 is 10–50 times more luminous in [O iii] than other low-scattering AGN. This could simply be a consequence of poor statistics at the high-luminosity end or a consequence of variability (the [O iii] X-ray relation decorrelates more easily than the MIR relation under AGN flickering because of the slower [O iii] response). We conclude that for our particular two targets, the [O iii] X-ray relation does not bear conclusive evidence for AGN flickering.

Zoom In Zoom Out Reset image size