Bipolar ionization cones in the Extended Narrow-Line Region of nearby QSO2s
Thaisa Storchi-Bergmann, Bruno Dall'Agnol de Oliveira, Luis Felipe Longo Micchi, Henrique Roberto Schmitt, Travis Cody Fischer, Steven Kraemer, Michael Crenshaw, Peter Maksym, Martin Elvis, Giuseppina Fabbiano, Luis Colina
DDraft version October 16, 2018
Typeset using L A TEX twocolumn style in AASTeX61
BIPOLAR IONIZATION CONES IN THE EXTENDED NARROW-LINE REGION OF NEARBY QSO2S
T. Storchi-Bergmann,
1, 2
B. Dall’Agnol de Oliveira, L. F. Longo Micchi, H. R. Schmitt, T. C. Fischer, S. Kraemer, M. Crenshaw, P. Maksym, M. Elvis, G. Fabbiano, and L. Colina Departamento de Astronomia, Universidade Federal do Rio Grande do Sul, IF, CP 15051, 91501-970 Porto Alegre, RS, Brazil Harvard-Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, MA 02138, USA Institute for Astrophysics and Computational Sciences, Department of Physics, The Catholic University of America, Washington,DC 20064, USA Naval Research Laboratory, Washington, DC 20375, USA 0000-0001-7376-8481 Goddard Space Flight Center, 8800 Greenbelt Rd, Greenbelt, MD 20771, USA Department of Physics and Astronomy, Georgia State University, Astronomy Offices, 25 Park Place, Suite 600, Atlanta, GA 30303,USA Centro de Astrobiologa (CAB, CSIC-INTA), Carretera de Ajalvir, 28850 Torrejn de Ardoz, Madrid, Spain
ABSTRACTWe have used narrow-band [OIII] λλ α +[NII] λλ ,
84 Hubble Space Telescope (HST) images of9 luminous (L[OIII] > erg s − ) type 2 QSOs with redshifts 0 . < z < . α +[NII]) excitation maps indicating that the torussurvives these luminosities, allowing the escape of ≈
10 times higher ionizing photon rates along the ionization axisthan perpendicularly to it. The exceptional HST angular resolution was key to our success in arriving at theseconclusions. Combining our measurements with previous ones based on similar HST data, we have revisited therelation between the ENLR radius R maj and L[OIII] over the range 39 < log(L[OIII]) < . − ): log(R maj ) =(0 . ± .
03) log(L[OIII]) − . ± .
98. The radius of the ENLR keeps increasing with L[OIII] in our data, implyingthat the ENLR can extend to distances beyond the limit of the galaxy if gas is present there – e.g. from AGN outflowsor interactions, seen in 6 objects of our sample. We attribute the flattening previously seen in this relation to thefact that the ENLR is matter-bounded, meaning that ionizing photons usually escape to the intergalactic medium inluminous AGN. Estimated ionized gas masses of the ENLRs range from 0.3 to 2 × M (cid:12) , and estimated powers forassociated outflows range from < .
1% to a few percent of the QSO luminosity.
Keywords: quasars: emission lines – galaxies: active — galaxies: Seyfert – galaxies: jets
Corresponding author: T. [email protected] a r X i v : . [ a s t r o - ph . GA ] O c t Storchi-Bergmann et al. INTRODUCTIONThe physical processes that couple the growth of su-permassive black holes (SMBH) to their host galaxiesoccur in the vicinity of the galaxy nucleus ( ≈ inner kpc)(Hopkins & Quataert 2010) when it becomes active dueto mass accretion to the SMBH (Ferrarese & Ford 2005;Kormendy & Ho 2013). The radiation emitted by ActiveGalactic Nuclei (AGN) works as a flashlight that illumi-nates and ionizes the gas in the vicinity of the nucleus,forming the Narrow Line Region (NLR). AGN-drivenwinds (Elvis 2000; Ciotti et al. 2010; Storchi-Bergmannet al. 2010; Barbosa et al. 2014) interact with the gasand produce outflows reaching velocities of hundredsof km s − (Fischer et al. 2013; Storchi-Bergmann et al.2010). Relativistic jets emanating from the AGN also in-teract with the gas of the NLR (Wang et al. 2009; Coutoet al. 2017). Both types of outflow produce feedback,which has been an important ingredient in galaxy evo-lution models to avoid producing over-massive galaxies(Fabian 2012). Inflows have also been observed in thisregion (Riffel et al. 2008, 2013; Crenshaw et al. 2010;Schnorr-M¨uller et al. 2014a). The importance of theNLR stems from the fact that it is spatially resolved, asit extends from hundreds to thousands of parsecs fromthe nucleus and exhibits strong line emission. Theseproperties of the NLR allow the observation of the in-teraction between the AGN and the circumnuclear gas inthe galaxy: (1) via the observation of the NLR geometryand excitation properties that constrain the AGN struc-ture and nature of the ionizing source; (2) via the gaskinematics that maps the processes of the AGN feedingand feedback.Over the last 30 years, narrow-band imaging of theNarrow Line Region (NLR) became an important toolin the study of their ionization mechanism, morphol-ogy and implied geometry for the AGN. Starting withground based observations of Seyfert (Sy) galaxies(e.g. Pogge 1988; Haniff et al. 1988; Wilson & Tsve-tanov 1994), emission-line images of [OIII] λ α +[NII] showed that the NLRs of several Sy galaxieshave biconical shapes, with the apex at the nucleus.These observations were of key importance in validat-ing the Unified Model of AGN (Antonucci 1993), inwhich both Sy types have a SMBH fed by an accre-tion disk, surrounded by a torus of molecular gas anddust that collimates the ionizing radiation leading tothe observed biconical shape. However, although theseground-based narrow-band images of Sy galaxies haveallowed the detection of shadowing of the nuclear ion-izing radiation, they could resolve only the nearest andbrightest sources. For fainter and more distant sources,the NLR angular diameters decrease below the resolu- tion limit of ground-based telescopes ( ∼ (cid:48)(cid:48) ). The use ofthe Hubble Space Telescope (hereafter HST) to imagethe NLR was crucial for the further development of thefield (e.g. Wilson et al. 1993; Capetti et al. 1996; Falckeet al. 1998; Ferruit et al. 2000; Schmitt et al. 2003a).The high spatial resolution of these observations allowedbetter constraints on the Unified Model, and the studythe NLR structure in detail. Some results from thesestudies include the alignment of the radio jet and NLR,which indicate that the torus and accretion disk areconnected; that the radio jet can significantly influencethe NLR emission; and that Sy 1 NLRs are statisti-cally more circular and concentrated than Seyfert 2’s,which can be attributed to foreshortening in the former(Schmitt et al. 2003b).Another important result from these studies – in par-ticular of the snapshot survey of Schmitt et al. (2003a,b)– was the detection of a relation between the NLR ra-dius R NLR and its luminosity L([OIII]). For a sampleof 60 AGN with L([OIII]) < erg s − , these authorsfound the relation R NLR ∝ L[OIII] . . This was inter-preted as due to the fact that the ionization parameteris not constant along the NLR, and that most of the[OIII] emission comes from a low density region. A con-stant ionization parameter had been suggested by previ-ous observations of the NLR of a sample of QSO’s thatimplied a slope of ∼ z ≥ .
1, the limited angular resolution and smearingby the seeing precludes firm constraints on the NLR‘(orENLR) morphology. In order to provide such con-straints at higher redshifts and AGN luminosities, wehave obtained narrow-band [OIII] λ α +[NII]images with the Hubble Space Telescope (HST) of a flux-limited sample of 9 luminous type 2 AGN (or QSOs, dueto their high luminosites), with redshifts in the range0 . ≤ z ≤ .
5. We use these observations to obtainexcitation maps, through the ratio [OIII]/(H α +[NII]),ionized gas masses from the H α luminosities and ex- arrow-Line Region of QSO2s maj andL[OIII], from 39 < log(L[OIII]) < − ).This paper is organized as follows: in Sec.2 we presentthe sample, in Sec. 3 we discuss the observations, datareduction and measurements, in Sec. 4 we present theimages and ionization maps, as well as total gas massesand spatial profiles of the ionized gas mass surface den-sity. In Sec. discussion we discuss our results and inSec. 6 we present our conclusions. SAMPLEOur sample was selected from the catalogue of QSO2galaxies of Reyes et al. (2008), that comprises 887 galax-ies with 0 . < z < .
8, thus just beyond the red-shift range of the HST imaging survey of Schmitt et al.(2003a). The selection was done according to the follow-ing criteria: (1) the galaxies should have [OIII] fluxeslarger than 1.5 × − ergs cm − s − in order to as-sure enough signal-to-noise ratio in the narrow-bandimages; (2) they should have redshifts in the range0 . < z ≤ .
5, thus allowing to probe more lumi-nous AGNs than in the previous study of Schmitt et al.(2003a); the luminous sources at z ≈ ∼ . − ).The list of the sample galaxies, together with theirredshifts, angular scales (obtained from the angular dis-tances) and luminosity distances ( D L ) are shown in Ta-ble 1. The listed redshifts and the other derived quanti-ties were obtained from the NASA/IPAC ExtragalacticDatabase (hereafter NED). The angular and luminos-ity distances were corrected to the Cosmic MicrowaveBackground Radiation reference frame. The uncertain-ties shown in the quantities listed in the Table 1 corre-spond to the maximum errors resulting from the uncer-tainties in H and z. The spatial scales range from 1.9to 5.6 kpc arcsec − , allowing a spatial resolution at thegalaxies of 95 pc to 280 pc at the galaxies with HST.We show in Fig. 1, the SDSS spectra of the samplegalaxies together with curves showing the narrow fil-ters passbands (see next section) used to obtain the lineemission images centered on [OIII] λ α +[NII],as well as the broad-band filter passbands used to obtainimages in the continuum. For a few galaxies, the SDSSspectra did not cover the H α +[NII] emission lines, astheir redshifts put these lines beyond the upper wave-length limit of the SDSS spectra. The passbands wereobtained with the Python package Astrolib PySynphot,which produces a synthetic throughput of the filter. Table 1.
Sample propertiesID Name z Scale D L (1) (2) (3) (4) (5)1 J082313.50+313203.7 0.433 5.46 ± ± ± ±
243 J085829.58+441734.7 0.454 5.61 ± ± ± ±
265 J110952.82+423315.6 0.262 3.91 ± ±
586 J113710.77+573158.7 0.395 5.16 ± ±
957 J123006.79+394319.3 0.407 5.26 ± ±
938 J135251.21+654113.2 0.206 3.26 ± ±
479 J155019.95+243238.7 0.143 2.42 ± ± Note — (1) galaxy identification number used througout the pa-per; (2) galaxy identification (without
SDSS prefix); (3) red-shift; (4) scale (in kpc/ (cid:48) (cid:48) ); (5) luminosity distance (in Mpc).3.
OBSERVATIONS AND DATA REDUCTIONThe sample galaxies were observed with the HSTAdvanced Camera for Surveys (ACS) using linearramp filters centered on the redshifted [OIII] λ α +[NII] λ imalign andcombined using crrej , which allows cosmic rays removal,and subsequently divided by the exposure time (headerkeyword EXPTIME ). Then, the image was multipliedby the header keyword
PHOTFLAM , to transform itsunits to erg cm − s − ˚A − . In order to allow for a pre-cise continuum subtraction from the emission line im-ages, the line images were aligned to the continuum im-ages. In order to do this, the emission line images wererotated using the rotate task, matching the continuum ORIENTAT header keyword, and subsequently alignedwith the continuum image with the task imalign . Af-ter this, the sky contribution – obtained as the center
Storchi-Bergmann et al. of a gaussian fit to the flux histogram distribution of aregion devoid of galaxy contribution – was subtractedfrom the image. For each image pixel, we assume a fluxuncertainty equal to 3 times the standard deviation ofthis gaussian ( σ sky ).As can be observed in Figure 1, in a number of cases,the wide/median band filter used to obtain the contin-uum images includes also some contribution of emissionlines. In order to correct for this contamination, we haveused the SDSS spectra to calibrate our continuum im-ages, by matching the continuum fluxes in the images tothat of the SDSS spectrum (that corresponds to a circu-lar aperture with diameter of 3 arcsec). This was done asfollows. We fitted the SDSS continuum under the emis-sion line [OIII] λ λ σ of the average in the region. We then ob-tained the corresponding value from the HST continuumimage ( C circ, [ OIII ] ), by integrating the flux within a cir-cular aperture with diameter of 3 (cid:48)(cid:48) to match the SDSSaperture. The ratio S C, [ OIII ] =C SDSS, [ OIII ] /C circ, [ OIII ] gives the value used to scale the continuum image beforesubtracting it from the [OIII] λ λ C, ([ NII ]+ Hα ) .In the cases for which these lines were not present inthe spectra, we fitted a large region of the continuum inthe red end of the spectrum (close to the missing lines).These values are presented in Table 3, where the uncer-tainties in the SDSS continuum values were adopted asthe standard deviation of the data used in the fit. Thesedata are highlighted as yellow dots in Figure 1.Table 3 shows that the smallest scale factors used tocorrect the continuum images for the emission-line con-tribution – which correspond to the largest correctionsdue to emission-line contamination – were obtained fortargets 7, 1, 6 and 3 (from the largest to the lowest cor-rections). In these cases, the continuum filter includesan important contribution from the [OIII] emission lines,ranging from 60% to 20% of the total flux.As mentioned above, the [OIII] filter did not cover en-tirely the λ λ λ that would be obtained using the [OIII] fil-ter. Next, we modified the filter throughput, stretchingthe profile from the center by 50˚A toward smaller wave-lengths, in order to cover both emission lines. Now,the new integrated flux (F[OIII] ) corresponds to whatwould be expected if the filter covered both emissionlines. The ratio r [OIII] =F[OIII] /F[OIII] gives the mul- tiplication factor used to make the narrow-band [OIII]images to include the sum of the [OIII] λ λ α ) was obtained as F=F ∗ -S C F C ∗ , where F ∗ isthe flux before the continuum subtraction, F is the fluxafter the subtraction, and F C ∗ is the continuum fluxbefore the multiplication by the S C scale. In the case ofthe [OIII] image, the emission-line flux was in additionmultiplied by r [ OIII ] .The resulting flux uncertainty per pixel in eachline l is the quadrature sum of each contribuition σ = σ ∗ + (S C , l σ C ∗ ) + ( σ C C , l F C ∗ ) , where σ l ∗ and σ l are uncertainties before and after the continuum sub-traction, σ S C , l is the continuum scale uncertainty, and σ C ∗ is the continuum uncertainty. Both σ l ∗ and σ C ∗ areadopted as 3 σ sky of the respective image. In the case of l =[OIII], the uncertainty in r [OIII] was also propagated.When a flux integration is performed, the errors of eachpixel are added in quadrature.In order to check the flux calibration after the abovecorrections, we integrated the resulting [OIII] HST im-age flux over the SDSS circular aperture (F[OIII] circ )and compare it with F[OIII] SDSS , the flux of both [OIII]emission lines in the SDSS spectra. These values aredisplayed in Table 5, and a comparison between them isshown in Figure 2, along with a dashed line showing theloci of equal values. It can be seen that the fluxes agreewith each other within the uncertainties, supporting therobustness of our reduction and calibration processes.The bandwidths (∆ l ) of each filter were obtained fromAstrolib PySynphot. It corresponds to the width ofa box-like throughput curve, with the same area ofthe bandpass profile and heigth equal to the through-put at the average wavelength of the bandpass ( avg-wave in PySynphot). The continuum and the [OIII]and H α +[NII] narrow-band images were multiplied bythe corresponding bandwidth when fluxes in units oferg cm − s − were required. RESULTSAs we have used the SDSS spectra to correct our con-tinuum images from the contribution of the emission-lines as well as to correct for the usually partial coverageof the [OIII] λ (cid:48)(cid:48) diameter.4.1. SDSS spectra emission-line fluxes
The emission lines of the SDSS spectra were fit-ted using fourth-order Gauss-Hermite polynomials. In arrow-Line Region of QSO2s R e l a ti v e f l ux J082313.50+313203.7
F775WFR931NFR716N
J084135.04+010156.3
FR647MFR716NFR551N R e l a ti v e f l ux J085829.58+441734.7
F775WFR931NFR716N
J094521.34+173753.3
FR647MFR716NFR551N R e l a ti v e f l ux J110952.82+423315.6
F775WFR853NFR656N
J113710.77+573158.7
F775WFR931NFR716N R e l a ti v e f l ux J123006.79+394319.3
F775WFR931NFR716N
Wavelength (Å)
J135251.21+654113.2
FR647MFR782NFR601N
Wavelength (Å) R e l a ti v e f l ux J155019.95+243238.7
FR647MFR782NFR551N
Figure 1.
SDSS spectra of the sample galaxies (normalized by the maximum flux). The colored curves illustrate the filterpassbands used to obtain the corresponding HST images: [OIII] (blue curves), H α +[NII] (red curves) and continuum (greencurves). The yellow dots are the data used in the continuum fit for the calculation of S C, [ OIII ] and S C, [ NII ]+ Hα . Storchi-Bergmann et al.
Table 2.
Observation logID Name Date Filter Exptime Region(1) (2) (3) (4) (5) (6)1 J082313 2015-01-01 F775W 200 Cont.2014-12-31 FR931N 2542 H α +[NII]2015-01-01 FR716N 1866 [OIII]2 J084135 2015-02-16 FR647M 200 Cont.2015-02-16 FR716N 2506 H α +[NII]2015-02-16 FR551N 2020 [OIII]3 J085829 2015-02-08 F775W 200 Cont.2015-02-08 FR931N 2600 H α +[NII]2015-02-08 FR716N 1927 [OIII]4 J094521 2015-04-22 FR647M 200 Cont.2015-04-22 FR716N 2516 H α +[NII]2015-04-22 FR551N 2031 [OIII]5 J110952 2015-02-16 F775W 200 Cont.2015-02-16 FR853N 2600 H α +[NII]2015-02-16 FR656N 2029 [OIII]6 J113710 2015-09-08 F775W 200 Cont.2015-09-08 FR931N 2754 H α +[NII]2015-09-08 FR716N 2081 [OIII]7 J123006 2015-07-04 F775W 200 Cont.2015-07-04 FR931N 2562 H α +[NII]2015-07-04 FR716N 1889 [OIII]8 J135251 2015-05-13 FR647M 200 Cont.2015-05-13 FR782N 1786 H α +[NII]2015-05-13 FR601N 1976 [OIII]9 J155019 2015-06-17 FR647M 200 Cont.2015-06-17 FR782N 2516 H α +[NII]2015-06-17 FR551N 1875 [OIII] Note — (1) ID used in the paper (2) galaxy name; (3) observa-tion date; (4) HST ACS filter; (5) exposure time (in seconds);(6) spectral region where Cont. means continuum. the case of the [OIII] λ λ h and h coefficients,and the two lines were constrained to have a flux ra-tio F[OIII] λ λ F[OIII] circ F [ O III ] S D SS Figure 2.
Relation between the flux F[OIII]
SDSS obtainedfrom the SDSS spectra and F[OIII] circ , obtained from a cir-cular aberture of 3 (cid:48)(cid:48) in our HST [OIII] images. The dashedline corresponds to unity. fit the line profile, additional components were included(keeping the same constrains above between them). Thefits were obtained via Chi-Square minimization usingthe SLSQP Method available at Python library SciPy.The same procedure was adopted in the fit of the [NII]doublet lines. H β was fitted using one or more Gauss-Hermite profiles.Objects 4, 5 and 8 show a broad base in the [NII]+H α emission lines, which could indicate the presence of abroad H α component. However, we were able to fitthese profiles similarly well without the need of a broadH α component, by assuming that both the H α and [NII]profiles match that of [OIII] λ α component can-not be discarded. The total fluxes and widths of theemission lines in the SDSS spectra are displayed in Ta-ble 4. Uncertainties in the fit of the [OIII] λ th column of Table 4, the valueof the ratio η =H α /([NII]+H α ) obtained from the SDSSspectra, that we have used to estimate the contributionof H α to the narrow-band HST images that contain boththe H α and [NII] emission lines. In the cases of thegalaxies for which we could not obtain η from the SDSSspectra (because these lines are beyond the observedspectral range), we have used the average of the values arrow-Line Region of QSO2s Table 3.
Parameters used in the calibration of the dataID Name ∆ C ∆ [NII]+H α ∆ [OIII] S C , [OIII] S C , [NII]+H α r [OIII] (1) (2) (3) (4) (5) (6) (7)1 J082313 1454.59 200.44 136.96 0.46 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Note — (1) galaxy identification in the paper; (2) galaxy name; (3) continuum band-width (in ˚A); (4) [NII]+H α filter bandwidth (in ˚A); (5) [OIII] filter bandwidth (in˚A); (6) Continuum scale factor used for [OIII]; (7) Continuum scale factor used for[NII]+H α ; (8) Correction factor for the partial coverage of [OIII] λ obtained for the other galaxies, and list these values inboldface in the table.In the last column of Table 4), we list the velocityv =W / .
3, where W is the width of the [OIII] λ HST images
The corrected continuum and emission-line images areshown in Figures 3–11. In the top left panel we showthe continuum images, in the top right the continuum-subtracted [OIII] images, in the bottom right thecontinuum-subtracted H α +[NII] images and in the bot-tom left panel an excitation map obtained as the ratiobetween the [OIII] and [NII]+H α images. The imageshave been multiplied by the bandwidth, and [OIII] refersto the sum of the λ λ (cid:48)(cid:48) , and shows the scale of theimages in kpc. The black cross marks the position of thegalaxy nucleus, adopted as corresponding to the peakof the continuum flux.The flux levels shown in the [OIII] and [NII]+H α im-ages of Figures 3-11 range from 1 σ sky to the maximumflux value in the galaxy image (F max ). Four equallyspaced contours, ranging from 3 σ sky to F max , have beenoverplotted on theses images. The contours were ob- tained after smoothing the images by a Gaussian filterwith 1 pixel standard deviation.In the case of the excitation maps, only pixels forwhich both images have values higher then 3 σ sky wereconsidered. The values of the remaining pixels wereset to zero. We have introduced a contour in thesemaps to highlight regions (inside the contour) with thehighest excitation values, typical of Sy galaxies, whiilelower values are typical of LINERs. This level was cho-sen at [OIII] λ β = 5 .
25, and the conversion to[OIII]/([NII]+H α ) was calculated using the correspond-ing fluxes of [NII]+H α in the SDSS spectra (further dis-cussed in Sec. 4.3). Dashed contours were used for galax-ies for which the [NII]/H α ratio could not be measuredbecause these lines are beyond the spectral range of theSDSS spectrum, and have thus been estimated using av-erage data from the other galaxies.4.2.1. Continuum
Although the continuum images have been correctedfor the average contamination of the emission lines viathe S C, [ OIII ] and S C, ([ NII ]+ Hα ) scale factors, the imagesin the continuum show some features that may be dueto the contamination from the emission lines, as dis-cussed in Sec. 3. The objects showing the largest cor-rections ( >
20% of the continuum flux) due to emission-line contaminations are the targets 1, 7, 6 and 3 (fromthe largest to the lowest corrections). We discuss each
Storchi-Bergmann et al.
Table 4.
SDSS spectra measurementsID Name F(H β ) F[OIII] F(H α ) F[NII] λ W [OIII] W H α η v (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)1 J082313 0.290 4.595 - - 375 - Note — (1) galaxy identification in the paper; (2) galaxy name; (3)-(6) Fluxes (in unitsof 10 − ergs s − ); in the case of [OIII], it is the sum of the λλ , − ; (9) ratio η = Hα/ ([ NII ]+ Hα ),where the bold highlight indicates that the line was missing in the SDSS spectra and thevalue thus corresponds to the average of the ratios obtained for the other galaxies. (10) v = W / .
3, where W (in km s − ) is the width of [OIII] λ of these cases below, and the possible consequences forthe origin of the features seen in the continuum.Target 1 (Fig. 3) shows in the continuum image a pos-sible companion to the south-east, although its contin-uum flux is low, while in the [OIII] image its emissionis much stronger, even after subtracting the continuumwith significant emission-line contribution.This featureis thus most probably a detached cloud of gas and nota companion galaxy. In the case of target 7 (Fig.9), theextended feature seen in the continuum image to thesouth-east is most probably also due to contaminationfrom the [OIII] emission lines, as seen by the strong cor-responding feature in the [OIII] image. In the case oftarget 6 (Fig. 8), the extended appearance in the con-tinuum from the south-east to the north-west may havesome contamination from the strong emission seen in the[OIII] line image. A compact object to the north-westis faint in the emission-line images and is most probablya companion in this case. Finally, in the case of target3 (Fig. 5), the continuum shows what seems to be a sec-ond nucleus, as its brightness approaches that of the nu-cleus (adopted as the brightest peak). Due to the [OIII]emission-line contamination in the continuum filter, onemay think that this feature is due to this contamination.But, if this is the case, then the [OIII] image should havea strong knot of emission at this location, even if thereis some contamination from the line to the continuum.But what we see instead is even a small depression in the emission-line images at this location, which implies thatthere is not much [OIII] emission in this knot, and thefeature in the continuum is most probably a secondarynucleus, from a possible recent capture of a small galaxy.In the other targets, where the contamination of thecontinuum by the emission lines is lower than 20% itcan be concluded that: target 2 is in obvious interactionwith a companion; target 4 shows some assymetry inthe continuum but other than that no signature of on-going interaction; target 5 is essentially round, with nosignature of interaction in the continuum; target 8 seemsto be an on-going merger and target 9 shows an object2 (cid:48)(cid:48) ( ≈ α +[NII] images in the case ot target 5, but, as thestructure of the continuum image is in most cases morecompact than that of the emission-line images, it doesnot affect the properties of the most extended gas, andin particular the measured extent of the ENLR, one ofthe main results of the present paper.4.2.2. Emission-line images
All galaxies show [OIII] and H α +[NII] extended emis-sion up to several kpc in the plane of the sky. Elongatedstructures clearly defining ionization axes are observed arrow-Line Region of QSO2s Excitation maps
The bottom-left panels of Figs. 3-11 show the excita-tion maps, obtained as the ratio between the [OIII] and[NII]+H α images. As discussed above, we added a con-tour to these maps as a reference for the location ofthe regions with the highest excitation (inside the con-tours). The contour adopted to separate these valuescorresponds to [OIII] λ β = 5 .
25, a value closeto the separation between LINERs (below this value)and Seyferts (above this value) in the BPT Veilleux& Osterbrock (1987); Baldwin et al. (1981) [OIII]/H β vs. [NII]/H α diagnostic diagram, considering values of[NII]/H α ≈ . − − β value to [OIII]/([NII]+H α ) – the actualratio shown in our excitation maps, was obtained un-der the assumptions [OIII] λ λ α /H β = 3. This H α /H β ratio was adopted consid-ering that, although its value is on average ≈ ≈
3. In any case, this con-tour is just a reference for the excitation level. In orderto obtain the value of the [OIII]/([NII]+H α ) ratio, wehave used the η value listed in Table 4). Considering theabove, we have:[ OIII ][ N II ]+ Hα = 4 η (cid:18) [ OIII ] λ Hβ (cid:19) (1)These reference contours are shown as continuous linesfor galaxies with available η values, and as dashed linesfor the galaxies with estimated η values (shown in bold-face in Table 4). It is also important to point out that weare using a constant η ratio througout the ENLR, deter-mined from the SDSS spectra, although this value mostprobably varies from pixel to pixel. Figs. 3-11 show thatthe contours in most cases delineate a (patchy) approx-imately biconical structure for the highest ionization re-gions (orange) surrounded by lower ionization regions(purple). In the case of targets 1 and 7, these contoursseem to be beyond the region separating the highestfrom the lower excitation, but the biconical structure isdelineated by the orange regions surrounded by purpleregions with the separation between them appearing athigher line ratios than that of the adopted contour.4.3.1. [OIII] λ and [OIII]/H β spatial profiles In order to better quantify the excitation along theENLR, we have used the η values listed in Table 4 toobtain the line ratio values [OIII] λ β (hereafter[OIII]/H β ), and construct one-dimensional spatial pro-files for the [OIII] fluxes and [OIII]/H β ratios throughthe nucleus along the ionization axis and perpendicu-lar to it within pseudo slits with a width correspond-ing to 0 . (cid:48)(cid:48)
05 ( ≈ −
300 pc at the galaxies).These one-dimensional profiles are shown in Figure 12. The [OIII]profiles are only shown along the ionization axis.4.4.
Extent of the ENLR
We have measured the total extent of the ionized gasemission in the [OIII] images along the ionization axis,up to an emission level corresponding to the 3 σ sky con-tour. The σ sky value of the [OIII] image is listed in thelast column of Table 5. The orientation of the ionizationaxis – also listed under PA in Table 5 – was adopted asthat corresponding to the longest axis of symmetry ofthe excitation maps in Figures 3-11. Figure 13 shows theadopted extents over the [OIII] images for each galaxy.The uncertainty in this extent was estimated as the dif-ference between the values measured using the contoursat 2 σ sky and 4 σ sky .The radius of the ENLR – which we hereafter call R maj – was adopted to be half the value in kpc cor-responding to the above angular extents (followingSchmitt et al. 2003a), with an uncertainty propagatedfrom those in the angular extent and correspondinggalaxy scale. The R maj values are listed in Table 5 andrange from 4 to 19 kpc.4.5. Emission-line luminosities and gas masses
The total emission-line luminosities L[OIII] andL([NII]+H α ) as derived from the images were obtainedby integrating the flux values above 3 σ sky using thedistances listed in Table 1. The resulting values arelisted in Table 5, where L[OIII] λ was obtained fromthe images after multiplying them by 0.75 in order toeliminate the contribution of the [OIII] λ Total ionized gas masses
We have estimated the total ionized gas mass of theENLR from L([NII]+H α ), for case B recombination (Os-terbrock & Ferland 2006) following Peterson (1997).The total luminosity of H β , originating from cloudswithin a total volume V c , is L ( Hβ ) = n e n p α eff Hβ hν Hβ V c ,where α eff Hβ and ν Hβ are the effective recombinationcoefficient and the rest frequency of H β , respectively,and n e and n p are the number densities of electronsand protons. We consider completely ionized hydro-gen clouds, thus n e ≈ n p . The H α luminosity can be0 Storchi-Bergmann et al. ( " ) log(F775W) (erg cm s ) log([OIII]) (erg cm s )21012345 (") ( " ) [OIII]/([NII]+H ) 21012345 (") log([NII]+H ) (erg cm s ) -17.8-17.1-16.5-15.8-15.1 -18.2-17.5-16.9-16.3-15.6-18.4-17.5-16.7-15.9-15.00.82.03.24.55.7 Figure 3.
SDSS-J082313.50+313203.7 (target 1). Top left: continuum image; top right: [OIII] narrow-band image; bottomright: H α +[NII] narrow-band image; bottom left: excitation map. Contours in the two top panels and in the bottom rightpanel range from 3 σ sky to the maximum value. In the bottom left panel, the contour separates the regions of highest excitation(inside the contour), corresponding to [OIII]/H β = 5 .
25 and [OIII]/([NII]+H α )=1.3. North is up and East to the left in allpanels. In this galaxy, the continuum image has contamination from the [OIII] emission lines, and the detached structure tothe south-east may be a gas cloud appearing in the continuum due to this contamination (see text). written as L ( Hα ) = ( j Hα /j hβ ) L ( Hβ ), where j Hα /j hβ is the ratio between H α and H β emissivities. Assum-ing L(H α )= η L([NII+H α ]) and the same density for allclouds – n p m p , where m p is the proton mass, the totalionized mass is M ENLR = ( n p m p ) V c . Using equationsabove: M ENLR = m p η L ([ N II ]+ Hα ) n e ( j Hα /j hβ ) α effHβ hν Hβ (2)= 0 . η L ([ N II ]+ Hα ) M (cid:12) , (3)where L ([ N II ]+ Hα ) is in units of 10 erg s − . As wedo not have resolved density values from our images, wehave adopted a constant density of 100 cm − thoughoutthe ENLR as a compromise. The value of the densityshould be higher in the inner kpc or so (thus we areoverestimating the mass for this region) but should belower for the outer kpcs (thus we are underestimatingthe mass for the external regions). The values for the emission coefficient and recombination ratio have beenobtained from Osterbrock & Ferland (2006) for n e =100 cm − and T= 10 K . The resulting M ENLR massesare shown in the 9 th column of Table 5.4.5.2. Ionized gas mass profiles
We can use the ionized gas masses per pixel calcu-lated as above to obtain the ionized gas surface massdensities. As we have adopted a constant gas density,the corresponding maps are very similar to those of theH α +[NII]. We show instead a one-dimensional profile forthe mass distribution binning the data within square re-gions corresponding to 1 kpc at the galaxies. We thenbuilt a cummulative mass profile along the ionizationaxis by adding all masses perpendicularly to this axiswithin each kpc. These one-dimensional cummulativemass distributions, shown in Fig. 14, present a “centralpeak” that reaches values of 10 . M (cid:12) in the inner kpcand decrease down to 10 M (cid:12) at about 5 kpc from thenucleus for targets 4, 5, 7 and 9, that show an approxi- arrow-Line Region of QSO2s ( " ) log(FR647M) (erg cm s ) log([OIII]) (erg cm s )54321012 (") ( " ) [OIII]/([NII]+H ) 54321012 (") log([NII]+H ) (erg cm s ) -17.3-16.7-16.2-15.6-15.0 -17.8-17.2-16.6-15.9-15.3-17.9-17.2-16.5-15.7-15.00.51.42.43.34.3 Figure 4.
As in Fig. 3 for SDSS-J084135.04+010156.3 (target 2). mately symmetric mass distribution relative to the nu-cleus. Assymmetric (more extended to one side) andmore extended mass distributions are observed for theremaining objects, whose values of cummulative masseswithin each kpc along the ionization axis range from ∼ M (cid:12) to ∼ M (cid:12) at distances between 10 and 20 kpc.Targets 1, 6 and 8 show the most extended mass distri-butions. All objects with assymetric mass profiles cor-respond to galaxies with signatures of interactions, theonly exception being target 9 that, although seeming tohave a symmetric mass profile, shows a companion. DISCUSSIONOne recent development in the study of AGN was therealization that AGN flicker on scales ≤ yr (e.g. No-vak et al. 2011; Hickox et al. 2014), what is supportedby observations of even our home galaxy by the discov-ery of the “Fermi bubbles” in the vicinity of the galacticcenter (Su et al. 2010). Further evidence of this pastactivity of the galaxy center has also been seen in theform of a “fossil imprint” of a past powerful flare (1–3 Myr ago) on the Magellanic Stream at a distance of50–100 kpc from the galactic center (Bland-Hawthornet al. 2013). In addition, two absorbing structures withvelocities of −
235 and +250 km s − in the light of back- ground quasars have been recently observed between thegalactic nucleus and these ionized regions of the Mag-ellanic Stream (Fox et al. 2015), in line with an originin the front and back walls of a bipolar outflow due thispast nuclear activity.In light of this known intermitency of the nuclear ac-tivity in galaxies, it is somewhat unexpected that theline emission from the ENLR is as continuous as ob-served in our sample. Eventual patchiness can be at-tributed to irregularites in the mass distribution, butthere are no clear signatures of light echoes. An ex-planation for this has been previously proposed (Cren-shaw et al. 2003; Sharp & Bland-Hawthorn 2010): dueto the low gas density of the most external regions of theENLR ( ≤ − ), the recombination time becomes ofthe order of the flicker time of ≈ − yr. In any case, atime-dependent analysis of the AGN photoionisation issomething we plan to do in the near future (e.g. Bland-Hawthorn et al. 2013) using resolved spectroscopy, thatis being acquired for the galaxies of our sample.5.1. The Ionization Cones
Inspection of Figs. 3-11 shows that most narrow-bandimages present an elongated morphology indicating col-limation of the ionizing radiation. The exceptions are2
Storchi-Bergmann et al. ( " ) log(F775W) (ergcm s ) log([OIII]) (ergcm s )1.00.50.00.51.0 (") ( " ) [OIII]/([NII]+H ) 1.00.50.00.51.0 (") log([NII]+H ) (ergcm s ) -17.5-16.8-16.1-15.4-14.7 -18.1-17.4-16.8-16.2-15.6-18.2-17.4-16.6-15.8-15.00.11.63.14.56.0 Figure 5.
As in Fig. 3 for SDSS-J085829.58+441734.7 (target 3). In this galaxy, the continuum image has contamination fromthe [OIII] emission lines, but the secondary nucleus seen in the continuum image is not due to this contamination (see text). targets 3 and 5, that show a more compact gas distri-bution. One possibility is that the orientation of theionization axis is closer to “face-on” in these cases. Inthe case of target 5 there is evidence that this is in-deed the case. This target shows the broader base inthe H α +[NII] of the SDSS spectrum. Although we fa-vored a fit without a broad H α component, the datais also consistent with the presence of such componentwithin the fitting uncertainties, in which case the obser-vation of a Seyfert 1 nucleus would be consistent with amore pole-one visualization of the AGN, explaining the“rounder” ENLR in this galaxy. A “pole-on” view of theENLR and the possible associated outflow is also con-sistent with the high value of v ∼ − , listedin Table 4 for this galaxy, possibly due to a high outflowvelocity component, as expected if the direction of theoutflow is close to the line-of-sight.The excitation maps in Figs. 3-11 reveal ionization bi-cones in 6 of our 9 targets, namely targets 1, 2, 4,6, 8 and 9, with the highest excitation levels – with[OIII]/(H α +[NII]) in the range ≈ α +[NII]) ≤
1– shown as violet in the figures. Lower excitation is also observed surrounding the apex of the cones, perpendic-ularly to the ionization axis, supporting an obscurationof the nucleus along our line of sight due to a collimat-ing structure – presumably the dusty torus postulatedin the Unified Model (Antonucci 1993).Our results thus do not support the disappearenceof the torus for high-luminosity AGN as postulated bysome studies (e.g. Elitzur et al. 2014) at least up to lumi-nosities of L[OIII]= 2 . × erg s − (the highes AGNluminosity in our sample). Although 2 of our targetsshow a “rounder” morphology for the ENLR, this canbe attributed to orientation effects, at least in the caseof target 5, as discussed above. We thus argue that thepreferred rounder morphologies for the ENLR found inprevious ground-based studies (e.g. Liu et al. 2013) maybe due to the lack of angular resolution to resolve thebiconical morphology. For example, in the case of target2 (Fig. 4), the region with low excitation perpendicularto the ionization axis, that defines the biconical mor-phology, has an angular width of only ≈ . (cid:48)(cid:48)
2, which isnot resolved by previous ground-based studies withoutadaptative optics. The spatial resolution of the HST ob-servations have been fundamental to reveal the presenceof the torus at these high luminosities. arrow-Line Region of QSO2s ( " ) log(FR647M) (erg cm s ) log([OIII]) (erg cm s )6543210123 (") ( " ) [OIII]/([NII]+H ) 6543210123 (") log([NII]+H ) (erg cm s ) -17.3-16.7-16.1-15.5-14.9 -17.9-17.0-16.2-15.4-14.5-17.9-16.9-16.0-15.1-14.20.51.11.82.53.1 Figure 6.
As in Fig. 3 for SDSS-J094521.34+173753.3 (target 4). ( " ) log(F775W) (ergcm s ) log([OIII]) (ergcm s )101 (") ( " ) [OIII]/([NII]+H ) 101 (") log([NII]+H ) (ergcm s ) -17.4-16.6-15.9-15.1-14.3 -18.3-17.4-16.5-15.6-14.7-18.2-17.2-16.3-15.3-14.40.51.11.62.22.7 Figure 7.
As in Fig. 3 for SDSS-J110952.82+423315.6 (target 5).
In order to further explore the ionization cones, wenow discuss spatial one-dimensional line-ratio profilesobtained through the ENLR. 5.1.1. [OIII]/H β spatial profiles Fig. 12 shows spatial profiles of [OIII]/H β passingthrough the nucleus along the ionization axis (blue andturquoise symbols) and perpendicularly to it (red and4 Storchi-Bergmann et al. ( " ) log(F775W) (erg cm s ) log([OIII]) (erg cm s )21012 (") ( " ) [OIII]/([NII]+H ) 21012 (") log([NII]+H ) (erg cm s ) -17.5-17.0-16.4-15.9-15.3 -18.2-17.6-17.0-16.4-15.7-18.3-17.5-16.7-15.9-15.10.51.62.63.74.7 Figure 8.
As in Fig. 3 for SDSS-J113710.77+573158.7 (target 6). In this galaxy, the continuum image has contaminationfrom the [OIII] emission lines, but the detached structure to the north-west may be a companion as it does not appear in theemission-line images (see text). orange symbols) in logarithmic units. Also shown arethe corresponding spatial profiles of the [OIII] fluxesalong the ionization axis (black and grey lines). Atrend can readily be seen for the line ratios: whilealong the ionization axis (blue symbols) the values arethe highest, in the range 0.9 ≤ log([OIII])/H β ≤ ≤ [OIII]/H β ≤
16, along the per-pendicular direction (the torus) they are lower, inthe range 0.5 ≤ log([OIII])/H β ≤ ≤ [OIII]/H β ≤
7. Although we defer a more in-depth analysis of the ex-citation structure of the ENLR to future studies via re-solved spectroscopy of our targets, preliminary resultscan be obtained by comparing the typical line ratioslisted above for the ionization cone and the values per-pendicular to it with those of photoionization models.We have used as reference the work of Kreimeyer &Veilleux (2013), and their photonization models con-structed to reproduce the line ratios of the very extendedemission nebula around the quasar MR 2251-178, that arrow-Line Region of QSO2s ( " ) log(F775W) (ergcm s ) log([OIII]) (ergcm s )1.00.50.00.51.0 (") ( " ) [OIII]/([NII]+H ) 1.00.50.00.51.0 (") log([NII]+H ) (ergcm s ) -17.7-17.0-16.3-15.6-14.9 -18.2-17.4-16.7-15.9-15.2-18.3-17.4-16.4-15.5-14.60.41.52.63.74.8 Figure 9.
As in Fig. 3 for SDSS-J123006.79+394319.3 (target 7). In this galaxy, the continuum image has contamination fromthe [OIII] emission lines, and the elongation to the south-east may be due to this contamination (see text). ( " ) log(FR647M) (erg cm s ) log([OIII]) (erg cm s )765432101234 (") ( " ) [OIII]/([NII]+H ) 765432101234 (") log([NII]+H ) (erg cm s ) -17.4-16.9-16.5-16.0-15.6 -17.9-17.3-16.7-16.1-15.5-17.9-17.2-16.6-15.9-15.20.20.81.52.12.8 Figure 10.
As in Fig. 3 for SDSS-J135251.21+654113.2 (target 8). shows an ionization cone extending to ≈
90 kpc from the nucleus. Adopting the [NII]/H α ratios of our tar-6 Storchi-Bergmann et al. ( " ) log(FR647M) (erg cm s ) log([OIII]) (erg cm s )321012 (") ( " ) [OIII]/([NII]+H ) 321012 (") log([NII]+H ) (erg cm s ) -17.3-16.8-16.2-15.7-15.1 -17.9-17.2-16.4-15.6-14.8-17.8-17.0-16.2-15.4-14.60.51.62.83.95.0 Figure 11.
As in Fig. 3 for SDSS-J155019.95+243238.7 (target 9). gets from the SDSS spectra – in the range 0.3–1, thus − . ≤ log([NII])/H α ) ≤ ≈ −
3, while for the line ratiosalong the ionization cone, log(U) ≈ − π R n e c ), where Q is therate of ionizing photons from the AGN, R is the dis-tance from the nucleus and n e is the electronic density,and that for the inner region of the cones and the per-pendicular direction n e and R should be similar, thedifference in U is reflecting the difference in Q. This im-plies that the rate of ionizing photons escaping along thecones is ≈
10 times the rate escaping perpendicularly tothe cone, supporting the presence of an obscuring andcollimating structure such as the torus.Besides decreasing along the direction perpendicularto the ionization axis due to dilution of the radiation bythe torus, the [OIII]/H β ratios also decreases in certainregions due to other factors. In the case of target 3,there is a sharp decrease at ≈ σ sky . One possibility is the presence ofa region of recent star formation. A similar effect – the presence of regions of recent star formation, althoughsmaller – could explain small decreases in the [OIII]/H β ratio along the ENLR in other objects. Alternatively –as we have estimated H β from H α , this decrease may bedue to reddening, decreasing the [OIII] flux but less sothe H α flux, leading to an overestimation of H β relativeto [OIII].5.1.2. The case for matter-bounded ENLR
The [OIII]/H β ratios decrease abruptely beyond thelimits of the ENLR, where the [OIII] fluxes become lowerthan 3 σ sky , while the [OIII]/H β ratios are remarkablyconstant througout the nebulae where the [OIII] is above3 σ sky as illustrated by the blue points in Fig. 12, sug-gesting a constant ionization parameter U. This can beunderstood if Q is constant and n e decreases with thedistance from the nucleus as R − . Resolved measure-ments of the gas density along the ionization cone in thenearby Seyfert 2 galaxy NGC 3281 (Storchi-Bergmannet al. 1992) show exactly this behaviour for the gas den-sity, supporting an approximately constant value of Qup to the limits of the nebulae, that we can define ascorresponding to where the [OIII] flux decreases below3 σ sky . This result suggests that the ENLR in all QSO2s of this study are matter bounded. arrow-Line Region of QSO2s
30 20 10 0 10181716151413 l og ([ O III])
J082313
J084135
J085829 l og ([ O III])
J094521
J110952
30 20 10 0 10 20
J113710
10 5 0 5 10 (kpc) l og ([ O III])
J123006
20 10 0 10 20 (kpc)
J135251 (kpc)
J155019 l og ([ O III] / H ) l og ([ O III] / H ) l og ([ O III] / H ) Figure 12.
Spatial profiles of: (1) [OIII] flux (black and grey lines); (2) [OIII]/H β ratio (blue and torquoise symbols) alongthe ionization axis (blue line in the inserts at the left corner of the panels); (3) [OIII]/H β ratio (red and orange symbols) alongthe direction perpendicular to the ionization axis (red line in the inserts). Black and gray lines correspond to [OIII] flux valuesabove 3 σ sky and between 3 σ sky and 1 σ sky , respectively. Similarly, for line fluxes above 3 σ sky , [OIII]/H β line ratios are shownin blue and red, and between 3 σ sky and 1 σ sky , in turquoise and orange. Typical error bars are shown close to the right verticalaxes. Circles represent line ratios for which η could be obtained, triangles represent cases in which an average value was used.Units are logarithmic, shown in the left axis for the [OIII] fluxes and in the right axis for the line ratios. The abrupt decrease in the [OIII]/H β line ratios at theborders of the ENLR was also pointed out by Liu et al.(2013) in their GMOS-IFU data, arguing that this mayimply that the ENLR is matter-bounded by the lack ofgas at larger distances, including being bounded by theextent of the disk of the galaxy. This limit – the galaxyradius – should apply if there are no external sources ofgas, but it could increase in the cases of availability of gas at larger distances, as seems to be the case in oursample.Further evidence that the ENLR of the QSOs of thepresent study are matter bounded are present in at leastthree targets of our sample: (1) in QSO 1 the ionizingradiation escapes the host galaxy and reaches a cloudto the south-east (Figs 3 and 12) ionizing the gas there.This implies that the apparent decrease in the gas exci-tation between the galaxy and the cloud is not due to8 Storchi-Bergmann et al. ( " ) J082313.50+313203.7
J084135.04+010156.3
J085829.58+441734.7 ( " ) J094521.34+173753.3
J110952.82+423315.6
J113710.77+573158.7 (") ( " ) J123006.79+394319.3 (")
J135251.21+654113.2 (")
J155019.95+243238.7 -18.4-17.5-16.7-15.8-15.0 -17.9-17.2-16.5-15.7-15.0 -18.2-17.4-16.6-15.8-15.0-17.9-16.9-16.0-15.1-14.1 -18.2-17.2-16.3-15.3-14.4 -18.3-17.5-16.7-15.9-15.1-18.3-17.4-16.4-15.5-14.6 -17.9-17.2-16.5-15.9-15.2 -17.8-17.0-16.2-15.4-14.5
Figure 13.
The adopted extent of the ENLR is shown with a red arrow in the [OIII] images of the galaxies. R maj was adoptedas the linear size corresponding at the galaxy to half the extent of the red arrow. The orientation chosen to measure the extentcorresponds to the approximate ionization axis of the bicone as observed in the excitation maps of the Figures 3-11 and listedas P.A. in Table 5. The contour values range from 3 σ sky to F max . the lack of photons, but to the lack of matter; (2) inQSO 2 the nuclear radiation ionizes gas up to ∼
10 kpcto the west (right in Figs. 4 and 12), by only to ∼ The incidence of galaxy interactions arrow-Line Region of QSO2s Table 5.
Measurements from the HST images circ
L[OIII] λ PA θ R maj L(H α +[NII]) M ENLR σ sky, [ OIII ] (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)1 J082313 4.57 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Note — (1) identification number in the paper; (2) galaxy name (3) [OIII] flux measured in the HST image within anaperture of 3 (cid:48)(cid:48) (in units 10 − erg s − ); (4) total [OIII] λ σ sky (in units 10 erg s − );(5) PA of ionization axis (in ◦ ); (6) ENLR angular extent (in (cid:48)(cid:48) ); (7) ENLR radius (in kpc); (8) H α +[NII] total luminosity,integrated above 3 σ sky (in units 10 erg s − ); (9) total ionized gas mass (in units 10 M (cid:12) ); (10) standard deviation ofthe sky value in the [OIII] image in units of 10 − erg cm − s − .
30 20 10 0 10 20 x (kpc) l o g ( M g a s )( M )
123 456 789
Figure 14.
Cummulative gas mass profiles obtained by binning the masses into 1 kpc at the galaxies, rotating the surfacemass density distributions so that the ionization axis is oriented along the x-axis and summing the mass contributions alongthe perpendicular direction within each kpc. The continuum images of the QSO hosts of our sampleshow signatures of interactions in 6 out of our 9 QSOs.In target 1 (Fig. 3), there is a structure with angularsize approximately 1 (cid:48)(cid:48) × (cid:48)(cid:48) in the continuum image atabout 15 kpc to the southeast of the AGN, but, due tothe large contamination from gas emission in the contin-uum image of this galaxy (see Sec. 4.2.1) we favor thatthis structure is a gas cloud only appering in the con-tinuum image because of the [OIII] emission-line con-tribution. In target 2 (Fig. 4), the QSO host is in clearinteraction with a big companion closer than ≈
10 kpc tothe southwest. In target 3 (Fig. 5), the continuum image shows what seems to be a secondary nucleus ≈ . (cid:48)(cid:48) ≈
20 kpc tothe northwest, with the QSO host showing an asymetricmorphology elongated in this direction, that could be aresult of an interaction. Target 8 (Fig. 10) seems to be ina late stage of a merger with what seems to be remains0
Storchi-Bergmann et al. of a galaxy ≈ ≈ . erg s − ) and galaxy mergers, assuggested by previous studies (e.g. Treister et al. 2012;Glikman et al. 2015).5.3. Relation between R maj and L[OIII]
The analysis of the [OIII] images from the HST NLRsnapshot survey of Schmitt et al. (2003a,b) revealed arelation between the NLR (or ENLR) extent R
NLR andthe AGN luminosity L([OIII]. For a sample of 60 AGNwith L([OIII]) < ergs s − , they have found the rela-tion R NLR ∝ L[OIII] . . This was interpreted as dueto the fact that the ionization parameter is not constantalong the NLR (in which case the slope should be 0.5),and that most of the [OIII] emission comes from a lowdensity region.A constant ionization parameter had been suggestedby previous observations of the NLR of a sample ofQSO’s that implied a slope of ∼ α =0.22. This has been interpreted as indicating that theNLR size is limited by the density and ionization stateof the NLR gas rather than the availability of ionizingphotons.With our measurements we can now revisit the rela-tion between the ENLR extent and the [OIII] luminos-ity. Combining our measurements of R maj with thoseof lower redshift and luminosity targets from Bennertet al. (2002), Schmitt et al. (2003b) and Fischer et al.(2018) we now span a much larger luminosity range –39 ≤ log(L[OIII]) ≤ − – from whichwe can derive a new relation. Although this relation has been recently investigated and expanded to higher lumi-nosities by other authors (Greene et al. 2011; Liu et al.2013; Hainline et al. 2013) as discussed above, thesestudies used ground-based heterogeneous data, that in-clude long-slit spectroscopy and images obtained by dif-ferent authors wiht different instruments. At the largestluminosities (and distances) of QSOs, the derived diam-eters in ground-based studies may be of the order ofthose of the PSFs, and thus carry large uncertainties.The higher luminosity data we have put together toderive a new relation were all obtained with HST, thuswith at least 10 times better angular resolution thanthose from ground-based studies. This new relation isshown in Fig. 15, where the colors separate AGN type1 (blue) and type 2 (orange), and our sample is repre-sented as stars. The QSOs in our sample that appear tobe in interaction are highlighted with black contours.Fig. 15 shows that R maj continues to grow with L [ OIII ] until the maximum luminosity of our sample –10 . erg s − . Linear least square regressions were ap-plied to the data using the function curve fit – availablein the Scipy library – which also returns a covariancematrix whose diagonal value gives the standard devia-tion of the parameters fitted. We have obtained threeregressions in order to verify also if we could detect anyvariation according to the AGN type. The first wasrestricted to type 1 AGN, returning the relation:log( R maj ) = (0 . ± .
05) log L [ OIII ] λ − . ± . , (4)shown as a blue line in Fig. 15. The second was re-stricted to type 2 AGN:log( R maj ) = (0 . ± .
03) log L [ OIII ] λ − . ± . , (5)shown as an orange line in Fig. 15. And finally, usingall AGNs we obtain:log( R maj ) = (0 . ± .
03) log L [ OIII ] λ − . ± . , (6)shown as a green line in Fig. 15.These regressions indicate that the dependency ofthe extent of the ENLR on L[OIII] is somewhat lesssteep in AGN type 2 than in AGN type 1, although,as there are only a few type 1 AGN with luminositiesL[OIII]] > . erg s − , this result is not very robust forhigh luminosities. The last fit, which uses all the data,returns a power of 0 . ± .
03, which is in agreementwith the other two fits within the uncertainties, suggest-ing that there is no significant difference of this relationfor the two types of AGN. arrow-Line Region of QSO2s R maj values change,respectively, from 18.8 kpc to 9.2 kpc, from 8.4 kpc to3.4 kpc, from 17.1 kpc to 14.0 kpc and from 6.0 kpc to3.9 kpc. A new fit, considering these new values resultin: log( R maj ) = (0 . ± .
03) log L [ OIII ] λ − . ± . , (7)thus consistent with the previous fit within the uncer-tainties. We prefer to keep the previous fit because de-tailed imaging as we have in the present study is notalways available, and in most cases of similar studies itwill not be obvious if the extended gas emission is dueto interactions or not.
39 40 41 42 43 log(L[OIII] ) (erg s ) l og ( R m a j )( p c ) type1: Schmitt etal. (2003)type2: Schmitt etal. (2003)type1: Bennert etal. (2002)type2: Fischer etal. (2017)type2: Thispapermerger: Thispaper type1: R maj L type2: R maj L all: R maj L t x Figure 15.
Relation between the extent of the ENLR R maj and log(L[OIII]), using data of three samples besides ours(stars): Schmitt et al. (2003b) (diamonds), Bennert et al.(2002) (circles) and Fischer et al. (2018) (squares), distin-guishing type 1 (blue) and type 2 (orange) AGN, togetherwith the corresponding fits (see text). The AGN in our sam-ple that appear to be in interaction are highlighted with blackcontours.
The relation we have obtained for the whole sampleis approximately the R maj ∝ L[OIII] / λ previously ob-tained by Bennert et al. (2002) for type 1 quasars at sim-ilar distances to those of our sources (0.1 ≤ z ≤ . erg s − . This also seems to imply that the ionization parameter U may beconstant, suggesting that the gas density n e decreaseswith distance from the nucleus (R) as ∼ R − . This ra-dial dependence for the gas density has been previouslyobserved in NGC 3281 (Storchi-Bergmann et al. 1992),as already pointed out, and is close to the ∼ R − . de-pendence also found for the NLR of NGC 4151 (Kraemeret al. 2000), but is distinct for that observed in othercases in which the kinematics is clearly disturbed dueto outflows (e.g. Revalski et al. 2018). It may apply,though, to most of the ENLR, as the outflows seem tobe restricted to the inner kpc or so (Fischer et al. 2018).As pointed out above, Netzer et al. (2004) have ar-gued for a limit in the size of the gaseous region ionizedby the AGN corresponding to the extent of the galaxy,what makes sense if there is no gas beyond this limit.However, if there is gas spread beyond the galaxy asa consequence of outflows, interactions or mergers, theextent of the ENLR could still increase with the lumi-nosity, provided that there are enough ionizing photonsreaching these outer regions, what seems to be the caseof our sample. Our results show that, as the AGN lumi-nosity increases, the extent the ionized region continuesto increase if there is gas to be ionized.ENLRs extending even beyond those studied herehave been reported in the literature, as is the case of thequasar MR 2251-178 (Kreimeyer & Veilleux 2013), dis-cussed in Sec. 5.1.1, in which the ENLR reaches ≈
90 kpcfrom the nucleus. In addition, the authors conclude thatthe ENLR is still matter-bounded, as they estimate that65%-95% of the quasar ionizing radiation makes its wayout of the system. As pointed out by these authors, thisfinding highlights the importance that quasar radiativefeedback may have on the intergalactic medium and theneed to for more similar studies.In a long-slit spectroscopic study of a similar sampleto ours, of 12 nearby (z ∼ .
1) QSOs, Sun et al. (2017)report, for 7 galaxies of their sample, NLR extents rang-ing from 13 to 19 kpc, overlaping our largest R maj valuesand reaching the same maximum value. But puttingtogether their data with those from previous works oftheir group, they argue that the relation between thesize of the NLR and the AGN luminosity flattens out atR
NLR ∼
10 kpc. We have not found this flattening inour data.Alternatively, it may be argued that our sample issomehow “special” due to the high occurrence of inter-actions, that provide gas to be ionized beyond the extentof the galaxies, even though they were selected solely onthe basis of distance and luminosity. In addition, it isnot obvious, at the corresponding galaxy distances or2
Storchi-Bergmann et al. even larger, how to recognize if the gas origin is previ-ous or on-going interactions.5.4.
Total gas masses and surface mass densities
The total masses of ionized gas are very similar amongthe galaxies, ranging from 0 . × M (cid:12) to 2 × M (cid:12) .These masses are 2–3 orders of magnitude larger thanthose we have been obtaining in integral-field spectro-scopic studies of nearby Seyfert galaxies (about 2 ordersof magnitude less luminous) in the optical and near-infrared, although in these nearby cases the field-of-viewcorrespond to smaller regions at the galaxies, of just afew kpcs (e.g Riffel et al. 2015, 2018).On the other hand, the ionized gas masses of our sam-ple are similar to those obtained by Harrison et al. (2014)for a sample of 16 type 2 AGN of similar luminositiesand redshifts to those of our targets and by Tadhunteret al. (2014); Couto et al. (2017) for radio galaxies atsimilar redshifts, consistent with then having a similarorigin, argued to be gas-rich mergers in Tadhunter et al.(2014). 5.5. Implications for Feedback
Our narrow-band images show extents for the ENLRranging from 4 kpc to 19 kpc. These can be comparedto those of Fischer et al. (2018), for a similar sample ofactive galaxies also selected from Reyes et al. (2008), atsomewhat lower redshifts and luminosities. Fischer et al.(2018) have also obtained HST [OIII] narrow-band im-ages, as in our study, but instead of H α +[NII] images,obtained long-slit STIS spectroscopy in order to investi-gate the [OIII] gas kinematics. The elongated morpholo-gies of the ENLR are similar to those of our targets, butwith somewhat lower R maj values, close to about 5 kpc.The long-slit spectroscopy of Fischer et al. (2018), al-though obtained only along the ionization axis, revealeda disturbed kinematics for the gas suggesting outflows– e.g. high velocity dispersion, double of triple com-ponents – only within the inner kpc. Beyond this re-gion, the gas kinematics, instead of outflows, revealed inmost cases apparent rotation in the galaxy plane, withblueshifts observed to one side and redshifts to the otherside of the nucleus. They measured the distance fromthe nucleus within which they have observed the signa-ture of outflows, calling it R out , finding that, on average,R out /R maj = 0 .
22, where R maj is their adopted maxi-mum extent of the ionized gas region (we have adoptedthe same notation as in their study). Typical outflowvelocities are in the range 250–500 km s − .The fact that the outflow is more compact than theionized gas region in active galaxies is a common resultalso of previous studies of local AGN, such as those of the AGNIFS (AGN Integral Field Spectroscopy) group:Schnorr-M¨uller et al. (2014b), Lena et al. (2015), Rif-fel et al. (2015), Couto et al. (2017), as has also beenpointed out by Sun et al. (2017), Fischer et al. (2017)and Revalski et al. (2018). This can be interpreted asdue to the fact that most of the extended [OIII] emissioncomes from gas that is not outflowing, but from gas thatis usually rotating in the plane of the host galaxy.Although we do not have yet observed the gas kine-matics of our sample, we can use the information wehave gathered via the images and SDSS spectroscopy toestimate the mass outflow rate and the range of powersof the presumed outflows in our objects. In order to dothis, we adopted the following assumptions.(i) Regarding the extent of the outflow: our sampleincludes AGNs with higher luminosities than those ofFischer et al. (2018), and it may well be that the out-flow extends to larger distances than R out . We thusconsidered two limits for the extents: an upper limit ofR maj (which we called model a ) and a lower limit of0.2 R maj (which we called model b ).(ii) Regarding the outflowing gas mass: we have as-sumed that all gas mass within the considered radius( R maj or R out = 0 . R maj ) is outflowing.(iii) Regarding the velocity of the outflow: as the onlyspectroscopic data we have so far is the SDSS-III spec-tra, we have used the [OIII] λ v = W / . W is the width of the line profile that comprises80% of the line flux, in the profile wings, thus probingthe highest velocities, most probably due to the out-flows, and the factor 1.3 is adopted to consider projec-tion effects (Sun et al. 2017). This velocity is listed inthe last column of Table 4, and we have adopted it asthe velocity of the outflow. In addition, we use the ap-proximation that all spaxels have the same velocity, aswe do not have spatial information on the kinematics sofar.We have used three methods to calculate the massoutflow-rate, and have used this rate to calculate alsothe power of the outflow in order to compare it withthe QSO bolometric luminosity L Bol . L
Bol was calcu-lated using the relation between the reddening-correctedL[OIII] and L
Bol of Trump et al. (2015). We have cor-rected L[OIII] for reddening according to equation 1 ofLamastra et al. (2009), in which the average reddeningvalue was obtained from the SDSS-III spectra H α /H β ratio assuming an intrinsic value of 3.0 (Osterbrock &Ferland 2006). In order to compare the power of theoutflows with L Bol , we have also corrected for redden-ing, in the calculations below, L(H β ) (assumed to havethe same correction as [OIII] due to the proximity in arrow-Line Region of QSO2s α and M ENLR . Method 1 –
The mass outflow rate was estimated fromthe ratio between the reddening-corrected mass of theENLR M and the time t = R/v it took the gas at theborder of the ENLR to reach the maximum distance Rfrom the nucleus (R maj for model a and 0.2 R maj formodel b).The power of the outflow dE/dt was calculated as: dEdt = 0 . M v t (8)where the outflow velocity is assumed constant at v = v , and M is the value of M ENLR from Table 5corrected for reddening.
Method 2–
We use the same equation above to calcu-late the power of the outflow but, instead of calculatingthe mass-outflow rate as
M/t , we use ( dM ) / ( dt ) calcu-lated as: dMdt = m p n e v f A (9)where n e is the electronic density assumed to be100 cm − (as a typical value for extended emission-line regions, but see discussion below for the effect of apossibly higher density), m p is the mass of the proton, f is the filling factor, v is the gas velocity, adopted asabove ( v ), and A is the cross-section of the outflow.We have then used the H α luminosity (obtained as η L([NII]+H α ), integrated within the radius considered,R maj or R out (for models a and b , respectively) andcorrected for reddening to derive f : L ( Hα ) = j Hα π f V (10)where j Hα is the H α emissivity as defined in Osterbrock& Ferland (2006), and V is the volume of the emittingregion. We have first adopted as the geometry of theoutflow a cone with base area A and height R , assumedto be R maj in model a and R out in model b . But we havenevertheless concluded that the geometry of the outflowcancels out when you calculate the filling factor, thus anoutflow through a spherical surface with radius R wouldresult in the same expression, as follows:˙ M = 3 m p v L ( Hα ) (cid:16) j Hα j Hβ (cid:17) n e α eff Hβ hν Hβ R , (11) with the emissivities j and recombination coefficient α effHβ values obtained from Osterbrock & Ferland (2006)for n e = 100 cm − and T = 10000 K . We note that,in this method, the mass-outflow rate results inverselyproportional to the gas density. Thus, if the the gasdensity is, for example, 5 times larger (500 cm − ), themass outflow rate will be 5 times smaller. But there isno assuption regarding the filling factor, which is calcu-lated from the data. Method 3 –
We have also estimated what can be con-sidered a “maximum” mass-outflow rate, using eq. 4 ofSun et al. (2017), that assumes a density, filling factorand a spherical geometry for the flow: dMdt = m p n e v π R (12)where Sun et al. (2017) adopt a constant n e = 0 . − ,which is equivalent, for example, to the product of adensity of n e = 100 cm − times a filling factor of 0.005(a typical value, similar to what we have obtained fromprevious studies), or to a somewhat larger density of n e = 500 cm − times a smaller filling factor of 0.001.The power of the outflow is obtained using againthe equation 5.5 replacing M/t by dM/dt calculated asabove.The resulting mass outflow rates and powers areshown in Table 6 for the three methods above for eachQSO. In the first line we show the power calculated as-suming that the extent of the outflow is R maj and in thesecond line assuming it is R out . We have also includedin the table an estimate of the ratio between the out-flow powers and the bolometric luminosity of the AGN˙ E/L bol , where L bol was calculated from the L[OIII] valuecorrected for reddening using the bolometric correctiongiven by equation 9 of Trump et al. (2015).The values of the mass outflow rates ˙ M range from ∼ ∼
200 M (cid:12) yr − for the first method and from ∼
15 to ∼
600 M (cid:12) yr − for the second method. A higher densityof 500 cm − will result in a mass outflow rate 5 timeslower, thus from ∼ ∼
40 M (cid:12) yr − for the first methodand ∼ ∼
100 M (cid:12) yr − for the second method. Thesevalues range from being of the same order up to twoorders of magnitude larger than those of our previousstudies of closer and less luminous AGN (e.g. Storchi-Bergmann et al. 2010; Riffel et al. 2013; Barbosa et al.2014). In the work of Revalski et al. (2018), in which theauthors calculate the mass-outflow rate as a function ofdistance from the nucleus in the NLR of the Seyfert 2galaxy Mrk 573, the highest value of 3 . ± . (cid:12) yr − is obtained at 210 pc from the nucleus.4 Storchi-Bergmann et al.
The corresponding estimated powers of the outflows(for methods 1 and 2) are in the range 10 . < log( ˙ E ) < . (in erg s − ), and correspond to a range of − < log( ˙ E/L
Bol ) < −
2, with only three cases in which it is larger than the0.5% threshold corresponding to a signficant impacton the host galaxy (e.g. Hopkins & Elvis 2010). These˙
E/L
Bol values are ploted as a function of L
Bol in Fig. 16in yellow and green symbols for methods 1 and 2, respec-tively. They increase between L
Bol = 0 . × erg s − and L Bol = 3 . × erg s − but then decrease forhigher luminosities.It is important to point out that our calculations referonly to ionized gas, and do not consider the possiblepresence of neutral and molecular gas. Recent studies(e.g. Fiore et al. 2017) suggest that molecular gas isthe dominant gas phase surrounding nearby AGN nuclei,with masses that can be ∼ α ) and the geometry is assumed tobe spherical, the mass outflow rates reach values 2 to3 orders or magnitude higher than those in Methods 1and 2, corresponding thousands of solar masses per year.These mass outflow rates lead to estimated outflow pow-ers that are above the 0.5% threshold for most QSOs.The ˙ E/L
Bol values for Method 3 are ploted as a func-tion of L Bol in Fig. 16 in red symbols. There seems to beno relation between these two quantities. As the massoutflow rate in this case is proportional to R , the re-sulting power is also proportional to R , and thus thosecalculated using R maj as the extent of the outflows are5 times higher than those calculated using R out , whilethis does not happen in the cases of Methods 1 and 2. CONCLUSIONSWe have presented HST continuum, [OIII] λ α +[NII] narrow-band images, as well as [OIII]/H β ex-citation maps and have obtained the extent, morphol-ogy and masses of the ionized gas in the ENLR of 9type 2 AGN classified as QSO2 due to their high lumi-nosities (logL[OIII] > .
5, L in ergs s − ). Under theassumption that the ionized gas is outflowing, we havealso estimated the corresponding mass outflow rates andpowers. Our main findings are: • The ENLR is spatially resolved in all galaxies andits extent ranges from 4 to 19 kpc; the largest ex-tents may be related to the fact that 6 of the 9galaxies seem to be in interaction, with gas emis- sion extending well beyond the body of the galaxyseen in the continuum maps; • The ENLR morphology is clearly elongated andbipolar in 6 of the 9 galaxies, and excitationmaps [OIII]/(H α +[NII]) show a biconical struc-ture, with the highest excitation gas along the ion-ization axis and decreasing with increasing angu-lar distance from it; the exceptional HST angularresolution was key to reveal this structure; • The high emission-line ratios obtained along theionization axes ([OIII]/H β ≥ β ≤
7) imply that the rate of ion-izing photons escaping along the ionization axis is ≈
10 times higher than that escaping perpendicu-larly to it; • The above result supports the presence of an ob-scuring torus that can thus survive the correspond-ing AGN luminosities, up to log(L[OIII])=43.3 (Lin ergs s − ); • We have re-visited the relation between the ex-tent of the ENLR (R maj ) and L[OIII], find-ing that R maj increases with L[OIII] over therange 39 . < log[OIII] < . − ), as:log(R maj ) = (0 . ± .
03) log(L[OIII]) − . ± . • We do not find a flattening in the above relationup to L[OIII]=43.3 (L in erg s − ), implying thatthe ENLR in luminous AGN can extend to dis-tances beyond the limit of the galaxy if there isavailability of gas, e.g. from outflows or interac-tions, as seems to be the case of most galaxies ofour sample; • We attribute the flattening of the above relationreported in previous studies to the fact that theENLR is matter bounded in most AGN hosts; theconstancy of the [OIII]/H β along the ENLR ob-served in our targets support that it is matter-bounded even in them, implying that the AGN ra-diation is still escaping to the intergalactic mediumand do so in most luminous AGN; • The ionized gas masses of the ENLR range from0.4 to 2 × M (cid:12) ; much of it may have been ac-quired in an interaction, and thus seem to be morerelated to the feeding of the AGN than to its feed-back, but this can only be concluded via resolvedspectroscopy of the ENLR to probe its resolvedkinematics what will be our next step followingthe present study; arrow-Line Region of QSO2s Table 6.
Mass outflow rates and outflow power according to three different methods.Method 1 Method 2 Method 3 / L bol ) log( ˙M) log( ˙E) log( ˙E / L bol ) log( ˙M) log( ˙E) log( ˙E / L bol )1 R maj out maj out maj out maj out maj out maj out maj out maj out maj out Note — (1): R is the adopted extent of the outflow, ˙ M is the mass outflow rate in units of M (cid:12) yr − and thepower of the outflow ˙ E is in units of erg s − . • Although we are still in the process of acquiringresolved spectroscopy of our targets, we have esti-mated the NLR outflow velocities from the widthsW of the [OIII] emission-line profiles in SDSSspectra, combined with the above ENLR massesand extents to estimate the mass outflow rates andpower of the outflows using 3 different methods, asfollows; • For the first two methods in which the ionized gasmasses are constrained by the measured H α lumi-nosity, we obtain mass outflow rates in ionized gasin the range ∼ (cid:12) yr − ; • The corresponding powers of the outflows arelarger than 0.5% L
Bol – the usually adoptedthreshold for significant impact on the host galaxy– only for 4 QSOs of the sample; on the otherhand, we point out that our calculations do nottake into account other gas phases, like neutraland molecular gas (that seems to be the dominant phase in many nearby AGN), that will lead tolarger feedback powers, if present; • Using a third method in which the gas masses arenot constrained by the H α luminosity, and the as-sumed outflow geometry is spherical, we obtainmuch larger mass-outflow rates, that reach thou-sands M (cid:12) yr − , with corresponding outflow pow-ers exceeding 0.5% L Bol for most QSOs.We plan to revisit the above calculations in a forth-coming paper using resolved spectroscopy of our targets,from which we will obtain gas densities and tempera-tures and model the ionization structure of the ENLR.We are now in the process of acquiring such data.ACKNOWLEDGEMENTSWe thank the referee, Joss Bland-Hawthorn, for thenumerous valuable suggestions that led to a much im-proved discussion of our results. This work has used6
Storchi-Bergmann et al. L bol (10 erg s ) l o g ( E / L b o l ) (1, maj)(1, out) (2, maj)(2, out) (3, maj)(3, out) Figure 16.
Ratio between the outflow powers ˙ E and the AGN bolometric luminosity L bol as a function of L bol for methods 1(yellow), 2 (green) and 3 (red) used to calculate ˙ E (see text). The filled circles correspond to an extent of the outflow of R maj while the star symbols correspond to an extent of R out = 0 . × R maj (see text). Antonucci, R. 1993, ARA&A, 31, 473Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981,PASP, 93, 5Barbosa, F. K. B., Storchi-Bergmann, T., McGregor, P.,Vale, T. B., & Rogemar Riffel, A. 2014, MNRAS, 445,2353Bennert, N., Falcke, H., Schulz, H., Wilson, A. S., & Wills,B. J. 2002, ApJL, 574, L105Bland-Hawthorn, J., Maloney, P. R., Sutherland, R. S., &Madsen, G. J. 2013, ApJ, 778, 58Capetti, A., Axon, D. J., Macchetto, F., Sparks, W. B., &Boksenberg, A. 1996, ApJ, 469, 554Ciotti, L., Ostriker, J. P., & Proga, D. 2010, ApJ, 717, 708 Couto, G. S., Storchi-Bergmann, T., & Schnorr-M¨uller, A.2017, MNRAS, 469, 1573Crenshaw, D. M., Kraemer, S. B., Schmitt, H. R., et al.2010, AJ, 139, 871Crenshaw, D. M., Kraemer, S. B., Gabel, J. R., et al. 2003,ApJ, 594, 116Elitzur, M., Ho, L. C., & Trump, J. R. 2014, MNRAS, 438,3340Elvis, M. 2000, ApJ, 545, 63Fabian, A. C. 2012, ARA&A, 50, 455Falcke, H., Wilson, A. S., & Simpson, C. 1998, ApJ, 502,199Ferrarese, L., & Ford, H. 2005, SSRv, 116, 523 arrow-Line Region of QSO2s27