Mon.Not.R.Astron.Soc.000,1–21(2011) Printed6October 2011 (MNLATEXstylefilev2.2) The Galaxy Zoo survey for giant AGN-ionized clouds: past and present black-hole accretion events William C. Keel1,2,3⋆, S. Drew Chojnowski2,3,4, Vardha N. Bennert5,6, Kevin Schawinski7,8,9, Chris J. Lintott10,11, Stuart Lynn2,10, Anna Pancoast5, Chelsea Harris5, A.M. Nierenberg5, Alessandro Sonnenfeld5, & Richard Proctor12 1Department of Physics and Astronomy, University of Alabama, Box 870324, Tuscaloosa, AL 35487, USA 2Visiting Astronomer, Kitt Peak National Observatory,operated by AURA, Inc. under contract to the US National Science Foundation. 3 SARA Observatory 4TexasChristian University, Forth Worth, TX 76129 USA 5 Department of Physics, Universityof California, Santa Barbara, CA 93106 USA 6 PhysicsDepartment, California Polytechnic State University,San Luis Obispo, CA 93407, USA7Department of Physics, Yale University,NewHaven, 8Yale Centerfor Astronomy and Astrophysics, Yale University,P.O.Box 208121, NewHaven, CT06520, USA 9EinsteinFellow 10Astrophysics, Oxford University 11Adler Planetarium, 1300 S. Lakeshore Drive,Chicago, IL 60605 12Waveney Consulting, Wimborne, Dorset BH21 3QY ABSTRACT Some active galactic nuclei (AGN) are surrounded by extended emission-line regions (EELRs), which trace both the illumination pattern of escaping radiation and its history over the light-travel time from the AGN to the gas. From a new set of such EELRs,wepresentevidencethattheAGNinmanySeyfertgalaxiesundergoluminous episodes 0.2–2×105 years in duration. Motivated by the discovery of the spectacular nebula known as Hanny’s Voorwerp, ionized by a powerful AGN which has appar- ently faded dramatically within ≈105 years, Galaxy Zoo volunteers have carried out both targeted and serendipitous searches for similar emission-line clouds around low- redshiftgalaxies.Wepresenttheresultinglistofcandidatesanddescribespectroscopy identifying 19 galaxies with AGN-ionized regions at projected radii r > 10 kpc. proj This searchrecoveredknown EELRs(such as Mkn 78, Mkn 266,and NGC 5252)and identified additional previously unknown cases, one with detected emission to r =37 kpc.OnenewSy2wasidentified.Atleast14/19areininteractingormergingsystems, suggestingthat tidal tails are a prime source ofdistant gas out of the galaxy plane to be ionized by an AGN. We see a mix of one- and two-sided structures, with observed ◦ cone angles from23–112 . We consider the energy balance in the ionized clouds, with lower and upper bounds on ionizing luminosity from recombination and ionization- parameter arguments, and estimate the luminosity of the core from the far-infrared data. The implied ratio of ionizing radiation seen by the clouds to that emitted by the nucleus, on the assumption of a nonvariable nuclear source, ranges from 0.02 to >12; 7/19 exceed unity. Small values fit well with a heavily obscured AGN in which only a small fraction of the ionizing output escapes to be traced by surrounding gas. However, large values may require that the AGN has faded over tens of thousands of years,giving us severalexamples of systems in which such dramatic long-periodvari- ation has occurred; this is the only current technique for addressing these timescales in AGN history. The relative numbers of faded and non-faded objects we infer, and the projected extents of the ionized regions, give our estimate (0.2–2×105 years ) for the length of individual bright phases. Key words: galaxies: Seyfert — galaxies: ISM — galaxies: active ⋆ E-mail:[email protected] (cid:13)c 2011RAS 2 W.C. Keel et al. 1 INTRODUCTION catalogued AGN, since the most interesting objects - those which have faded dramatically - may no longer appear as The compact sizes of the central engines of active galactic spectroscopically classified AGN. nuclei (AGN) have long driven study of their distant sur- Hanny’s Voorwerp was first noted by Dutch teacher roundings for clues to their geometry and interaction with Hanny van Arkel in the course of the Galaxy Zoo project thesurrounding galaxy. Observationsof gas seen many kpc (Lintott et al. 2008), on the basis of its unusual structure from the AGN itself have proven fruitful in offering views and colour. In view of the interest of similar ionized clouds of the core from different angles, and implicitly at different for studyof both thehistory and obscuration of AGN,par- times. ticipantsintheGalaxy Zooproject havecarried out awide Narrowbandimagesrevealedextendedemission-linere- search for such clouds using data from the Sloan Digital gions (EELRs) around some luminous AGN, particularly SkySurvey(SDSS).TheyexaminedbothknownAGNhosts radio-loud QSOs as well as radio galaxies, as reviewed by and galaxies not known to have AGN, using the distinc- Stockton,Fu,&Canalizo(2006).Similarstructuresinlower- tive colour of highly-ionized regions across the SDSS gri luminosity Seyfert galaxies often appearas single or double filtersasafirstselection criterion.Wepresenttheresultsof trianglesinprojection(Ungeretal.1987,Tadhunter&Tsve- further analysis of the SDSSimages, narrow-band imaging, tanov1989),generallyinterpretedasionizationcones.When and spectroscopy, yielding a list of 19 galaxies with AGN- small-scale radiojetsarepresent,theyliewithintheioniza- photoionized clouds detected to beyond 10 kpc from the tioncones.However,inmanycases,thegasmustbeionized nuclei (many of which are newly identified). We consider by radiation from the nucleus rather than direct interac- constraints on changes in ionizing luminosity for these, and tion with a jet or outflow, as seen from narrow linewidths identifyseveralasthemostlikelycandidatesforthekindof and(particularlydiagnostic)modestelectrontemperatures, long-term fading seen in IC 2497 and Hanny’sVoorwerp. bothofwhichwouldbemuchlargerinthepresenceofshocks fast enough tomatch theobserved ionization levels. This is particularly true for very large EELRs, where interaction withtheradiojetoranorigininoutflowsalonebecomeless 2 SEARCHES FOR EMISSION-LINE CLOUDS andless likely.Infact, thebest-definedionization conesare seen in radio-quiet objects (Wilson 1996). The Galaxy Zoo search for giant AGN-ionized clouds com- This is one line of evidence linking large-scale struc- bined both targeted and serendipitous approaches, to com- turestothesmall-scaleobscuringregions(“tori”)impliedby bineacomplete examination of knownAGNhostswith the otherargumentsfor aunification scheme(Antonucci1993), possibility of finding ionized clouds around AGN which are in which Seyferts of types 1 and 2 are part of a single par- yet unknown or in fact optically unseen. In the targeted entpopulation,appearingdifferentbasedonhowourlineof search, we formed a sample of potential AGN at z < 0.1. sight passes this torus. The emission-line structures can be This combined all galaxies whose SDSS pipeline emission- largeandwell-resolved,offeringawaytomeasuretheopen- lineratiosputthemineithertheAGNorcompositeregions ingangleoverwhichionizingradiationescapes.Someprevi- oftheBaldwin-Phillips-Terlevich (BPT)diagram (Baldwin, ous studies have also noted that these emission-line clouds Phillips, & Terlevich 1981, asrevised byKewley et al. 2001 provideaviewtotheimmediatepastoftheAGN,vialight- andKauffmannetal.2003)using[OIII]/Hβand[NII]/Hα, travel time to the cloud and then toward us (Dadina et al. andalladditionalobjectslistedintheVeron-Cetty&Veron 2010). catalog (Veron-Cetty & Veron 2010) at z < 0.1 falling Using extended emission-line clouds as probes of AGN within the SDSS data release 7 (DR7) area. This addition history came of age with the discovery of Hanny’s Voor- accounted for AGN with no SDSS nuclear spectrum, either werp, a high-ionization region extending 45 kpc in projec- becausetheyarerelativelybrightor,moreoften,becausefi- tion from the LINER IC 2497 at z = 0.05 (Lintott et al. brecollisions orsampling rulespreventedtheirselection for 2009). Linewidths and electron temperature indicate that spectra, and type 1 AGN where the pipeline spectroscopic the gas is photoionized rather than shock-excited, while a classificationislessreliablethanfornarrow-lineobjects.The combination of ionization-parameter and recombination ar- merged AGN sample, designed to err on the side of inclu- guments bound the required nuclear ionizing luminosity to sion in borderline cases, included 18,116 objects. With a be1−4×1045 ergs−1.However,X-rayspectroscopyshows web interface designed by RP, 199 participants examined the nucleus of IC 2497 to be only modestly absorbed, with all of these within a 6-week period in 2009, marking each ionizing luminosity only ≈ 1042 erg s−1 (Schawinski et al. ascertain,possible,orlackinganextendedemission region. 2010a).Itisdifficulttoavoidtheconclusionthatthenucleus These emission regions have distinctive signatures in both of IC 2497 was in fact a QSO (thenearest known luminous morphology and colour from the SDSS data. They do not QSO) until roughly 105 years before our current view, and followtheusualspiralorannulardistributionsofstarforma- hasfadeddramaticallyintheinterim;radioandHSTobser- tion in disc galaxies. Such regions show unusual colours in vations offerhintsthat some of its energy output may have the SDSS composite images, which map gri bands to blue, switched to kinetic forms over this timespan (Josza et al. green,andred(Luptonetal. 2004). Hencestrong[OIII]at 2009,Rampadarathetal.2010,Schawinskietal.2010a,Keel low redshift is rendered as a pure blue, as in the discovery et al. 2011). The unlikeliness of the nearest QSO showing of Hanny’s Voorwerp. A combination of strong [O III] and highlyunusualbehavioursuggeststhatsuchvariationsmay significantHα+[NII]appearspurple;beyondaboutz =0.1, be common among AGN, prompting us to re-examine the [OIII]falls inthegap betweeng andr filters,sooursearch incidence and properties of extended ionized clouds around techniquelosesutilityuntil[OIII]iswellwithintherband, nearbyAGN.Suchanexaminationshouldnotbeconfinedto when the galaxies have much smaller angular sizes. This (cid:13)c 2011RAS,MNRAS000,1–21 Giant AGN clouds 3 subproject was known as the “voorwerpje hunt”, using the Dutchdiminutiveform of Voorwerp. Eachgalaxy wasexaminedbyatleasttenparticipants; 199 Zoo volunteers participated in this program, seven of g r g-0.75r whom examined the entire sample. The final average num- berofvoteswas11.2perobject.Afterthisscreeningprocess, a straightforward ranking was by relative numbers of “yes” (weight=1), “maybe” (weight=0.5), and “no” (weighted zero) votes. The most interesting results of such a search would be galaxies with prominent AGN-ionized clouds in which we don’t see the AGN, either because of strong obscuration or dramatic variability during the light-travel time from the nucleus to the clouds. These would not be found by tar- geting known AGN,and neitherwould cloudsaround AGN whichdonothavecatalogued spectralinformation.Accord- ingly, we also posted a request on the Galaxy Zoo discus- sion forum, with examples of confirmed AGN clouds and Figure 1. Linear combination of SDSS images to iso- various kinds of similar-appearing image artifacts. Partici- late candidate [O III] emission regions, shown with SDSS pants were invited to post possible cases from among the J143029.88+133912.0(theTeacupAGN).Theregionshownspans galaxies they saw in the ordinary course of the Galaxy Zoo 64×128SDSSpixels,or25.3×50.6”,withnorthatthetop.Each classification programs (Lintott et al. 2008), and some ac- image is displayed with a logarithmic intensity mapping, with a tiveusers reposted examples from other discussion threads. smallzeropointoffsettoreducetheeffects ofnoisearoundzero. The resulting followups1 provided an additional sample for At the redshiftz =0.085 of this galaxy, Hα and [N II] emission investigation;WCKalsocheckedallthethreadswithimage fallredwardoftherfilterband(response0.2%ofpeak)sother discussionformoresuchobjects,earlyexamplesofwhichin- imageisusedasacontinuum estimate. stigatedthissearchonthefirstplace.Toreducethenumber offalsepositivescausedbyextendedstar-formingregionsor starburstwinds,objectswereremovedfromconsideration if galaxy image are more vulnerable to this artifact than nor- an SDSS spectrum shows emission lines characteristic of a mal galactic nuclei. Since many candidates (includingsome starburst.Remainingcandidateswereexaminedfirstonthe withspectroscopic confirmation) have“purplehaze”on the SDSS composite images for appropriate colour and geome- SDSSimages,whichcouldeitherbegenuinelyextendedand try, and the most promising ones were carried forward for somewhat amorphous [O III] and Hα or an artifact, this furtheranalysis. was a helpful step. We adopted a tomographic approach, Both targeted and serendipitous lists overlap for many taking one of the SDSS bands free of strong emission lines objects with bright emission-line structures, and recover (r or i, depending on redshift) as an estimate of the struc- suchwell-studiedcasesfromtheliteratureasMkn266,NGC tureofstarlightinthegalaxy.Thiswasscaledtomatchthe 5252 and Mkn 78; we observed these so as to have a con- largest part of the g structure, iteratively when necessary. sistent set of spectra for comparison. The entire list of can- This is illustrated in Fig. 1, isolating the emission-line loop didates is given in Table 8. In the Survey column, S or T in SDSS 1430+13 (nicknamed the Teacup AGN because of denotes whether the object was found in the serendipitous thisstructure).ChojnowskiandKeelinspectedthebestsub- survey, the targeted survey of known AGN, or both. The traction among various scalings (often a compromise, due typeofnuclearopticalspectrumislisted asSy1/1.5/1.9/2, to colour gradients within the galaxy) to assess the reality LINER, SB for starburst, or nonAGN for an ordinary stel- of extended emission-line features not associated with clear lar population. The final column indicates which Galaxy spiralarmsorstellarrings.Theseresultsletusrankthecan- Zoo participant (by user name) first posted objects in the didatelistsfrombothtargetedandserendipitoussearchesin serendipitous survey. orderofsignificanceoftheemission-linestructuresbasedon theSDSSimages themselves.Weusedtheseresultstolimit thenumberofcandidatesfromthetargetedsearchtothetop 2.1 SDSS image analysis and new images 50;belowthistherewerenoconvincingcandidatesbasedon more detailed analysis of the SDSSimages. Forbothsubsamples,furtherwinnowinghadthesamesteps. Where appropriate filters were available for [O III] or Mostimportantly,wereanalyzedtheSDSSimages,toverify Hα at a galaxy’s redshift, some candidates were imaged at that the features do not have continuum counterparts, and theremote SARA1m (Kitt Peak) and 0.6m (Cerro Tololo) eliminate artifacts caused by imperfect registration of the telescopes. For [O III], we used a filter centered at 5100 ˚A imageswhenformingthecolourcomposites.Thiseffectisof withhalf-transmission width100˚A,useablefortheredshift particularconcernforSeyfert1nuclei,wherethePSFofthe rangez=0.009−0.025.AtHα,bothtelescopeshavestepped bright nucleus can produce a decentered colour signature if sets of filters 75 ˚A apart with FWHM ∼75 ˚A . Continuum one of theconstituent images is slightly misregistered; Sy 1 was taken from V,R or g,r, appropriately scaled for sub- tractiontoshownetemission-linestructures.Thesedataare particularlyhelpfulintracingtheemission-linestructuresof 1 inhttp://www.galaxyzooforum.org/index.php?topic=275014.0 UGC 7342 (Fig. 2) and SDSS2201+11 (Fig. 3). (cid:13)c 2011RAS,MNRAS000,1–21 4 W.C. Keel et al. run, so that the 3m spectra could be concentrated on the most interesting galaxies. Total exposures ranged from 30 minutes,for initial reconnaissance to see whether an object might host AGN clouds, to 2 hours for weaker lines in con- firmed targets. Either night-sky line or interspersed lamp observations were used to track flexure, as needed. Reduc- tion used the longslit package in IRAF2 (Tody 1986), and included rebinning to a linear wavelength scale, sky sub- traction,andfluxcalibration. Spectrophotometricstandard starswereobservedtosetthefluxscale;inafewcaseswhere passingcloudswereanissue,thespectrawerescaledsothat the nucleus within a 2×3” region matched the flux of the SDSSspectra. Our identification of these extended regions as being photoionized by AGN rests on three results - location in thestrong-lineBPTdiagram,strengthofthehigh-ionization species He II and [Ne V], and electron temperature consis- tentwithphotoionizationbutnotwithshockionization.We Figure 2.TheextendedcloudsinUGC7342.Left,thestarlight classify emission regions based on the “BPT” line-ratio di- continuum in a band at 6450 ˚A from the SARA 1m telescope. agrams pioneered for galactic nuclei by Baldwin, Phillips, Center,anestimated[OIII]imagefromtheSDSSdataasinFig. & Terlevich (1981) and refined by Veilleux & Osterbrock 1. Right, continuum-subtracted Hα image from the SARA 1m. (1987), with caution based on the possibility that some of North is at the top and east to the left; the field shown spans the external gas could have much lower metal abundances 97×150arcseconds. thanfoundingalactic nuclei(asseen inHanny’sVoorwerp; Lintott et al. 2009). Abundance effects in gas photoionized by AGN, as manifested in the BPT diagrams, have been considered in calculations by Bennert et al. (2006a). The largest effect is higher equilibrium temperature at lower O abundance, since it is an important coolant, which drives stronger forbidden lines and higher ionization levels until verylowlevels(0.1solar)arereached.Inanycase,theabun- dance changes are not large enough to move these clouds across the empirical AGN/starburst ionization boundary. Furthermore, in the galaxies where we have data covering the red emission lines, the clouds’ locations in the (essen- tiallyabundance-independent)auxiliaryBPTdiagramof[O III]/Hβversus[OI]/Hαalsoindicatedphotoionizationbyan AGN continuum.The various BPT diagrams are compared for Points along the slit in each of the clouds we classify as AGN-ionizedinFig. 4.Thisclassification isexamined more closely in the context of its radial behavior in the next sec- tion. Independent of these line ratios, strong He II λ4686 or Figure3.TheextendedcloudsinSDSS2201+11.Left,agimage [Ne V] λλ3346,3426 indicate photoionization by a harder fromtheSARA-S0.6mtelescope,showingthedustydisk.Right, continuumthanprovidedbyyoungstars,andresolvedemis- continuum-subtracted [O III] image from the SARA-N 1m tele- sion from these species is immediately diagnostic of AGN scope, smoothed by a Gaussian of 2.0” FWHM. North is at the photoionizationinthiscontext.Forsomeobjects,wedonot topandeasttotheleft;thefieldshownspans64×84arcseconds. havereddata;inthese,weclassifythecloudasAGN-ionized based on the presence of the high-ionization lines or conti- nuity of line ratios with the nucleus. Line ratios in the ex- 2.2 Spectroscopy treme blue may be affected in subtle ways by atmospheric ToconfirmthatregionsareinfactionizedbyAGN,andde- dispersion(Filippenko1982);theschedulingofourobserva- rivediagnosticemission-lineproperties,wecarriedoutlong- tionsforcedustoobservemosttargetsathourangleswhich slit spectroscopy for thehighest-priority candidates. Obser- did not allow us to put the slit simultaneously along the vations used the GoldCam spectrograph at the 2.1m tele- structuresofinterestandclosetotheparallacticangle.The scopeofKittPeakNationalObservatoryandtheKastdou- extended regions we observe are generally wider than the ble spectrograph at the 3m Shane telescope of Lick Obser- vatory. Table 1 compares the setups used for each session. The slit width was set at 2” for all these observations, and 2 IRAF is distributed by the National Optical Astronomy Ob- thespectrographswererotatedtosamplethemostextended servatory, which is operated by the Association of Universities known structures of each galaxy. Scheduling allowed us to forResearchinAstronomy(AURA)undercooperativeagreement reduce the Kitt Peak data before the first Lick observing withtheNationalScienceFoundation. (cid:13)c 2011RAS,MNRAS000,1–21 Giant AGN clouds 5 Figure 4. Summary Baldwin-Phillips-Terlevich (BPT) diagrams for the Lick spectra, where we measured the requisite red emission lines. Circles indicate points along the slit for extended clouds classified as AGN-ionized and the host nuclei. Gray-scale background showsthedensityofpointsfromalllow-redshiftSDSSgalacticnuclei,asinSchawinskietal.(2010b).Thedividinglinesbetweenregions photoionized byAGNandbyhotstarsareshownasgivenbyKewleyetal.(2001; Ke01)andbyKauffmannetal.(2003;Ka03). slit; to first order line intensities are not affected by atmo- remnants (including some with lower ionization levels than sphericrefraction, since we calibrate with standard stars at in these clouds) range from 20,000-69,000 K (e.g., Fesen et lowairmass.SomeoftheLickbluespectrahaveatmospheric al. 1982, Wallerstein & Balick 1990, Morse et al. 1995). In dispersioncontributingasmuchas3”ofoffsetalongtheslit addition,veryhighshockvelocities≈400kms−1areneeded from red to blue ends of the spectrum, important only for toproducesignificant[NeV]emission Dopita&Sutherland thenuclei and corrected in extracting their spectra. (1996). This is far in excess of the local velocity ranges we observe (section 6); even though we would not necessarily The BPT diagrams are designed to separate common observe material on both sides of a shock in the same ion, sourcesofphotoionizationingalaxies;temperatureandkine- it is difficult to envision a situation with large-scale shocks maticdataarealsoimportanttounderstandwhethershocks ofthisvelocity without observablevelocity widthsor struc- pay a significant role. In a few cases, the [O III] λ4363 line turesexceeding 100 k s−1. was measured in theextended clouds with sufficient signal- to-noise ratio for a measurement of the electron tempera- While not the main thrust of our survey, it is worth tureviaitsratiotothestrongλλ4959,5007lines.Usingthe noting that we find a few instances of either double AGN IRAFtasktemden,whichimplementsthealgorithmofShaw in interacting systems, or AGN in the fainter member of &Dufour(2007),andconsideringne <100cm−3,wefindTe a close pair (Mkn 177, Was 49, possibly SDSS J111100.60- valuesof18,600±1000 intheSDSS2201+11, 13,300±300 005334.9 and SDSS J142522.28+141126.5). These may be forMkn266,and15,400±500fortheTeacupsystem.These worth deeper spectroscopy in the context of mapping AGN confirmthatthegasisphotoionizedratherthanshocked;for obscuration; if a high-ionization component can be isolated comparison, temperatures in the [O III] zone of supernova in the gas of the other galaxy, its distribution could show (cid:13)c 2011RAS,MNRAS000,1–21 6 W.C. Keel et al. where ionizing radiation escapes any circumnuclearabsorb- 3 ENERGY BUDGET IN EXTENDED ingstructure.Thisoffersadistinctwayoftracingtheioniz- CLOUDS: OBSCURATION VERSUS ing radiation even in the absence of extensive gaseous tidal VARIABILITY features, in an approach that has been discussed for Was 49ab by Moran et al. (1992). Seeing the effects of radiation from an AGN on gas tens of kpcfromthenucleusallowsusthepossibilityoftracingdra- matic changes in core luminosity. One straightforward way to approach this question is a simple energy balance. The Table4liststheresultsofourspectroscopy.Confirmed, spectragiveusupperandlowerboundsontherequiredion- resolvedcloudsionizedbytheAGNareseparatedfromother izing luminosity. To probe themost extreme conditions, we results(unresolvedAGNemission,extendedstar-formingre- analyzegalaxiesinwhichwedetectionizedgasatprojected gionsdenotedasHII,andsoon).Theinstrumentsusedare distances r>10 kpc.For all distances and luminosities, we denotedbyGCam(KittPeakGoldCam)andLick(Lick3m use the WMAP “consensus” cosmological parameters, with with Kast spectrograph). New redshifts and spectral classi- H =72 km s−1 Mpc−1 (Spergel et al. 2007). fications are marked with asterisks. We separate the AGN 0 clouds of most interest based on the detected extent of [O Thelowerboundcomesfromthehighestrecombination- III]λ5007;ourspectrahavealowerdetectionthresholdthan line surface brightness we observe; the central source must ourimagesforthis,roughly10−16 ergcm−2s−1arcsec−2 for provideatleastenoughionizingphotonstosustainthisover emission regionsafewarcsecondsinsize.Spectraofthenu- periodslongerthantherecombinationtimescale(whichmay clei and representative cloud regions are shown in Figs. 5 be as long as 104 years at these low densities). This is a and 6. Table 5 lists emission-line ratios and selected fluxes lowerlimit,sincetheactualemission-linesurfacebrightness forthesameregionsplottedinthesefigures.Fluxesaregiven of some regions may be smeared out by seeing, and we do bothfor[OIII]λ5007andHα,sincethesewereusuallymea- notknowthatagivenfeatureisopticallythickattheLyman sure with different gratings and detectors. For some of the limit. This limit depends only very weakly on the slope of nuclei, correction of the Hβ flux for underlying absorption the ionizing continuum, since helium will generally absorb inthestellarpopulationwassignificant;wehaveappliedan mostoftheradiationshortwardofitsionizationedgeleaving approximatecorrectionbasedontypicalvaluesforsynthetic onlythe13.6-54.4 eVrangetoconsiderforhydrogenioniza- stellar populations from Keel (1983). tion.Webaseourboundsonthehighestimpliedluminosity among structures at various projected radii in a given sys- tem, with no correction for projection effects. This makes The upper part in Table 4, with AGN-ionized gas de- our limits conservative, since a given cloud will always lie tectedmorethan10kpcfromthenucleus,formsthesample farther from the nucleus than our projected measurement. for our subsequent analysis. As a sign of completeness, of Inessence, thisargumentisbased onthesurfacebrightness these 19, 14 were found in both targeted and serendipitous inarecombinationline;weuseHβ sincewehavethesedata searches. SDSS J095559.88+395446.9 was newly identified for the whole sample. In a simple approximation, we take asatype2Seyfertinourspectrum,afterhavingbeenfound the surface brightness in the brightest portion of a cloud, in the “blind” search of galaxies independent of prior clas- assuming this to be constant across the slit. We take the sification as an AGN host (so it was not included in the region sampled in this way to be circular in cross-section targeted sample). Of the remainder, Mkn 78 and Mkn 463 as seen from the nucleus, so its solid angle is derived from were selected in thetargeted AGNsample, whileMkn 1498 the region subtended by the slit. We then see this region and UGC 11185 were recognized only in the serendipitous occupying a small angle α = 2arctan(slit halfwidth/r) as survey. It may be relevant that both Mkn 78 and Mkn 463 seen projected at angular distance r from theAGN,there- have ionized regions with relatively small projected extent, quiredionizingluminosity isgivenfrom observedquantities easilylost against thegalaxystarlight(whichinMkn463is as Lion = 1.3×1064z2F(Hβ)/α2 for α in degrees. The de- morphologically complex). rivedvaluesarelistedinTable6,alongwithcomplementary quantitiesrelatedtothenuclearluminosity(ascollectedbe- low).Thederivedionizingluminositiesarelowerlimits,since The [S II] λ6717,6731 lines are particularly important, theremaybeunresolvedregionsofhighersurfacebrightness, tracing electron densities and thereby providing one esti- andwedonotknowwhetheragivencloudisopticallythick mate of the intensity of the impinging ionizing radiation. in the Lyman continuum. Higher-resolution imaging in the Since the densities in these extended clouds are low, and emission lines could this increase these values. the ratio is generally near its low-density limit, where the Upper limits to the incident ionizing flux come from mapping from line ratio to density is highly nonlinear, we a complementary analysis using the ionization parameter haveexaminedtheerrorsinmeasuringthelineratioclosely. (U, the ratio of ionizing photons to particles), since these We generated multiple realizations of pixel-to-pixel noise, emission-linefeaturesallhave[SII]lineratiosnearthelow- and each was scaled to four representative fractions of the density limit. Our density results from the λ6717/λ6731 [S strongerlinepeak.Thiswasaddedtolinepairs,modeledto II] line ratio are given in Table 7. Values are listed only for match the line widths and pixel separation of the red Lick objectswithusefulmeasuresfarfromthecore.Ineachcase, data. Gaussian fitting of the lines with added noise gave a weevaluatedthedensityatatypicaltemperatureof104 K, relation between the peak signal-to-noise and error of the andatthehighertemperature1.3×104 KfoundinHanny’s fitted ratio which we adopted; we use ±2σ error bounds to Voorwerp(Lintottet al.2009) andinourdataforMkn266 deriveboundsonthedensity.Densityvalueswerecalculated and SDSS 2201+11, as set by thermal equilibrium for sub- using theIRAFtask temden. stantiallysubsolaroxygenabundance.Wequotetheextreme (cid:13)c 2011RAS,MNRAS000,1–21 Giant AGN clouds 7 [Ne V] 3346[Ne V] 3426 [O II] 3727 [Ne III] 3869 [Ne III] 3968[S II] 40Hδ6 94102 Hγ 434[1O III] 4363 [He II] 4686 Hβ 4861 [O III] 4[95O 9III] 5007[O I] 6300[O I] 6364 Hα 65[6N3 II] 6584 [S II] [6S 7I1I]6 3675300 800 nucleus cloud cloud: 6.3" (6.6 kpc) wsw nucleus Mkn 78 300 600 (Lick - Kast) 250 400 200 200 110500 0 50 −200 0 3500 3700 3900 4100 4300 4500 4700 4900 5100 6600 6800 7000 11802000 nucleus cloud cloud: 21.9" (12.5 kpc) north nucleus M(Lickk n- K 2as6t)6 111024000 60 80 40 60 20 40 0 20 −20 0 3500 3700 3900 4100 4300 4500 4700 4900 5100 6500 6700 6900 235000 nucleus cloud cloud: 8.4" (6.4 kpc) east nucleus Mkn 883 2350 200 (Lick - Kast) 20 150 100 15 50 10 172(10erg/cm/s/)x Å− −−−2111145422068024500000000000000 35003600 nnuucclleeuuss37030cc8ll0oouu0dd 39040000 cc4lloo10uu402dd0::0 71.40."0 "( 7(.884.3 5k04 04pk0cp0)c )n nnnww 45406000 47nn408uu00cc0lleeuuss 49500000 51502000 MN((LLiiG6cck67kk6 0nC--00 KK 0 1aa5ss4tt9))9782 66980000 717000000501234511100000024000 Cloud Flux (10erg/−17 Flu 100 6800 cm Nucleus 3520005 3500 nucleus 37cl0o0ud 3900 clou4d10: 0 5.7" (5.6 k4p3c00) sse 4500 nu47c0l0eus 4900 5100 6S50D0SS 09556700 6900 0242000 Å/s/)2 20 (Lick - Kast) 15 15 150 10 0 5 −5 0 3600 3800 4000 4200 4400 4600 4800 5000 5200 6600 6800 7000 40 40 30 nucleus cloud cloud: 4.2" (4.3 kpc) ene nucleus S(LDickS - SKa s1t)005 3305 20 25 20 10 15 0 510 −10 0 3600 3800 4000 4200 4400 4600 4800 5000 5200 6700 6900 7100 100 50 6800 nucleus cloud cloud: 11.7" (7.0 kpc) nne nucleus S(LDickS - SKa s2t)201 3400 40 20 20 0 10 −20 0 3500 3700 3900 4100 4300 4500 4700 4900 5100 6500 6700 6900 Observed Wavelength (Å) Figure 5. Lick spectra for nuclei and associated AGN-ionized clouds. Small insets at left show the Hβ+[O III] region for nuclei and clouds,scaledtothepeakofλ5007emission.Panelsontherightshowthe[OI]-[SII]regionatthesamefluxscaleasthebluespectra. He II and [Ne V] emission,especially important as indicators of a hard ionizingradiation field, are indicated by red dotted lines when clearlydetectedinclouds.Nuclearspectrarepresent2×3.1-arcsecondareas,andcloudspectraaresummedover2×6.2-arcsecondareas. Distances and directions ofcloudrelative tonuclei areindicated asshown. Threespectra havegaps inthe blueregion,sincethey were takenwiththedichroicsplittingredandblueoptical trainsnear4600˚A. rangeofdensityvaluesbetweenthesetwocases(allowingin aregiveninTable7.Itisreassuringthattheupperlimitsto theTeacup an upperbound on theelectron densityas high ionizingluminosityderivedfromU andne alwaysfallabove as240cm−3,andinsomecaseslimits <10cm−3),sincethe thelower limits from recombination balance. temperature-sensitive[O III]λ4959+5007/λ4363 lineratio The lower limits from recombination-balance are inde- isnot well-enough measured in most of theseobjects touse individualTe values. WederiveU from the[O II]λ3727/[O pendent of assumptions about thelocal density ne, making it more robust than ionization-parameter arguments when III]λ5007ratiousingthepower-lawcontinuummodelsfrom we have no independent tracer at these low densities. Fig. Komossa&Schulz(1997),andtheanalyticfitsfromBennert 7 shows several of our objects in one of the “BPT” dia- (2005) as interpolation tools. For fully ionized hydrogen at grams, going beyond their initial use to classify the gas as a distance d from the AGN, thephoton flux in theionizing continuumisQ=4πd2neU/c.Forobjects withredspectra, AGN-ionized to examine changes with projected distance from the nuclei. Some of these, such as Mkn 1498 and the givingdensitiesfromthe[SII]lines,limitstotheluminosity Teacup 1430+13, show a phenomenon remarked earlier in, (cid:13)c 2011RAS,MNRAS000,1–21 8 W.C. Keel et al. [Ne V] 3346[Ne V] 3426 [O II] 3727 [Ne III] 3869[Ne III] 3968[S II] 4H0δ6 94102 Hγ 43[41O III] 4363 [He II] 4686 Hβ 4861 [O III] [49O 5I9II] 5007[O I] 630[0O I] 6364 Hα 6[56N 3II] 6584 [S II][ S6 I7I1] 66730 50 120 nucleus cloud cloud: 6.6" (11.3 kpc) ne nucleus Teacup 100 40 (Lick - Kast) 80 30 60 ux 40 20 C Fl 20 10 lo us 0 0 ud ucle 150 3700 nucleus 39c0l0oud 4100 cl4o3u00d: 6.8" (545.600 kpc) eas4t700 490n0ucleus 5100 5300 UG690C0 111857100 7300 50 Flu N 100 (Lick - Kast) 40 x 30 50 20 0 10 −50 0 3500 3700 3900 4100 4300 4500 4700 4900 5100 6600 6800 7000 Observed Wavelength (Å) Figure 5–continued for example, NGC 5252 (Dadina et al. 2010) - the ioniza- tween 42-122µ. Numerically, tniaotnurbaalllyanecxepslatianyesdriofutghhelychcaornascttaenrtistwicithdernasditiyusn,ew∝hicrh−2is. FIR(W m−2)=1.26×10−14(2.58f60+f100) This might occur naturally for gas in the host galaxy; tidal forIRASfluxesinthe60and100µbandsgiveninJy(mul- streamsofgaswouldnotbelikelytomatchtheextrapolated tipliedby107foraresultinergscm−2s−1).IRASdatawere behavior of gas within the galaxy and indeed we see some supplemented,where possible, by Akari data (Murakami et cases (Mkn 266, NGC 5972, SDSS 2201+11) with substan- al. 2007, Kawada et al. 2007, Yamamura et al. 2010) of tial radial changes in U. However, for Seyfert narrow-line higher accuracy. The positions of all these galaxies were regions, Bennert etal. (2006b) findashallower densitygra- covered in the IRAS survey,so we can assign typical upper dient ne ∝ R−1.1, which would imply U ∝ R−0.9 for gas limitstonondetectionsdependingoneclipticlatitude;Akari which is optically thin (or has a small covering fraction). addedtwoadditionaldetectionsnotfoundintheIRASdata, Theseobjectshaveheterogenousbehavior;Intheionization using only quality 1 (confirmed detection) fluxes. For non- coneofNGC7212, Cracco etal.(2011) findnoradialtrend ULIRG objects (LFIR < 1045 erg s−1), we can reproduce of ne. the IRAS FIR parameter from Akari 90µ fluxes and mean Similar conclusions come from the more limited blue- colours via line diagram also considered by Baldwin, Phillips, & Ter- FIR(W m−2)≈5.0×10−14f levich (1981), which we can apply to the objects for which 90 we have only blue-light spectra from KPNO. Some of ob- with 30% accuracy (±0.15 dex), and we use this to fill in jects in this diagram as well as in Fig. 7 show systematic FIR luminosities for the objects detected only by Akari. changes in ionization level with radius, manifested as off- Theinputvaluesand results of thisenergy-balancetest are setsfromupperleft(higherionization)tolowerright(lower showninTable6.Withinoursample,Mkn273andMkn266 ionization). We show this behaviorin Fig. 8. are classic ultraluminous infrared galaxies (ULIRGs), with We use far-IR data to estimate (or limit, for nondetec- LFIR = 1−5×1045 erg s−1. Some of the others remain tions) the amount of AGN radiation absorbed (and reradi- undetectedin both the IRASand Akari surveys,leading to ated)bynearbydensematerial,whetherinanAGN“torus” limits typically < 1044 erg s−1. An index of whether the or in the inner parts of the host interstellar medium. The extendedcloudscanbeionizedbyanobscuredAGNispro- FIR luminosity is conservatively high as an estimate of vided by the ratio of required ionizing luminosity to FIR thepotentialobscured AGN luminosity,since theremay be luminosity,tabulatedinTable??.Thesevaluesarealllower a nontrivial contribution from star formation in the host limits,sincetheionizingluminosityisalowerlimit Thisra- galaxy as well as the AGN, and in some cases companion tio ranges from 0.02 to values > 12 (FIg. 9). Low values galaxies might blend with the target in the FIR beam. In clearly represent AGN which are strongly obscured along a simple picture where a fraction f of the AGN radiation ourlineofsightbutnottowardtheEELRclouds.Largera- is absorbed by nearby dust and reradiated, the FIR lumi- tiosindicatelong-term fadingof theAGN,aspectral shape nosity will beof order Lionf/4π, with an additional scaling strongly peaked in theionizing UV, or very specific geome- factorofafewtoaccountfornon-ionizingradiationheating tryforobscuringmaterial,andthusindicateobjectsworthy the grains (which we omit at this point for the sake of a of close attention. conservative calculation). For convenience, we approximate Arguments for long-timescale variations in the central thetotalfar-IRoutputbytheFIRparameterintroducedfor sources here depend on our having estimates for their total IRASpoint-sourcecatalogdata(Fullmer&Lonsdale1989), luminosity as seen directly, which could in principle fail ei- alinearcombinationoffluxvaluesinthe60and100µbands theriftheirionizingradiationwerecollimatedbysomething whichgivesareasonableapproximationtothetotalfluxbe- other than obscuration, or the spectral shapes in the deep (cid:13)c 2011RAS,MNRAS000,1–21 Giant AGN clouds 9 [Ne V] 3426 [O II] 3727 [Ne III] 386[9Ne III] 3968[S II] 4Hδ0 649102 Hγ 4[34O 1III] 4363 [He II] 4686 Hβ 4861[O III] [4O 9I5II9] 5007 [Ne V] 3426 [O II] 3727 [Ne III] 386[9Ne III] 3968[S II] 4Hδ0 649102 Hγ 4[34O 1III] 4363 [He II] 4686 Hβ 4861[O III] [4O 9I5II9] 5007 250 200 120 80 IC 2637 cloud: 5.6" (3.3 kpc) sw nucleus 100 Mkn 273 cloud: 9.8" (7.6 kpc) sw nucleus 70 200 (KPNO - GCam) 150 (KPNO - GCam) 60 80 150 50 100 60 40 100 40 30 50 50 20 20 10 0 0 0 0 3700 3900 4100 4300 4500 4700 4900 5100 3600 3800 4000 4200 4400 4600 4800 5000 5200 100 140 100 800 Mkn 463 cloud: 10.9" (11.0 kpc) south nucleus 120 Mkn 739 cloud: 10.6" (6.4 kpc) nnw nucleus 600 (KPNO - GCam) 80 100 (KPNO - GCam) 80 )Å 400 60 80 60 C 172(10erg/cm/s/us Flux − −12250000000000 N(K3G7P0NC0O - 4GC3a8m8)3900 clo41u0d0: 12.9" (243.40 0kpc) ssw4500 4700nucleus4900 5100 024681100000200 2231140520605000000000 N3(6K0GP0NCO - 5GC2a5m23)800 c4lo0u00d: 14.4" (462.070 kpc) sou4t4h00 4600nucleus4800 5000 02446810000000loud Flux (10erg/cm/−172 Nucle 0 2400 500 20 Ås/) −500 −50 0 0 3500 3700 3900 4100 4300 4500 4700 4900 3600 3800 4000 4200 4400 4600 4800 5000 60 40 50 50 SDSS 1510 cloud: 7.3" (6.8 kpc) south nucleus 35 60 SDSS 1524 cloud: 12.8" (9.6 kpc) sse nucleus 40 40 (KPNO - GCam) 30 (KPNO - GCam) 30 25 40 30 20 20 10 15 20 20 −100 510 0 10 0 0 3700 3900 4100 4300 4500 4700 4900 5100 3600 3800 4000 4200 4400 4600 4800 5000 5200 50 Observed Wavelength (Å) 150 UGC 7342 cloud: 19.5" (18.7 kpc) sw nucleus 40 (KPNO - GCam) 100 30 50 20 0 10 0 3700 3900 4100 4300 4500 4700 4900 5100 Observed Wavelength (Å) Figure 6.KPNOGoldCamspectrasamplingnucleiandassociatedAGN-ionizedclouds.Spectraarescaledtoshow[OIII]λ4959,with nuclearspectrarepresenting2×3.1-arcsecondareasandcloudspectrarepresenting2×6.2-arcsecondareas.HeIIand[NeV]emission, especially important as indicators of a hard ionizing radiation field, are indicated by red dotted lines when clearly detected in clouds. Distancesanddirectionsofcloudrelativetonucleiareindicatedasshown. ultravioletdifferfromourexpectationsbasedontheUVand (Condon et al. 1998). All but two objects (SDSS 1510+07 X-raybehavioroffamiliar AGN.Collimation byrelativistic and1005+28)weredetectedabovethe2.5mJysurveylimit; beamingdoesnotfitthecombinationsofopeningangleand thesource luminosity L(1.4 GHz) ranges from <1.3×1022 flux ratio is required (as found for Hanny’s Voorwerp; Keel WHz−1 to2.0×1024,thelatterforthedoublesourceasso- et al. 2011). A spectral solution to the behavior would re- ciated with NGC 5972 and comprising 94% of the galaxy’s quireanextreme-ultraviolet’bump”dominatingtheionizing total flux. Eight of the galaxies qualify as radio-loud if one flux from 13.6-54 eV by more than an order of magnitude. uses a simple, representative division at 1023 W Hz−1, and However, known AGN do not provide evident examples of onlyoneliesabove1024.Thisone-NGC5972-representsan either solution; the most straightforward interpretation of interesting departurefrom theusualalignment of emission- the data suggests that some of these clouds are ionized by line and radio structure(section 7). AGNwhichhavefadedoverthedifferentiallight-traveltime between ourviews of theclouds and nuclei. 5 HOST AND CLOUD MORPHOLOGY The examples of Hanny’s Voorwerp (Josza et al. 2009) and 4 NUCLEAR AND EXTENDED RADIO NGC5252(Prieto&Freudlng1996)suggestthatacommon EMISSION source of extraplanar gas at large radii is tidal debris. The To further characterize the AGN in these galaxies, we col- host morphologies of the galaxies where we find extended lectedradiofluxesat1.4GHzfromtheNVSSsourcecatalog ionized clouds support this notion. Table 8 lists morpho- (cid:13)c 2011RAS,MNRAS000,1–21 10 W.C. Keel et al. Figure 7.Baldwin-Phillips-Terlevich(BPT)diagramsforspatialslicesinLickdata. EachrowshowsthethreeclassicalBPTdiagrams foreach object, highlightingradial ionizationbehaviors. Thenucleus isindicated bycrosshairs,withcolors changing fromwhitetored with increasing distance from the nucleus. The greyscale background and dividing lines are the same as in Figure 4; these show only the region around the AGN loci in each case for discriminationof detail. Nearlyall measurements liefirmlyin the AGN domain, with possible exceptions in some regions of SDSS 1005 and SDSS 0955. The starburst/AGN ionization boundary from Kewley et al. (2001) is shownas the redfullcurve, whilethe boundary fromKauffmann et al. (2003) isthe black dashed curve. Allthese measurements lie firmlyinthe AGN domain, with the possible exception of two regions inSDSS 1005 and the nucleus of Mkn 883, whose red spectrum showsabroad-lineregionandstrong[OI].Thegreyscaleshowthedensityofpointsrepresentinglow-redshiftgalacticnucleiintheSDSS, fromSchawinskietal.(2010b), whichwealsofollowinadoptingthestraightlineastheLINER/Seyfertboundary (cid:13)c 2011RAS,MNRAS000,1–21