Fading AGN Candidates: AGN Histories and Outflow Signatures^∗

It has long been known that some active galactic nuclei (AGNs) are accompanied by extended emission-line regions (EELRs), zones of ionized gas spanning galaxy scales or even larger. Such regions can trace the geometry of ionizing radiation escaping the AGN and host galaxy, and at least implicitly give hints to the luminosity history of the AGN. EELRs occur around Seyfert nuclei, QSOs, and radio galaxies.

A very luminous and extensive EELR was found in the course of the Galaxy Zoo project (Lintott et al. 2008) near the spiral galaxy IC 2497. Known after its discoverer as Hanny's Voorwerp, this object shows high-ionization emission lines with ratios essentially identical to the narrow-line region of an AGN, spanning a projected range from 15 to 35 kpc from the galaxy nucleus, which fails by at least two orders of magnitude in bolometric output to match the ionization requirements of the cloud (Lintott et al. 2009; Keel et al. 2012b). The cloud's electron temperature, narrow line widths, and quiescent velocity field indicate photoionization rather than shocks as the energy source. H i mapping by Józsa et al. (2009) shows this to be the ionized part of a 300 kpc trail of otherwise neutral gas, suggesting a strong galaxy interaction gigayears ago. The energy mismatch between the nucleus of IC 2497 and the ionization requirements in Hanny's Voorwerp indicates that the nucleus faded from a luminosity associated with QSOs to a modest Seyfert or LINER level within 10⁵ yr.

Starting with the unusual morphology and Sloan Digital Sky Survey (SDSS) gri colors of Hanny's Voorwerp, a targeted search by Galaxy Zoo volunteers found 19 additional EELRs, including hitherto unknown examples, of which eight showed energy shortfalls similar to that in IC 2497 (Keel et al. 2012a). This sample was selected in a homogeneous way, requiring detection of emission features showing AGN-like line ratios more than 10 kpc in projection from the nucleus. We have pursued a detailed study of this subset of eight (plus IC 2497/Hanny's Voorwerp) to learn more about AGN variations on the otherwise inaccessible range of timescales of 10⁴–10⁵ yr. Analogous cases of extended ionized gas near energetically inadequate AGNs have been reported at higher redshift and luminosity by Schirmer et al. (2013, 2016), their "Green Bean" systems, and at low luminosity in the local universe in the merger remnant NGC 7252 (Schweizer et al. 2013) and as one interpretation of an off-nuclear ionized region in the spiral galaxy NGC 3621 (Menezes et al. 2016).

Interest in the history of accretion onto nuclear black holes has also been renewed by evidence for an eventful history of outbursts within 10⁶ yr by the currently quiescent black hole at the Galactic Center, from X-ray echoes (Muno et al. 2007), the ionization structure of the Magellanic Stream (Bland-Hawthorn et al. 2013), and possibly the "Fermi bubbles" (Zubovas et al. 2011; Guo & Mathews 2012).

The variations in AGN luminosity we infer from these objects, spanning scales up to 10⁵ yr, connect implicitly with the variations on timescales we can observe directly. "Changing-look" AGNs can have broad-line emission essentially vanish over ≈10 yr in the emitted frame (MacLeod et al. 2015; Ruan et al. 2015; Runco et al. 2016; Runnoe et al. 2016), while the narrow-line region is often so large that light-travel-time smearing leaves its emission nearly constant; similar interpretations were discussed in the context of spectral variations in Seyfert nuclei at least as early as Penston & Perez (1984). The temporal spectrum of variations in AGN output clearly spans many orders of magnitude, and affects our understanding not only of their physical structure but also of the demographics of accretion in the galaxy population. Seeing dim episodes so long that they appear in surrounding ionized gas when smeared by the recombination timescales of thousands of years, as well as geometric projection factors, means either that the AGNs stay in a very low state for these long times or that excursions to a high state fill only a small fraction of the time. Understanding this history is key to understanding the broad idea of duty cycles in AGN accretion; X-ray surveys in particular suggest a duty cycle connected to the Eddington limit and evolving with cosmic time (e.g., Shankar et al. 2009), but not how many episodes of what durations are involved.

In a similar sense, radio observations suggest very episodic production of jets, a major form of kinetic energy output from accretion. The identification of radio jets and a circumnuclear outflow in IC 2497 (Józsa et al. 2009; Rampadarath et al. 2010; Keel et al. 2012b) led us to speculate that some of the extreme variability needed to explain the ionization of Hanny's Voorwerp might come not solely from a drop in the accretion-driven luminosity of the AGN in IC 2497, but from some of the "missing" luminosity switching to a kinetic mode. This idea is supported by a "bubble" of hot gas around the fading AGN of IC 2497 seen with Chandra (Sartori et al. 2016).

All these factors motivated more detailed study of our fading-AGN candidates with a variety of techniques. Paper I (Keel et al. 2015) presented results of Hubble Space Telescope (HST) imaging concerning the host galaxies and the origin of the ionized gas. The hosts are bulge-dominated, and every one shows signatures of ongoing or past galaxy interactions, 1.5–3 Gyr ago for systems with favorable geometry to estimate this timescale. The extended gas has modestly subsolar metallicity, and is kinematically rotation-dominated; outflows have only a very localized role, in contrast to most QSO EELRs (Stockton et al. 2006). These are externally illuminated tidal debris; merger remnants are particularly good environments for H i to occur far from the nucleus and out of any host disk plane, and act as a screen to show the escaping ionizing radiation. This appears to be such a strong selection factor that we do not know whether high-amplitude variability on scales of 10⁴ yr is confined to merger aftermaths or not.

In this paper, we describe additional data, including Gemini integral-field spectroscopy, and present implications of the entire data set for the properties and history of these AGNs. We use the peak surface brightness in recombination lines at various projected radii to reconstruct the luminosity history of the AGNs, compare structures seen in the emission-line images and line ratios to address the role of ionization cones, and examine the occurrence of outflows near the nuclei to trace the role and extent of feedback near the epochs of radiative dimming.

In evaluating sizes and luminosities, we compute distances using a Hubble constant of 73 km s⁻¹ Mpc⁻¹.

2.1. Galaxy Sample

2.2. HST Imaging

2.3. WISE Data

2.4. Spectroscopic Mapping

Some of these cloud systems show plausible ionization cones, briefly mentioned in Paper I. In Figure 1 we show line-ratio images of [O iii]/Hα, which often map these structures more clearly than [O iii] intensity alone. Among our galaxies, some show cone-like structures in line intensity; we see a few where the line ratio maps ionization in a clearer structure. Most notably, in the Teacup AGN there is a reasonably well-defined triangular region of higher ionization (where the "hole" lies within it). In others, such as UGC 7342 and UGC 11185 (Hainline et al. 2016), the entire emission region could plausibly show a biconical structure without any internal differences in ionization. In these two systems, radial changes in the line ratio are greater than azimuthal variations at fixed projected radius.

Zoom In Zoom Out Reset image size

Broadly, all these emission complexes show a twofold symmetry, as sketched in Keel et al. (2012a) from early data. When their outer parts resemble ionization cones, we often see much broader regions near the core (often completely encompassing the AGN), whose ionization may have a distinct mechanism.

The emission region in Mkn 1498 is highly elongated but not biconical. Among the remainder, most of the opening angles implied by the HST data are close to our initial estimates, except for the Teacup AGN where we now recognize an inner structure traced by the [O iii]/Hα ratio for which each half has a full angle of 37°. In SDSS 2201+11, we derive 25°, although this may be affected by obscuration within the host galaxy on one side of each cloud. The rest are broader, with NGC 5252 at 64°, UGC 7342 at 75°, and NGC 5972 at 96° to encompass the outer structures. The broadest are seen in UGC 11185 (116°) and SDSS 1510+07 (126°). Projection effects in a conical distribution make this angle appear larger than its three-dimensional value.

Side-to-side asymmetries in ionization (as in BPT plots, or in reconstructed ionization history) could in principle record front–back differences in light travel time when seen at a sharp angle to the axis of ionization cones or radially elongated clouds. We do not actually see any such differences; our selection for large transverse extent favors cloud systems viewed near the plane of the sky.

In SDSS 2201+11, a dust lane is continuous with the emission-line material into the northern cloud, showing that the northern cloud must be at the front of the system in order to give significant dust attenuation.

Near the nuclei, and outside the ionization cones, several of these galaxies show gas with low ionization as seen in [O iii]/Hα. One goal of the GMOS IFU spectra was to measure additional line ratios in these regions to probe their ionization mechanism. We will explore this issue elsewhere, in conjunction with recent spectra from the HST Space Telescope Imaging Spectrograph that isolate key regions near selected nuclei.

Overall, the occurrence of ionization cones pointing to the nucleus suggests that the ionized gas has distributions that are mostly radial, rather than, for example, ringlike. This means that projected distance from the nucleus can serve as a useful proxy for physical separation from the AGN in these clouds.

4.1. Method: Recombination Balance

We explore here the use of recombination-balance arguments to provide lower limits to the ionizing flux reaching various parts of the AGN emission regions. Except in areas where we see the superposition of multiple emission regions along the line of sight, or similarly, elongated filaments seen end-on, the measured luminosity in a recombination line gives a lower limit to the ionizing luminosity seen by that volume of material. The brightest pixels at each projected radius give the most stringent limits. Even in these locations, the derived luminosity will be an underestimate because of the combined effects of optical depth and covering fraction within the region spanned by each pixel on the object; both quantities are likely to have complex and patchy distributions. Suitably located gas must be present to apply this approach, of course; a strongly flaring AGN would look the same as one without suitable neutral gas around the host system.

Despite these clear limitations, this approach could yield useful information on, for example, the history of AGN luminosity, if the amount of underestimation is comparable for the highest pixels at various radii (if not necessarily between objects). Regions farther from the AGN are physically unlikely to intercept progressively more radiation, since the density in emission regions general falls with distance from the core, and the changes we will infer span orders of magnitude in the opposite sense to what is expected from this. Two immediate checks on this approach are available—the photoionization modeling for spectra of various regions around the Teacup AGN (Gagne et al. 2014), and narrowband [O iii] structures of a set of Palomar–Green (PG) QSOs.

We employ the ionization balance as follows. We use the highest surface brightness observed in Hα as a function of projected distance from the AGN to estimate the required (isotropic) emission rate of ionizing photons Q_ion needed to power the observed emission, making the broad assumption as noted above that the densest gas clumps are similar at various projected distances from the AGN within each system. In practice, we use the required photon rate between the H and He ionization edges, since the photoionization cross section of helium is large enough to dominate the absorption immediately above 54.6 eV. As expected, the HST images reveal numerous small regions of high surface brightness, so that the lower limit we derive in this way is often much higher than estimated from ground-based spectra (Keel et al. 2012a). Our calculation incorporates the fraction of recombinations leading to an Hα photon (Spitzer & Greenstein 1951; Nussbaumer & Schmutz 1984). In "Case B," where the gas is optically thick to Lyα, the fraction of recombinations leading to an Hα photon ranges roughly from 0.25 to 0.34 depending on the trapping properties of Lyα. The other limit, case A where the nebula is optically thin in Lyα, has these fractions lower by a factor 1.6 (these values are for T_e = 10⁴ K, with recombination coefficients as collected by Osterbrock & Ferland 2006). If we denote the fraction of recombinations leading to an Hα photon by f, Q_ion can be derived conveniently from the area-normalized count rate N. This value represents the arrival rate (Hz cm⁻²) of line photons in the imaging data, corrected for the quantum efficiency η (≈0.39 of the HST+ACS+filter system at Hα; for NGC 5972, where Hα was observed with WFC3 and the F673N filter, the value is 0.22; for NGC 5252 with the WFPC2 ramp filter it is 0.119). The current HST pipeline calibration corrects for the gain (yielding results in electrons per second, N, which is to say detected photons) for ACS and WFC3 data, but not for WFPC2 data (so the gain factor must be applied to these images); the incident flux of Hα photons is then NA/η for unobscured telescope area A, so the number emitted per second in the emitter's frame is $4\pi {D}^{2}(1+z){NA}/\eta$ for source distance D, now including the time dilation $(1+z)$ in photon arrival rate. Neglecting projection effects, which are poorly known, material in each pixel of observed solid angle a would intercept a fraction of the emitted radiation of the order of a/4πR² where R is the projected distance from the nucleus to the area of interest; if R is in pixels, a = 1 for unsmoothed data.

Collecting these components, the count rate N per unit area in a single pixel implies

$$\begin{eqnarray*}&&{Q}_{\mathrm{ion}}=(1+z)(N/\eta \,f)4\pi {D}^{2}4\pi {R}^{2}\end{eqnarray*}$$

for a region projected R pixels from the AGN. For the ranges of density and Lyα optical depth that matter here, as noted above, f ranges from 0.16 to 0.34. This value of Q_ion is a lower limit, since we then have no constraint on how much radiation escapes the system. To be conservative, we discuss values using f = 0.29, in the middle of the range of Case B photoionization. This was cast in a somewhat different way in Keel et al. (2012a), using the angular width of a spectrograph slit about the nucleus at a given R.

The ACS ramp-filter passband has a nearly flat profile near the peak wavelength, falling below 50% of peak transmission well outside the span of the [N iii] lines when centered on Hα at these redshifts. We make a correction for the adjacent [N iii] emission based on our slit spectroscopy (Keel et al. 2012a), assuming the behavior of the line ratio to be circularly symmetric about the nuclei on each side. We fit this correction factor as a polynomial in radius for computational convenience.

For optically thin gas, or material with a small covering factor (when shadowing by absorbers closer to the nucleus is not important), there will be an upper envelope in the diagram of Hα surface brightness-projected radius, which will follow a 1/r² form for constant source luminosity as long as various regions in emission do not have radically different projection factors. Departures from this behavior can be evidence of long-term changes in Q_ion; neglecting projection factors, one can in principle extract the source history from such a diagram.

4.2. Consistency Tests

We show one set of data for this calculation suitable for an external test in Figure 2 for the Teacup AGN, where the surface brightness is in electrons pixel⁻¹ s⁻¹ and the radius is in arcseconds. Transforming these into the physically meaningful units of Q_ion and projected distance in light-years gives our estimates of the minimum Q_ion at each pixel.

Zoom In Zoom Out Reset image size

This conversion is illustrated in Figure 3, showing Q_ion converted into ionizing luminosity for each pixel above a 3σ threshold for the Teacup AGN. The upper envelope is the most significant feature, since we argue that the densest regions are likely to be comparable at various radii. For this object, we can compare our results to the detailed photoionization modeling using the CLOUDY 90 code (Ferland et al. 1998) carried out by Gagne et al. (2014) using ground-based spectra spanning its emission regions. Those results parallel the upper envelope of our reconstructed points, systematically ≈25% higher depending on how the envelope is defined, which supports both our results and the basic assumption of our analysis, that the most optically thick regions are comparable at all radii in the clouds (so we tentatively assume this to be the case for the other galaxies in our sample). In particular, our analysis reproduces the same change in luminosity, declining by a factor 10^1.03 over a time ΔT = 26,500 yr between the peaks in reconstructed luminosity. It is also reassuring that the peaks in reconstructed luminosity approach the spectroscopic results closely but do not exceed them at any point.

Zoom In Zoom Out Reset image size

As a second test of our approach, we use HST data on low-redshift QSOs for their high spatial resolution, since numerical tests on our images show that the derived peak ionizing flux drops rapidly with lower image quality. (If this behavior proves to be consistent enough, an analogous study could be done on the emerging body of integral-field spectroscopic mapping of AGNs with EELRs, such as Liu et al. (2013) or Husemann et al. (2016).) We compare with the narrowband [O iii] data shown for seven PG QSOs by Bennert et al. (2002), which use the WFPC2 linear-ramp filters and continuum subtraction incorporating stellar point-spread functions. The scaling of their implied ionizing luminosities is only approximate; we used a mean value for [O iii]/Hα = 4, lacking independent narrow-line ratios at the core. That program was targeted at the sizes of narrow-line regions, with no selection for extent of EELRs, and only three of the PG QSOs have detected emission beyond the 10 kpc radial bound used to select our fading-AGN sample. We masked regions around obvious companion galaxies, but not a bright emission region within 1'' of PG 0157+001 since its nature is ambiguous. The implied histories are shown in Figure 4. These show brightening, fading, and near-constant episodes. PG 0157+001 and PG 1012+008 show fading by nearly an order of magnitude before brightening close to the observed epoch, while PG 0953+414 has brightened by nearly this much. Brightening episodes will be harder to detect at large radii because the signal drops more strongly near the detection threshold. This test broadly confirms that our sample of fading candidates does indeed show distinct behavior using our technique of reconstructing luminosity from recombination balance. In particular, none of the PG QSOs shows the rapid drops in the last 20,000 yr that are ubiquitous in our fading-AGN sample.

Zoom In Zoom Out Reset image size

4.3. Luminosity Histories of Fading AGNs

The two external tests outlined above suggest that, indeed, we can reconstruct at least qualitative histories for AGN luminosity using pixel-by-pixel recombination balance. Figure 5 shows these pixel-by-pixel results for the other seven objects in our sample plus IC 2497 and Hanny's Voorwerp, where similar data were described by Keel et al. (2012b). Figure 6 compares the upper envelopes for all objects, connecting the brightest pixels in bins of 1000 lt-yr.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Recombination timescales for hydrogen approach 10⁴ yr in the outer parts of these clouds, so only changes on longer terms can reflect variations in the AGN output; along with line-of-sight depth of the cloud systems, this means that the temporal resolution degrades with projected radius. We illustrate this in Figure 7, which smooths the derived luminosities by a range 0.14 times the radius, a value that matches the best-constrained timescale in the outer regions of Hanny's Voorwerp and roughly accounts for a decrease in density with radius matching the limits from Lintott et al. (2009). The points at ΔT = 0 in both versions of this figure are from the Hα fluxes at the nuclei (operationally within a projected radius of 1'' from slit spectroscopy), in the approximation that they represent full covering of the instantaneous ionizing flux (strictly, a lower limit since we see that ionizing photons do escape the nuclear region). The flux values are taken from our spectroscopy in Keel et al. (2015), as given in Table 5, including only the narrow component of the composite profile in Mkn 1498 for consistency. As a rough numerical description of these reconstructed histories, we collect in Table 4 a set of properties fit exponentially to segments of the smoothed version in Figure 7, showing not only fading periods but constant and brightening episodes.

Zoom In Zoom Out Reset image size

Table 4.  Timescale Estimates for Variability Episodes

Galaxy	Data Span (yr)	Time Range (yr)	Behavior	e-folding Timescale (yr)
IC 2497	65,000	111,000–102,000	brighten 1.0 dex	4000
		102,000–60,000	constant	⋯
		60,000–46,000	fade 2.0 dex	3000
Mkn 1498	32,000	32,000–14,000	brighten 0.6 dex	13,000
		14,000–2000	fade 0.5 dex	10,000
		2000–0	fade 1.6 dex	550
NGC 5252	40,000	40,000–14,000	constant	⋯
		14,000–2000	fade 1.2 dex	4300
		2000–0	fade 1.4 dex	620
NGC 5972	55,000	55,000–15,000	fade 0.3 dex	5800
		15,000–5000	fade 1.0 dex	4300
		5000–3000	brighten 0.5 dex	1700
		3000–0	fade 1.3 dex	1000
SDSS 1510	60,000	60,000–18,000	constant	⋯
		18,000–2000	fade 1.8 dex	3900
		2000–0	fade 0.6 dex	1500
SDSS 2201	45,000	45,000–10,000	fade 1.2 dex	13,000
		10,000–3000	fade 0.8 dex	3800
		3000–0	fade 0.6 dex	2200
Teacup	55,000	55,000–36,000	brighten 1.0 dex	8000
		36,000–1000	fade 1.2 dex	12,000
(Gagne et al.)	39,000	39,000–0	fade 1.7 dex	9800
UGC 7342	120,000	120,000–50,000	constant	⋯
		50,000–34,000	fade 1.03 dex	6700
		34,000–20,000	fade 0.1 dex	60,000
		20,000–11,000	fade 0.40 dex	9700
		11,000–2000	fade 0.6 dex	6500
		2000–0	fade >1.0 dex	900
UGC 11185	55,000	55,000–45,000	brighten 1.0 dex	4300
		45,000–4000	constant	⋯
		4000–0	fade 2.2 dex	800

Download table as:  ASCIITypeset image

We considered several alternative approaches to calculating ionizing history, including integrating the line flux in slices across the putative ionization cones. This was defeated by the very patchy structure of the ionized features; tracing only the peak intensity reduces sensitivity to the larger-scale structure.

We can estimate the values of Q_ion predicted from the complementary assumption that the AGN is unseen along the direct line of sight due to dust obscuration. In this case, we take the total infrared output to be the total AGN luminosity, and a typical fraction of the bolometric luminosity 0.16 to be in ionizing radiation from 13.6 to 54 eV (typical of composite spectral energy distributions, although this is often the least well-determined piece of the spectrum). For a wide range of spectral index in this range, the mean ionizing energy is close to 27 eV (Netzer 1990, p. 57). This is intended to be conservative with respect to the possible role of AGN observation, making no correction for host-galaxy IR contribution or the possibility that significant IR reprocessing occurs far from the AGN and thus may not reflect its luminosity at the same epoch as our direct view. The resulting Q_ion values or limits for most of these AGNs (listed in Table 5) fall well below even the lower limits imposed by recombination balance, adding to the case for drops in luminosity. Finally, the spatially resolved drops in required ionizing flux seen within each object's clouds add to the evidence that the energy mismatch between nuclei and extended clouds really results from variability rather than unusually hard FUV spectral slopes or localized extinction around the AGN. The shortfalls for Q_ion derived from the infrared are less extreme, indicating modest local obscuration, but in most cases still very large.

Table 5.  Nuclear Hα Fluxes and Ionizing Rates

Galaxy	Hα flux	Qion	MIR+FIR Qion
	(erg cm−2 s−1)	(photons s−1)	(photons s−1)
IC 2497	1.3 × 10−15	2.6 × 1050	2.9 × 1054
Mkn 1498	5.5 × 10−14	3.3 × 1052	9.3 × 1054
NGC 5252	4.5 × 10−13	8.4 × 1052	8.3 × 1053
NGC 5972	2.6 × 10−14	8.4 × 1052	8.1 × 1053
SDSS 1510	5.9 × 10−14	4.4 × 1052	<1.4 × 1054
SDSS 2201	1.1 × 10−13	3.2 × 1052	<2.2 × 1053
Teacup	7.9 × 10−14	2.0 × 1053	9.5 × 1054
UGC 7342	1.3 × 10−13	1.0 × 1053	1.8 × 1053
UGC 11185	4.9 × 10−14	2.9 × 1052	1.2 × 1054

Download table as:  ASCIITypeset image

Most of our sample would not have shown enough fading to satisfy our selection criteria based only on the early parts of the reconstructed histories. The strongest, characteristic fading took place within the last 20,000 yr before the epoch when we observe the nuclei, except for IC 2497/Hanny's Voorwerp, where it occurred in a similar span about 50,000 yr before the direct view and remained dim. This distinctive fading is evidence (internal to each galaxy) for actual drops in luminosity rather than alternative interpretations of the mismatch in energy budget in these objects, such as obscuration from dust far from the nucleus (so there is no necessary connection between the obscuration and mid-IR emission) or spectral energy distributions unusually strongly peaked in the ionizing UV.

In IC 2497, host galaxy of Hanny's Voorwerp, HST observations reveal an expanding loop of low-ionization gas extending ≈500 pc to one side of the nucleus (Keel et al. 2012b). This, along with evidence for an oppositely directed radio jet (Józsa et al. 2009; Rampadarath et al. 2010) and bubble-like structure in the X-ray gas (Sartori et al. 2016), invites speculation that these objects are not necessarily undergoing a near-shutdown of accretion, but may (also) be switching modes of energy output incidental to the accretion, to become kinematically dominated (sometimes known as radio-mode). In X-ray binaries, the switch from radiative to kinetic mode is associated with a change in accretion state (and thus a decrease in the luminosity). It is possible that the same is happening also in AGNs (Sartori et al. 2016). There are additional AGNs with well-studied bubbles or loops attributed to outflows, such as recent results on NGC 3393 by Maksym et al. (2017). This motivated us to look for similar loops of emission, and seek kinematic evidence as to whether they represent outflows or something else.

Morphologically similar loops to one side of nuclei, or encircling them, are common in this sample (Figure 8), and could by themselves be signatures of outflow episodes. However, the GMOS IFU data give a more complex picture both kinematically and in emission-line ratios.

Zoom In Zoom Out Reset image size

Kinematic signs of outflow do appear in several nuclei, although some of the loops are more nearly in rotation than radial motion. We can evaluate this using the Gemini IFU spectroscopy, as well as the wider-field Fabry–Perot velocity fields for [O iii] obtained with the 6 m BTA (Paper I). Spectroscopic signatures of outflow include multiple-peaked or asymmetric line profiles, and departures from large-scale rotational velocity fields even for locally narrow and symmetric profiles. Such features are seen in confined regions of some of our target galaxies. Figure 9 overlays the [O iii] profiles for the Gemini IFU data on HST narrowband images in [O iii], with the line profiles representing averages of the data over 0.5 × 05 regions.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Mkn 1498 the ringlike emission features that are prominent at Hα are dominated kinematically by rotation, possibly with one outflowing feature adjacent to the nucleus (Figure 10). The implied rotation axis is close to the minor axis of the largest emission-line structures, making it almost parallel to the large-scale radio structure. In the line profiles, red wings occur along the outer ring to the south, and about 2'' east of the nucleus along the other ring (but not the rest of it). The inner arc structure to the northwest of the nucleus has narrow line profiles, but departs from the kinematic pattern of the velocity peaks elsewhere in the ring.

Zoom In Zoom Out Reset image size

As seen in Figure 9, UGC 7342 shows substantial regions with double emission-line peaks, likely evidence of outflow in at least one component. We examine this in more detail by fitting sets of two Gaussian [O iii] profiles where this improves the fit, constraining them to have equal widths for consistency across a range of S/Ns. Similar behavior is seen in Hα and Hβ, but Hβ is weaker and Hα is affected by overlap with [N iii] components. The results are shown in Figure 11. Only peak separations >100 km s⁻¹ are detected, in view of the typical line width of 230 km s⁻¹ FWHM. Component separations range up to 270 km s⁻¹. The pattern of component velocities is not a simple two-sided outflow, and may have more than one origin. The outer part of the filamentary emission region about 35 northeast of the nucleus shows velocity splitting. This cloud hosts the most intense emission seen in the redder component (whose radial velocity changes from place to place), while its outer part shows additional blue wings (Figure 9). In addition, perpendicular to the main axis of the emission regions, there are extensive regions with split line profiles; much of this emission is too diffuse to be well detected in the HST images.

Zoom In Zoom Out Reset image size

Using the BTA Fabry–Perot global velocity map from Paper I as a guide to the overall velocity field, which is well fitted by a rotating and slightly warped disk, we find that the redder component is the more quiescent one to the south of the nucleus, while the bluer component fills this role to its north and northwest (where large velocity splittings occur in diffuse gas). East of the nucleus, the HST images do show a faint filament that could be associated with one of these velocity structures. The areas with large local velocity dispersion σ_v in the BTA data just to the north of the nucleus are a good match to regions of line splitting in the GMOS data cubes.

We are left with a complex empirical picture for kinematics in the inner few kiloparsecs of UGC 7342. Emission-line components suggest outflow, or at least motions departing from the overall rotation pattern by as much as 260 km s⁻¹, in regions of diffuse, low-surface-brightness emission perpendicular to the main emission regions, with the strongest off-rotational emission in the outer parts of a bright cloud, part of which also shows blue wings beyond a simple two-component line profile. We might speculate that this could be entrainment of denser gas by a low-density outflow such as seen west of the nucleus. Curiously, the brightest [O iii] structure, resembling an inclined ring just southwest of the AGN position, has no distinct signature in radial velocity.

The Teacup AGN was shown to have resolved outflow signatures by Harrison et al. (2015). We find double or asymmetric profiles extending 2'' north, 4'' west, >15 south, and 2'' east in a roughly circular zone. What appears as an [O iii] peak in the center of the loop on the underlying HST image may be affected by residuals from cosmic-ray events. The Gemini data are better at detecting emission that is diffuse and at low surface brightness, as in the center of the loop, and show no distinct emission-line component there. Blue wings dominate east of the nucleus, red wings to the west; in many locations they extend to 1000 km s⁻¹ from the systemic velocity.

NGC 5972 shows significant velocity structure, and a nearly constant ionization level (Figure 12). It shows double and asymmetric peaks near the dust lane, which can occur strictly from obscuration in a continuous velocity field, and in the inner part of the emission-line loop to its east. Overall, the emission-line loop to the east of the nucleus shows no discernible velocity signature; it does have narrower line profiles than its surroundings, perhaps because its stronger emission dominates the more diffuse gas elsewhere along the line of sight. There is larger-scale velocity structure in the GMOS field spanning ≈100 km s⁻¹, not matching the loop morphology or location; this is part of the departures from a fit to circular motions seen near the nucleus in the wide-field BTA Fabry–Perot map (Keel et al. 2015). The [O iii]/Hα ratio declines with radius from the nucleus at the same rate in the loop and its surroundings; the loop does not stand out in this parameter. It has lower [N iii]/Hα and [S ii]/Hα ratios than its surroundings, but not [O iii]/Hα or the density-sensitive [S ii] λ6717/λ6731 ratio. The loop region 20 east of the nucleus has [N iii] λ6583/Hα = 0.33 (0.71 in the surroundings) and ([S ii] λλ6717 + 6731)/Hα = 0.36 (0.70). Both line ratios are more sensitive to abundances than to ionization parameter under AGN-like photoionization, unlike [O iii]/Hα (Storchi-Bergmann et al. 1998). Very broadly, the line ratios show lower abundances in this small-scale loop of emission than in its surroundings, which could suggest, contrary to its appearance, that it consists of infalling or recently accreted gas. Together, these properties seem more like infall than an AGN-driven outflow; the alternative would require a fairly contrived set of circumstances, such as motion in the plane of the sky and the outflow itself being unseen (for example, too hot for optical emission) and entraining material that originated in a low-luminosity companion disrupted by the obvious merger history. Young stars might give the observed line ratios, in which case it would be a coincidence that the [O iii]/Hα ratio matches the surrounding AGN-ionized gas so closely, but no luminous star clusters appear in WFC3 continuum images (Keel et al. 2015).

Zoom In Zoom Out Reset image size

We can compare the changes in luminosity we measure to simulations such as Novak et al. (2011) to see how common such dramatic changes are predicted to be for various feedback properties. Existing simulations have time resolution of the order of 10⁴ yr, so there may be just enough information to tell what behavior is expected on timescales close to 10⁵ yr. Figure 13 shows how common various levels of drops in accretion luminosity (derived from the simulations as L/L_Edd) are predicted to be when sampled on various timescales. We emulate the measurements by smoothing the starting luminosity at each time step by 30% of the timespan being sampled, which mimics the effects of finite depth along the line of sight and increased recombination timescale at the typically lower densities farther from the AGN. Although the time resolution is not really adequate to test what happens when sampling a few 10⁴ yr, it is encouraging that this simulation shows a few per cent of its time steps that would satisfy our selection criteria for AGN fading if suitable surrounding gas were present (Table 6). The table lists the fraction of time steps where the luminosity drops at least as steeply as the listed ratios, evaluated for various values of ΔT according to $L(T)/\overline{L(T-{\rm{\Delta }}T)}$ where the averaging in the denominator is over 0.3ΔT as above. As Novak et al. (2011) note, their simulations show long periods of near-constant accretion, and other periods showing many cycles of nearly periodic changes.

Zoom In Zoom Out Reset image size

Some simulations indicate that smaller linear scales for changes in the inflow rate (from clumpy accretion or changes due to feedback from the AGN) yield faster changes in accretion rate (Gabor & Bournaud 2013).

We present HST narrowband imaging and Gemini integral-field optical spectra for a set of fading AGNs, selected for shortfalls in the energy budget between giant ionized clouds and the nuclei themselves, suggesting significant reduction in ionizing photons within the last ≈2 × 10⁴ yr. In particular, we use recombination balance to estimate the history of radiative output, finding that the feature in common is a radial drop in luminosity within 20,000 yr before our direct view of the nucleus. For earlier times, we see a range of behaviors—constant, brightening, fading—which are echoed in similarly reconstructed histories of PG QSOs where [O iii] HST imaging is available.

In contrast to many of the extended emission clouds around radio-loud AGNs, these are rotationally dominated and show only very localized outflow velocities greater than ≈100 km s⁻¹. These EELRs are mostly externally illuminated tidal debris rather than wind material (Keel et al. 2015).

As in Keel et al. (2012a), these results continue to support the idea that AGNs with extended emission regions are bright for periods of 10⁴–10⁵ yr at a time, with substantially fainter episodes interspersed (which fits with the idea that the opposite behavior—brightening—can be seen in the number of X-ray-bright AGNs without substantial narrow-line regions; Schawinski et al. 2015). Most galaxies do not have large reservoirs of (extraplanar) cold gas to show such behavior, so we can trace it preferentially in interacting or merging systems where warped disk gas or tidal tails provide a screen to be ionized by escaping ionizing photons.

Among EELR hosts from the Galaxy Zoo sample, fading AGNs make up ≈40% of the total, leading to the simple estimate of long accretion timescales in Keel et al. (2012a). This leaves open the possibility that fading cases could be a specific subset of the whole population. For example, inspiral of bound binary supermassive black holes could disturb a pre-existing accretion disk, while significant changes in the direction of ionizing radiation could also result from the kinds of tilted disk discussed by Lawrence & Elvis (2010). In that case, shadowing by the disk itself would leave ionization "cone" shapes that are set by the intersection of two cones of different center directions and opening angles.

This work was supported by NASA through STScI grants HST-GO-12525.01-A and -B. Some of the data presented in this paper were obtained from the Mikulski Archive for Space Telescopes (MAST). This research has made use of NASA's Astrophysics Data System, and the NASA/IPAC Extragalactic Database (NED), which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. We thank Linda Dressel for advice on setting up the HST observations, especially reducing intrusive reflections. Identification of this galaxy sample was possible through the efforts of nearly 200 Galaxy Zoo volunteers; we are grateful for their contributions, and thank once more the list of participants in Keel et al. (2012a). This work was partly supported by a grant from the President of the Russian Federation (MD-3623.2015.2). The observations obtained with the 6 m telescope of the SAO of the RAS were carried out with the financial support of the Ministry of Education and Science of the Russian Federation (contracts no. 16.518.11.7073 and 14.518.11.7070). We benefited from conversations with Hai Fu, Larry Rudnick, Kelly Holley-Bockelmann, Tamara Bogdanovic, and Stephanie Juneau. Greg Novak kindly provided detailed results of his numerical simulations. We thank Dara Norman and Kathy Roth for key help in setting up the Gemini observations, James Turner for sharing experience in subtleties of their processing, and Gemini staff observers M. Hoenig, A.-N. Chene, J. Chavez, M. Pohlen, L. Fuhrman, and J. Ball for obtaining these data in queue mode.

C. J. Lintott acknowledges funding from The Leverhulme Trust and the STFC Science in Society Program. W.P.M. is grateful for support by the University of Alabama Research Stimulation Program. V.N.B. acknowledges assistance from a National Science Foundation (NSF) Research at Undergraduate Institutions (RUI) grant AST-1312296. Note that findings and conclusions do not necessarily represent views of the NSF. A.M. is also grateful for the financial support of the "Dynasty" Foundation. K.S. was supported by a NASA Einstein Fellowship at Yale, and gratefully acknowledges support from Swiss National Science Foundation Grant PP00P2 138979/1. Galaxy Zoo was made possible by funding from the Jim Gray Research Fund from Microsoft and The Leverhulme Trust.

This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration.

Facilities: HST (ACS - , WFC3 - , WFPC2) - , BTA - , Gemini Gillette - , WISE. -