Giant scattering cones in obscured quasars
Georges Obied, Nadia L. Zakamska, Dominika Wylezalek, Guilin Liu
SSubmitted to MNRAS Sept 19, 2015
Preprint typeset using L A TEX style emulateapj v. 01/23/15
GIANT SCATTERING CONES IN OBSCURED QUASARS
Georges Obied , Nadia L. Zakamska , Dominika Wylezalek , Guilin Liu Submitted to MNRAS Sept 19, 2015
ABSTRACTWe analyze
Hubble Space Telescope observations of scattering regions in 20 luminous obscuredquasars at 0 . < z < .
65 (11 new observations and 9 archival ones) observed at rest-frame ∼ −
10 kpc-scale scattering regions in almost all cases. The median scatteringefficiency at this wavelength (the ratio of observed to estimated intrinsic flux) is 2.3%, and 73% of theobserved flux at this wavelength is due to scattered light, which if unaccounted for may strongly biasestimates of quasar hosts’ star formation rates. Modeling these regions as illuminated dusty cones, weestimate the radial density distributions of the interstellar medium as well as the geometric propertiesof circumnuclear quasar obscuration – inclinations and covering factors. Small derived opening angles(median half-angle and standard deviation 27 o ± o ) are inconsistent with a 1:1 type 1 / type 2 ratio.We suggest that quasar obscuration is patchy and that the observer has a ∼
40% chance of seeinga type 1 source even through the obscuration. We estimate median density profile of the scatteringmedium to be n H = 0 . − . /r ) cm − , depending on the method. Quasars in our samplelikely exhibit galaxy-wide winds, but if these consist of optically thick clouds then only a small fractionof the wind mass ( (cid:46) Subject headings: galaxies: ISM – polarization – quasars: general – scattering INTRODUCTION
The unification model proposed by Antonucci (1993)holds that active galactic nuclei (AGN) are intrinsicallythe same and that differences in their observed spectralproperties are due to different relative orientations of thenucleus and surrounding opaque dust with respect to theobserver. An active nucleus possesses a compact strongradiation source (presumably an accretion disk aroundthe supermassive black hole and a surrounding corona),as well as an emission-line region with characteristic ve-locities of a few thousand km s − . If we have a directview to the central engine, then we observe an unob-scured, “type 1” source, with characteristic strong ultra-violet, optical and X-ray radiation from the accretiondisk and broad emission lines. If the direct view to thecentral engine is blocked by circumnuclear dust, then ahidden “type 2” nucleus may be identified using indirectsignatures, such as strong infrared radiation producedby the circumnuclear dust or narrow emission lines pro-duced in photo-ionized regions outside obscuration whichhave a direct view to the nucleus.The unification model was developed and tested fornearby relatively low-luminosity AGN – Seyfert galax-ies – using imaging, polarimetry and spectropolarimetryof scattered-light regions (Antonucci & Miller 1985; Bai-ley et al. 1988; Miller & Goodrich 1990; Miller et al.1991; Tran et al. 1992; Tran 1995a,b,c; Capetti et al.1995; Kishimoto 1999). Even if the direct view to thenucleus is obscured, the quasar light can escape alongother unobscured directions, scatter off surrounding ma- Department of Physics & Astronomy, Johns Hopkins Uni-versity, Bloomberg Center, 3400 N. Charles St., Baltimore, MD21218, USA Department of Physics, Harvard University, 17 OxfordStreet, Cambridge MA 02138, USA Department of Physics, Virginia Tech, Blacksburg, VA24061, USA terial and reach the observer. The scattered componentcan be then identified either via its polarization signatureor its morphology in imaging observations, as illumina-tion of extended material by a light source blocked alongsome directions produces a characteristic conical shape.Whether the same orientation-based unification modelis directly applicable to quasars – high-luminosity( L bol (cid:38) erg s − ) active nuclei – is still unclear. Eventhe basic measurement of the ratio of obscured to unob-scured quasars remains problematic (Lawrence & Elvis2010). For quite a while, few obscured quasars wereknown, leading to suggestions that the powerful radia-tion of a luminous quasar obliterates obscuring dust outto large distances and that the obscured fraction mightas a result decline with luminosity (so-called ‘recedingtorus’ model). These ideas found support in X-ray sur-veys (Ueda et al. 2003; Steffen et al. 2003; La Francaet al. 2005; Barger et al. 2005; Hasinger 2008) which un-covered few obscured quasars at high luminosities and,indirectly, in the spectral energy distribution of luminousquasars (Treister et al. 2008). At the same time, wide-field infrared, optical and radio surveys uncovered largenumbers of obscured quasar candidates whose numberdensities are similar to those of unobscured quasars at thesame luminosity (Zakamska et al. 2003; Lacy et al. 2004;Stern et al. 2005; Mart´ınez-Sansigre et al. 2006; Glikmanet al. 2007; Reyes et al. 2008; Donley et al. 2012; Assefet al. 2015; Lacy et al. 2015). Thus a full accounting ofthe obscured quasar population is still incomplete.Furthermore, the physical nature and the dynamicalstate of the obscuring material remains poorly under-stood (Krolik 2007). Despite many questions aboutthe origins and the long-term stability of an ‘obscuringtorus’, this concept remains popular, in part because itfinds strong support in observations. Indeed, classicalpolarimetric observations of low-redshift luminous ob-scured quasars confirm that some of these objects would a r X i v : . [ a s t r o - ph . GA ] D ec Obied et al.be seen as unobscured type 1 sources if viewed along an-other direction, and a number of very extended ( (cid:38) h =0.7, Ω m =0.3, Ω Λ =0.7 cosmology throughout this pa-per. Although both the Hubble Space Telescope and theSloan Digital Sky Survey use vacuum wavelengths, weuse air wavelengths for designating emission lines follow-ing long-standing convention. SAMPLE SELECTION, OBSERVATIONS AND DATAREDUCTIONS
Type 2 quasars studied in this paper are drawn fromthe optically-selected type 2 quasar candidates presentedby Zakamska et al. (2003) and Reyes et al. (2008). Thislarge parent sample is selected from the first generationof the spectroscopic part of the Sloan Digital Sky Survey(SDSS; York et al. 2000) based on emission-line prop-erties. Briefly, type 2 quasar candidates at z (cid:46) < − ) emission lines with line ratios character-istic of photo-ionization by a hidden quasar continuum.Because the full range of standard emission line diagnos-tics (Baldwin et al. 1981) is inaccessible in the opticalspectrum for z (cid:38) .
3, we require a high [OIII] λ β ratio (at least 3, but in practice ∼
10 for the luminoustype 2 quasars discussed here) accompanied by addi-tional signatures to distinguish these objects from low-metallicity star forming galaxies (e.g., high-ionizationemission lines such as [NeV] λλ erg s − )infrared luminosities presumably due to the thermal re-radiation of the quasar emission by the obscuring dust(Zakamska et al. 2004, 2008; Mateos et al. 2013; Zakam-ska et al. 2015). The fraction of radio-loud objects (Za-kamska et al. 2004; Lal & Ho 2010; Zakamska & Greene2014) is similar to that of unobscured quasars, suggestingthat the incidence and detectability of powerful jets fromsupermassive black holes is roughly independent of thegeometry of obscuration. Finally, spectropolarimetry oftype 2 quasar candidates reveals polarized broad emis-sion lines typical of unobscured (type 1) quasars, directlyconfirming that these objects would be seen as type 1salong some other lines of sight (Zakamska et al. 2005).Type 2 quasars are ideally suited for studies of hostsof luminous quasars as the circumnuclear obscurationprovides a natural coronagraph and an opportunity toobserve the host galaxy without the bright glare of thequasar itself. We therefore conducted follow-up HubbleSpace Telescope imaging of some of type 2 quasar can-didates on two occasions. In 2003 – 2004, nine radio-quiet type 2 quasars were selected on the basis of theirhigh [OIII] luminosity from the original parent sampleof about 150 type 2 quasars of Zakamska et al. (2003);these
HST observations (GO-9905, PI Strauss) were firstpresented by Zakamska et al. (2005, 2006) and are re-analyzed here.As the SDSS progressed and the sample was expandedto ∼
900 sources (Reyes et al. 2008), we were able toextend the luminosity range to higher values of L [OIII].A further subsample of 11 sources with high [OIII] lu-minosities was extensively studied using Gemini (Liuet al. 2013a,b), and new HST observations of these ob-jects (GO-13307, PI Zakamska) were conducted in 2013– 2014 and are presented here. We analyze the 20 ob-jects from programs GO-9905 and GO-13307, measuringtheir scattering geometry and estimating scatterer den-sities. Finally, the most [OIII]-luminous type 2 quasar inthe Reyes et al. (2008) catalog – IRAS 09104+4109 – isone of the first type 2 quasar candidates known (Klein-mann et al. 1988). It was observed using the
HST andanalyzed by Hines et al. (1999), and we use some of theirmeasurements in this paper. Thus the total sample sizeis 21 sources, comprised of 11 new observations and 10archival ones. The [OIII] and infrared luminosities ofour targets (as measured using
Wide-Field Infrared Sur-vey Explorer , Wright et al. 2010) are shown in Figure 1and the sources are tabulated in Table 1.The new observations with the
HST – which we discussbelow in detail – are similar in concept to the archivalones presented by Zakamska et al. (2006), with differ-ences in filter selection due to differences in redshift. Inthe new program, each target is imaged for one orbitblueward of the 4000˚A break, with a typical effectivewavelength near 3000˚A (hereafter described as rest-frame U -band observations) and for one orbit in the rest-frameiant scattering cones 3 Fig. 1.—
Left: Line luminosities and 12 µ m infrared luminosities (from WISE ) for the archival
HST sample (open circles, from Zakamskaet al. 2003, 2005, 2006), new
HST sample (filled squares, from Liu et al. 2013a,b) and IRAS 09104+4109 (filled circle, from Hines et al.1999) compared to SDSS type 2 quasars (grey points; Reyes et al. 2008). Right: scattering efficiencies at 3000˚A – the ratio of observed toestimated intrinsic luminosity at this wavelength. The points show the scattering efficiencies calculated based on our best stellar subtraction,whereas the error bars show the upper limits on the scattering efficiencies if no stellar subtraction is performed and all of the observed U -band flux is attributed to scattered light. yellow (between V and R ), with both bands carefullychosen to sample the continuum and to avoid strongforbidden emission lines. If the scattering efficiency isroughly wavelength-independent (Kishimoto 1999), thenthe spectral energy distribution of the scattered light isthe same as that of an unobscured quasar – i.e., this com-ponent is blue – whereas in the absence of strong starformation the host galaxy is expected to be red. There-fore, by observing blueward of the 4000˚A break we max-imize sensitivity to the scattered light and by observingredward of the break we concentrate on the stellar com-ponent of the host galaxy. In practice, both componentscontribute to each band, and we perform detailed anal-ysis of the stellar component and isolate the scatteredcomponent as described in Section 3.1.In this paper we present U -band-based measurementsof the scattered light, such as the opening angles of thescattering cones and their surface brightness profiles, andwe estimate the inclination angle of cone axes to the lineof sight (Section 3). Furthermore, we use the measure-ments of the surface brightness of the scattered light toestimate the density of the scattering particles (electronsor dust), which is possible because the observed surfacebrightness of scattered light is proportional to the col-umn density of scatterers along the line of sight, andwe discuss the implications of our results in Section 4.The yellow-band images are used to investigate galaxymorphology and the immediate environments of the hostgalaxies and are presented in the companion paper byWylezalek et al. (2015).The choice of instrument is driven by the availablefilters, their widths and throughput, and by the re-quirement to avoid strong narrow emission lines such as[OII] λ λλ β and H α . Thehigh throughput of blue filters on the Advanced Cam-era for Surveys (ACS; Sirianni et al. 2005) determinesthe choice of the camera, so we use ACS F475W for ob-jects with z < . λ α depending on the redshift.For each target, the ACS Wide Field Channel consistsof four exposures obtained by using the default acs-wfc-dither-box pointing pattern designed for optimal half-pixel sampling. To optimize the quality of the data prod-ucts, we reprocess the data using the AstroDrizzle taskin the software package DrizzlePac distributed throughPyRAF. As our targets are single compact sources, weset the size of the shrunk pixels (“drops”) in the drizzlealgorithm to be half of the native plate scale (0.05 (cid:48)(cid:48) ), fol-lowing the suggestion in the HST
DrizzlePac Handbook(Gonzaga et al. 2012). The adopted drizzle pattern alsofacilitates rejection of cosmic rays and detector artifacts.For the purpose of quality control, we have verified thatstatistics performed on the drizzled weight images yielda r.m.s.-to-median ratio of ∼ .
1, satisfying the < . HST
Dither Handbook (Koekemoer & et al.2002). Accordingly, the final pixels of our drizzled imagesare resampled to 0.025 (cid:48)(cid:48) . We re-reduce archival data fromGO-9905 (Zakamska et al. 2006) using the same methods. MEASURING AND MODELING GIANT SCATTEREDLIGHT NEBULAE
Host galaxy subtraction and identification ofscattering cones
We make use of the rest-frame yellow observations toremove contamination by the stellar component of thehost galaxy from the U -band. Both U -band and yellow-band images are a combination of the light from the starsin the host galaxy and of the scattered light, with pos-sibility of dust extinction affecting both of these emis-sion components. Although the yellow-band images aredominated by the stellar light, they might also contain acontribution from the scattered light of the quasar, andtherefore we cannot simply subtract a scaled version of Obied et al.the yellow images from the U -band images.Since even small offsets of a few pixels in the coordinatesystems of the yellow and U -band images may impactthe interpretation of the cleaned U -band images, we firstcarefully align the images in the two bands. We use irafimexam to measure the centroids of two or three stars inthe field of view where the U -band and yellow imagesoverlap and then use iraf imalign to shift the U -bandimages and align them with the yellow images.With astrometric adjustment in hand, we use galfit (Peng et al. 2002) to model the stellar component in theyellow-band images. We fit the two-dimensional surfacebrightness distributions in the yellow-band images withone or two Sersic components (Wylezalek et al. 2015).The galfit models are our best estimates of the dis-tribution of the stellar component, and to subtract thiscomponent from the U -band image we only need appro-priately normalize the model.To this end, we measure the U-band and yellow-bandfluxes within the same 1 × region offset from thecenter of the quasar by 1 arcsec and we use the ratio f of U -band to the yellow-band fluxes within this region toscale the galfit model derived from yellow-band imagesfor subtraction from the U -band images. By performingthis subtraction procedure, we assume in effect that thecolor of the stellar component of the host galaxy is thesame as in the chosen off-center ( ∼ galfit normalization would lead to an overestimation ofthe stellar contribution to the U -band image. We thenmultiply the model yellow images by the derived scalingfactor f and subtract them from the U -band images. Al-most no negative residuals result from this subtraction,which we will refer to as ‘optimal subtractions’ hereafter.If slight negative residuals arise from this subtraction,these host galaxies typically show morphological distur-bances, i.e. are undergoing or have currently undergonea merger, with dust lanes and tidal tails leading to thisslight oversubtraction (see e.g. SDSS J0842+3625 orSDSS J1040+4745 in Figure 2). In a successful optimalsubtraction (e.g., in SDSS J0841+2042), the scatteringcone stands out nicely after the scaled host galaxy con-tribution is removed.The optimal subtraction method is quite conservativein that it is optimized to subtract the contribution ofthe old stellar population as probed by the yellow-bandimages. Therefore, if there is a strong gradient of the stel-lar populations within the galaxy or other unaccountedfor sources of U -band light, positive residuals emerge asa result of our subtraction in the locations in the hostwhere stellar populations are younger. A particularlydifficult host galaxy subtraction example is SDSS J0149-0048 (Figure 2). Even after the scaled galfit modelis subtracted from the U -band image, a smooth roundcomponent remains in the image. A similar componentis seen in IRAS 09104+4109 by Hines et al. (1999).Aside from the morphology (featureless and centeredon the quasar) and the color (bluer than the outskirtsof galaxies used for galfit normalization), we know lit-tle about this component. One possibility is that it is due to circumnuclear star formation, as suggested byHines et al. (1999). In SDSS J1039+4512, where theoptimal subtraction leaves a strong smooth component,the infrared-based measurement of star formation rate is ∼
40 M (cid:12) yr − (Wylezalek et al. 2015), so it may wellhost a significant unobscured young stellar populationfor which the optimal subtraction method is not able toaccount. An alternative possibility for the origin of theunsubtracted blue component – quasar light percolatingthrough patchy obscuration and then scattered off theinterstellar medium – is discussed in Section 4.1.In some cases even with subtraction of the scaled galfit model it is challenging to identify the scatteredlight component. Therefore, we also perform an ‘ex-treme’ subtraction, where we subtract 6 × f × yellow-band model image from the U -band image. Thesetypically yield strong negative residuals (because muchof the stellar component of the host galaxy has nowbeen over-subtracted), but often make the scatteringcones stand out clearly, like in SDSS J0149-0048 orSDSS J1039+4512 (Figure 2). For measurements of thescattered light geometry described in subsequent sectionswe use both subtractions, which provides us with an es-timate of the systematic uncertainties involved.The total flux in the U -band and the flux after the ‘op-timal’ stellar component subtraction are listed in Table1. We find that the the median correction for the hostgalaxy contribution to the U -band is only 27% and thattherefore most of the flux (73% ± U -band images is due to the scat-tered light from the quasar. If not properly accountedfor, this component can strongly bias the star formationrates of quasar host galaxies as measured from the U -band images. Scattering efficiency
We define scattering efficiency as the fraction of theintrinsically emitted quasar radiation at ∼ U -band emission pro-duced by the quasar is not known, so we estimate itusing the rest-frame 12 µ m monochromatic luminosity νL ν [12 µ m] as measured from WISE data. We assumethat although the observed 3000˚A flux is suppressed byextinction, νL ν [12 µ m] is unaffected by it, so we use thelatter and the average quasar spectral energy distribu-tion from Richards et al. (2006) to estimate the intrinsic3000˚A luminosity. The scattering efficiency is then theratio between the actual 3000˚A luminosity of our ob-jects (k-corrected as necessary using F ν ∝ ν − . , Van-den Berk et al. 2001) to the estimated intrinsic value.The median scattering efficiency in our sample is 2.3%,with a standard deviation of 0.4 dex (Figure 1, right). If12 µ m luminosities are in fact affected by extinction, assuggested by the red mid-infrared colors of the quasars inour sample (Liu et al. 2013b), then the observed 12 µ mluminosities underestimate the true luminosities of ourquasars and the measured scattering efficiencies serve asupper limits on the actual values. The scattering effi-ciency can be related to the geometry of the scatteringregions and to the total amount of scattering mass; fur-ther discussion of scattering efficiency and the implica-tions of the 2.3% average value is presented in Sectioniant scattering cones 5 a b c d SDSS J0149-0048 SDSS J0759+1339SDSS J0210-1001 SDSS J0841+2042SDSS J0319-0019 SDSS J0842+3625SDSS J0319-0058 SDSS J0858+4417SDSS J0321+0016 SDSS J1039+4512SDSS J1040+4745
Fig. 2.—
In each strip, we show (from left to right) 4 . × . postage stamps of the (a) yellow-band HST images, (b) U -band HST images, (c) host galaxy subtracted U -band images and (d) extreme host galaxy subtracted U -band images. In (c), we subtract thescaled galfit fit to the yellow-band image from the blue image to remove the extended stellar component, however a centrally concentratedquasi-spherical component still remains. In (d), we show extreme subtractions in which we over-subtract the stellar component to bringout the scattered light detection. Depending on redshift, the sizes of these images correspond to physical sizes of 25 to 30 kpc on the side.The magenta lines indicate the orientation of the forward-facing scattered light cone, and the projected angle (as shown) is used in ourmodeling as an upper limit on the deprojected (intrinsic) opening angle of the cone. Estimates of extents, projected opening angles andinclination angles
We identify scattered light regions primarily by mor-phology in the ‘optimal’ and ‘extreme’ host-subtracted U -band images. We look for conically shaped featureswith apex coinciding with the center of the galaxy. Iden-tification is aided significantly by the polarimetric ob-servations available for approximately half of the sample (Zakamska et al. 2005, 2006): the polarization positionangle of scattered light is expected to be orthogonal tothe position angle of the major axis of the scatteringcone in the plane of the sky, and indeed this relationshipis borne out in our previous observations. To identify theprojected axis of the cone, we look at the surface bright-ness distribution along annular sections taken around thegalaxy center and find the peak surface brightness whichcorresponds to the ‘spine’ (thickest part) of the cone.Because of the forward-scattering nature of dust (Section Obied et al.4.3), if the cone axis is not exactly in the plane of the skythe observer-facing cone appears brighter and we may ormay not be able to see the backward-facing cone. Wefind that in ∼
14 objects the second peak correspondingto the backward-facing counter-cone is also visible.After identifying the directions of the cone, we tracethe extents of the cones down to a limiting surface bright-ness λ (cid:48) I λ (cid:48) (cid:39) × − erg sec − cm − arcsec − at ob-served wavelength λ (cid:48) . These maximal projected extentsare listed in Table 2. In the majority of objects the conesare detected out to at least 5 kpc from the nucleus, andin a few out to (cid:38)
10 kpc.In our subsequent calculations we assume that theapex of the scattering cone coincides with the hid-den quasar, which in turn we assume is at the ex-act center of the stellar component of the galaxy de-rived from galfit yellow-band fits. We made < . ∼ . o , standard deviation in the sample 8 o )by evaluating the lateral extent of the cone in compari-son to the distance from the quasar at which the lateralslice is taken. Because the cone axis is not necessarilyin the plane of the sky, the opening angles as seen inprojection on the plane of the sky are larger than the in-trinsic ones. In the next Section, we present full conicalscattering models which can be used to deproject thesevalues and calculate the actual opening angle of the scat-tering cones. The most uncertain scattering cone identi-fications are in SDSS J0210-1001, SDSS J0319-0019 andSDSSJ 0858+4417. As can be seen in Figure 2, theseare also the most compact. SDSS J0210-1001 has a lu-minous unsubtracted U -band component, and the othertwo objects lack the characteristic triangular shape evenin the ‘extreme’ subtractions.Dust is strongly forward-scattering (Draine 2003).Thus if an intrinsically symmetric bicone is not lying ex-actly in the plane of the sky, the cone pointed toward theobserver would appear brighter than the other one, andthe brightness ratio of the two can yield the inclinationangle relative to the line of sight i . We detect unambigu-ous counter-cones in ∼ ∼ ∼ o . Full three-dimensional model
We model the scattering region as a cone of half-angle θ inclined at an angle i from the line of sight (Figure 4),filled uniformly with scatterers (electrons or dust par-ticles) whose number density declines as a function ofdistance from the radiation center r , n ( r ) = n (cid:16) r r (cid:17) m , (1)and the normalization is given as n at some fixed dis-tance r (set equal to 1 kpc in our calculations) from thecenter. The observed surface brightness is then obtainedby integrating the radiation scattered into our directionby every volume element of the cone along the line ofsight ζ : I λ (cid:48) = (206265) (1 + z ) (cid:90) L λ πr n ( r ) d σ dΩ d ζ, (2)where the integral ranges from the back wall of the scat-tering cone to its front wall. Here L λ is the intrinsic lu-minosity density of the quasar at rest-frame wavelength λ (in units of erg s − ˚A − ) and I λ (cid:48) is the measured sur-face brightness (in units of erg s − cm − arcsec − ˚A − )at the observed wavelength λ (cid:48) = (1 + z ) λ . Other pa-rameters include r – the physical distance between thescattering volume element and the radiation center, andd σ/ dΩ – the differential cross section of scattering incm /sr. The factor (1 + z ) reflects cosmological surfacebrightness dimming of the luminosity density, and thefactor (206265) converts steradians into arcsec . Thefirst crude scattering model for SDSS J1039+6430 waspresented by Greene et al. (2011) to estimate the den-sity of scattering gas. Here we take into account the fullgeometry of scattering and appropriate scattering cross-sections for the first time.In a partly or fully ionized medium with the stan-dard gas-to-dust ratio, dust scattering dominates overelectron scattering by a factor of (cid:29)
10 at ultra-violetand optical wavelengths (Weingartner & Draine 2001;Draine 2003). Therefore, in our models we focus ondust scattering, and we present an extensive discussionof the scattering mechanism in Section 4.3. We use theSmall Magellanic Cloud dust with reddening parameter R V ≡ A V /E ( B − V ) = 2 .
87 whose scattering cross sec-tions and phase functions are given by Weingartner &Draine (2001) and Draine (2003): (cid:18) dσd Ω (cid:19) dust = C sca × p ( α ) . (3)Here α is the scattering angle between the observer’s lineof sight ( ζ -axis) and the initial radiation direction fromthe quasar accretion disk to the scattering event (Figure4) and p ( α ) is the phase function (angular dependence) ofscattering (Draine 2003). The wavelength dependence oftotal scattering cross-section C sca for various dust com-positions is tabulated by Weingartner & Draine (2001),and we take values appropriate for the rest-frame of ourobservations. Weingartner & Draine (2001) and Draine(2003) present scattering cross-sections normalized periant scattering cones 7 Distance (kpc) D i s t a n ce ( k p c ) SDSS J103951.49+643004.2 − − −
50 0 50 100 150
Angle (degrees) N o r m a li ze d B r i g h t n e ss Cone ModelDisk Model
Distance (kpc) D i s t a n ce ( k p c ) SDSSJ0841+2042 − − −
50 0 50 100 150
Angle (degrees) N o r m a li ze d B r i g h t n e ss Cone ModelDisk Model
Fig. 3.—
Surface brightness profiles along a circular aperture in SDSS J1039+6430 and in SDSS J0841+2042. Left: optimally-subtracted U -band images with the location of the circular aperture (blue circle) and the direction of the spine of the scattering cone (purple line).Right: the observed surface brightness profiles (shown with crosses) are fitted with cone models (dotted lines) and disk models (dashedlines) discussed in Section 4.2, both convolved with the Airy function to represent the blurring by the point-spread function. Disk modelsshow a more abrupt cut-off in surface brightness than is observed, and cone models provide a better fit to the data. Angles are countedcounter-clockwise from the positive horizontal direction; counter cones are detected at ∼ o in SDSS J1039+6430 and possibly from ∼ − o to ∼ − o in SDSS J0841+2042 (the morphology of this component is ambiguous and it could be due to under-subtracted starformation in the nucleus of the galaxy). Fig. 4.—
Parameters of the scattering cone model. θ is half ofthe opening angle of the cone, i is its inclination relative to the lineof sight ζ , and ξ − η is the plane of the sky. The scattering angle α determines the phase function of scattering. hydrogen nucleus, so instead of the density of scatteringdust particles we derive the density of hydrogen nuclei at1 kpc from the quasar n H , under the assumption thatthe gas-to-dust ratio in our objects is similar to that ofthe Small Magellanic Cloud. We use the Airy disk with a full width at half max-imum of 0.1 arcsec as a point-spread function. This iscomparable to the width of the numerically calculatedACS point-spread function obtained from TinyTim . Atwo-dimensional map of the model surface brightnessis convolved with the Airy function before taking sec-tions to obtain the model one-dimensional brightnessprofiles. These are then the models we fit to the ra-dial and lateral brightness profiles taken from the data.As the first approximation, we are interested in esti-mating the typical densities of the interstellar mediumand the radial density distributions in our galaxies. Wetherefore decided against a full two-dimensional imagefitting procedure because in many objects the scatteredlight regions are very clumpy (e.g., SDSS J0321+0016,SDSS J1040+4745). The use of one-dimensional radialand lateral brightness profiles gives us the freedom toavoid such irregularities by, for instance, taking sectionsthat do not pass through regions of unusually low/highdensity scattering medium. As we discuss below, thereare other important systematic uncertainties in our de-rived densities, so they should be considered only asorder-of-magnitude estimates.We use the lateral and the radial surface brightnessprofiles to simultaneously fit eight parameters. The fourphysically meaningful parameters are density normaliza-tion n H , , half-opening angle of the cone θ , inclination an-gle of the cone axis from the line of sight i , and half-slopeof the scatterers’ density profile m . These parameterscompletely determine the geometry of the cone and thedensity profile and are listed in Table 2. Two more pa-rameters are background surface brightness values for the Obied et al.lateral and radial profiles and the remaining two param-eters are the centroids of the surface brightness peaks inthe one-dimensional lateral and radial profiles. In the ra-dial profiles, our power-law analytical approximation forthe density breaks down near the nucleus of the galaxywhere circumnuclear obscuration can suppress scatteredlight, so we mask the nuclear points from the fits. Insome objects we see clumping of the scattering gas; thestrongest clumps are masked from the fit as well, whichleads to an underestimate of the density of the scatteringmedium in these cases.This fit is performed subject to several constraints.First, the opening angle of the cone must be smaller thanor equal to the projected opening angle measured in theprevious section (shown in magenta in Figure 2). Sec-ond, the inclination angle is greater than or equal to theinclination angle derived in the previous section. Therationale for using the measured inclination angle as thelower limit rather than as an actual measurement is thatany dust within the galaxy would obscure the backward-facing cone more strongly, though this is a small ef-fect: the inclination angles from model fits and from thebrightness ratios agree to within 10 degrees in all but 3objects. Third, we require that the inclination angle begreater than half of the opening angle so that we see theobject as a type 2 quasar, i.e., our line of sight cannotpass within the cone. Without constraints on the angles,and especially the inclination angle from the brightnessratios of the cone to counter-cone, the inclination angleand the opening angle are quite degenerate with one an-other: a narrower cone pointing close to the line of sighthas observed surface brightness profile similar to a widercone closer to the plane of the sky.Fitting is done using the python scipy.optimize package and example fits are shown in Figure 5. Oneof the sources of systematic uncertainties in our fittingprocedure is due to the uncertainties in subtraction of thehost galaxy. While in Table 2 we report the results of fit-ting the ‘optimally’ subtracted images, we carry out thefitting procedure for the ‘extreme’ subtractions as well.The fitted densities agree between the two subtractionswithin 0.25 dex (standard deviation), the fitted openingangles agree within 8 o , the inclination angles within 8 o and the half-slopes m within 0.2. There are no signifi-cant systematic offsets between the fitted parameters inthe two subtractions.It is clear from equation (2) that the apparent surfacebrightness constrains the product of n H , L λ and we needan independent estimate of the intrinsic luminosity L λ toconstrain n H , . Our default method is to use the [OIII]luminosity to derive the 2500˚A luminosity density fromthe empirical relationship known for type 1 quasars andpresented by Reyes et al. (2008), which we then adjustto the effective rest wavelength of our observations usingthe average quasar spectrum L ν ∝ ν − . (Vanden Berket al. 2001). The density normalizations obtained usingthis method are shown in Table 2 and in Figure 6.An alternative method is to start from 12 µ m luminosi-ties (Table 1) and augment them to total infrared lumi-nosities using average unobscured quasar spectral energydistributions from Richards et al. (2006). The specificmultiplicative factor that we use is 3.4. Then we as-sume that the infrared luminosity is due to re-radiation of the optical luminosity intercepted by the obscuringmaterial whose covering fraction is cos θ , so that L opt = L IR / cos θ . We then use again spectral energy distribu-tions from Richards et al. (2006) to connect the total op-tical luminosity to the monochromatic luminosity at therest-frame of our observations ( νL ν [3000˚A] (cid:39) L opt / . ).This gives us another set of density normalizations, alsoshown in Table 2. While the two sets of densities arewell correlated with one another, the second set is signif-icantly higher, by a factor of (cid:38)
10. We suspect that theintrinsic luminosities are significantly underestimated inthe second method; among other factors, cos θ signifi-cantly overestimates the obscuration covering fraction asdiscussed in Section 4.1.For this reason we somewhat prefer the first set of den-sities, but in Sec. 4.4 we use both sets of values to bracketthe likely range of scattering gas masses. The comparisonbetween the two sets clearly demonstrates the severity ofsystematic uncertainties in our density calculation, so weassume that our derived densities are known to no betterthan an order of magnitude. RESULTS OF THE MODELING SCATTERED LIGHT
Opening angles of scattering regions
The mere detection of scattering cones constitutesproof of one of the key components of the classical uni-fication model of active galactic nuclei (Antonucci &Miller 1985; Antonucci 1993): the objects in our sam-ple would be seen as normal unobscured quasars if wewere positioned along the lines of sight within the cones.This is demonstrated by the color of the scattered light(blue, reflecting the incident quasar spectrum) and bythe spectropolarimetric observations which indicate thatscattered light shows the classical broad lines character-istic of unobscured quasars (Zakamska et al. 2005).The average (median) and the standard deviation ofthe half-opening angles in our sample is θ = 27 o (27 o ) ± o .In principle, the opening angle is related in a straight-forward manner to the ratio of type 1 to type 2 quasarswithin the quasar population: in the case of axisymmet-ric toroidal obscuration, the probability of a line of sightto the observer to be located within the opening of thecone (and therefore for the object to be seen as a type1 quasar) is P type1 = 1 − cos θ , while the probability tosee the object as a type 2 quasar is P type2 = cos θ . Forthe observed average opening angle, the implied type 1fraction in the population is 11%. Although the type 1/ type 2 ratio remains somewhat controversial, this mea-surement is inconsistent with the typical type 1 / type 2ratio measured from quasar demographics ( ∼ .
0, Reyeset al. 2008; Lawrence & Elvis 2010) and that measuredfrom the infrared-to-optical ratios ( ∼ .
0, Treister et al.2008).One reason for the low calculated fraction of type 1quasars in the population is that a type 2 sample selectedfor follow-up observations would naturally be biased to-ward lower opening angles, because the probability ofselecting a type 2 quasar with a half-opening angle of θ increases as θ declines, so such objects would be over-represented in our sample. We construct simple modelsof this bias that take into account the observed θ distri-bution and calculate a bias-corrected type 1 fraction inthe population to be ∼ Distance/kpc D i s t a n ce / k p c SDSS J1039+4512
Distance/kpc D i s t a n ce / k p c Distance/kpc N o r m li ze d B r i g h t n e ss Distance/kpc N o r m a li ze d B r i g h t n e ss Distance/kpc D i s t a n ce / k p c SDSS J0149-0048
Distance/kpc D i s t a n ce / k p c Distance/kpc N o r m li ze d B r i g h t n e ss Distance/kpc . . . . . N o r m a li ze d B r i g h t n e ss Distance/kpc D i s t a n ce / k p c SDSS J0842+3625
Distance/kpc D i s t a n ce / k p c Distance/kpc − N o r m li ze d B r i g h t n e ss Distance/kpc − . − . . . N o r m a li ze d B r i g h t n e ss Distance/kpc D i s t a n ce / k p c SDSS J0319-0058
Distance/kpc D i s t a n ce / k p c Distance/kpc N o r m li ze d B r i g h t n e ss Distance/kpc N o r m a li ze d B r i g h t n e ss Distance/kpc D i s t a n ce / k p c SDSS J0321+0016
Distance/kpc D i s t a n ce / k p c Distance/kpc − N o r m li ze d B r i g h t n e ss Distance/kpc − N o r m a li ze d B r i g h t n e ss Fig. 5.—
Examples of fitting the cone-scattering model to lateral and radial surface brightness profiles: two good-quality fits on top andthree mediocre fits on the bottom. In the two left columns, we show the ‘extreme’ subtraction of the stellar component and the ‘optimalsubtraction’ of the stellar component. The blue arrow originates at the assumed center and points along the cone spine and marks theextraction direction of the radial profile. The orthogonal line marks the location of the lateral profile. In the third column we show theobserved radial profile (crosses) and the model fit (dotted line) and in the fourth column we show the same for the lateral profile. Filledpoints indicate data masked during the fitting process.
Fig. 6.—
Distribution of fitted density parameters: half of the power-law slope m (left panel) and the normalization of hydrogen densityat 1 kpc (right panel). We show the densities estimated using [OIII]-based bolometric luminosity n H , with the solid histogram and thedensities estimated from the infrared luminosities n ∗ H , with the dashed lines histogram. The means (medians) and standard deviations ofthe distributions are shown at the top. iant scattering cones 11by comparison with the expected ∼ −
70% value. Asan extreme version of the bias, a dust-free quasar wouldnot enter into a type 2-selected sample at all, and ourbias correction procedure would not be able to accountfor these objects, but there are only 10% of dust-poorquasars among optically-selected type 1 nuclei (Hao et al.2011), so they do not fully resolve the discrepancy either.To produce at 1:1 ratio of type 1 to type 2 quasars,the average half-angle of scattering cones would need tobe 60 o . We do not have a single object with θ above40 o . (The potential counter-cone in SDSS J0841+2042shown in Figure 3 comes close, but it is unclear whetherthe observed feature is in fact a counter-cone or anunder-subtracted blue stellar component.) One intrigu-ing possibility is that the toroidal obscuration is patchyor porous, to the extent that an observer has a reason-able chance (30 − U -band light.It is becoming increasingly clear that quasar obscura-tion is clumpy (Nenkova et al. 2002, 2008; Schartmannet al. 2008; Zakamska et al. 2008; Nikutta et al. 2009; Deoet al. 2011; Markowitz et al. 2014), and the probabilityof observing a source through obscuration as a type 1 ob-ject is determined by the covering factor of these clumps.The sizes and numbers of clumps might ultimately beconstrained by spectral fitting of mid-infrared emissionor by modeling of the ‘changing-look’ active nuclei (whichare interpreted as being temporarily blocked by a singlecloud). It will be interesting to see whether statistics ofclumps from these observations are in agreement withour scattered light data. Illuminated disk vs filled cone
In our modeling of the scattered light regions we haveassumed a cone filled with scattering particles – or rather,we assume that the entire galaxy is filled with scatterersand the cone is produced where they happen to be illumi-nated by the quasar. Another possibility for explainingthe observed triangular morphology of scattered light isthat there is a large-scale gas disk (e.g., a disk compo-nent of the quasar’s host galaxy) which is illuminatedby the quasar, and the circumnuclear obscuration deter-mines the pattern of illumination. Such component isseen in a nearby active galaxy (Lena et al. 2015) whereit is distinct from the outflowing gas component. It isimportant for us to make a distinction for volume-filling,potentially dynamically disturbed gas (possibly in an or-ganized quasar-driven outflow from the galaxy) and apassively illuminated rotating galaxy disk.Most of our objects are elliptical galaxies (Wylezaleket al. 2015) without signs of large-scale galactic disks,so producing the observed scattered light components inpassively illuminated disks is unlikely because we do notsee a corresponding stellar component. In one of the few objects with disks (SDSS J0149 − The nature of the scattering medium.
So far we have assumed that dust scattering dom-inates significantly over free electron scattering. Al-though the profiles of the scattered light regions can beequally well modeled assuming either electrons or dust,we have adopted dust scattering as our primary assump-tion throughout this paper, for reasons we discuss in thissection.Electron scattering is characterized by the classicalThompson cross section and phase function: (cid:18) d σ dΩ (cid:19) el = (cid:18) e m e c (cid:19) α . (4)Dust scattering is more complex because in a typi-cal astrophysical medium dust particles of many sizesare present. Weingartner & Draine (2001) and Draine(2003) use multi-wavelength observations of absorption2 Obied et al.and emission of dust in the Milky Way and in the Mag-ellanic Clouds to derive dust size distributions and tocalculate the resulting cross-section and phase functionof dust scattering in these objects, normalized to the hy-drogen mass (equation 3).For a normal dust-to-gas ratio typical of the MilkyWay or the Magellanic Clouds, dust scattering is muchmore efficient than electron scattering even when hydro-gen is fully ionized. Taking for example the scatteringcross-sections of Small Magellanic Cloud dust at 3000˚A(Weingartner & Draine 2001), we find that at scatteringangles of α = 20 ◦ , ◦ , ◦ , ◦ , the minimal ratio of dust-to-electron cross section (when all hydrogen is ionized) is d σ dust dΩ / d σ el dΩ = 220 , , ,
42. This ratio is even larger formore dust-rich galaxies like the Large Magellanic Cloudor the Milky Way whose dust-to-gas ratios are ∼ ∼
10 times higher than that of the Small MagellanicCloud, respectively (Roman-Duval et al. 2014).Therefore, dust scattering should dominate over elec-tron scattering unless dust is efficiently destroyed bythe radiation of the quasar or by collisions with ther-mal electrons (sputtering). Radiative evaporation isunlikely outside of the dust sublimation radius r ∼ . L bol / erg s − ) . pc (Barvainis 1987), and dustsputtering is inefficient at the typical temperatures ofthe narrow-line regions T ∼ , (cid:29) K (Zubovas & King 2012;Faucher-Gigu`ere & Quataert 2012; Nims et al. 2015),at which dust sputtering is effective (Draine & Salpeter1979) and is known to occur in hot atmospheres of galaxyclusters (McGee & Balogh 2010). Thus dust could bedestroyed in the bulk of the volume of the host galaxy,although most of dust mass should still survive in thedenser phases of the interstellar medium (warm ionizedand cold neutral clouds). Indeed, despite shock signa-tures in quasar narrow emission line regions (Zakam-ska & Greene 2014), line ratios are remarkably uniformand similar to those seen in low-luminosity active galax-ies, suggesting that they are dusty (Dopita et al. 2002;Groves et al. 2004). An additional complication is thatthe narrow-line region clouds are optically thick to ultra-violet radiation, and thus the bulk of their mass does notcontribute to either dust or electron scattering. Thus thequestion of whether electron scattering might dominateboils down to whether the average dust-to-gas ratio ofthe gas exposed to the ultra-violet emission of quasars islowered by two orders of magnitude due to sputtering.While such calculation is outside the scope of the pa-per, we have several indirect arguments in favor of dust scattering from the
HST and polarimetric observations.Electron scattering is wavelength independent, while wesee some wavelength variations of scattering efficiency inthe polarized spectrum of SDSS J1039+6430 (Zakamskaet al. 2005). Furthermore, the number of free electrons inthis source is strongly constrained by the observed fluxof the recombination lines and is insufficient to accountfor the observed scattering efficiency (Zakamska et al.2005). Finally, scattering by electrons in the hot low-density phase of the interstellar medium is ruled out bythe lack of resulting kinematic broadening in the scat-tered spectrum (Zakamska et al. 2005). These are allstrong arguments in favor of dust scattering, but they re-quire very high quality spectropolarimetric observationsavailable for only three of the sources discussed here,SDSS J1039+6430, SDSS J0842+3625 (Zakamska et al.2005) and IRAS 09104+4109 (Hines & Wills 1993; Hineset al. 1999).A larger number of objects in our sample present an-other signature of dust scattering – brightness contrastbetween two scattering cones. Here we have to assumean axisymmetric structure for the circumnuclear obscu-ration, which would naturally result in two centrally sym-metric scattered light bicones. If scattering is dominatedby electrons, then regardless of the inclination angle ofthe main axis to the line of sight the two cones are ex-pected to have the same brightness. Indeed, the conepointed toward the observer scatters at sharp angles α ,and the cone pointed away from the observer scatters atobtuse angles 180 o − α , but the phase function of electronscattering (eq. 4) is the same for these scattering direc-tions. In contrast, dust is strongly forward-scattering(Draine 2003), and the cone pointed toward the observeris expected to be brighter.Out of the 21 objects in our combined sample, we find ∼ ∼ ∼ (cid:38)
2. This strongly suggests that dust scatteringdominates. The only alternative is electron scatteringcombined with dust extinction of the backward-pointingcone within the galaxy, but that requires so much dust( N H = 8 × cm − and the Small Magellanic Clouddust-to-gas ratio to get the median brightness ratio of ∼ .
5) that again dust scattering would dominate overthe electron scattering in such a galaxy.The assumption of dust scattering (as opposed toelectron scattering) has critical implications for our de-rived values of the interstellar medium density. Becausedust scattering is more efficient (by about two ordersof magnitude), a smaller amount of interstellar mediumis required to account for a given surface brightness ofscattered light under the assumption of dust scatteringthan what we would calculate assuming electron scat-tering. We specifically choose Small Magellanic Clouddust because Hopkins et al. (2004) suggest that it is agood fit to the observed extinction in reddened quasars.Choosing another dust scattering curve would not no-ticeably affect the quality of our fits, but would affectour derived density normalization. For other types ofdust like in the Large Magellanic Cloud or the MilkyWay, the dust-to-gas ratio is a factor of 3 −
10 timeshigher than for the Small Magellanic Cloud, so usingthese curves to fit the observed scattered surface bright-iant scattering cones 13ness we would derive densities smaller by that factor.Self-absorption within the scattering cones is negligible;the optical depth to dust extinction within the cones is τ ∼ n H , ζC ext = 0 .
01 for our median n H , (Figure 6), ζ = 1 kpc and C ext = 1 . × − cm .In principle, polarimetric measurements can help dis-tinguish between electron and dust scattering becausepolarization due to electron scattering can reach 100%(when light is scattered at α = 90 o ), whereas the theo-retical limit for dust-induced polarization is much lower( ∼
20% at α ∼ o at 3000˚A, Draine 2003). In practice,many factors act to lower the observed polarization frac-tion to typical values of a few per cent. Deviations ofthe inclination of the scattering cone away from the op-timal directions lead to steep decline of the polarizationfraction, wide opening angles of scattering region lead togeometric cancellation of the polarized signal (a centrallysymmetric scattering nebula has zero net polarization),and most importantly the unpolarized light from the hostgalaxy dilutes the weak polarized signal.The two objects with the highest observed levels ofpolarization in our sample are SDSS J1039+6430 andSDSS J0842+3625, both with P = 16 .
5% at 3000˚A andlikely negligible host galaxy dilution at this wavelength(Zakamska et al. 2005). Taking the geometric param-eters of our best-fit scattering cones in these two ob-jects, we derive model polarization fractions of ∼ ∼ ∼ o , whichsuppresses polarization, and the resulting model polar-ization is inconsistent with the observed value. Whileelectron scattering could easily bring the model valueup to the levels consistent with observations, this ob-ject presents the strongest multi-wavelength case for dustscattering in our sample (Zakamska et al. 2005) and itsinclination angle is well-determined from the detectionof the counter cone (Figure 3). The only way to re-solve this inconsistency is to postulate that the dust sizedistribution in this object is not well described by thedistributions considered by Draine (2003) for the MilkyWay and the Magellanic Clouds which all produce < Implications of the density distribution estimates
The 11 objects with new
HST observations were stud-ied using Gemini integral field spectroscopy by Liu et al.(2013a,b) who uncovered galaxy-wide ionized gas out-flows in these objects. If the gas detected within thescattering regions is outflowing with the typical velocitiesseen in the spectroscopic observations, then we can calcu-late the mass outflow rate: ˙ M ( r ) = Ω r n H ( r ) m H v ( r ) /X .Here n H ( r ) is the density of the interstellar medium thatwe obtain from our scattered light measurements, m H is the mass of a hydrogen atom, X (cid:39) . v ( r ) canbe measured in integral-field unit observations of quasar-driven winds, and in observations to date it appears al-most constant as a function of distance (Liu et al. 2013b; Harrison et al. 2014), with typical magnitude v (cid:39) − .While we have no direct measurement of the velocityof the gas which is responsible for the scattered lightseen in the HST images, we have some evidence that thescattering medium is co-spatial with the kinematicallyidentified wind. Wylezalek et al. (2015) compared themorphologies of the U -band images, yellow-band imagesand the kinematic maps from Liu et al. (2013b) for thesame 11 objects analyzed here. They find that the scat-tering cones are aligned with the direction of the velocitygradient, as expected if both the scattering cones and thephoto-ionization pattern of the emission-line gas are de-termined by the same illumination geometry.If only the illuminated gas is outflowing, then our esti-mates of the opening angles suggest Ω (cid:39) .
1, whereas the1:1 obscured-to-unobscured ratio found in quasar demo-graphic studies (Lawrence & Elvis 2010) suggests Ω (cid:39) π – half of the sky, as seen from the quasar (this interest-ing discrepancy is discussed in Sec. 4.1). However, thecircumnuclear material does not necessarily collimate theoutflow toward the unobscured directions. (Wagner et al.2013) show that dense circumnuclear clouds disperse anddeflect AGN outflows. The clouds receive most of the mo-mentum of the outflowing gas which propagates throughthe low-density channels of least resistance between thedense clouds. As a result of this interaction, any direc-tionality of the original wind vanishes on larger scales,with the outflow proceeding in all directions and curvingaround dense clouds and larger obstacles. In this case,the outflow would have a covering solid angle of Ω (cid:39) π ,but only part of it would be illuminated by the quasarand be detectable as scattered light, resulting in a bi-conical appearance.The estimate for n H ( r ) is a direct result of the obser-vations presented in this paper. From our scattered lightfits, we find that the median density at r = 1 kpc fromthe quasar is n H , (cid:39) .
04 cm − and that the medianpower-law profile of the density distribution is close to r − . Assuming volume-filling geometry for the outflowsuggested by the scattered light observations, the esti-mated density slopes imply that ˙ M ( r ) is nearly constantas a function r , so that the outflowing mass does notconcentrate at any one distance, which is consistent witha steady-state process of gas removal. Supplying now allthe estimates into the equation for mass outflow rate, wefind˙ M (cid:39) . M (cid:12) year × (cid:18) Ω4 π (cid:19) (cid:16) n H , . − (cid:17) (cid:16) v − (cid:17) . (5)These outflow rates are much smaller than those derivedfrom the ionized gas observations (Liu et al. 2013b) –˙ M ionized (cid:39) M (cid:12) yr − .The outflow mass rate is an important value for un-derstanding what role quasar-driven winds might play inthe evolution of their host galaxies, and unfortunatelyhas proven to be difficult to obtain. The mass measure-ment from Liu et al. (2013b) is subject to many possibleuncertainties, such as the electron density in the warmionized clouds at 7 kpc from the quasar which is uncon-strained from the current ionized gas observations. Inthe ˙ M ionized estimate above, we have assumed radiationpressure confinement for the narrow-line gas in agree-4 Obied et al.ment with Dopita et al. (2002) and Stern et al. (2015).The chief reason that the ionized gas outflow rate de-rived by Liu et al. (2013b) is so high is that it is mea-sured at the location where the ionized gas clouds tran-sition from being optically-thick to ionizing radiation tobeing optically-thin (from ionization-bounded to matter-bounded), which was deduced from the observations ofLiu et al. (2013a) of the increase in the HeII λ β ratio at this distance. In other words, over a small rangeof distances from the quasar (at ∼ r (cid:38) r (cid:46) ∼ .
5% at 1 kpc from the quasar, as suggested byeq. 5) participates in the scattering.Can we construct a crude model of a clumpy narrow-line region which would be consistent with both the newscattered light observations and the previous emission-line observations (Liu et al. 2013a,b)? Assuming thatclouds are radiation-pressure confined (Dopita et al.2002; Stern et al. 2015) and that they are propagatingoutward in a mass-conserving fashion, we find that theircolumn density depends on the distance from the quasaras N H ∝ r − / . Clouds transition from being ionization-bounded to being matter-bounded at hydrogen columndensity N H = 10 cm − (Stern et al. 2015). If thistransition happens at 7 kpc (Liu et al. 2013b), then at1 kpc from the quasar – the typical distance probed byour scattered light observations – the column density ofnarrow-line clouds is N H ∼ × cm − . Ultra-violetphotons can only penetrate through and escape from anarrow layer on the surface with optical depth τ , withcolumn density N H , scattered (cid:39) . × ( τ / .
2) cm − ,where we have used the appropriate extinction cross-section from the Small Magellanic Cloud dust opacitycurve (Weingartner & Draine 2001). This suggests thatat 1 kpc 10 −
15% of the mass of the narrow-line cloudsshould be visible in scattered light observations.Therefore, clumping of the narrow-line region explainssome (though not all) of the tension between the massestimates produced by the two different methods. Pos-sibly, the densities n H, derived from scattered light ob-servations are too low. As we discussed in Sec. 3.4,the derived densities are degenerate with the assumedintrinsic luminosity at 3000˚A. The upper limits on thedensities n ∗ H, derived from infrared luminosities are anorder of magnitude higher; using these values would re-sult in an estimated mass outflow rate of 200 M (cid:12) yr − ,which in combination with clumping would be consistentwith ionized gas observations. Another source of uncer-tainty in our estimates is the assumption of the single matter-bounded transition boundary. It is much morelikely that a wide range of cloud sizes is present in thenarrow-line region and they transition into the matter-bounded regime at different distances. We are currentlydeveloping such models and are aiming to include themin our future analyses of quasar-driven winds.Furthermore, most of the outflow mass could be in theform of neutral or even molecular gas (Morganti et al.2005; Feruglio et al. 2010; Veilleux et al. 2013; Ciconeet al. 2014; Sun et al. 2014). This phase would be in-visible to optical emission line observations and only thethin outer layers of such clouds would participate in scat-tering because the molecular gas presumably would beconcentrated in very dense individually optically thickclouds. Neither ionized gas observations nor scatteredlight observations would capture this component of theoutflows.We can now use the derived density profiles n ( r ) ∝ r − to relate the scattering efficiency ε = (cid:82) d V (d σ/ dΩ) n ( r ) /r (Zakamska et al. 2005) to the totalmass of gas in the galaxy out to r max from the center: ε = (cid:28) d σ dΩ (cid:29) P type1 M H m H r max r min . (6)Scattering is dominated by small distances from thequasar ( r min ) where the quasar radiation is least di-luted. To make an estimate of the total hydrogen mass,we adopt d σ/ dΩ = 3 × − cm /sr, r max = 10 kpc, r min = 0 . P type1 = 0 . M H = 10 M (cid:12) .This is the lower limit on the actual gas mass since muchof the mass might be invisible in scattered light as dis-cussed above. Furthermore, this gas is not concentratedin a disk but rather is distributed through the volume ofthe host galaxies. The combination of moderately highgas mass and the spatial distribution of this gas makesthe host galaxies of luminous obscured quasars highlyunusual among early-type galaxies (Serra et al. 2012),supporting the hypothesis that the gas we see in scat-tered light is associated with quasar outflows also seenin kinematic measurements (Wylezalek et al. 2015).To summarize, the picture we propose on the basisof the scattered light observations involves many smallclouds filling the host galaxy. When the clouds are il-luminated by the quasar, they are partly photo-ionizedand produce the observed optical emission lines, whiletheir surfaces contribute to the observed scattered light.An alternative geometry often discussed in the contextof galactic winds is that of overpressured bubbles (MacLow et al. 1989; Greene et al. 2012; Harrison et al. 2015)which expand into the interstellar medium and plow ashell of gas off to the sides and into the intergalacticspace. Such shells may be responsible for the discrete-velocity features seen in absorption-line observations ofquasar-driven outflows, and this geometry is often as-sumed in modeling these data (Arav et al. 2008; Moeet al. 2009; Borguet et al. 2012).In our sample SDSS J0319-0019 shows potential bubblewalls in the emission line observations (Liu et al. 2013b;Wylezalek et al. 2015), and in the HST image we see a cir-cular shell-like feature whose left wall (as seen in Figure2) is co-spatial with one of them. As mentioned in Sec.3.3, the putative scattered component near the nucleus iscompact and does not follow the regular triangular mor-iant scattering cones 15phology. Although the pressure inside the bubble maybe very high, the density of the bubble medium can below as it is shock-heated to high temperatures. There-fore, in the case of an illuminated bubble we might pref-erentially see only the bubble walls both in the ionizedgas and in the scattered light observations, in qualitativeagreement with our observations of SDSS J0319-0019. Aweaker bubble candidate is SDSS J0210-1001. It, too,has a shell-like feature in the U -band image (Figure 2).This feature corresponds to a region of very low velocitydispersion, as discussed by Liu et al. (2013b) and there-fore could be attributed to an illuminated dwarf galaxycompanion or tidal debris.The observed morphology of the scattered light de-pends both on the underlying density distribution of thegas and on the illumination pattern. For objects whichare undergoing mergers, the scattered light morphol-ogy is affected by the merger (e.g., SDSS J0842+3625,SDSS J0858+4417). For objects with candidate wind-driven bubbles (SDSS J0319-0019 and SDSS J0210-1001), the scattered light morphology is determined bythe presence of the bubbles. Therefore, modeling ourscattered light regions as cones filled with scatterers ofdeclining density is only an approximation. Figures 3and 5 support the use of this approximation for estimat-ing the surface brightness distribution of scattered lightin most of the sample and indicate that this approxi-mation is preferable to some other geometries (e.g., thindisk). CONCLUSIONS
In this paper we investigate
HST images of 20 power-ful obscured quasars selected based on the [OIII] emissionline luminosities. The images probe continuum emissionat ∼ ∼ λ (cid:48) I λ (cid:48) (cid:39) × − erg sec − cm − arcsec − at observedwavelength λ (cid:48) (cid:39) (cid:38) (cid:38)
10 kpc from the nucleus in 3 objects. While signaturesof quasar-driven outflows are now detected out to sev-eral kpc via a variety of methods (e.g., Nesvadba et al.2008; Feruglio et al. 2010; Borguet et al. 2012; Greeneet al. 2012), only a handful of giant scattering nebulae ofcomparable extents had been seen before (Hines & Wills1993; Hines et al. 1999; Schmidt et al. 2007). As thesescattering regions predominantly have conical / triangu-lar appearance, they can be identified based on morphol-ogy alone. Furthermore, the orientation of the scatteredlight regions is in excellent agreement with ground-basedpolarimetric measurements available for half of this sam-ple (Zakamska et al. 2005, 2006), in that the measuredpolarization position angles are orthogonal to the axes ofthe scattering cones as seen in the plane of the sky. The mere detection of conical scattered light regions in thesesources implies that these sources would be seen as nor-mal blue quasars along some lines of sight, in agreementwith the foundational principle of the geometric unifica-tion model of the active galactic nuclei (Antonucci 1993).We estimate that 2.3% of the intrinsic luminosityof the quasar is scattered off the interstellar mediumand reaches the observer. Furthermore, at 3000˚A scat-tered light constitutes about 73% of the total emission.This fraction includes both the conically shaped brightscattered-light regions and the faint quasi-spherical com-ponent which remains after subtraction of scaled yellow-band stellar models. The origin of this component is notfully understood; it may be due to star formation in thenucleus of the host galaxy, but it can also be due to scat-tered light produced when quasar radiation percolatesthrough patchy obscuring material.Failing to correct for the scattered light component inobservations of hosts of luminous quasars would lead toa dramatic overestimate of the star formation rates inquasar host galaxies. The median apparent 3000˚A lu-minosity of the objects in our sample is νL ν [3000˚A]=10 . erg s − . If we did not know that most of this lu-minosity is due to scattered light, we would have derivedthe rate of unobscured star formation of ∼ M (cid:12) yr − (Iglesias-P´aramo et al. 2004), having corrected the ap-parent luminosity from 3000˚A to 2000˚A using f λ ∝ λ − (Meurer et al. 1999). But because most of the ultra-violet emission is due to scattered light, the actual ratesof unobscured star formation are closer to ∼ M (cid:12) yr − .Our results demonstrate the key difficulty of usingquasar hosts’ colors and luminosities for deriving starformation rates and stellar masses, as is commonly donefor type 1 (unobscured) quasars (S´anchez et al. 2004;Schramm et al. 2008). Extended scattered light is ex-pected to be present in quasars of both types, though itis difficult to say how much of a contribution it makesto the extended emission in type 1s (Young et al. 2009).On the one hand, in type 1s, which face the observer,one might expect that the scattered light regions wouldappear more compact because of projection effects andthus would be absorbed into the point-spread functioncomponent associated with the quasar itself, in whichcase scattered light would not present much of a prob-lem for the derived host values. On the other hand, be-cause of the forward-scattering nature of dust particlesone might expect the scattering efficiencies to be higherin type 1s than in type 2s, which would make scatteringeffects stronger in type 1s. Morphological identificationof scattered light regions would be next-to-impossible intype 1 sources, both because of the bright quasar andbecause the scattering cone is facing the observer and isthus lacking the characteristic triangular shape we see intype 2s. Stellar population decomposition of off-nuclearspectra which take into account the possibility of scat-tered light may be a more reliable procedure for calcu-lating the star formation rates of quasar hosts (Liu et al.2009; Canalizo & Stockton 2013).Two cones are detected in 14 of the objects, pointingin roughly opposite directions from the nucleus. This isconsistent with axisymmetric toroidal obscuration pos-tulated by the classic geometric unification model. Thebrightness ratios of the two cones and the absence of a6 Obied et al.second cone in the 6 remaining sources strongly suggestthat dust (which is more efficient in the forward directionthan backwards) is responsible for scattering. By com-paring our models of dust-scattered cones with observa-tions, we find tentative evidence that dust in quasar hostsis a more efficient polarizer at 3000˚A than dust in Magel-lanic Clouds and the Milky Way (Draine 2003). Modelswith dust concentrated in a disk are inconsistent withobservations; volume-filled cones produce better fits.The measured opening angles of scattering cones allowus to calculate the type 1 / type 2 ratio from the prob-ability that the observers’ line of sight lies within thescattering cones, assuming that those correspond exactlyto the opening angles of the obscuring material. The re-sulting type 1 quasar fraction in the population is only ∼ (cid:38)
50% at these redshifts and luminosities. This is aninteresting discrepancy which is not easily discounted asdue to problematic opening angle measurements, as wedo not see a single scattering cone with the half openingangle of 60 o expected for a 1:1 ratio of type 1 to type 2quasars. We suggest that the actual type 1 fraction canbe higher than that derived from the opening angles ifthe obscuring material is patchy and if 30 −
50% of thelines of sight going through it nonetheless result in a type1 appearance. This hypothesis would not only reconcilethe measured opening angles with the studies of quasardemographics, but would explain the excess spherically-symmetric U -band emission left over after stellar sub-traction and the morphology of the emission-line nebu-lae which are more isotropic than would be expected inconical illumination (Liu et al. 2013a,b).Detection of extended scattered light offers a uniqueopportunity to estimate some properties of the inter-stellar medium of the host galaxies of very luminousquasars. From modeling scattered light regions we find that the typical profile of the density of the interstellarmedium responsible for the observed scattered light is n H ( r ) (cid:39) . − . − × ( r/ − . There are largesystematic uncertainties in the normalization of this den-sity and it should be regarded as an order-of-magnitudeestimate. The slope of this density profile is consistentwith that established in a steady-state, constant velocityoutflow. If this gas participates in an outflow with typicalvelocities v (cid:39)
800 km s − suggested by the emission-lineobservations (Liu et al. 2013b), then the mass outflowrate of the interstellar medium seen in scattered lightis ˙ M ∼ − M (cid:12) yr − . This is only 1 . −
20% ofthe previously measured outflow rates of the narrow-line-emitting gas. We suggest that if the outflow is primarilyin the form of dense clouds (either in the warm ionizedphase or in the cold neutral or molecular phase) thenthese clouds are likely to be optically thick to ultra-violetemission and the bulk of their mass is not participatingin scattering, so only thin outer shells of the clouds facingthe quasar would contribute to scattered light.The authors are grateful to Julian Krolik and theanonymous referee for useful discussions. Based on ob-servations associated with programs GO-13307 and GO-9905 made with the NASA/ESA
Hubble Space Tele-scope , obtained at the Space Telescope Science Institute(STScI), which is operated by the Association of Univer-sities for Research in Astronomy, Inc., under NASA con-tract NAS 5-26555. Support for program GO-13307 wasprovided by NASA through grant HST-GO-13307.01-A from the STScI. G.O. acknowledges support by theProvost’s Undergraduate Research Award at Johns Hop-kins University. D.W. acknowledges support by Akbari-Mack Postdoctoral Fellowship.
REFERENCESAlexandroff, R., Strauss, M. A., Greene, J. E., et al. 2013,MNRAS, 435, 3306Antonucci, R. 1993, ARA&A, 31, 473Antonucci, R. R. J., & Miller, J. S. 1985, ApJ, 297, 621Arav, N., Moe, M., Costantini, E., et al. 2008, ApJ, 681, 954Assef, R. J., Eisenhardt, P. R. M., Stern, D., et al. 2015, ApJ,804, 27Bailey, J., Axon, D. J., Hough, J. H., et al. 1988, MNRAS, 234,899Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, PASP, 93, 5Barger, A. J., Cowie, L. L., Mushotzky, R. F., et al. 2005, AJ,129, 578Barvainis, R. 1987, ApJ, 320, 537Borguet, B., Hutsem´ekers, D., Letawe, G., Letawe, Y., & Magain,P. 2008, A&A, 478, 321Borguet, B. C. J., Edmonds, D., Arav, N., Dunn, J., & Kriss,G. A. 2012, ApJ, 751, 107Canalizo, G., & Stockton, A. 2013, ApJ, 772, 132Capetti, A., Macchetto, F., Axon, D. J., Sparks, W. B., &Boksenberg, A. 1995, ApJ, 452, L87Cicone, C., Maiolino, R., Sturm, E., et al. 2014, A&A, 562, A21Deo, R. P., Richards, G. T., Nikutta, R., et al. 2011, ApJ, 729,108Donley, J. L., Koekemoer, A. M., Brusa, M., et al. 2012, ApJ,748, 142Dopita, M. A., Groves, B. A., Sutherland, R. S., Binette, L., &Cecil, G. 2002, ApJ, 572, 753Draine, B. T. 2003, ApJ, 598, 1017Draine, B. T., & Salpeter, E. E. 1979, ApJ, 231, 77Elvis, M., Marengo, M., & Karovska, M. 2002, ApJ, 567, L107 Faucher-Gigu`ere, C.-A., & Quataert, E. 2012, MNRAS, 425, 605Feruglio, C., Maiolino, R., Piconcelli, E., et al. 2010, A&A, 518,L155Glikman, E., Helfand, D. J., White, R. L., et al. 2007, ApJ, 667,673Gonzaga, S., Hack, W., Fruchter, A., & Mack, J. 2012, TheDrizzlePac Handbook (Baltimore, STScI)Greene, J. E., Zakamska, N. L., Ho, L. C., & Barth, A. J. 2011,ApJ, 732, 9Greene, J. E., Zakamska, N. L., & Smith, P. S. 2012, ApJ, 746, 86Groves, B. A., Dopita, M. A., & Sutherland, R. S. 2004, ApJS,153, 75Hao, H., Elvis, M., Civano, F., & Lawrence, A. 2011, ApJ, 733,108Harrison, C. M., Alexander, D. M., Mullaney, J. R., & Swinbank,A. M. 2014, MNRAS, 441, 3306Harrison, C. M., Thomson, A. P., Alexander, D. M., et al. 2015,ApJ, 800, 45Hasinger, G. 2008, A&A, 490, 905Hines, D. C., Schmidt, G. D., Smith, P. S., Cutri, R. M., & Low,F. J. 1995, ApJ, 450, L1Hines, D. C., Schmidt, G. D., Wills, B. J., Smith, P. S., &Sowinski, L. G. 1999, ApJ, 512, 145Hines, D. C., & Wills, B. J. 1993, ApJ, 415, 82Hopkins, P. F., Hernquist, L., Cox, T. J., et al. 2006, ApJS, 163, 1Hopkins, P. F., Strauss, M. A., Hall, P. B., et al. 2004, AJ, 128,1112Iglesias-P´aramo, J., Buat, V., Donas, J., Boselli, A., & Milliard,B. 2004, A&A, 419, 109 iant scattering cones 17
Jia, J., Ptak, A., Heckman, T., & Zakamska, N. L. 2013, ApJ,777, 27Kishimoto, M. 1999, ApJ, 518, 676Kleinmann, S. G., Hamilton, D., Keel, W. C., et al. 1988, ApJ,328, 161Koekemoer, A. M., & et al. 2002, HST Dither Handbook(Baltimore, STScI)Krolik, J. H. 2007, ApJ, 661, 52La Franca, F., Fiore, F., Comastri, A., et al. 2005, ApJ, 635, 864Lacy, M., Ridgway, S. E., Sajina, A., et al. 2015, ApJ, 802, 102Lacy, M., Sajina, A., Petric, A. O., et al. 2007, ApJ, 669, L61Lacy, M., Storrie-Lombardi, L. J., Sajina, A., et al. 2004, ApJS,154, 166Lal, D. V., & Ho, L. C. 2010, AJ, 139, 1089Lansbury, G. B., Alexander, D. M., Del Moro, A., et al. 2014,ApJ, 785, 17Lansbury, G. B., Gandhi, P., Alexander, D. M., et al. 2015, ApJ,809, 115Lawrence, A., & Elvis, M. 2010, ApJ, 714, 561Lena, D., Robinson, A., Storchi-Bergman, T., et al. 2015, ApJ,806, 84Liu, G., Zakamska, N. L., Greene, J. E., Nesvadba, N. P. H., &Liu, X. 2013a, MNRAS, 430, 2327—. 2013b, MNRAS, 436, 2576Liu, X., Zakamska, N. L., Greene, J. E., et al. 2009, ApJ, 702,1098Mac Low, M.-M., McCray, R., & Norman, M. L. 1989, ApJ, 337,141Markowitz, A. G., Krumpe, M., & Nikutta, R. 2014, MNRAS,439, 1403Mart´ınez-Sansigre, A., Rawlings, S., Lacy, M., et al. 2006,MNRAS, 370, 1479Mateos, S., Alonso-Herrero, A., Carrera, F. J., et al. 2013,MNRAS, 434, 941McGee, S. L., & Balogh, M. L. 2010, MNRAS, 405, 2069Meurer, G. R., Heckman, T. M., & Calzetti, D. 1999, ApJ, 521, 64Miller, J. S., & Goodrich, R. W. 1990, ApJ, 355, 456Miller, J. S., Goodrich, R. W., & Mathews, W. G. 1991, ApJ,378, 47Moe, M., Arav, N., Bautista, M. A., & Korista, K. T. 2009, ApJ,706, 525Morganti, R., Tadhunter, C. N., & Oosterloo, T. A. 2005, A&A,444, L9Nenkova, M., Ivezi´c, ˇZ., & Elitzur, M. 2002, ApJ, 570, L9Nenkova, M., Sirocky, M. M., Nikutta, R., Ivezi´c, ˇZ., & Elitzur,M. 2008, ApJ, 685, 160Nesvadba, N. P. H., Lehnert, M. D., De Breuck, C., Gilbert,A. M., & van Breugel, W. 2008, A&A, 491, 407Netzer, H., & Laor, A. 1993, ApJ, 404, L51Nikutta, R., Elitzur, M., & Lacy, M. 2009, ApJ, 707, 1550Nims, J., Quataert, E., & Faucher-Gigu`ere, C.-A. 2015, MNRAS,447, 3612Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W. 2002, AJ,124, 266Ptak, A., Zakamska, N. L., Strauss, M. A., et al. 2006, ApJ, 637,147Reyes, R., Zakamska, N. L., Strauss, M. A., et al. 2008, AJ, 136,2373Richards, G. T., Lacy, M., Storrie-Lombardi, L. J., et al. 2006,ApJS, 166, 470 Roman-Duval, J., Gordon, K. D., Meixner, M., et al. 2014, ApJ,797, 86Ross, N. P., Hamann, F., Zakamska, N. L., et al. 2015, MNRAS,453, 3932S´anchez, S. F., Jahnke, K., Wisotzki, L., et al. 2004, ApJ, 614, 586Schartmann, M., Meisenheimer, K., Camenzind, M., et al. 2008,A&A, 482, 67Schmidt, G. D., Smith, P. S., Hines, D. C., Tremonti, C. A., &Low, F. J. 2007, ApJ, 666, 784Schramm, M., Wisotzki, L., & Jahnke, K. 2008, A&A, 478, 311Serra, P., Oosterloo, T., Morganti, R., et al. 2012, MNRAS, 422,1835Sirianni, M., Jee, M. J., Ben´ıtez, N., et al. 2005, PASP, 117, 1049Smith, P. S., Schmidt, G. D., Hines, D. C., & Foltz, C. B. 2003,ApJ, 593, 676Steffen, A. T., Barger, A. J., Cowie, L. L., Mushotzky, R. F., &Yang, Y. 2003, ApJ, 596, L23Stern, D., Eisenhardt, P., Gorjian, V., et al. 2005, ApJ, 631, 163Stern, J., Faucher-Giguere, C.-A., Zakamska, N. L., & Hennawi,J. F. 2015, arXiv: 1510.07690Sun, A.-L., Greene, J. E., Zakamska, N. L., & Nesvadba, N. P. H.2014, ApJ, 790, 160Tran, H. D. 1995a, ApJ, 440, 565—. 1995b, ApJ, 440, 578—. 1995c, ApJ, 440, 597Tran, H. D., Cohen, M. H., & Villar-Martin, M. 2000, AJ, 120,562Tran, H. D., Miller, J. S., & Kay, L. E. 1992, ApJ, 397, 452Treister, E., Krolik, J. H., & Dullemond, C. 2008, ApJ, 679, 140Tsai, C.-W., Eisenhardt, P. R. M., Wu, J., et al. 2015, ApJ, 805,90Ueda, Y., Akiyama, M., Ohta, K., & Miyaji, T. 2003, ApJ, 598,886Vanden Berk, D. E., Richards, G. T., Bauer, A., et al. 2001, AJ,122, 549Veilleux, S., Mel´endez, M., Sturm, E., et al. 2013, ApJ, 776, 27Vignali, C., Alexander, D. M., Gilli, R., & Pozzi, F. 2010,MNRAS, 404, 48Wagner, A. Y., Umemura, M., & Bicknell, G. V. 2013, ApJ, 763,L18Weingartner, J. C., & Draine, B. T. 2001, ApJ, 548, 296Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010,AJ, 140, 1868Wylezalek, D., Zakamska, N., Liu, G., & Obied, G. 2015,submitted, 000, 00York, D. G., et al. 2000, AJ, 120, 1579Young, S., Axon, D. J., Robinson, A., & Capetti, A. 2009, ApJ,698, L121Zakamska, N. L., G´omez, L., Strauss, M. A., & Krolik, J. H.2008, AJ, 136, 1607Zakamska, N. L., & Greene, J. E. 2014, MNRAS, 442, 784Zakamska, N. L., Lampayan, K., Petric, A., & et al. 2015,MNRAS, 000, 000Zakamska, N. L., Strauss, M. A., Heckman, T. M., Ivezi´c, ˇZ., &Krolik, J. H. 2004, AJ, 128, 1002Zakamska, N. L., et al. 2003, AJ, 126, 2125—. 2005, AJ, 129, 1212—. 2006, AJ, 132, 1496Zubovas, K., & King, A. 2012, ApJ, 745, L34
TABLE 1
HST observations of quasar scattered light nebulae ID z L [OIII] λ eff F ν subtracted F ν L [12 µ m]SDSS J014932.53 − − − − − − − − − Note . — Summary of the 21 sources discussed. Top 11 sources are new
HST obser-vations conducted in 2013-2014 (GO-13307, PI Zakamska); middle 9 sources are archivalfrom
HST observations conducted in 2003-2004 (GO-9905, PI Strauss); last source wasanalyzed by (Hines et al. 1999) and we use their results in this paper. L [OIII] is given inunits of log( L [OIII] , erg s − ), F ν is the flux density in the U -band HST image before andafter ‘optimal subtraction’ in µ Jy. λ eff is the rest-frame effective wavelength of our U -bandobservations. L [12 µ m] is given in units of log( νL ν [12 µ m], erg s − ). TABLE 2Model Fit Parameters ID n H , (cm − ) n ∗ H , (cm − ) θ ( ◦ ) i ( ◦ ) m max. extent (kpc) cone-to-counterconeSDSS J014932.53 − > − > − > − > − > − − − − > − − − − − Note . — Parameters of the density profiles and geometric parameters of the scattering cones derived from our dustscattering models. θ are half opening angles of scattering cones, i are inclination angles measured relative to the line ofsight, m are half-slopes of the density power law profiles, and n H , is the density at 1 kpc normalized assuming [OIII]-derivedbolometric luminosities, whereas n ∗ H , values are obtained from 12 µ m-derived bolometric luminosities. Maximal projectedextents are estimated by looking for U -band emission inconsistent with stellar distribution down to limiting surface-brightnessof λ (cid:48) I λ (cid:48) (cid:39) × − erg sec − cm − arcsec −2