Optical--to--X-ray emission in low-absorption AGN: Results from the Swift-BAT 9 month catalogue
R. V. Vasudevan, R. F. Mushotzky, L. M. Winter, A. C. Fabian
aa r X i v : . [ a s t r o - ph . H E ] J u l Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 25 October 2018 (MN L A TEX style file v2.2)
Optical–to–X-ray emission in low-absorption AGN: Results from theSwift-BAT 9 month catalogue
R. V. Vasudevan R. F. Mushotzky , L. M. Winter and A.C. Fabian Institute of Astronomy, Madingley Road, Cambridge CB3 0HA Laboratory for High Energy Astrophysics, NASA/GSFC, Greenbelt, MD 20771, USA Center for Astrophysics and Space Astronomy, University of Colorado at Boulder, 440 UCB, Boulder, CO 80309-0440, USA25 October 2018
ABSTRACT
We present simultaneous optical–to–X-ray spectral energy distributions (SEDs) from
Swift ’sX-ray and UV–optical telescopes (XRT and UVOT) for a well-selected sample of 26 low-redshift ( z < . ) active galactic nuclei (AGN) from the Swift/Burst Alert Telescope (BAT)9-month catalogue, the largest well-studied, hard X-ray selected survey of local AGN to date.Our subsample consists of AGN with low intrinsic X-ray absorption ( N H < cm − ) andminimal spectral complexity, to more accurately recover the intrinsic accretion luminosity inthese sources. We perform a correction for host galaxy contamination in all available UVOTfilter images to recover the intrinsic AGN emission, and estimate intrinsic dust extinctionfrom the resultant nuclear SEDs. Black hole mass estimates are determined from the host-galaxy 2MASS K-band bulge luminosity. Accretion rates determined from our SEDs are onaverage low (Eddington ratios λ Edd . . ) and hard X-ray bolometric corrections cluster at ∼ λ Edd < . ) objects is presented, with and without correction for extinction.Significant dust reddening is found in some objects despite the selection of low N H objects,emphasising the complex relationship between these two types of absorption. We do not finda correlation of optical–to–X-ray spectral index with Eddington ratio, regardless of the opticalreference wavelength chosen for defining the spectral index. An anti-correlation of bolomet-ric correction with black hole mass may reinforce ‘cosmic downsizing’ scenarios, since thehigher bolometric corrections at low mass would boost accretion rates in local, lower massblack holes. We also perform a basic analysis of the UVOT-derived host galaxy colours forour sample and find hosts cluster near the ‘green valley’ of the colour-magnitude diagram,but better quality images are needed for a more definitive analysis. The low accretion ratesand bolometric corrections found for this representative low-redshift sample are of particularimportance for studies of AGN accretion history. Key words: black hole physics – galaxies: active – galaxies: Seyfert
Active Galactic Nuclei (AGN) are known to emit radiation overthe whole range of available energies observable using current de-tectors. Characterising their spectral energy distribution (SED) istherefore important for understanding the different physical pro-cesses at work. The accepted paradigm behind the radiation outputof AGN is accretion onto a supermassive black hole. The emissiondirectly due to accretion emerges primarily in the optical, UV andX-ray regimes by a combination of thermal emission from an ac-cretion disc and inverse-Compton scattering of UV disc photonsby a corona above the disc. A dusty torus partially absorbs the UVphotons from the accretion disc, re-emitting them more isotropi-cally in the infrared. In this study, we aim to recover the true accre-tion luminosity in a representative sample of AGN, and use these results to identify trends between different SED parameters. How-ever, recovering the true accretion luminosity can be complex, andprevious studies have found a wide spread in the conversion fac-tors (bolometric corrections) between the X-ray luminosity and thetotal accretion luminosity. This is, in a large part, due to intrinsicvariation between objects in the whole AGN population, but is alsocomplicated by numerous systematics and biases in the samplesused.The pioneering study of Elvis et al. (1994) (hereafter E94)presented radio–to–X-ray SEDs for 47 quasars and has provided auseful template for determining bolometric corrections and iden-tifying trends in the SED shapes across a range of AGN prop-erties. Recent studies have confirmed the usefulness of the av-erage bolometric corrections and SED parameters from E94, butthe sample of quasars used was predominantly X-ray bright and c (cid:13) R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian their bolometric luminosities included the re-processed infraredemission. Richards et al. (2006) present infrared–to–X-ray SEDsfor 259 quasars selected by a combination of optical and mid-infrared colour selection criteria, and highlight the spread in SEDparameters within the sample. This large spread in SED parame-ters was known from E94, prompting refinements such as the tem-plate SED of Marconi et al. (2004) who employ the observed cor-relation between X-ray–to–optical spectral index and X-ray lumi-nosity to construct a luminosity-dependent SED, and importantly,they exclude the reprocessed infrared emission to avoid double-counting of part of the accretion energy budget. A similar tem-plate is presented by Hopkins et al. (2007) in their study on thebolometric quasar luminosity function. These SED templates al-lowed the diversity of SED shapes to be taken into account whencalculating the supermassive black hole mass density from the X-ray background and AGN luminosity functions. More recently,Vasudevan & Fabian (2007) (VF07 hereafter) presented optical–to–X-ray SEDs for a sample of AGN observed by the Far Ultra-violet Spectroscopic Explorer (
FUSE ), which yielded interestingtrends between SED shape and Eddington ratio, confirmed later byVasudevan & Fabian (2009) (VF09 hereafter) using simultaneousdata and reverberation mapping mass estimates to improve accu-racy of the bolometric luminosities and accretion rates. However,the sample of VF07 is by necessity UV-bright, as it is selectedby
FUSE and the sample of VF09 is restricted to those which arebright enough in the optical/UV for reverberation mapping to becarried out.These considerations emphasise the pressing need for a rep-resentative sample of AGN from which conclusions can be drawnabout accretion properties of the wider AGN population. One ma-jor factor influencing the selection of AGN for surveys is the natureof their absorption. It has been known for some time that X-ray andoptical surveys detect different segments of the underlying AGNpopulation (Mushotzky 2004), with X-ray surveys able to probeto higher absorbing column density. The standard physical mech-anisms for the generation of X-ray emission in AGN produce apower-law, but often AGN exhibit X-ray spectra deviating signif-icantly from this, in part due to absorption. The geometry of theabsorption in X-rays is the subject of much debate, and the com-plex spectra in some sources can be accounted for with either ‘par-tial covering’ scenarios (Gierli´nski & Done 2004) or strong reflec-tion from the accretion disc due to light bending (Miniutti & Fabian2004), and warm absorbers can also complicate our view of the in-trinsic emission (Reynolds et al. 1997, Blustin et al. 2005). Manyof the reverberation mapped AGN in VF09 showed signs of suchspectral complexity, and the authors identified the difficulties in re-gaining the true accretion luminosity in these cases.The Burst Alert Telescope (BAT) on board the
Swift satel-lite proves invaluable for addressing many of these considera-tions. Canonical levels of absorption (column densities N H < cm − ) imprint signatures on the 0.1–10 keV X-ray band, sothe very hard X-ray sensitivity of the instrument (14–195 keV) pro-vides the capability to observe AGN in a bandpass relatively unaf-fected by such absorption. The 9-month catalogue of BAT-detectedAGN (hereafter the Swift /BAT catalogue, Tueller et al. 2008) there-fore provides an unprecedented level of completeness (with respectto absorption) when surveying the AGN population as it is unbi-ased to all but the most heavily obscured sources. The study ofWinter et al. (2009) presents a comprehensive overview of the X-ray spectral properties of the 153 sources in the
Swift /BAT cata-logue and allows us to select an appropriate sample from whichwe can be confident of calculating accurate luminosities and SED
Filter Central wavelength ( ˚ A )V 5468B 4392U 3465UVW1 2600UVM2 2246UVW2 1928 Table 1.
Central frequencies of the UVOT filters. shapes. Most importantly, they present values of absorbing col-umn density N H for each source, determined from fits to the X-ray data from the literature and new fits by the authors themselves,and identify those sources in which significant spectral complexityis present. In this work, we preferentially select those with lowerabsorption and minimal spectral complexity, facilitating a morestraightforward calculation of the accretion luminosity. We assumethat for such sources, the accretion luminosity excludes the repro-cessed IR emission and is principally seen in the optical–to–X-rayregime (as in VF07 and VF09). Additionally, AGN with more X-ray absorption are expected to display higher levels of optical–UVreddening, the calculation of which depends on the precise form ofthe extinction curve and is difficult to account for when calculatingthe total luminosity. These considerations motivate our selection ofobjects with low N H .Variability in AGN can produce large changes in luminosity,over timescales from hours to years. In order to catch an accuratesnapshot of the total energy budget in an AGN at a given time, itis therefore highly desirable to use simultaneous data. The poten-tial inaccuracies in SED parameter values from non-simultaneousSED data are discussed in detail in VF07 and VF09, with thelatter following the approach of Brocksopp et al. (2006) by usingcontemporaneous optical–to–X-ray data from XMM-Newton’s PNand Optical Monitor instruments. In this work, we make use of Swift ’s own co-aligned UV–optical telescope (UVOT) and X-raytelescope (XRT) to determine simultaneous SEDs. Many of ourtarget subsample from the
Swift /BAT catalogue have simultaneousUVOT and XRT observations available in the archives; typicallythe UVOT data span six filters between 5468 ˚ A and 1928 ˚ A , provid-ing a window onto a substantial fraction of the disc emission. Thecentral frequencies of the UVOT filters are given in Table 1.The large wavelength coverage, field of view and spatial reso-lution of the UVOT data also allow a correction for the host galaxyflux to be estimated for each AGN. We employ GALFIT for thispurpose, a program for performing 2D fitting of PSFs and variousanalytical profile forms to galaxy images (Peng et al. 2002). Fora detailed discussion of the requirements of GALFIT, the readeris directed to peruse the accompanying documentation . We alsopresent basic results on the host galaxies from the GALFIT fittingprocess. The UVOT data used in this study also clarify how the de-gree of host galaxy contamination varies with wavelength (UVOTfilter).Understanding variations in the AGN SED is important, andone of the main drivers of these variations appears to be the ac-cretion rate (parameterised as the Eddington ratio, L bol /L Edd for an object with bolometric luminosity L bol and Eddington lu-minosity L Edd = 1 . × [ M BH /M ⊙ ]ergs − for a blackhole of mass M BH ). The recent Principal Component Analysis of http://users.ociw.edu/peng/work/galfit/galfit.htmlc (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN Kuraszkiewicz et al. (2008) on a sample of red 2MASS AGN SEDsidentify the Eddington ratio as the ‘eigenvector 1’ for these AGN(being responsible for most of the variation in the SED shapes).Both VF09 and VF07 highlight substantial variations in the bolo-metric output (in comparison to the X-ray output) across the ob-served range of Eddington ratios, and numerous studies explorethese themes further (Shemmer et al. 2008, Kelly et al. 2008). Ed-dington ratios require accurate estimates of the black hole massin AGN, but since the most accurate estimates from reverbera-tion mapping are only available for 35 AGN, here we employ the M BH − L bulge correlation to calculate black hole masses. The com-bination of robust determinations of the bolometric luminosity andsensible estimates of the black hole mass provide great opportu-nities for insight into the accretion process. In this work, we dis-cuss the analysis of the Swift data, the construction of SEDs, thederivation of SED parameters, the identification of correlations andtrends, and finally the host galaxy properties.
We firstly apply a cut in intrinsic column density N ( int )H , select-ing only those with log( N ( int )H ) < from the list of objects inthe 9-month BAT catalogue as presented in Winter et al. (2009).This is the suggested crossover point between ‘absorbed’ and ‘un-absorbed’ classes of AGN in their study, which also classifies theobjects in the sample into two broad categories based on X-rayspectral properties: ‘simple’ (S) and ’complex’ (C). The S classhave X-ray spectra which are best fit by a simple power law withintrinsic and galactic absorption (sometimes with a soft excess),whereas the C class display more complex X-ray spectra which arebetter fit by models such as double-power laws or partial cover-ing (these models are illustrative in that they highlight the spectralcomplexity; more detailed fits including reflection may also be pos-sible). We also eliminate any objects from the C class, as they willpresent difficulties when determining the true accretion luminosity(as discussed in VF09).In order to facilitate profile fits to optical and UV images ofthe host galaxies, we also impose a redshift cut, requiring z < . for our selection. For z < . , the angular resolution of the UVOT( ∼ . arcsec) allows separation of the nucleus from the galaxyto a physical length scale of 3 kpc, thus allowing a reasonablegalaxy–AGN separation while still yielding 54 potential objectsfor study from the sample. We then identify those for which Swift
XRT and UVOT data are available from the High Energy Astro-physics Science Archive Research Center (
HEASARC ), yielding33 objects. For estimating black hole masses, we also obtain K-band magnitudes from the Two-micron All-Sky Survey (2MASS)catalogues. The key requirement for calculating black hole massesis an estimate of the bulge luminosity. The K-band is least subjectto the effects of Galactic reddening and predominantly traces theolder stars in the bulge over the stellar populations of the galaxydisc, motivating its selection over the J and H bands also avail-able in the 2MASS catalogues. The 2MASS data were gatheredwith two ground based telescopes: the Whipple Observatory in Ari-zona, USA and the Cerro Tololo telescope at La Serena, Chile. Thelimitations introduced by ground-based observations (most impor-tantly, the level of seeing for each observation) need to be takeninto account when attempting to recover the bulge luminosity. This http://heasarc.gsfc.nasa.gov/cgi-bin/W3Browse/swift.pl is discussed at greater length in § Swift observations used.
We download the optical–UV and X-ray data for our selection fromHEASARC. When multiple observations were available, thosewith the maximum UVOT exposure time and maximum numberof UVOT filters were preferentially selected. Pipeline-processed‘level 2’
FITS files are readily available from HEASARC. The datafor the 26 sources identified in section 2 were then processed ac-cording to the procedure outlined in the following sections.
We employ the custom-built software tools designed specificallyfor processing UVOT data where possible. Each individual UVOTfilter data file in general contains a number of exposures whichwere summed using the tool
UVOTIMSUM . The
UVOTSOURCE toolwas then used to extract magnitudes from simple aperture pho-tometry. Source and background regions were created for this pur-pose, with the position of the source region being obtained fromthe NASA Extragalactic Database (NED) in the first instance, fol-lowed by fine adjustment of the source region position if necessary.The required source region size for
UVOTSOURCE is 5 arcsec, andthe background region can be any size (at the time of writing): weused background regions with radii from 5 arcsec to ∼
30 arcsec de-pending on the frequency of other foreground sources in the image.The magnitudes from
UVOTSOURCE provide a useful first-orderestimate of the nuclear flux, and were saved for a later compari-son with the more carefully determined nuclear fluxes using pointspread function (PSF) fitting.The images were then prepared for use within GALFIT. Oneof the requirements for accurately determining the nuclear flux isto have suitable PSFs available for each image. Preliminary stud-ies of the UVOT PSF in the six filters show that the PSF shapemay depend on a variety of factors including position on the de-tector, countrate of the source and the filter being used. AveragePSF full-width half-maxima range between 1.7–2.5 arcsec, whenconsidering all six filters. As discussed by Kim et al. (2008a) inthe context of Hubble Space Telescope (
HST ) images, the choice c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian
AGN redshift
Swift observation ID Total UVOT exposure time (ks) XRT exposure time (ks)UGC 06728 0.006518 00035266001 2.1 6.33MCG-06-30-15 0.007749 00035068003 3.19 8.72NGC 4593 0.009 00037587001 1.83 4.88Mrk 766 0.012929 00030846039 1.34 3.71ESO 548-G081 0.01448 00035250002 2.8 6.33Mrk 352 0.014864 00035243002 6.38 15.91NGC 7469 0.016317 00031245005 1.38 4.27NGC 5548 0.017175 00030022062 1.84 5.13WKK 1263 0.02443 00035268002 1.45 8.94ESO 490-G026 0.02485 00035256001 2.87 8.51Mrk 590 0.026385 00037590001 1.43 4.461RXS J045205.00+493248 0.029 00035281002 0.65 1.99SBS 1301+540 0.0299 00035269001 2.3 7.85Mrk 279 0.030451 00037591001 1.75 5.22MCG +04-22-042 0.032349 00035263001 3.0 9.11Ark 120 0.032713 00037593003 1.44 4.11IRAS 05589+2828 0.033 00035255001 2.68 5.913C 120 0.03301 00036369001 2.1 6.37Mrk 509 0.034397 00035469003 2.5 6.8Mrk 841 0.036422 00035468002 3.33 8.4Mrk 1018 0.042436 00035166001 1.41 4.53NGC 985 0.043143 00036530005 3.16 8.653C 390.3 0.0561 00037596001 2.25 6.46IRAS 09149-6206 0.0573 00035233002 1.87 5.04SBS 1136+594 0.0601 00035265001 3.22 9.172MASX J21140128+8204483 0.084 00035624002 1.72 5.09
Table 2.
Details of UVOT observations used for the objects in our sample. of PSF can significantly affect PSF fitting results, since a poorlychosen PSF can give profile parameters that deviate dramaticallyfrom their ‘true’ values (also see Simmons & Urry 2008 for an-other discussion of the factors affecting AGN–host-galaxy decom-position). To mitigate these problems and minimize PSF mismatch,we adopted the following approaches. The first and preferred ap-proach was to generate a unique PSF for each filter image from afew known guide stars in the image, requiring the stars selected tobe within a certain range of the count rate of the source of inter-est (as determined from simple aperture photometry with
UVOT - SOURCE ). Firstly, the UVOT utility
UVOTDETECT was used to pro-vide a list of detectable sources in each image (
UVOTSOURCE es-sentially uses the S
EXTRACTOR package for this purpose). Thesewere then cross-checked against published catalogues of knownguide star positions. The
XBROWSE tool in the
HEASOFT suite ofutilities provides access to many such catalogues, such as the
USNaval Observatory (USNO) and
HST guide star catalogues. Typi-cally, the USNO catalogue was used to obtain a list of such stars,requiring them to be within 13 arcmin of the AGN of interest toensure they were within the UVOT field of view. The guide starswere selected such that between 3-20 stars were used to generateeach PSF image. For the lowest energy V-band, this translated intoa requirement that the stars were typically within 0.1–0.2 dex ofthe target AGN count rate; for the highest energy UVW2 band, thelimits were typically within 1.5-2.0 dex of the target AGN countrate to obtain a similar number of guide stars. Finally, the regionsof the images containing these stars were then summed to forma final PSF image, using IDL code by A. A. Breeveld designedspecifically for generating PSFs from UVOT images. The resulting PSFs were used in GALFIT to model the central AGN in the galaxyprofiles.When this approach was not possible due to lack of guidestar detections within reasonable count-rate limits (or lack of pointsources which were also identified in the guide star catalogue), thefilter image was viewed using the
XIMAGE package and a ‘King’profile model was fit to an identifiable point source which lookedsimilar in form and intensity to the AGN nucleus. The parametersfrom the fit were then used as a model for the PSF in GALFIT (viathe ‘Moffat’ model option), in lieu of a real PSF image. In the vastmajority of cases, a real PSF image could be extracted from the im-ages and a ‘King/Moffat’ model PSF was rarely required. In onlyone filter image for one object, a Gaussian model was found to fitthe nucleus better than other available models, with the Gaussianwidth fixed at the average FWHM of the UVOT PSF in that filteras reported in the CALDB documentation .The PSF or PSF model thus generated was then fit along withother galaxy profile components in GALFIT. We performed teststo ascertain the level of detail discernible within the UVOT im-ages, starting initially with a four-component model consisting ofa constant sky background, a central nuclear point source, a Ser-sic profile for the galaxy bulge and an outer exponential disk. Suchtests overwhelmingly indicated that four components incorporatedtoo high a degree of degeneracy, due to the large UVOT PSF size.The PSF of the UVOT instrument is typically the same size as thegalaxy bulge would be in higher resolution images of galaxies nearredshift ∼ http://heasarc.nasa.gov/docs/heasarc/caldb/swift/docs/uvot/c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN the PSF, and would ‘absorb’ some of the flux which would other-wise been reported as coming from the PSF component. Therefore,three components were adopted as sufficient to give sensible modelreconstructions of the profiles seen in the images (sky + PSF + ex-ponential disk). To get as accurate a representation as possible ofthe nuclear flux, the following iterative algorthim was employedusing GALFIT:(i) All bright foreground stars were excluded from the fit by cre-ating a bad pixel mask which covered these unwanted objects.(ii) The sky background was fit independently to a blank regionof the sky close to the AGN of interest.(iii) The central point source and sky were fit together, keep-ing the sky parameters constant from the previous step. The centralpoint source magnitude was seeded with that obtained from UVOT - SOURCE aperture photometry, to provide a sensible initial guess ofthe nuclear flux.(iv) The sky, central source and exponential disk componentswere fitted together, keeping the sky and central source parametersfixed at the values from the previous step.After these steps, the resultant model fits were compared withthe original images and the residual images obtained from subtract-ing the model from the data were inspected. In the first instance,the positions of the different components were left free to fit. If anyobvious signs of PSF mismatch or other problems with the fit wereevident, these positions were frozen manually by inspection andthe fit was re-evaluated. In some cases, such as for point-like AGNat the higher end of our redshift range, a simpler model consistingof just a PSF and a constant background was adopted, if the first-pass attempt at fitting an exponential disk did not show a significantgalaxy disk component. Where the PSF image displayed obviousPSF mismatch, a ‘King’ profile fit was also attempted. The best fitobtained from all these approaches, determined by visual inspec-tion of the residual image, was chosen to obtain the final nuclearmagnitudes. An example of the sky-nucleus-disk decomposition isshown in Fig. 2. The choices of profiles used for each filter in eachobject are given in table 3.
The detailed study of Bentz et al. (2006) highlights the need to takeinto account host galaxy contamination when calculating nuclearluminosities from optical HST images, specifically in the contextof determining the radius–luminosity relationship for the broad lineregion (BLR) in AGN. They find that even for an aperture of 1 arc-sec, there can be significant host galaxy contamination. However,the HST Advanced Camera for Surveys (ACS) offers a significantadvantage with a PSF of FWHM 0.0575 arcsec in contrast to thetypical FWHM of ∼ DS A and 3353 ˚ A respectively). We extract counts from the nu-cleus with a circular region with radius 0.2 arcsec, which shouldprovide an upper limit on the nuclear flux and contain >
90 per centof the counts from the nucleus. We also determine the total countsfrom a 5 arcsec region, as used with
UVOTSOURCE for comparison.In the F550M filter (closest to the UVOT V-band in wave-length), we identify a difference in fluxes of a factor ∼ UVOTSOURCE magnitude andthat from PSF fitting to the UVOT image is a factor ∼ ∼ . / . too large. Theproblem is less acute in the F330W filter (closest to the UVOT U-band), with an identical analysis yielding that the nuclear flux fromthe UVOT image is a factor ∼ . / . of that from the HST im-age. There is also the issue of optical AGN variability, which couldbe artificially increasing these ratios (the continuum near ˚ A is known to vary by factors of ∼ in the case of NGC 5548; seePeterson et al. 1999), but it is difficult to disentangle this from theerror intrinsic to the instrument, and the factors we calculate hereare therefore probably upper limits. This reflects a fundamentallimitation of the resolution of the UVOT images, but the subse-quent analysis shows that in many of the objects seen, attemptingto remove the host galaxy does produce an optical–UV SED shapecloser to that expected for an AGN despite possible overestimatesof the nuclear flux. A more detailed analysis using observationsfrom the Kitt Peak National Observatory (KPNO) 2.1m telescopeshould provide a much more detailed galaxy profile decomposition(Koss et al. in prep), which will make use of the smaller PSF avail-able with the KPNO (FWHM ∼ For AGN with very high count rates, the fluxes obtained fromUVOT images are highly susceptible to coincidence losses andtherefore often appear lower than their true values. Coincidenceloss refers to the phenomenon where multiple photons arrive at thesame location on the detector during a single frame, and is dis-cussed in detail in Poole et al. (2008). They the theoretical correc-tion to be applied to the measured count rate for individual pixels,but also point out the need to account for the spread of a pointsource over many pixels. They also provide an empirical correc-tion which takes the latter effect into account. We correct all ofour magnitudes from GALFIT for coincidence loss manually usingthese expressions; the
UVOTSOURCE aperture photometry magni-tudes are automatically corrected for coincidence loss.There is some scope for error in the coincidence loss cor-rection when the PSF image is not quite a point source, as thecoincidence loss correction only holds good for point sources.This problem is especially pronounced for very bright sources. Inthese sources, the coincidence-corrected fluxes are sometimes evengreater than those obtained from
UVOTSOURCE , which should the-oretically provide an upper limiting nuclear flux. In cases wherethis phenomenon is very pronounced, we do not use the magni-tudes from GALFIT and instead employ the 5.0 arcsec aperturephotometry magnitudes in any further analysis.After calculation of magnitudes from both GALFIT and c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800100 200 300 400 500 600 700 800900 100 200 300 400 500 600 700 800900 100 200 300 400 500 600 700 800900100 200 300 400 500 100 200 300 400 500 100 200 300 400 50050 100 150 200 50 100 150 200 50 100 150 20010 20 30 40 10 20 30 40 10 20 30 4020 40 60 20 40 60 20 40 60
Figure 1.
Example GALFIT profile fitting results for MCG-06-30-15. The first column in each row shows the original data image, the second shows the modelfit and the third shows the residual obtained from subtracting the model from the data. The rows, from top to bottom, show the results for the V, B, U, UVW1,UVM2 and UVW2 bands, in that order. In the V and B band, the residuals imply that a bulge component may also be discernible in this object, but the UVOTdata did not allow robust fits including a bulge component for many of our objects. A simple analysis using DS UVOTSOURCE , these magnitudes were corrected for Galactic ex-tinction using the values for E ( B − V ) Gal from the NASA/IPACInfrared Science Archive and the Galactic extinction curve ofCardelli et al. (1989). The final magnitudes were turned into XSPEC
PHA files using the
FLX XSP utility, to facilitate fitting along withthe contemporaneous XRT data using the
XSPEC analysis software. http://irsa.ipac.caltech.edu/applications/DUST/ The pipeline-processed event files from the XRT detector were pro-cessed using the
XSELECT package, as directed in the Swift XRTuser guide. Source regions of 50 arcsec were used, with largeraccompanying background regions (average radius ∼
150 arcsec).Background light curves were determined from the event files andinspected for flaring, but this was not found to be a problem in anyof our observations. Source and background spectra were extracted,and the source spectra were grouped with a minimum of 20 countsper bin. c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN AGN V B U UVW1 UVM2 UVW21RXS J045205.00+493248 P P P P – –2MASX J21140128+8204483 P+E P P+E P P+E P+E3C 120 P+E P+E P+E P+E M+E P+E3C 390.3 P+E P+E P P P+E P+EArk 120 P+E P+E P+E P+E P+E P+EESO 490-G026 P+E P+E P+E P+E P+E P+EESO 548-G081* P+E P+E P+E P+E M+E M+EIRAS 05589+2828 P P P P M GaussianIRAS 09149-6206 P+E M P+E P+E P+E P+EMCG +04-22-042 P+E P+E P+E P+E M+E P+EMCG-06-30-15 P+E P+E P+E P+E P+E P+EMrk 1018 P+E P+E P+E P+E M+E M+EMrk 279 P+E P+E P+E P+E P+E P+EMrk 352 P+E P+E P+E P+E P+E M+EMrk 509 P+E P+E P+E P+E P+E –Mrk 590 P+E P+E P+E P+E P+E P+EMrk 766 P+E P+E P+E P+E P+E P+EMrk 841 P+E P+E P+E P+E M+E P+ENGC 4593 P+E P+E P+E P+E P+E P+ENGC 5548 P+E P+E P+E P+E P+E P+ENGC 7469 P+E P+E P+E P+E P+E P+ENGC 985 P+E P+E P+E P+E P+E P+ESBS 1136+594 P P P+E P+E P+E P+ESBS 1301+540 P+E P P+E P P P+EUGC 06728 P+E P+E P+E P+E – P+EWKK 1263 P+E P+E P P P P
Table 3.
Table showing the model components used in GALFIT with each filter image for each AGN. ‘P’ signifies a PSF generated from guide stars withinappropriate count-rate ranges,from each image, ‘M’ signifies a Moffat (King) profile with parameters estimated from a single point source in the image, and‘E’ signifies an exponential disk model component.. *For ESO 548-G081, an extremely bright foreground star was located within ∼
30 arcsec of the AGN,rendering the photometry obtained susceptible to large errors.
Figure 2.
Comparison between images from HST (left) and UVOT (right) for the source NGC 4593. A 5 arcsec circular region is shown for reference on bothimages.
The M BH − L bulge correlation for galaxies provides a useful wayof estimating their central black hole masses. The key challenge isto obtain an accurate estimate of the bulge luminosity. The study ofMarconi & Hunt (2003) presents a decomposition of images from2MASS using GALFIT to obtain the bulge luminosity and corre-late it with the black hole mass from direct determination meth-ods (stellar or gas kinematics, maser kinematics, etc.). However,the redshifts of the objects under scrutiny here are generally sig-nificantly higher than those in Marconi & Hunt (2003), presentingproblems for performing a full nucleus-bulge-disc decomposition.Mushotzky et al. (2008) present a method for calculation of black hole mass for the Swift-BAT catalogue AGN from 2MASS K-bandtotal source magnitudes, which involves subtracting the central nu-clear component from the extended source flux to estimate thebulge luminosity. We attempt to refine this method here by incorpo-rating information on the expected angular size of the bulge on thesky and the resolution limitations of the 2MASS data. The crucialfactor is the degree of seeing affecting the 2MASS observations;typically the seeing takes a value of 2.6 arcsec. For bulges withtypical size ∼ . − kpc (Graham & Worley 2008), the bulgebecomes unresolvable at redshifts of about z ≈ . (assuming H = 71 km s − Mpc − , Ω M = 0 . and a flat Universe). There-fore, for the overwhelming majority of objects in our sample, thebulge should be unresolved. Inspection of the atlas images from2MASS for the lowest redshift objects in our sample confirms that c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian the visually identifiable bulge component is not resolved. The bulgelight is therefore mixed in with the nuclear light in these sourcesand identified as a ‘point source’. The point sources identified by2MASS have been collated, along with their magnitudes, in the ; similarly extended sourceshave been catalogued in the . Based on our assumption that the bulge is unresolved, weinitially download only the magnitudes from the PSC for furtheruse, as they will contain the bulk of the bulge light we wish to re-cover.By way of calibrating our attempt to obtain black hole masses,we first turn to the reverberation mapping (RM) sample of AGNpresented by Peterson et al. (2004), representing the most securemass determinations for AGN to date (augmented with the refinedresult from Denney et al. 2006 for NGC 4593). There are uncer-tainties in the RM method, connected to assumptions about the ge-ometry of the broad-line region and whether it is gravitationallybound, but the ∼ AGN for which RM mass estimates exist nev-ertheless constitute the sample of AGN with the most carefully de-termined black hole masses. Initially, we download the 2MASS at-las images for those RM AGN below a redshift of 0.01, in whichthe bulge should be theoretically resolved and perform a GALFIT3-component decomposition using a gaussian PSF for the nuclearpoint source, a Sersic bulge and an exponential disc. Our GALFITanalysis yields that there is a significant degree of uncertainty inthe identification of the bulge component in these AGN, and themagnitude ascribed to the bulge is heavily dependent on the priorconstraints imposed. We explore a number of different constraintson the model profiles; for example, the Sersic index n is often con-strained to an upper limit to avoid the fitting routine confusing itwith the nuclear point source, and often requires a lower limit toavoid confusion with the disc component. Despite such strategies,in some sources the bulge and disc profiles acquire very similar oridentical radii and ‘share’ the disc luminosity, with the Sersic pro-file partially tracing the disc instead of the bulge. In such sources,three-component model fits do not yeild meaningful bulge parame-ters. Our attempts at constraining the bulge using GALFIT result inbulge luminosities that vary by up to an order of magnitude depend-ing on the choice of model and constraints imposed, and show thateven for the close by sources, the bulges are probably not resolved.In light of this, we then employ the 2MASS PSC cataloguemagnitudes to provide a simple estimate of the total bulge and nu-clear flux. In order to estimate the relative contributions of the nu-cleus and bulge in the PSC magnitudes, we employ the infraredSED templates presented in Silva et al. (2004). In their study, theypresent nuclear IR SED templates constructed using available near-to-mid IR data on 33 Seyferts, using a radiative transfer models fordust heating to interpolate between the wavelengths covered by thedata. They also present host galaxy SED templates by subtractingthe nuclear template contribution from the total photometry at eachavailable wavelength. They present a number of different nuclearSED templates appropriate for different levels of X-ray absorption,and host SED templates are presented for different intrinsic 2–10keV AGN luminosity ( L X ) regimes. Using the appropriate nuclearSED along with the corresponding host galaxy SED, it is there-fore possible to estimate the fraction of the total luminosity whichcan be accounted for by the host for an AGN with a particular L X .In the process we make the assumption that, in the case of the K-band, the host flux should be dominated by the bulge as discussedin §
1. We calculate the ratio of host to total luminosity (nuclearplus host luminosity) from these SED templates in the K-band forthe four different luminosity regimes ( . < log(L X ) < . , . < log(L X ) < . , . < log(L X ) < . , . < log(L X ) < . ). The fractions are plotted in Fig. 3 against thecentral luminosity of each bin. We also plot a logarithmic inter-polation between the points, to allow determination of the ratioof L K , host /L K , total at a general value of L X , assuming the ratiovaries continuously. For values of L X greater than those spannedby the Silva et al. (2004) study, we simply employ the fraction cal-culated for L X = 10 erg s − , extrapolating from the directionof the trend shown between the bins centred on log(L X ) = 43 . and . . For luminosities log(L X ) < . , we use the fractionevaluated at log(L X ) = 41 . (extrapolating from the trend wouldyield fractions larger than Unity). The high luminosity extrapola-tion is clearly not likely to be accurate for extremely powerful ob-jects ( log L X > ), but none of the reverberation mapped AGNor the AGN in our Swift-BAT subsample are in this regime. In theabsence of information on IR SEDs outside this luminosity rangewe do not attempt a more complex extrapolation.Before using these fractions to calculate bulge luminositiesfrom the 2MASS PSC magnitudes, we consider potential sourcesof bias. Since Silva et al. (2004) adopt a reprocessing scenario toaccount for the K-band continuum, it is possible that the model nu-clear K-band luminosity is a function of the Eddington ratio (sinceit is ultimately linked to the UV accretion disc luminosity, the dom-inant component of the total accretion luminosity). The X-ray lu-minosity can be recast as the product of the Eddington ratio, bolo-metric correction and black hole mass (via the Eddington ratio),so the trend seen in Fig. 3 could be the result of two functionsof Eddington ratio being plotted against each other. However, weagain note the low X-ray luminosities probed by the Swift/BAT9-month catalogue (up to a few times erg s − ), and find thatthe host-to-total fraction varies at most by a factor of two in thisregime. Therefore, the dependence is not strong and will not intro-duce significant biases compared to other sources of error, such asthe intrinsic spread in RM masses and the intrinsic dispersion ofthe M BH − L K , bulge relation. Regardless of the particular modelused to predict the nuclear K-band continuum, it is clear that thefraction of an unresolved bulge that is attributable to the nucleuswill decrease for lower luminosity AGN and vice versa; the simplerelationship in Fig. 3 provides an observationally-rooted estimateof this effect.We employ the values of L X from XMM-Newton reported inVasudevan & Fabian (2009) to calculate the K-band host-to-totalratio for each of the RM objects, and assume the Silva et al. (2004)Seyfert 1 IR SED template for all objects in the sample. In the ab-sence of a detailed, uniform study on the absorbing columns of theRM AGN sample this is a reasonable assumption, since the rever-beration mapping sample is heavily biased towards Seyfert 1 AGN;none of the RM AGN are classified as Seyfert 2s in NED. We scalethe PSC magnitude by the appropriate fraction, using the expres-sion M K , bulge = M K , PSC − . ( f bulge ( L X )) , (1)where the fraction f bulge = L K , host / L K , total is calculated asabove. The black hole masses are then calculated using the M BH − L K , bulge relation of Marconi & Hunt (2003). The resulting massestimates are compared with RM mass estimates in Fig. 4.A reasonable agreement between the two methods is seen, al-beit with some significant deviation at higher masses. The greyshaded area highlights the uncertainties in RM masses due to thelack of precise knowledge of the BLR geometry. The largest dis-crepancy is for the AGN with the highest mass, 3C 273, for which c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN f bu l g e = L K , ho s t / L K , t o t a l log(L X )
40 41 42 43 44 45 46
Figure 3.
Variation of L K , host /L K , total with X-ray luminosity, extrap-olated from the host and nuclear IR SED templates of Silva et al. (2004)(black filled points - determined using the four fiducial luminosities pro-vided in their paper; solid line - extrapolation between points as detailedin the text). These fractions are employed to calculate the contribution ofthe bulge to the 2MASS point source magnitudes, for determining blackhole masses. The curve for an AGN with intrinsic X-ray absorption N H =10 . cm − is shown (dashed line) for comparison (the curves for otherabsorptions above N H = 10 cm − are very similar; the significant dif-ference is between obscured and unobscured AGN). the K-band bulge luminosity estimate gives a black hole mass anorder of magnitude too high. This is known to be a very power-ful object, and it is possible that our simple approach to estimatingthe bulge fraction in the point source light has still underestimatedthe powerful nucleus. There is also likely to be a significant syn-chrotron (non-accretion) contribution to fluxes at all wavelengthsin this source, which further increases uncertainties. This AGN isfar more powerful than the majority of the Swift-BAT catalogueAGN, so such problems are not likely to affect the lower-powerobjects we consider here.Marconi et al. (2008) point out that the previous reverberationmass estimates have been calculated with the assumption that theBLR is gravitationally bound, and introduce a correction based onan estimator of the accretion rate (the ˚ A monochromatic lu-minosity) to provide revised mass estimates for the RM AGN cat-alogue. We also provide a comparison with their revised mass es-timates, in Fig. 5. The correlation between the two mass estima-tors is strengthened significantly if these first-order corrections tothe reverberation masses are applied, and particularly reduces thescatter seen for those objects with high accretion rates (for whichthe correction is maximal and works to increase the reverberationmass estimate). The offset is still present however, and a simple fitof the form log( M BH , L K , bulge ) = A + log( M BH , revamp ) yields A ≈ . in both cases (using reverberation masses with or with-out correction for radiation pressure), implying an offset of a factorof up to ∼ . in black hole mass, just outside the typical quotedtolerance for reverberation masses (a factor of ∼ ). We now sug- M B H fr o m M K ( bu l g e ) fr o m s ca li ng M A SS PS C m a gn it ud e s BH Figure 4.
Comparison between masses from reverberation mapping andmasses estimated from K-band bulge luminosities. The solid line representsthe desired one-to-one correspondence between the two methods and theshaded area depicts the intrinsic uncertainties in reverberation mapping. Thedashed line shows the best fit of the form log( M BH , L K , bulge ) = A +log( M BH , revamp ) . gest a possible contributor to this offset and present a correction forit. The Silva et al. (2004) host galaxy SEDs have been calcu-lated using large-aperture and extended-source data (such as the2MASS XSC catalogue magnitudes) which could include morethan the bulge light. If the galaxy disc is significant in some cases,the Silva et al. (2004) host-to-total ratios actually provide the frac-tion of bulge plus galaxy disc to the total luminosity rather thanjust the bulge, and therefore using them to scale the ‘nucleus plusbulge’ luminosities from the 2MASS PSC is not strictly correct. Itis therefore necessary to gauge the effect of neglecting the galaxydisc in this calculation. We do this by obtaining the 2MASS Ex-tended Source Catalogue magnitudes (matching the centroids ap-propriately as before), using these fluxes as estimates of the totalnuclear, bulge and disc light ( L total = L nuc + L bulge + L disc ).Employing the PSC magnitudes as the bulge and nuclear light asbefore ( L bulge + L nuc ), we then can utilise the fraction of nuclear-to-total luminosity predicted from the Silva et al. (2004) IR SEDtemplates ( f nuc = L nuc / [ L nuc + L bulge + L disc ] ) to obtain a the-oretically more accurate estimate of the bulge magnitude: M (2)K , bulge = M K , PSC − . (1 − f nuc M K , PSC − M K , XSC2 . )) , (2)This expression is obtained by solving for L bulge from the avail-able quantities. Put simply, it arises from using the nuclear frac-tion from the IR SED templates to estimate the nuclear flux fromthe XSC flux; this nuclear component is then subtracted fromthe PSC flux to obtain the bulge flux. This estimate is not avail-able for all of the objects available previously, as some objectsdo not have 2MASS XSC magnitudes. For a couple of sources, c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian M B H fr o m L K , bu l g e ( M a r c on i & H un t ) fr o m M A SS BH from reverberation mapping (corrected for radiation pressure effects)6 7 8 9 10 11 Figure 5.
Comparison between masses from RM (corrected for radiationpressure effects as discussed by Marconi et al. 2008) and masses estimatedfrom K-band bulge luminosities. The empty white circles show the resultsusing the reverberation masses before correcting for radiation pressure.Other key conventions as in Fig. 4 the predicted fraction f nuc = L nuc / ( L nuc + L bulge + L disc ) from the Silva et al. (2004) templates was larger than the fraction ( L nuc + L bulge ) / ( L nuc + L bulge + L disc ) found from the ratio of thePSC and XSC fluxes, leading to an impossible solution for L bulge .This indicates that the nuclear-to-total fraction obtained from theSilva et al. (2004) templates is not appropriate for that object, orthat the bulge is not completely contained in the PSC flux. Theseproblems limit the use of this method, but for those objects where acalculation is possible, we find that the new K-band mass estimatesare systematically shifted down by 0.14 dex (a factor . ) fromthe values from the simpler approach of using the PSC magnitudealone. We therefore apply this correction to all masses obtainedusing the PSC alone, which brings the best-fit line for the newmass estimates within the tolerance of the reverberation masses.The strength of the correlation is good, as indicated by the Pear-son correlation coefficients of 0.83 and 0.93 obtained (correlatingwith reverberation masses before and after radiation pressure cor-rection). This analysis also indicates that the galaxy disc contribu-tion in the host-to-total ratios is of the order ∼ per cent, and asimilar value applies to all AGN in the RM sample. There still re-mains some offset above reverberation and within the tolerance re-gion, but any further attempt to account for this is somewhat redun-dant in light of the intrinsic uncertainties in reverberation masses,so we do not make any other corrections to our approach. The off-set of virial mass estimates below the M BH − L bulge relation hasbeen noted and discussed in detail by Kim et al. (2008b); they findthat it is a complex function of a variety of parameters includingthe black hole accretion rate and host galaxy morphology.We finally calculate the masses using this approach for all ofthe z < . objects in the Swift-BAT 9-month catalogue for which2MASS PSC magnitudes are available. The K-band host-to-totalratios are calculated using the L X values reported in Winter et al. l og ( M B H ) fr o m L K , bu l g e ( M a r c on i & H un t ) BH ) from reverberation mapping6 7 8 9 10 11 Figure 6.
Comparison between masses from reverberation mapping andmasses estimated from K-band bulge luminosities (with offset due to un-derestimation of the galaxy disc contribution taken into account). Key as inFig. 5. (2009), and the N H values in the same paper are used to select be-tween the two host-to-total ratio curves for obscured or unobscuredAGN (Fig. 3) as required. By way of another check on the accu-racy of our mass estimates, the recent study of Wang et al. (2009)presents virial mass estimates for a sample of Seyferts which par-tially overlaps with Swift-BAT 9-month catalogue AGN, using thesingle-epoch H β linewidth-based black hole mass estimator (asgiven in Greene & Ho 2005). We plot a comparison between theirvalues and ours in Fig. 7 for the 24 overlapping objects. While therelation shows some scatter, we caution that the linewidth-basedblack hole mass estimation methods also show significant scatterthemselves when compared to RM. The two estimators still appearto trace each other (with a correlation coefficient of 0.80); addi-tionally, preliminary calculations of black hole masses from KittPeak National Observatory (KPNO) and
Sloan Digital Sky Survey (SDSS) spectra for the Swift/BAT 9-month catalogue AGN alsoseem to match our estimates well, with a scatter of a factor of ∼ (Winter et al., Koss et al. in prep). We see some indications thatthe local Swift-BAT 9-month catalogue AGN have predominantlylow black hole masses, as expected in ‘anti-hierarchical’ structureformation scenarios. The masses thus determined are used in allsubsequent analyses for calculation of accretion rates and for con-straining accretion disc models in SED construction. We fit the optical, UV and X-ray SED data points with a simplemulticolour disc
DISKPN and absorbed power-law model combi-nation using XSPEC, following the approach outlined in VF09. Inthis work, we extend the approach of VF09 by inclusion of a multi-plicative
ZDUST component on the
DISKPN model used for the UV c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN M B H fr o m M A SS PS C K - b a nd bu l g e e s ti m a t e BH from Wang et al. (2009)6 7 8 9 10 11 Figure 7.
Comparison between masses from H − β linewidths (Wang et al.2009) and masses estimated from K-band bulge luminosities. The solid linerepresents the desired one-to-one correspondence between the two meth-ods. Wang et al. (2009) do not provide error estimates on their black holemasses, and the typical (random) error in the K-band estimates can begauged from Figs 4 and 5 (it is typically dominated by uncertainties in-trinsic to the M BH − L bulge relation), although the discussion in the texthighlights that systematic errors are likely to dominate. bump, to estimate intrinsic reddening where possible. The presenceof data in at least four UVOT filters for all of the objects in the sam-ple makes this possible, and indeed since most of the objects haveUVOT data in all six filters, we are able to constrain the shape ofthe SED over a wider energy range than was generally possible inVF09. Signs of intrinsic reddening (such as an optical–UV spec-trum steeply declining with energy) can then be more unambigu-ously identified.The full model used in XSPEC consisted of the optical–UV reddening component multiplied by a thermal accretion diskmodel, and a broken power law in X-rays with both Galac-tic and intrinsic absorption components ( ZDUST ( DISKPN ) +
WABS ( ZWABS ( BKNPOWER ))). The normalisation of the
DISKPN component was determined as in VF09, assuming an inclinationangle of zero and a colour-to-temperature ratio of Unity. Unifica-tion scenarios for AGN (Antonucci 1993) suggest that AGN withlower gas column densities such as the ones in this sample shouldbe preferentially oriented with their accretion disks almost face-on,and therefore small inclination angles are appropriate (see VF09 fora discussion of the effect of varying inclination angle in the
DISKPN component on luminosities calculated from SEDs). We also freezethe inner radius of the disc at 6.0 gravitational radii, which is appro-priate for radiatively efficient accretion. The local Galactic absorp-tion was determined using the NH utility from the FTOOLS suiteand frozen in the fit. The low energy branch of the
BKNPOWER model was set to have negligible values within the energy rangeof the UV bump (as in VF09). We fit the XRT data in the range0.3–10.0 keV. While there is an interesting debate as to the pre-cise nature of the intrinsic reddening in AGN (Gaskell et al. 2004), we adopt the Small Magellanic Cloud (SMC) reddening curve forour ‘first-order’ estimation of intrinsic extinction E ( B − V ) andfreeze the ratio of total to selective extinction, R V , at 3.0. The in-trinsic extinction E ( B − V ) , the intrinsic column density N ( int )H ,the maximum temperature T max for the DISKPN model and the pa-rameters for the high-energy branch of the
BKNPOWER componentwere all left free in the fit. Occasionally, if the combined optical,UV and X-ray data resulted in a photon index in the X-ray regimewhich was influenced by the slope between optical and X-ray, thephoton index was frozen at the value obtained from just fitting theX-ray data alone.We freeze the normalisation of the UV bump
DISKPN compo-nent using the black hole mass estimate from the 2MASS K-bandbulge luminosity estimate (normalisation K = M cos ( i ) /D β for black hole mass M BH , inclination angle i , luminosity distance D L and colour-to-effective temperature ratio β ). We present theSEDs determined using this approach in Fig. 8. For a few objects,the DISKPN model thus constrained fits poorly to the data despitetheir relatively ‘blue’ optical–UV spectral shapes, thought to betypical for the lower-energy part of an un-reddened accretion discspectrum (i.e. Ark 120, Mrk 1018, Mrk 509 and NGC 985), whichis probably attributable to a slightly too large black hole mass. Thisis confirmed later in the case of Ark 120 for which a reverbera-tion mass estimate is available (see § DISKPN normalisation, butthe reverberation mass estimate produces a more plausible fit. The
DISKPN model fit for IRAS 09149-6206 also looks poor, althoughthis object possesses a rather unusual, flat optical–UV SED (notedinitially by Kirhakos et al. 1991) unlike the other AGN in the sam-ple, which may be due to some other process. These five casesmight point to a genuine offset between the reverberation massesand our 2MASS derived estimates, but the degeneracy of the masswith the inclination and other parameters in the
DISKPN normal-isation could indicate, for example, unusual inclination propertiesin these few AGN. The remaining 21 objects all exhibit sensiblemodel fits to the optical–UV points.We present the average model SED (before and after correc-tion for reddening) calculated for the 22 AGN with λ Edd < . (normalised at 1eV) in Fig. 9. This allows a useful comparison withthe average SEDs presented in VF07 and VF09 for high and lowEddington ratio AGN, and the mean quasar SED templates of E94;the latter are also reproduced in Fig. 9. While the E94 template isrepresentative of X-ray bright quasars, ours offers a similar tem-plate for local, X-ray unabsorbed and low accretion rate Seyferts,albeit calculated from the model fits instead of the data. Some dif-ferences between our average SED and those of E94 include theslightly reddened shape of the E94 SEDs in the optical–UV (theirSEDs trace our extincted SED more closely, at least at longer wave-lengths) and the differing X-ray spectral shapes. The radio-quietquasar X-ray spectrum in particular appears softer in the 0.5–2 keVband which would be expected for a brighter, higher accretion ratesample of AGN (according to the trends seen in e.g. Shemmer et al.2008); but this could also be in part due to the inclusion of someX-ray absorbed objects in the E94 composite, depressing the flux atthis part of the spectrum relative to the flux at higher energies. Thehard shape of the radio-loud template can be accounted for by theexpected synchrotron contributions from jets. The E94 SEDs alsoshow a prominent soft excess; due to the limited spectral resolutionof the XRT we do not consider the soft excesses in our SEDs here.We repeat the cautionary note from E94, that AGN properties de- c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian
ESO 490-G026 E N ( E ) / k e V s - c m - -3 -3 ESO E N ( E ) / k e V s - c m - -3 IRAS 05589+2828 E N ( E ) / k e V s - c m - -3 -3 IRAS 0 E N ( E ) / k e V s - c m - -4 -3 -3 M CG +04-22-042 E N ( E ) / k e V s - c m - -3 -3 M CG -06-30-015 E N ( E ) / k e V s - c m - -3 -3 S J045205.0+493248 E N ( E ) / k e V s - c m - -3 -3 M rk E N ( E ) / k e V s - c m - -3 -3 Mrk E N ( E ) / k e V s - c m - -3 -3 Figure 8.
A selection of AGN SEDs. The crosses represent the 5 arcsec aperture photometry results from
UVOTSOURCE , and the filled circles represent thedata points used for the fit. In the UV and optical regime, the black filled points represent the fluxes with the host galaxy component removed using GALFITas detailed in the text. Empty circles were not used in the fit in a few cases where significant coincidence losses risk affecting the accuracy of the fluxes. termined from the average SED do not necessarily reflect the fulldegree of variation in the AGN population.We follow the approach outlined in VF09 for calculating theunabsorbed hard X-ray (2–10 keV) and bolometric (0.001–100keV) luminosities and errors in
XSPEC , for use in determining ac-cretion rates, bolometric corrections and other SED parameters.The results are presented in Table A1. As discussed in §
1, we de-fine the bolometric luminosity as being the primary emission dueto accretion emerging in the optical–to–X-ray regime and assumethat the IR is reprocessed emission; it is therefore not included in L bol . The BAT data allows us to extend the SEDs into hard X-rays (14–195 keV). Four-channel BAT spectra are publicly available on theinternet , time-averaged over periods of many months. We iden-tify the four least variable objects in the 14–195 keV band using http://swift.gsfc.nasa.gov/docs/swift/results/bs9mon/ measurements of excess variance on the BAT lightcurves, in orderto maximize confidence in luminosities extrapolated from fits tothe combined X-ray the hard X-ray data. We present the SEDs forthese four sources (NGC 5548, NGC 7469, Mrk 279 and 2MASXJ21140128+8204483) including the 4-channel BAT data in Fig. 10.We also present the fits to the more robust 8-channel data (the datathemselves are not yet public at the time of writing) to illustrateany substantive differences between 4- and 8-channel datasets. Inall cases, the simple model fits ( POWERLAW or PEXRAV ) do not in-dicate a signifcant departure from the general shape implied by the4-channel data.In three out of four of these low-variability sources, it ap-pears that the BAT data may be capturing a reflection humppeaking at around ∼
30 keV, in accordance with standard sce-narios for reflection from the accretion disc. There does notseem to be any pronounced reflection signature in NGC 5548.We corroborate this suggestion by fitting the model combination[
WABS ( ZWABSPOWERLAW )+ PEXRAV ] to the XRT and BAT datatogether in XSPEC, linking the photon index of the illuminatingsource in
PEXRAV to that for the
POWERLAW component. Themodel fits (thin dot-dashed lines) agree well with the fits to the c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN M AS X J21140128+8204483 E N ( E ) / k e V s - c m - -3 Mrk E N ( E ) / k e V s - c m - -3 C 5548 E N ( E ) / k e V s - c m - -4 -3 E N ( E ) / k e V s - c m - -3 Figure 10.
Spectral energy distributions with BAT data included. Empty squares represent 4-channel BAT data (publicly available). Thick dashed linesrepresent fits to the 8-channel data (not public at the time of writing). The thin dot-dashed line represents a fit to the XRT and BAT data combined, using themodel [
WABS ( ZWABSPOWERLAW )+ PEXRAV ] to illustrate the plausibility of a reflection fit to the data. ν F ν ( a r b it r a r y un it s ) -3 Figure 9.
Average, reddening-corrected optical–to–X-ray SED for the 22low Eddington ratio ( λ Edd < . ) AGN, normalised at 1 eV (thick blacksolid line). The average fit (in the optical–UV regime) with reddening in-cluded is shown using the thick black dashed line. The mean SEDs forradio-quiet and radio-loud quasars from Elvis et al. (1994) are also repro-duced (thin black solid and dot-dashed lines, respectively). XRT alone at low energies, and show a plausible reflection humpfit to the BAT data at higher energies for NGC 7469, Mrk 279 and2MASX J21140128+8204483. There are hints of accompanying FeK- α lines at 6.4 keV in the XRT data, as expected from reflection models, but the quality of the XRT data does not lend itself to aconfident identification of this line. The superior quality XMM-PNdata presented for NGC 7469, Mrk 279 and NGC 5548 in VF09show a much more convincing Fe K- α line shape at 6.4 keV, al-though the line signature is less evident in NGC 5548, which wouldbe expected for weaker reflection. The indications of reflection inthese sources are confirmed by a more detailed analysis using XMM and BAT data (Mushotzky et al. 2009 in prep).
We briefly comment on the effect of emission lines in optical–UVspectra of AGN which have the potential to introduce systematic ef-fects in our SED measurements. The study of Francis et al. (1991)presents a high signal-to-noise composite quasar spectrum, allow-ing the parameters of such emission lines to be accurately deter-mined. We identify the emission lines with the largest widths andhighest peaks from Figs. 2 and 7 of their work and redshift them in-dividually for each AGN in our sample to ascertain whether emis-sion line fluxes are likely to contaminate the UVOT photometry.As a first-order attempt to quantify this, we ascertain whetherthe central wavelength of the redshifted line lies within the band-pass of each of the UVOT filters, as obtained from the CALDBdocumentation. The most common contaminant from this anal-ysis would seem to be the Mg II line at 2798 ˚ A (rest-frame),which is most likely to appear in the UVW1 filter, and oc-casionally in the U filter for two high-redshift objects. Al III(1958 ˚ A ) and C III] (1909 ˚ A ) are also likely to be common con-taminants in UVM2 and UVW2 bands. For the four highest red-shift sources (3C 390.3, IRAS 09149-6206, SBS 1136+594 and c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian A and 5007 ˚ A ) appear in the UVOT V-band and in the case of 2MASXJ21140128+8204483, the H β line (4861 ˚ A ) also. The presence ofthese lines may explain, in part, the ‘angular’ behaviour of theoptical–UV SEDs in these four sources. However, the V-band‘bumps’ in more local sources (ESO 548-G081, Mrk 279, NGC4593 and NGC 7469 for example) are not accounted for by thepresence of such lines and must have some other cause. In anycase, for the vast majority of objects in the sample, the SED pointsin most filters are highly unlikely to be affected by the presence ofany prominent emission lines. We present the values for intinsic extinction E ( B − V ) determinedfrom the SEDs against Eddington ratio in Fig. 11. This allows acomparison between the amount of estimated dust extinction in thesource and the accretion power available from the central AGN.One notable feature of the distribution of E ( B − V ) observed isthat despite the selection of objects with low N H , there appearsto be significant dust reddening in many of these sources, basedon the shape of their optical–UV spectrum. This was noted pre-viously by Kraemer et al. (2000) and provided the basis for their‘lukewarm absorber’ scenario to account for this discrepancy. Ourresults contrast with those of Maiolino et al. (2001), who find thatdust-to-gas ratios – E ( B − V ) /N H – are systematically lower intheir sample AGN than in the Galaxy. However, their sample isselected specifically to include higher N H objects with relativelyun-absorbed optical/UV/IR broad lines, and the data used are notsimultaneous. The results of both of these studies results imply thatobjects which are significantly absorbed in X-rays can show lit-tle sign of dust extinction in the optical–UV and vice-versa, andin general that classifications of AGN into ‘obscured’ or ‘unob-scured’ based on their optical properties can differ significantlyfrom X-ray classifications. Our sample of objects confirms that op-tical and X-ray obscuration scenarios can differ significantly in anobject, influenced as they are by different physical processes (assummarised in Crenshaw & Kraemer 2001). We caution of a de-generacy between the mass and E ( B − V ) in the model fits (not-ing in the case of Mrk 590 that using its RM mass gives a best-fit E ( B − V ) of . compared to ∼ . using the 2MASS massestimate), although the SEDs for ESO 490-G026, MCG-06-30-15,Mrk 590 and Mrk 766 show more unambiguous signs of significantintrinsic extinction. In support of this, the aforementioned study onreddened Seyfert 1 galaxies of Crenshaw & Kraemer (2001) con-firms that observations from the International Ultraviolet Explorer (IUE) also strongly suggest intrinsic reddening in MCG-06-30-15,Mrk 590 and Mrk 766. The high optical polarisation for Mrk 766discussed by Ulvestad et al. (1995) may also be an indicator of thedust responsible for reddening in this source. Encouragingly, themore detailed study of MCG-06-30-15 by Reynolds et al. (1997)reports a lower limit on E ( B − V ) very similar to the value of ∼ ZDUST model can provide reddeningestimates in line with more detailed studies.The studies of Fabian et al. (2008) and Fabian et al. (2009)highlight the connection between the radiation pressure exerted bythe AGN and the configuration of the surrounding absorbing ma-terial. They identify the effective Eddington limit for dusty gasin the N H − λ Edd plane; above this limit is a ‘forbidden region’in N H − λ Edd space within which absorbing dusty gas cloudsare unstable to radiation. They verify that most local and deep M CG -06-30-015 M r k 766NGC 7469NGC 5548Mrk 590 1RXS J045205.0+493248Mrk 279 Ark 120IRAS 05589+28283C 120Mrk 509Mrk 841NGC 985 3C 390.3 SBS 1136+5942MASX J21140128+8204483Mrk 1018IRAS 09149-6206NGC 4593 UGC 06728ESO 548-G081 Mrk 352WKK 1263ESO 490-G026 SBS 1301+540MCG +04-22-042 I n t r i n s i c r e dd e n i ng E ( B - V ) -3 Figure 11.
Intrinsic extinction E ( B − V ) against Eddington ratio λ Edd ,for those objects for which reddening was included in the model fit via the
ZDUST
XSPEC model. samples of AGN avoid the forbidden region, and the latter workin particular verifies this in the case of the unbiased
Swift /BATcatalogue. Objects near the effective Eddington limit are knownto exhibit AGN winds/warm absorbers, which would be expectedwhen radiation pressure locally exceeds gravity. Since the dustand gas are expected to be coupled, the dust extinction parame-terised by E ( B − V ) may also exhibit a similar distribution inthe E ( B − V ) − λ Edd plane. We see from Fig. 11 that the ma-jority of objects occupy the lower-left part of the plot, and theonly object with particularly high reddening – MCG-06-30-15 – isalso known from Winter et al. (2009) to have a relatively high N H ( ∼ . × cm − ) and λ Edd , placing it close to (but not within)the forbidden region. This object showcases a warm absorber, asexpected for an object near the effective Eddington limit. The ab-sorbing column reported from analyses of
BeppoSAX and
ASCA data is strongly model-dependent (see for example, Reynolds et al.1997 who find N H to be an order of magnitude lower than the Win-ter et al. result; or Morales et al. 2000, whose detailed multi-zonewarm absorber scenario gives a column density of × cm − ).These hints may suggest that an analogous ‘forbidden region’could be proposed in the upper-right portion of the E ( B − V ) − λ Edd plane. A more complete study of intrinsic reddening proper-ties for a larger sample of sources spanning a wider range of X-raycolumn densities would be valuable for exploring this possibilityfurther.
We present the bolometric corrections κ − = L bol /L − against Eddington ratio in Fig. 12. The bolo-metric corrections cluster between 10–20 with a low fraction ofobjects possessing higher bolometric corrections, and Eddingtonratios are overwhelmingly below < . . The low bolometriccorrections obtained are expected for low Eddington ratios (asfound in VF07 and VF09) and the few objects with high bolometriccorrections generally lie at higher Eddington ratios as expected. c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN Errors in luminosities take into account the errors in modelparameters obtained from fitting the data, and the Eddington ratioadditionally takes into account the random error in the black holemass estimate (extrapolated from the uncertainties in the K-bandmagnitudes). However, the potential systematic errors due tointrinsic disparities between RM and K-band mass determinationmethods also contribute an uncertainty on both axes which is notrepresented in the error bars.We therefore invoke RM mass estimates for nine sources com-mon to the objects discussed by Peterson et al. (2004) to performsome simple comparisons which provide an estimate of the magni-tude of this systematic effect. If the RM masses are used in the fit-ting process instead, we obtain bolometric corrections and Edding-ton ratios as given in Fig. 13. We also present the results from VF09for comparison, which employed simultaneous data from XMM-Newton. If reddening is not included, as in VF09, we obtain verysimilar bolometric corrections and Eddington ratios as seen withXMM (left panel, Fig. 13). However if reddening is taken into ac-count, this generally increases both bolometric corrections and Ed-dington ratios (right panel, Fig. 13) and in the case of objects suchas Mrk 590 and NGC 7469, this change is very pronounced. How-ever, in both scenarios, the general trend for bolometric correctionincreasing with Eddington ratio identified in VF07 and VF09 ispreserved.We also perform a comparison between RM masses andmasses from K-band luminosities using only the
Swift simultaneousdata, including intrinsic reddening effects in both cases (Fig. 14).Bolometric corrections and Eddington ratios are clustered at signif-icantly higher values for this subsample when using the RM massescompared to the K-band masses. There is an increase of around ∼ dex in Eddington ratios when RM masses are used, accompaniedby an increase in bolometric correction to an average of ∼ . Themost extreme changes are for NGC 7469, Mrk 590 and Mrk 279.These results illustrate the degree of change which could be seen ifthe 2MASS K-band mass estimates need to be scaled down to bringthem more in line with RM estimates, but given the BLR geome-try uncertainty intrinsic to the RM masses, the bolometric correc-tions and Eddington ratios calculated using better-calibrated RMmasses could also be reduced significantly. As depicted in Fig. 6,the two methods are broadly consistent within these uncertainties.We suggest that the true magnitude of any shift in results may liein between these two extremes; the 2MASS masses may representan upper limit whereas the reverberation estimates may represent alower limit. Indeed, the radiation pressure corrections may requirefurther calibration as discussed by Marconi et al. (2009), and couldbring these estimates closer together, for example. More work isrequired on calibration between different black hole mass estima-tors for AGN before this issue can be addressed more convincingly,and at this stage we are only able to provide a crude estimate of thelimiting uncertainties.These analyses suggest that the typical bolometric correctionsand Eddington ratios for this sample are low, even when uncer-tainties in the mass estimates are taken into account. The distribu-tion of Eddington ratios and bolometric corrections are presented inFig. 15, using the 2MASS K-band mass estimates. The distributionis similar to that presented for Seyfert 1s in Fig. 7 of Winter et al.(2009) shifted by approximately an order of magnitude, in line withthe relatively low bolometric corrections we find for the majority ofthe sample. NGC 746 GC Mrk 590 Mrk 279 Ark 1203C 120Mrk 509 3C 390.3NGC 4593 NGC 7469NGC 5548Mrk 590 Mrk 279Ark 120 3C 120Mrk 5093C 390.3NGC 4593 κ - e V Edd
Figure 14.
Hard X-ray (2–10keV) bolometric correction against Edding-ton ratio for the objects with reverberation mapping mass estimates. Herewe present a comparison between the values obtained using mass estimatesfrom 2MASS K-band bulge luminosities (blue empty squares) and reverber-ation mapping (black), using the UVOT and XRT data in both cases, withintrinsic reddening included in the fit.
The ionizing continuum in AGN is responsible for the productionof the optical and UV emission lines which are often an integralpart of selection criteria in optical surveys of AGN. Variations inthe ionizing fraction can therefore lead to pronounced changes inthe detection of these lines (VF07). We plot the ionizing fraction( L E> . /L bol ) against X-ray luminosity, bolometric luminos-ity and Eddington ratio in Fig. 16 to identify any possible trends.It appears that there is little correlation between L E> . /L bol and luminosity (either X-ray or bolometric), but there may be indi-cation of a more substantial correlation with Eddington ratio. Thisis in line with the variation in SED shape reported in VF07 andVF09, despite our sample spanning only the lower-Eddington ratioregime. The average ionizing fraction is ∼ b ion in Marconi et al.(2008), but there is a large spread; for low Eddington ratio objectsthis could drop to ∼ ∼ DISKPN big blue bump model fits, and that there may be significantcomponents contributing to this ionizing flux in the extreme-UVand soft X-ray regime, which our approach cannot quantify; theionizing luminosity fractions presented here may therefore repre-sent lower limits to the ‘true’ ionizing fractions. However, such atrend for increasing ionizing fraction with Eddington ratio is alsoexpected from combined accretion disk and corona models such asthe one by Witt et al. (1997). The distrubution of values seen for theionizing luminosity fraction may also need to be taken into accountin the radiation pressure corrections to virial masses as discussedby Marconi et al. (2009), in a refinement over their 2008 work.
The α OX – L ν (2500 ˚ A) correlation reported in the literature(Strateva et al. 2005, Steffen et al. 2006) has been widelyused to encapsulate a relationship between the UV accre- c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian M CG -06-30-015 M r k 766NGC 7469NGC 5548Mrk 590 1RXS J045205.0+493248Mrk 279 Ark 120IRAS 05589+2828 3C 120Mrk 509Mrk 841NGC 985 3C 390.3SBS 1136+5942MASX J21140128+8204483Mrk 1018IRAS 09149-6206NGC 4593 UGC 06728ESO 548-G081 Mrk 352WKK 1263ESO 490-G026 SBS 1301+540MCG +04-22-042 κ - k e V Edd -3 Figure 12.
Hard X-ray (2–10keV) bolometric correction against Eddington ratio. The mass estimates used for model fitting and calculation of Eddington ratiosare determined from the 2MASS bulge luminosity. C κ - e V Edd C κ - e V Edd
Figure 13.
Hard X-ray (2–10keV) bolometric correction against Eddington ratio for the objects with reverberation mapping mass estimates (black filledcircles).
Left panel:
Intrinsic reddening was not included in the fit, as was the case in VF09.
Right panel:
Intrinsic reddening was included in the fit, to contrastwith VF09 and assess the role that intrinsic extinction may play in shaping the SED for some of the reverberation mapped AGN. In both cases, the reverberationmapping masses were used in the fit and to determine the Eddington ratios. The results from XMM-Newton PN and Optical Monitor (VF09) are shown forcomparison (blue empty squares). c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN P r ob a b ilit y d e n s it y log(λ Edd ) -3 -2 -1 0 P r ob a b ilit y d e n s it y log(κ ) Figure 15.
Histograms of Eddington ratios (left panel) and bolometric correction (right panel) for the sample. L E > . e V / L bo l log(L )
42 42.5 43 43.5 44 L E > . e V / L bo l log(L bol ) L E > . e V / L bo l λ Edd -3 Figure 16.
Ionizing fraction L E> . /L bol . Left panel: L E> . /L bol against X-ray luminosity. Middle panel: L E> . /L bol against bolometricluminosity. Right panel: L E> . /L bol against Eddington ratio. tion disk emission and the coronal X-ray emission, and therelation between optical and X-ray emission in AGN hasbeen extensively studied observationally (e.g. the RIXOS cata-logue, Mason et al. 2000). The spectral index α OX is definedas − log(L ν (2500 ˚ A) / L ν (2keV)) / log( ν (2500 ˚ A) /ν (2keV)) andtherefore the optical flux used in its calculation is measured atan energy significantly lower than the peak of the big blue bump(in the vicinity of ∼ A for typical AGN black hole masses of ∼ − M ⊙ , accreting at ∼
10 per cent of the Eddington limit). Weassess how changing the optical reference point in the calculationof this optical–to–X-ray spectral index alters the correlations seen,and denote the new spectral indices as α BX , α UX and α UV W X (which use the rest-frame B, U and UVW2 bands as the opticalreference point instead of ˚ A ). It is evident from the center-left panel of Fig. 17 that this sample follows the established cor-relation from Steffen et al. (2006), within the expected scatter. Wepresent spectral indices calculated both from the raw data (by in-terpolating to the desired optical wavelength, black filled circles)and from the de-reddened model fit (white filled circles). In thecase of α BX , there seems to be little correlation as the B-band is at the low-energy end of the disk component, where variations in thedisk emission have relatively little effect. However the correlationslinking α UX and α UVW2X to their respective reference luminosi-ties seem more substantial for the raw, interpolated values: the cor-relation coefficients decrease monotonically from − . to − . as the reference wavelength moves from the B to the UVW2 band,demonstrating the stronger anti-correlations seen using a shorterreference wavelengths. However, the intrinsic spectral indices donot show such trends: the correlation coefficients actually increasefrom − . to − . going from B to UVW2, which on visual in-spection of these plots reveals a large degree of scatter between thereddening-corrected spectral indices and their reference luminosi-ties. This demonstrates the importance of accounting for redden-ing. Previous studies on α OX have attempted to minimize the ef-fects of reddening by carefully selecting the samples used; for ex-ample Strateva et al. (2005) require that their AGN are requiredto be bluer than a template AGN SED with fixed reddening; red-shifting this template then provides a redshift-dependent colourthreshold (see Richards et al. 2003) which they use to exclude ‘red’ c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian α B X -2-1.5-1-0.5 log(L B )
26 27 28 29 30 31 α O X -2-1 .5-1-0.5 log(L Å )
26 27 28 29 30 31 α U X -2-1 .5-1-0.5 log(L U ) α U V W X -2-1.5-1-0.5 log(L UVW2 )
26 27 28 29 30 31
Figure 17.
From left to right, top to bottom: Spectral indices connecting SED at the rest frame B-band, U-band, 2500 ˚ A ( α OX ) and UVW2-band with the SEDat rest frame 2keV. Each spectral index is plotted against the corresponding rest-frame reference luminosity in the optical/UV band (e.g. L B for α BX , L U for α UX etc.). The black filled circles represent the values obtained from interpolating at rest-frame energies from the data points (not corrected for intrinsicreddening) and the white filled circles show the indices calculated from the de-reddened SED model fit. The solid line and shaded area in the top right panelshows the best fit and spread obtained by Steffen et al. (2006) for a sample of 333 optically selected AGN. The correlation coefficients of de-reddened, intrinsicspectral index against luminosity are as follows: α BX − L B :-0.27, α UX − L U :-0.24, α OX − L ˚ A :-0.19, α UVW2X − L UVW2 :-0.16.
AGN. This should statistically exclude heavily reddened AGN,but we find from individually de-reddening our optical–UV AGNSEDs, that reddening effects could still be significant. The anti-correlations between the raw spectral indicies (uncorrected for red-dening) and their respective reference luminosities appear strongbecause reddening in an object produces a flatter optical–X-rayspectral index and reduces the UV luminosity; this will tend to pop-ulate the top-left regions of each plot with reddened objects whilethe lower-right regions will contain unreddened objects, produc-ing an artifically robust-looking relationship. This effect may con-tribute to the strength of the α OX − L ˚ A correlation in the liter-ature. From the perspective of understanding the accretion process,however, it is the intrinsic , de-reddened optical–X-ray spectral in-dices which are of interest. We only use a small sample here, butif individually de-reddened optical SEDs for AGN can be used toconstruct the α OX − L ˚ A relation anew with larger samples, itwill clarify whether the versions of this relation seen in the litera-ture can be used to robustly constrain the accretion process in AGN.When plotted against Eddington ratio instead of optical/UVreference luminosity, even the anti-correlations between the raw,uncorrected indices and optical/UV reference luminosities disap-pear (see Fig. 18). This phenomenon was also observed by VF07 and Shemmer et al. (2008). The latter study suggests that the lackof significant correlation between optical–X-ray spectral index andEddington ratio could be because of inaccuracies in mass determi-nation methods. They also discuss how the use of an X-ray selectedsample (such as the hard X-ray selected BAT catalogue used here)should alleviate some selection effects which artificially could blurany correlation with Eddington ratio, but this study seems to rein-force the lack of connection between optical–X-ray spectral indexand Eddington ratio. This seems to be the case for both the ‘raw’values of α OX etc. from the data and the ‘intrinsic’ values calcu-lated from de-reddened SED fits. This may imply that the opticalthreshold needs to be moved further into the far–UV before an Ed-dington ratio correlation is reported, if present.We also briefly discuss possible correlations of these spectralindices with mass. We remind the reader of the potential uncertain-ties with our K-band mass estimates discussed in § α band , X and the black hole mass, but find that the corre-lation coefficient increases with decreasing reference wavelength,increasing from 0.30 (B-band) to 0.61 (UVW2 band; see Fig. 19),if the outlier IRAS 09149-6206 (the AGN with the highest black c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN α B X -2-1.5-1-0.5 λ E dd -3 α OX -2-1.5-1-0.5 λ Edd -3 α U X -2-1.5-1-0.5 λ Edd -3 α UV W X -2-1.5-1-0.5 λ Edd -3 Figure 18.
From left to right, top to bottom: Spectral indices connecting SED at the rest frame B-band, 2500 ˚ A ( α OX ), U-band and UVW2-band with the SEDat rest frame 2keV. Each spectral index is plotted against Eddington ratio. All conventions are as in Fig. 17. hole mass in this sample) is excluded. This relation implies thatthe spectrum flattens for higher mass objects, presumably indicat-ing that the fraction of UV emission directly from the accretiondisc is less. A plot of bolometric correction against black hole mass(Fig. 20) displays this in a more pronounced fashion, confirmingthat the fraction of the luminosity appearing in X-rays, /κ − (i.e. from the corona rather than the disc) increases with mass. Thislends credibility to a scenario in which local, high mass black holesare radiating predominantly at low Eddington ratios: not only doesthe high mass lead to a low Eddington ratio, but lower bolometriccorrection reduces the bolometric luminosity further. Conversely,it is then expected that lower mass black holes exhibit higher Ed-dington ratios by virtue of their large bolometric corrections, in linewith ‘cosmic downsizing’ scenarios in which the bulk of accretionactivity shifts to lower mass objects at low redshift (Heckman et al.2004). Since we are using the bulge luminosity to calculate theblack hole mass, the possibility of a correlation of bolometric cor-rection with bulge luminosity is also interesting to consider. If real,it suggests that brighter bulges galaxies contain AGN accreting instates with a greater fraction of X-ray emission. This may indicate alink between the stellar populations of the host galaxy, the accretionstate of the central black hole and the fuelling process; these themesare discussed in more detail by Kauffmann & Heckman (2008).The data from VF09 on the reverberation mapped sample donot exhibit any similar trends between either α OX or bolometriccorrection and the black hole mass, despite the better quality mass α UVW2X α O X α UX α B X α UVW2X - l og( M B H ) best fit α b a nd - X -2- - - - l og( M B H /M ๏ ) Figure 19. α O , B , U , UVW2 − X against black hole mass. The key providesdetails of the symbols used. The best fit (solid line) excludes outlier IRAS09149-6206. estimates used. However, the reverberation sample is not well-selected and the quantities in VF09 do not include reddening andhost galaxy corrections in the optical–UV, so do not necessarilyprovide a useful comparison for our purposes. c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian l og ( κ - e V ) log(M BH /M ๏ ) Figure 20.
X-ray (2–10keV) bolometric correction against black holemass. The best fit (excluding the outlier IRAS 09149-6206) has the form log( κ − ) = 2 . − . BH ) / M ⊙ (the spread in bolomet-ric corrections is a factor of ∼ , by visual inspection). The host galaxy properties of the 9-month BAT catalogue are dis-cussed in some detail in Winter et al. (2009). Specific conclusionsfrom their study include the lack of a convincing relation betweenhost galaxy inclination and X-ray absorbing column N H ; the largefraction of interacting systems in the whole catalogue ( ∼
54 percent) and suggestions of evolution away from the elliptical, redhosts seen around AGN at z ∼ g − r colour inthe range ∼ . − . ) in the colour-magnitude diagram, mid-waybetween the late-type ‘red sequence’ ( g − r ∼ . − . ) and star-forming ‘blue cloud’ ( g − r ∼ . − . ) galaxies (see for exampleNandra et al. 2007, Hickox et al. 2009). Recently, Schawinski et al.(2009) present the colours for a small sample of 16 AGN fromthe BAT catalogue which have SDSS data and find that their hostsalso lie in this intermediate region. The studies of Silverman et al.(2008) and Westoby et al. (2007) are also of interest as they probehigh and low redshift populations (from the Extended Chandra
Deep Field South and the Sloan Digital Sky Survey, respectively).Silverman et al. (2008) find that, accounting for biases caused bylarge-scale structure, AGN hosts at z & . have bluer hosts( U − V < . ) but the lower redshift counterparts ( z . . ) tendto exhibit redder colours. Westoby et al. (2007) find a broad dis-tribution of colours for their composite AGN sample, towards thered end of the colour distribution for normal galaxies, and coveringthe green valley. We note that, in making our comparison with theresults of Silverman et al. (2008), we do not have any biases withrespect to large-scale structure due to the nature of the Swift -BATcatalogue, and additionally the only bias with respect to host galaxy properties would be the preferential selection of face-on galaxies(by virtue of our low N H selection).Using the prescription in the UVOT CALDB for convertingthe UVOT magnitudes into standard Johnson magnitudes, alongwith the prescription for calculating SDSS colours from Johnsoncolours, we present our colour-magnitude diagrams in Fig. 21. Wepresent both the Johnson U − V colour against absolute V-bandmagnitude M V and SDSS g − r colour against absolute SDSS r-band magnitude M ( r ) ; the g − r colour is calculated using thetransformation g − r = 1 . B − V ) + 0 . U − B ) − . (Karaali et al. 2005) and the r –band magnitude is interpolated us-ing the UVOT points. We provide U − V colours for 21 out of our26 objects for which both U and V host galaxy magnitudes wereavailable from our GALFIT analysis; 3 of these objects did nothave a B-band magnitude for the transformation to g − r , reducingour sample of objects to 18 for the latter plot.We see from Fig. 21 that the colours of our local sampleshow a similar distribution to the X-ray selected AGN in Fig. 3of Silverman et al. (2008), but without a significant noticeable redpopulation. In comparison with Westoby et al. (2007), we find adistribution similar to their composite AGN sample, with a handfulof very blue AGN. We therefore seem to corroborate the sugges-tions in Winter et al. (2009) of an evolution away from red typehost galaxies to bluer host galaxies, with the majority of objectsin the ‘green valley’. We also find somewhat bluer colours thanthose found for the 16 BAT AGN hosts in Schawinski et al. (2009)although there are only five objects overlapping between the twosamples (NGC 7469, NGC 5548, Mrk 766, MCG +04-22-042 andMrk 1018; this list was obtained by private communication with theauthor). We caution that four of the bluest objects in Fig. 21 (Ark120, Mrk 509, Mrk 841 and Mrk 352) have point-like morphologiesin the UVOT images, and despite our efforts to minimize PSF mis-match, if such mismatch is present the wings of the PSF could bemistaken for the host galaxy giving rise to artificial ‘hosts’ whichare influenced by the bluer colours of the AGN. These objects aredepicted using crosses in Fig. 21 for easy identification. This mayindicate that the blue-ward shift is less pronounced than it may ini-tially seem, but the colours for the rest of our sample do lie in theblue end of the green valley. The more accurate and complete anal-ysis Koss et al. (in prep) will clarify these issues.We also caution that the PSF size of the UVOT may im-pose limitations on the accuracy of our host galaxy magnitudes.In § F UV OT nuc = F HST nuc + F spurious (for nuclear flux from HST F HST nuc and spuri-ous flux F spurious ), we can solve for F spurious using the estimateof the error in the UVOT nuclear flux from § F HST galaxy = F UV OT galaxy + F spurious . The correction to the galaxy fluxemerges as F spurious = F UV OT nuc [1 − ( F HST nuc /F UV OT nuc )] . We em-ploy the fractions F HSTnuc /F UV OTnuc ≈ . / . for the V-band and ≈ . / . for the U-band based on the analysis of NGC 4593, andcalculate the corrections to the host galaxy magnitudes as follows: http://heasarc.nasa.gov/docs/heasarc/caldb/swift/docs/uvot/c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN U - V - - M V -24-22-20-18 g - r M(r) -24-22-20-18
Figure 21.
Left panel:
Host galaxy colour U − V against absolute V-band magnitude for the host galaxies in the sample. Right panel:
SDSS g − r colouragainst absolute r-band magnitude for the sample. Crosses represent objects with point-like morphologies, for which the host galaxy colours should be treatedwith caution. GALFIT returns errors of ± ∼ . mag on magnitudes, but systematic errors due to the resolution of UVOT are likely to dominate (see Fig. 22) ∆ M = − . UVOTnuc F UVOTgalaxy (1 − F HSTnuc F UVOTnuc ))) . (3)It is not straightforward to predict whether the colours shouldgenerally increase or decrease under this correction, since the cor-rection is dependent on two things: the ratio of host galaxy flux tonuclear flux, and the degree of error in the nuclear flux intrinsicto UVOT, in a particular band. We plot the corrected U-V coloursin Fig. 22. If the corrective factors assumed here are reasonable,we find that the U − V colours generally become bluer, in somecases substantially so. As expected, the corrected host galaxy V-band magnitudes are brighter due to the extra contribution from thenucleus. We do not make an attempt to correct the SDSS g − r magnitudes as these would require an estimate of the nuclear fluxoverestimate in the B-band as well, and the r-band magnitudes re-quire interpolation to wavelengths between the central UVOT filterwavelengths. It is therefore possible that better quality images mayreveal even bluer hosts than our results indicate, but the limitationsof the UVOT highlighted by this analysis suggest that the upcom-ing KPNO/SDSS work of Koss et al. (in prep) will provide moredefinitive estimates of the host galaxy colours. We have presented SEDs for a well-selected subsample of localAGN from the
Swift
BAT 9-month catalogue of AGN. This studymakes use of the unbiased and representative nature of the cat-alogue and offers numerous advancements over previous studiesof the AGN SED. The simultaneous optical–to–X-ray observationsfrom
Swift used here provide a more accurate snapshot of the ac-cretion energy budget in AGN, as pioneered by Brocksopp et al.(2006) and adopted by VF09. We are also able to increase confi-dence in measurements of the accretion luminosity by employinga subsample of AGN with minimal spectral complexity (no strongsigns of partial covering or strong reflection) and lack of heavyabsorption [log( N H / cm − ) <
22] which could complicate the pic-ture: the X-ray spectra for these objects can all be fit with absorbed U - V -1-0.500.511.5 M V -24-22-20-18 Figure 22.
Host galaxy colour U-V against absolute V-band magnitude,partially corrected for large PSF of UVOT. Key as in Fig. 21 power-laws according to previous studies (Winter et al. 2009), al-lowing straightforward calculation of absorption-corrected X-rayluminosities. The data from
Swift -UVOT typically cover at leastfour, if not all six available optical–UV filters on the instrument,and allow for a detailed reconstruction of the optical–UV SED.This is useful on two counts: firstly, in understanding the degree ofhost galaxy contamination in the observed SED for these objects;and secondly, in estimating how much dust extinction is responsiblefor the shape of the host-galaxy corrected nuclear SED. The inclu-sion of black hole mass estimates from K-band bulge luminositiesallows accretion rates to be calculated using a uniform method forthe sample.We consistently remove the host-galaxy contribution from theUVOT data points by fitting a ‘PSF+disk’ profile to the UVOT im-ages using GALFIT, and obtain a corrected nuclear flux in all avail-able filters. The host galaxy is often found to be a significant con- c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian taminant in the optical (V, B and U) bands, but far less significantin the UV bands. We also find that optical/UV emission lines areunlikely to contaminate the nuclear continuum significantly giventhe relatively large bandwidths of the UVOT filters (500–1000 ˚ A ).When fitted along with the simultaneous X-ray data with a sim-ple ‘multicolour accretion disc + power-law’ model, including bothGalactic and intrinsic absorption, we are able to recover many im-portant parameters of the AGN SED, and produce an average modelSED for lower-luminosity Seyferts with low accretion rates. TheGALFIT fitting on the UVOT images also provides potential forunderstanding the host galaxy population of the unobscured BATcatalogue AGN. • In this representative sample, the 2–10 keV bolometric correc-tions cluster around values of 10–20. These bolometric correctionsare essential for calculations of bolometric luminosities and accre-tion rates from X-ray observations, and therefore are of importancein the wider picture of scaling up the total X-ray energy density dueto SMBH accretion in the X-ray background. • The Eddington ratios for the sample are typically below 0.1.Low bolometric corrections are expected for these low Eddingtonratios (VF07, VF09), and furthermore, this may indicate that local,unobscured AGN are generally in an accretion state analogous tothe ‘low/hard’ state in Galactic black holes. This reinforces pre-vious findings that local Seyferts have low accretion rates, suchas in early work by Sun & Malkan (1989), who fit accretion discmodels to optical–to–IR SEDs for a sample of 60 bright quasarsand Seyferts. We arrive at the same conclusion using a differentapproach and employing a different sample: here we also use X-ray data in the calculation of accretion rates on a lower-luminositysample selected for minimal absorption. • Some objects possess significant optical–UV dust redden-ing, even though these objects were selected for low absorb-ing gas column. This is broadly consistent with the observationsof Kraemer et al. (2000), Barcons et al. (2003) and Mateos et al.(2005) amongst others, and adds to the body of work suggestingthat optical and X-ray obscuration classifications can show signif-icant discrepancies (Maiolino et al. 2001, Comastri et al. 2001 andSilverman et al. 2005 are examples of studies which find evidencefor the opposite effect in part or all of their samples; namely AGNwith X-ray absorption but negligible optical–UV obscuration sig-natures). If we adopt a threshold of E ( B − V ) ∼ . to divideheavily dust reddened and unreddened objects, ∼ − out of our26 AGN ( ∼ per cent) are heavily dust reddened, consistent withthe proportions of ‘anomalous’ AGN with mismatching optical andX-ray classifications found in some of the above studies, but thisobviously depends on the threshold adopted. • The reddening-corrected ionizing luminosity fraction in-creases with Eddington ratio, as expected from combined accretiondisc and corona models. • We confirm that spectral indicies linking the optical to the X-ray ( α OX etc.) show evidence for correlation with luminosity butnot with Eddington ratio, corroborating the findings of previousstudies. We find that extinction effects may need to be more care-fully considered when considering the correllation with luminosity,than has been done in previous works. De-reddening objects indi-vidually in constructing the α OX − L ˚ A relation yields a weakercorrelation, which may need to be investigated in larger samplesto check the robustness of this correlation and its use in constrain-ing the accretion process. The presence of a possible correlation of α UVW2X with mass, coupled with the attendant anti-correlation ofbolometric correction with mass, may reinforce cosmic downsizing scenarios where lower mass black holes dominate AGN activity inthe local universe by virtue of their larger accretion rates. • The host galaxies for this subsample generally lie in the greenvalley or blue-wards of it, in the standard colour-magnitude dia-gram. However, the UVOT images are of limited resolution and abetter quality analysis with KPNO/SDSS (Koss et al. in prep) willshed more light on the host galaxy properties. • Adding the hard X-ray BAT data (14–195 keV) to these ob-servations can help with ascertaining the presence of a reflectionhump in the spectrum. We selected four sources with the lowesthard X-ray variability, and find that three of these four showed asignificant excess peaking at around ∼
30 keV which can plausiblybe fit with a reflection hump.We have also discussed at length, the calibration of the blackhole mass estimation method used to obtain these conclusions. TheEddington ratios and bolometric corrections could be significantlyaffected by these uncertainties. We estimate the degree of this ef-fect by comparison with results using reverberation mass estimates.This analysis reveals somewhat higher bolometric corrections andEddington ratios, but the bulk of the sample lies consistently atlower Eddington ratios and exhibits low bolometric corrections.The bulge luminosity-based and reverberation mapping methodsrequire more accurate calibration before more definitive statementson the magnitudes of these uncertainties can be constrained.The hard X-ray selection criteria used to specify the BAT cat-alogue eliminate potential biases. As an illustration of this, we no-tice the presence of substantially more low accretion rate objects inthis sample than the samples of VF07 and VF09 (the former wasUV selected by
FUSE , the latter selected for being optically brightenough for reverberation mapping). The low accretion rates foundhere broadly corroborate the results of Winter et al. (2009) using X-ray luminosities (with no bolometric correction applied). Althoughthese objects have low-absorption and are situated at low redshift,they are similar in power to the obscured, higher redshift sourcesthought to be responsible for the bulk of the X-ray background.As suggested in Fabian (2004), we confirm that a lower bolomet-ric correction is appropriate for these objects, based on our multi-pronged approach to recovering the true bolometric luminosities inthese sources. Bolometric corrections of the magnitude seen hereare appropriate for reconciling the energy density from the X-raybackground with the local black hole density. It remains to be seenif such properties are displayed in large samples of obscured AGN,the dominant class of AGN contributing to the X-ray background.Useful extensions to this work would be to acquire simul-taneous
Swift observations for the remaining objects in the BATcatalogue, which would potentially confirm the trends we identifyhere, or highlight unknown biases in our chosen sample. What isof particular importance next is to understand the workings of thecentral engine of obscured AGN. In a companion paper (Vasude-van et al. in prep ) we hope to address this issue for both obscuredand unobscured AGN in the Swift/BAT catalogue, using the re-processed IR emission along with the hard X-ray BAT observa-tions to estimate their bolometric properties, analogous to the workdone by Pozzi et al. (2007) for eight higher redshift quasars. Thewell-studied 9-month catalogue again provides an excellent start-ing point for such work, and follow-up studies on the 22-monthcatalogue (Tueller et al. 2009) will offer new scope for constrain-ing the accretion physics of a larger, representative sample of AGN. c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN RVV acknowledges support fom the Science and Technology Fa-cilities Council (STFC) and ACF thanks the Royal Society for Sup-port. We thank the
Swift /BAT team for the 9-month AGN cataloguedata. We thank Stephen Holland for help with UVOT data analy-sis, and Alice Breeveld for kindly providing and customizing herPSF generation code for UVOT images. We thank Jack Tuellerfor the use of results derived from the 8-channel BAT data, KevinSchawinksi for providing host galaxy colours and magnitudes fromhis paper for comparison with our study. We thank Richard McMa-hon for help in understanding the correct use of the data in the2MASS catalogues. We also thank the anonymous referee for use-ful comments and suggestions which improved this work. This re-search has made use of the NASA Extragalactic Database (NED)and the NASA/IPAC Infrared Science Archive, which are operatedby the Jet Propulsion Laboratory, California Institute of Technol-ogy, under contract with the National Aeronautics and Space Ad-ministration.
REFERENCES
Antonucci R., 1993, ARA&A, 31, 473Barcons X., Carrera F. J., Ceballos M. T., 2003, MNRAS, 339,757Bentz M. C., Peterson B. M., Pogge R. W., Vestergaard M., OnkenC. A., 2006, ApJ, 644, 133Blustin A. J., Page M. J., Fuerst S. V., Branduardi-Raymont G.,Ashton C. E., 2005, A&A, 431, 111Brocksopp C., Starling R. L. C., Schady P., Mason K. O., Romero-Colmenero E., Puchnarewicz E. M., 2006, MNRAS, 366, 953Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ, 345, 245Comastri A., Fiore F., Vignali C., Matt G., Perola G. C., La FrancaF., 2001, MNRAS, 327, 781Crenshaw D. M., Kraemer S. B., 2001, ApJ, 562, L29Denney K. D. et al., 2006, ApJ, 653, 152Elvis M. et al., 1994, ApJS, 95, 1Fabian A. C., 2004, in Ho L. C., ed, Coevolution of Black Holesand Galaxies, p. 446Fabian A. C., Vasudevan R. V., Gandhi P., 2008, MNRAS, 385,L43Fabian A. C., Vasudevan R. V., Mushotzky R. F., ReynoldsL. M. W. C. S., 2009, ArXiv e-printsFrancis P. J., Hewett P. C., Foltz C. B., Chaffee F. H., WeymannR. J., Morris S. L., 1991, ApJ, 373, 465Gaskell C. M., Goosmann R. W., Antonucci R. R. J., WhysongD. H., 2004, ApJ, 616, 147Gierli´nski M., Done C., 2004, MNRAS, 349, L7Graham A. W., Worley C. C., 2008, MNRAS, 388, 1708Greene J. E., Ho L. C., 2005, ApJ, 630, 122Heckman T. M., Kauffmann G., Brinchmann J., Charlot S.,Tremonti C., White S. D. M., 2004, ApJ, 613, 109Hickox R. C. et al., 2009, ArXiv e-printsHopkins P. F., Richards G. T., Hernquist L., 2007, ApJ, 654, 731Karaali S., Bilir S., Tunc¸el S., 2005, Publications of the Astro-nomical Society of Australia, 22, 24Kauffmann G., Heckman T. M., 2008, ArXiv e-printsKelly B. C., Bechtold J., Trump J. R., Vestergaard M., Siemigi-nowska A., 2008, ApJS, 176, 355Kim M., Ho L. C., Peng C. Y., Barth A. J., Im M., 2008a, ArXive-prints Kim M., Ho L. C., Peng C. Y., Barth A. J., Im M., Martini P.,Nelson C. H., 2008b, ApJ, 687, 767Kirhakos S., Strauss M. A., Yahil A., Griffith M., Fisher K. B.,1991, AJ, 102, 1933Kraemer S. B., George I. M., Turner T. J., Crenshaw D. M., 2000,ApJ, 535, 53Kuraszkiewicz J., Wilkes B. J., Schmidt G., Smith P. S., Cutri R.,Czerny B., 2008, ArXiv e-printsMaiolino R., Marconi A., Salvati M., Risaliti G., Severgnini P.,Oliva E., La Franca F., Vanzi L., 2001, A&A, 365, 28Marconi A., Axon D., Maiolino R., Nagao T., Pietrini P., RisalitiG., Robinson A., Torricelli G., 2009, ArXiv e-printsMarconi A., Axon D. J., Maiolino R., Nagao T., Pastorini G.,Pietrini P., Robinson A., Torricelli G., 2008, ApJ, 678, 693Marconi A., Hunt L. K., 2003, ApJ, 589, L21Marconi A., Risaliti G., Gilli R., Hunt L. K., Maiolino R., SalvatiM., 2004, MNRAS, 351, 169Mason K. O. et al., 2000, MNRAS, 311, 456Mateos S., Barcons X., Carrera F. J., Ceballos M. T., Hasinger G.,Lehmann I., Fabian A. C., Streblyanska A., 2005, A&A, 444, 79Miniutti G., Fabian A. C., 2004, MNRAS, 349, 1435Morales R., Fabian A. C., Reynolds C. S., 2000, MNRAS, 315,149Mushotzky R., 2004, in Astrophysics and Space Science Library,Vol. 308, Barger A. J., ed, Supermassive Black Holes in the Dis-tant Universe, p. 53Mushotzky R. F., Winter L. M., McIntosh D. H., Tueller J., 2008,ArXiv e-printsNandra K. et al., 2007, ApJ, 660, L11Peng C. Y., Ho L. C., Impey C. D., Rix H.-W., 2002, AJ, 124, 266Peterson B. M. et al., 1999, ApJ, 510, 659Peterson B. M. et al., 2004, ApJ, 613, 682Poole T. S. et al., 2008, MNRAS, 383, 627Pozzi F. et al., 2007, A&A, 468, 603Reynolds C. S., Ward M. J., Fabian A. C., Celotti A., 1997, MN-RAS, 291, 403Richards G. T. et al., 2003, AJ, 126, 1131Richards G. T. et al., 2006, ApJS, 166, 470Schawinski K., Virani S., Simmons B., Urry C. M., Treister E.,Kaviraj S., Kushkuley B., 2009, ApJ, 692, L19Shemmer O., Brandt W. N., Netzer H., Maiolino R., Kaspi S.,2008, ApJ, 682, 81Silva L., Maiolino R., Granato G. L., 2004, MNRAS, 355, 973Silverman J. D. et al., 2005, ApJ, 618, 123Silverman J. D. et al., 2008, ApJ, 675, 1025Simmons B. D., Urry C. M., 2008, ApJ, 683, 644Steffen A. T., Strateva I., Brandt W. N., Alexander D. M., Koeke-moer A. M., Lehmer B. D., Schneider D. P., Vignali C., 2006,AJ, 131, 2826Strateva I. V., Brandt W. N., Schneider D. P., Vanden Berk D. G.,Vignali C., 2005, AJ, 130, 387Sun W.-H., Malkan M. A., 1989, ApJ, 346, 68Tueller J. et al., 2009, ArXiv e-printsTueller J., Mushotzky R. F., Barthelmy S., Cannizzo J. K., GehrelsN., Markwardt C. B., Skinner G. K., Winter L. M., 2008, ApJ,681, 113Ulvestad J. S., Antonucci R. R. J., Goodrich R. W., 1995, AJ, 109,81Vasudevan R. V., Fabian A. C., 2007, MNRAS, 381, 1235Vasudevan R. V., Fabian A. C., 2009, MNRAS, 392, 1124Wang J., Mao Y. F., Wei J. Y., 2009, AJ, 137, 3388 c (cid:13) , 000–000 R.V. Vasudevan, R. F. Mushotzky, L. M. Winter & A. C. Fabian
Westoby P. B., Mundell C. G., Baldry I. K., 2007, MNRAS, 382,1541Winter L. M., Mushotzky R. F., Reynolds C. S., Tueller J., 2009,ApJ, 690, 1322Witt H. J., Czerny B., Zycki P. T., 1997, MNRAS, 286, 848
APPENDIX A: RESULTS FROM SED FITTING c (cid:13) , 000–000 ptical–to–X-ray emission in low-absorption AGN AGN M BH /M ⊙ Γ L − α ox , A E ( B − V ) L bol α ox , B λ Edd κ − (1) (2) (3) (4) (5) (6) (7) (8) (9)1RXS J045205.0+493248 .
04 1 .
86 43 . (none) . . − .
29 0 . +0 . − . +18 − .
68 1 .
85 44 . − .
18 0 . +0 . − . . − .
18 0 . +0 . − . . +0 . − .
3C 120 .
35 1 .
78 43 . − .
32 0 . . − .
32 0 . +0 . − . . +1 . − .
3C 390.3 .
35 1 .
75 44 . − .
13 0 . +0 . − . . − .
21 0 . +0 . − . . +0 . − . Ark 120 .
54 1 .
90 43 . − .
47 0 . +0 . − . . − .
40 0 . +0 . − . . +0 . − . ESO 490-G026 .
05 1 .
91 43 . − .
979 0 . +0 . − . . − .
28 0 . +0 . − . . +1 . − . ESO 548-G081 .
74 2 .
03 42 . − .
25 0 . . − .
26 0 . +0 . − . . +0 . − . IRAS 05589+2828 .
64 1 .
61 43 . − .
36 0 . +0 . − . . − .
34 0 . +0 . − . . +1 . − . IRAS 09149-6206 .
45 1 .
74 44 . − .
54 0 . + − . − . . − .
72 0 . +0 . − . . +6 . − . MCG +04-22-042 .
09 1 .
94 43 . − .
30 0 . . − .
29 0 . +0 . − . . +0 . − . MCG -06-30-015 .
25 1 .
62 42 . − .
785 0 . +0 . − . . − .
46 0 . +0 . − . . +1 . − . Mrk 1018 .
21 1 .
95 43 . − .
29 0 . . − .
28 0 . +0 . − . . +0 . − . Mrk 279 .
42 1 .
88 43 . − .
27 0 . . − .
25 0 . +0 . − . . +0 . − . Mrk 352 .
93 1 .
68 42 . − .
26 0 . +0 . − . . − .
47 0 . +0 . − . . +6 . − . Mrk 509 .
56 1 .
83 43 . − .
39 0 . +0 . − . . − .
38 0 . +0 . − . . +0 . − . Mrk 590 .
30 1 .
88 42 . − .
13 0 . +0 . − . . − .
22 0 . +0 . − . . +1 . − . Mrk 766 .
53 1 .
76 42 . − .
07 0 . +0 . − . . − .
51 0 . +0 . − . . +5 . − . Mrk 841 .
17 1 .
89 43 . − .
36 0 . +0 . − . . − .
37 0 . +0 . − . . +2 . − . NGC 4593 .
51 1 .
62 42 . − .
18 0 . +0 . − . . − .
26 0 . +0 . − . . +0 . − . NGC 5548 .
00 1 .
51 43 . − .
16 0 . +0 . − . . − .
25 0 . +0 . − . . +1 . − . NGC 7469 .
16 1 .
98 43 . − .
32 0 . +0 . − . . − .
33 0 . +0 . − . . +0 . − . NGC 985 .
36 1 .
80 43 . − .
36 0 . +0 . − . . − .
32 0 . +0 . − . . +0 . − . SBS 1136+594 .
62 1 .
94 43 . − .
24 0 . +0 . − . . − .
29 0 . +0 . − . . +1 . − . SBS 1301+540 .
25 1 .
81 43 . − .
14 0 . +0 . − . . − .
20 0 . +0 . − . . +0 . − . UGC 06728 .
44 1 .
82 42 . − .
26 0 . +0 . − . . − .
53 0 . +0 . − . +23 − WKK 1263 .
67 1 .
68 42 . − .
38 0 . +0 . − . . − .
43 0 . +0 . − . +28 − Table A1.
Results from SED fitting.(1) - Log of central BH mass from 2MASS K-band magnitude. Errors from K-band magnitudes translate into errors of ± . in values for log(M BH / M ⊙ ) ,but we refer the reader to the systematics discussed in the text.(2) - Photon index from fit to 0.3–10keV regime.(3) - Log of 2–10keV luminosity from power-law fit to 0.3–10keV regime.(4) - Spectral index α ox calculated by interpolation between available UVOT data points to determine the 2500 ˚ A luminosity.(5) - Intrinsic extinction E ( B − V ) from fitting ZDUST ( DISKPN ) model combination to UVOT data.(6) - Log of bolometric (0.001–100keV) luminosity, corrected for X-ray absorption and optical–UV dust reddening.(7) - Spectral index α ox calculated from the full optical–to–X-ray model fit, corrected for optical–UV dust reddening.(8) - Eddington ratio.(9) - Hard X-ray 2–10keV bolometric correction L bol /L − .Random error estimates are provided on values for intrinsic extinction, Eddington ratio and bolometric correction; random errors on all other quantities areomitted as systematic errors generally dominate.c (cid:13)000