Super- and sub-Eddington accreting massive black holes: A comparison of slim and thin accretion discs through study of the spectral energy distribution

4.4 MB·

by  Nuria Castello-Mor

MNRAS000,1–21(2015) Preprint23February2016 CompiledusingMNRASLATEXstylefilev3.0
Super- and sub-Eddington accreting massive black holes: A
comparison of slim and thin accretion discs through study
of the spectral energy distribution.
N. Castello´-Mor1(cid:63), H. Netzer1 and S. Kaspi1,2
1School of Physics and Astronomy, Tel Aviv University, Tel Aviv 69978, Israel
2Wise Observatory, School of Physics and Astronomy, Tel Aviv University, Tel Aviv 69978, Israel
6
1
0 AcceptedXXX.ReceivedYYY;inoriginalformZZZ
2

b
e ABSTRACT
F WeemployopticalandUVobservationstopresentSEDsfortworeverberation-mapped
samples of super-Eddington and sub-Eddington AGN with similar luminosity distri-
1
butions. The samples are fitted with accretion disc models in order to look for SED
2
differences that depend on the Eddington ratio. The fitting takes into account mea-

] sured BH mass and accretion rates, BH spin and intrinsic reddening of the sources.
A All objects in both groups can be fitted by thin AD models over the range 0.2-1µm
G with reddening as a free parameter. The intrinsic reddening required to fit the data
are relatively small, E(B −V) ≤ 0.2 mag, except for one source. Super-Eddington
.
h AGN seem to require more reddening. The distribution of E(B − V) is similar to
p what is observed in larger AGN samples. The best fit disc models recover very well
-
the BH mass and accretion for the two groups. However, the SEDs are very different,
o
r with super-Eddington sources requiring much more luminous far-UV continuum. The
t exact amount depends on the possible saturation of the UV radiation in slim discs. In
s
a particular, we derive for the super-Eddington sources a typical bolometric correction
[ at 5100˚A of 60–150 compared with a median of ∼20 for the sub-Eddington AGN.

3 The measured torus luminosity relative to λLλ(5100˚A) are similar in both groups.
v The α distribution is similar too. However, we find extremely small torus cover-
OX
7 ing factors for super-Eddington sources, an order of magnitude smaller than those of
7 sub-EddingtonAGN.Thesmalldifferencesbetweenthegroupsregardingthespectral
1 range 0.2-22µm, and the significant differences related to the part of the SED that
7
we cannot observe may be consistent with some slim disc models. An alternative ex-
0
planation is that present day slim-disc models over-estimate the far UV luminosity of
.
1 such objects by a large amount.
0
6 Key words: accretion,accretiondiscs–galaxies:nuclei–galaxies:Seyfert–galaxies:
1 active
:
v
i
X
r 1 INTRODUCTION Eddingtonratios(L /L ≤0.3,Koratkar&Blaes1999;
a AGN Edd
Blaesetal.2001;Shangetal.2005;Davisetal.2007;Davis
Optically thick accretion flows in the vicinity of the central
& Laor 2011; Laor & Davis 2011; Jin et al. 2012; Slone &
black hole (BH) are believed to be the main power-house
Netzer 2012; Netzer & Trakhtenbrot 2014; Capellupo et al.
ofactivegalacticnuclei(AGN).Theemittedradiationfrom
2015).Athigheraccretionrates,thediscbecomes“thick”or
such systems is determined by the BH mass (M ), BH
BH “slim”and the nature of the accretion changes dramatically
spin (a ), and the mass accretion rate, M˙. The accretion
(cid:63) with processes like photon trapping and advection becom-
ratecanbeexpressedinnormalizedunits,orEddingtonra-
ingimportant(Abramowiczetal.1988;Sa¸dowskietal.2014,
tio,m˙ =L /L ,whereL isthebolometricluminos-
AGN Edd AGN see review by Wang et al. 2014b). Slim ADs are thought to
ity of the system. Optically thick geometrically thin accre-
have SEDs that are different from thin ADs, with an en-
tiondiscs(ADs)havebeenproposedtoexplaintheobserved
ergy cut-off that extends to higher energies and a strong
spectral energy distribution (SED) of many AGN with low
anisotropy of the emitted radiation. We follow Wang et al.
(2014a) and coin such objects“Super-Eddington Accreting
Massive Black Holes”(SEAMBHs).
(cid:63) E-mail:[email protected](TAU)
c 2015TheAuthors
(cid:13)
2 Castell´o-Mor et al.
TheradiationefficiencyofthinADsaroundBHs,η,de- temptstodiscovertheuniquespectralsignatureofthinADs
fined by L = ηM˙ c2, is obtained from the standard failed as a result of AGN variability and non-simultaneous
AGN BH
AD theory (e.g. Shakura & Sunyaev 1973; Thorne 1974). data that cover only a limited wavelength range. All the
η depends on the location of the innermost stable circular sources in the Capellupo et al. (2015) sample are relatively
orbit (ISCO) which in turn depends on the BH spin. For low accretion rate systems with 90% of the sources in the
maximally rotating BHs, with spin parameter a = 0.998, range 10 2 <m˙ <0.3 and not a single source with m˙ >1.
(cid:63) −
η=0.32 and for retrograde discs, with a =−1, η=0.038. Themostaccurateinformationaboutm˙ intype-IAGN
(cid:63)
This is not the case for slim ADs that are not well under- is obtained for sources with directly measured BH mass
stood. In such cases, there is an ill-defined radiation effi- through reverberation mapping (RM). Such information is
ciencythatmaynotdependontheBHspinandissuggested now available for about 40 objects with m˙ <0.1 (see Bentz
to be considerably smaller than the corresponding thin AD etal.2013)andfor15objectswithm˙ >0.1(Duetal.2015,
efficiency due to the so-called“photon trapping”. The issue theobjectswerefertoasSEAMBHs).Thecomparisonofthe
isparticularlyimportantforveryhighaccretionrateswhere two groups suggests that for a given optical luminosity, the
the theoretical models suggest that the emitted radiation emissivity weighted radius of the broad line region (BLR)
follows an expression of the type (Mineshige et al. 2000) is considerably smaller in SEAMBHs with differences that
can amount to a factor of ≈3. As explained in Wang et al.
(cid:104) (cid:105)
L ≈3×1038 1+lnM˙/M˙ M (1) (2014a) and Du et al. (2015), this can be interpreted as a
AGN crit BH
change in the nature of the power-house where the lower m˙
where M˙ is the dimensionless accretion rate defined as systemsarepoweredbythinADsandthehigherm˙ objects
m˙ = ηM˙ and M˙ ≈ 20 (Du et al. 2015). Such theo- by slim ADs.
crit
reticalapproximationsareyettobeconfirmedbynumerical There are several other well known difference between
simulations. For example, the recent numerical simulations high and low m˙ systems. The first is the nature of their X-
of such objects by Sa¸dowski & Narayan (2015) suggest a ray continuum manifested by higher variability amplitude,
relatively small drop in efficiency up to extremely high ac- steepersoftX-rayslopeandlargerα (theslopeconnect-
OX
cretion rates. Finally, all AGN are known to be powerful ing the flux at 2500˚A and 2 keV) for the higher m˙ objects.
X-ray sources with L luminosity that, in most ob- These differences have been studied, extensively, in numer-
2 10keV
served cases, is consid−erably below the integrated optical- ouspapers(Pogge2000;Steffenetal.2006;Gallo2006),es-
UVluminosity(e.g.Marconietal.2004;Steffenetal.2006; pecially among lower luminosity AGN. Clear spectroscopic
Vasudevan & Fabian 2007, 2009; Grupe et al. 2010; March- differencesarealsofoundatoptical-UVwavelengthswhere,
ese et al. 2012). The dimension and variability of the X-ray foragivenmonochromaticluminosity,thehigherm˙ sources
source suggest that it is intimately related to the accretion show narrower broad emission lines and much stronger Feii
process, thus the emitted X-ray radiation is most probably lines. Such objects have been named“narrow line Seyfert 1
drawnfromtheaccretionprocessitself.Thismustbetaken galaxies”(NLS1s).Theirspectraldifferencesarewellcharac-
into account when comparing the radiation efficiencies of terizedinthe“eigen-vector1”scheme(Sulenticetal.2000).
thin and slim ADs. Thegoalofthepresentworkistomakeadetailedcom-
The observed optical-UV SEDs of some AGN resem- parisonoftheoptical-UVSEDsofsub-Eddingtonandsuper-
blethepredictionofthethinADmodel.Inparticular,they Eddington AGN. The number of accurately measured BH
show a prominent bump in the optical-ultraviolet, the Big mass in SEAMBHs has reached a stage where such a com-
Blue Bump (BBB), which peaks at a BH-mass dependent parisonisfeasibleandcanbeusedtotestvariousmodelsand
frequencyanddeclinesathigherenergies.Unfortunately,the scenariosrelatedtotheAGNpower-houseandtheaccretion
comparison with theoretical AD SEDs has been hampered process itself. In particular, we want to test whether thin
bythelackofsimultaneousobservationsofthesehighlyvari- and slim AD models reliably explain the observed optical-
able sources, and the limited wavelength range of most ob- UV SED of such objects and the properties of their dusty
servations. Another source of uncertainty is intrinsic red- tori,andtestthedifferences,ifany,betweenthetwogroups.
dening due to dust in the host galaxy of the AGN. Because An important aim of the present work is to improve the X-
of this, many earlier studies failed to reach a conclusion re- ray-to-optical SED measurements and hence the estimates
garding the origin of the observed SED with only marginal ofthebolometricluminosityofAGNinsourcespoweredby
indications for a disc-like spectrum (see Koratkar & Blaes thinorslimADs.ThepioneeringstudyofElvisetal.(1994)
1999; Davis et al. 2007). Moreover, an empirical SED made presented radio-to-X-ray SEDs for quasar, is restricted to
of a broken power-law with slopes −0.5 < α < 1.5, where brightX-raysourcesandcoverboththeintrinsicAGNcon-
Fν ∝ν−α, have been shown to give better fits to many ob- tinuumandtheprocessedtorusemission.Unfortunately,the
served, limited wavelength spectra (Zheng et al. 1997). processedradiationisaddedtotheintrinsicemissionwhich
The recent work of Capellupo et al. (2015) shed new results in double counting and too large bolometric correc-
lightonthisissue.TheworkisbasedonVLT/Xshooterob- tion factors. In later works, double counting was avoided,
servationsof39type-IAGNwithz≈1.55andalargerange but intrinsic reddening and host galaxy contribution was
of L and m˙. They provide simultaneous information nottakenintoaccount(seee.g.Marconietal.2004;Hopkins
AGN
on the rest-frame wavelength range of 1100-9200˚A which is etal.2004;Richardsetal.2006;Vasudevan&Fabian2007).
large enough to test the AD model predictions. More than This affects the bolometric luminosity measurements, un-
90%ofthesourcesinthissamplearewellfittedwithathin derestimatingitsvalue.Morerecentworks(Brocksoppetal.
AD model. The fitting requires moderate intrinsic redden- 2006;Vasudevan&Fabian2009,Vasudevanetal.2009here-
ing in ∼30% of the sources with extinction in the range afterV09,Jinetal.2012,amongothers),presentedoptical-
0.1 < A < 0.5mag. This indicates that many earlier at- to-X-ray SED emphasizing interesting trends between SED
V
MNRAS000,1–21(2015)
super-Eddington AGN SED 3
signatureandEddingtonratio.However,mostsampleswere dington ratio sample consists of 16 radio-quiet AGN that
restricted to low accretion rate AGN. In particular, there are basically all the potential SEAMBH candidates listed
was a lack of super-Eddington sources with reliable, RM- in the Du et al. (2015) sample. The BLR sizes of 13 of
basedBHmassestimatessincesuchobjectswerenotknown the sources were measured in a two-year RM-campaign on
at the time. theLijiang2.4mtelescopebythe“SEAMBHCollaboration”.
ThesampleofAGNpresentedhereisrestrictedtothose Theobjectswereselectedbytheirspectroscopicproperties.
AGNwhicharebrightenoughintheoptical/UVandallhave TheyareallNLS1swithsuspectedhighEddingtonratio,m˙.
directmassmeasurements.Thereare29objectsselectedby Theselectionassumesaconservativemethodtoestimatethe
their spectroscopic properties with 16 of those showing ex- normalized accretion rate M˙ ≥3. The value was chosen to
tremelyhighaccretionrates(referredtoassuper-Eddington make sure that m˙ for the lowest efficiency discs (those with
AGN). This makes our approach different than all previous spin -1 and η = 0.038) exceeds 0.1 (Du et al. 2014; Wang
studies. From all previous studies, the work of V09 is the et al. 2014a; Du et al. 2015). Three additional SEAMBHs
one most similar to ours. It takes into account reddening candidates,PG2130+099,PG0844+349andMrk110,with
andhostgalaxycorrectionsintheoptical/UV,butfocuson similar properties, were selected from the general catalogue
low accretion rate sources and does not discuss the torus of ∼50 RM-AGN. As of early 2015, these are the best can-
properties. Below we show the similarity between V09 and didates for being super-Eddington accretors. We coin this
our study regarding sub-Eddington AGN, and the missing sample the“super-Eddington group”.
informationaboutthetorus,andmakedetailedcomparison We have also selected a control sample of 13 type-I
with the new group of super-Eddington sources. The most AGN from the general group of ∼50 RM objects with a
recent works of Marchese et al. (2012) and Fanali et al. much smaller Eddington ratio (defined in the same way)
(2013) are other attempts to correlate accretion disc and andM˙ <3.Thisgroup,namedassuper-Eddington,wasse-
X-ray properties of luminous AGN. There are several fun- lectedtooverlapinopticalluminosity,definedasλL atthe
λ
damental differences between these papers and the present rest-frame wavelength 5100˚A, the luminosity of the super-
work. First, the BH mass estimate is based on single-epoch Eddington group. The luminosity distributions of the two
(virial) method. Second, the accretion disc model is not a groupsarecomparedinFigure1.AlthoughtheKolmogorov-
fit to individual SEDs but rather the same model for all Smirnov test can not reject the hypothesis that both lu-
sources with an assumed stationary BH but an efficiency η minosity distributions are drawn from the same distribu-
of0.1whichismoreappropriatetorotatingBH.Third,and tion, there is a factor two on its average rest-frame 5100˚A
mostimportant,thethreebandsusedfortheestimatebolo- luminosity which can not be improved due to the limited
metric luminosity (X-ray XMM-Newton, UV GALEX and RM-AGNsample (1044 erg/s and4.8×1044 erg/s for super-
opticalSDSS)arenotobtainedsimultaneouslywhich,aswe Eddington and sub-Eddington groups, respectively). The
showbelow,introducesanuncertaintywhichisimpossibleto redshift distribution of the sub-Eddington group is quite
estimate.Becauseof thesedifferenceswedonotattemptto similar to the range covered by our super-Eddington ob-
comparetheresultsobtainedinthesepaperstotheonepre- jects. The median redshifts are 0.054 and 0.031 for the
sentedbelow.Recently,Jinetal.(2016)presentedanewre- super-Eddington and the sub-Eddington samples, respec-
sultforasuper-EddingtonsourceRXJ1140.1+0307,where tively. Figure 1 also shows the distribution of the virial BH
upperlimitontheBHmasswasobtainedthroughRM(the massinthetwosamples.Duetotheselectionmethodwhich
upperlimitisontheHβ timelag).Jinetal.(2016)present is based on source luminosity, the typical BH mass in the
several SED fits based on accretion disc models using non- control sample is much larger.
simultaneousdata.Therearealsodifficultiesinhostgalaxy
subtractionandreddeningcorrection.Jinetal.(2016)found
thatthebest-fitSEDoverestimatethemassoftheblackhole 2.1 Photometric and Spectroscopic Data
byanorderofmagnituderelativetotheRMresults.Because
Wehavecollectedoptical-UVphotometricandspectroscopic
of these and the fact that the source is in the lowest black
datafromvarioussurveysanddatabases.Thedataareused
holemassregimewedonotattempttocomparetheirresults
todefineandseparatestellarfromnon-stellarsourcesofcon-
to the ones presented here.
tinuumradiationandtoconstrainthepropertiesoftheSEDs
Thestructureofthepaperisasfollows.InSection2we
of all sources. The optical spectrum of 17 RM objects were
describethesampleselectionandtheobservationaldata.In
takenfromtheSloanDigitalSkySurvey(SDSS)DR10.For
section3wedescribetheintrinsicdiscSEDthatweuseand
the remaining 12 sources, we used a mean optical spectrum
thefittingprocedureispresentedinsection4.Theprincipal
from previous reverberation mapping campaigns. From the
results of the thin AD modelling and some additional spec-
AGN Watch database1 we selected the mean spectrum of
tral properties are presented in Section 5. Finally, in Sec-
Mrk 279, Mrk 509, NGC 7469 and Fairall 9. The mean op-
tion 6, we summarize our main conclusions from this work.
ticalspectrumofMrk79andMrk817weregivenbyPeter-
ThroughoutthispaperweassumeaΛCDMcosmologywith
sonM.B.(privatecommunication,Petersonetal.1998).The
Ω =0.7, Ω =0.3, and H =70 km/s/Mpc.
Λ m 0 mean optical spectrum of the Palomar Green (PG) quasars
(PG 2130+099, PG 0844+349, and PG 1617+175) were
2 SAMPLE SELECTION
1 The AGN Watch is a consortium of astronomers who
Thegoalofthisworkistocomparevariouspropertiesofhigh
have studied the inner structure of AGN through continuum
Eddington ratio AGN with RM-measured BH mass with a and emission-line variability. See http://www.astronomy.ohio-
similar group of low Eddington ratio AGN. Our high Ed- state.edu/ agnwatch/
∼
MNRAS000,1–21(2015)
4 Castell´o-Mor et al.
10 10 10
8 8 8
6 6 6
N N N
4 4 4
2 2 2
0 0 0
42 43 44 45 6 7 8 9 0.00 0.05 0.10 0.15 0.20
logλLλ(5100A˚) logMBH z
Figure 1. From left to right: Distributions of logλLλ(5100˚A), virial black hole mass, and redshift for both samples. Filled (blue)
histogramrepresentssuper-EddingtonAGNandhatched(grey)histogramsub-EddingtonAGN.Thepointwitherrorbarsonthetopof
eachpanelisthemedianandthe16th and84th percentilesforeachgroup.
taken from Kaspi et al. (2000). Finally, the mean optical AGN. For this we need to correct for non-simultaneous ob-
spectrumofMrk486,Mrk335andMrk1044arethosepub- servations,tosubtractthestellarandemissionlinecontribu-
lishedbythe“SEAMBHCollaboration”(Duetal.2014).All tions, and to consider the possibility of intrinsic reddening
objectsinoursample,exceptforMrk79andPG1617+175, ofthesources.AsforGalacticinterstellarreddening,thisis
have high quality optical spectra with broad optical wave- done assuming the Cardelli et al. (1989) extinction law us-
length range. ingtheGalacticextinctioncolourexcessE(B−V)obtained
All objects in this work, except for Mrk 486, from the NASA/IPAC Infrared Science Archive2.
PG 0844+349 and SDSS J080101.14+184840.7, have NUV
and FUV photometry (2306˚A and 1551˚A, respectively) ob-
tainedbytheNASAGalaxyEvolutionExplorer(GALEX). 3.1 Long term optical-UV variability
TheGALEXfluxeswerecollectedfromtheonlinedatabase
The presence of variability in the optical/UV continuum of
and all image were inspected visually. We found 23 ob-
allthesourcesinoursampleiswelldocumented.Whilstwe
jects with observations in the all-sky imaging survey (AIS).
canexpectvariabilityoforderafewpercentintheoptical-
In addition, three objects (Mrk 1044, IRAS 04416+1215,
UVcontinuumonshort-timeofdaystoaweek,muchlarger
Mrk 590) are also included in the medium imaging survey
factorsareexpectedontimescalesofmonthsandyears.For
(MIS). No discrepancy was found between both observa-
example, Santos-Lleo´ et al. (1997) found a factor of ∼2 for
tions (a difference between AIS and MIS photometry less
Fairall 9 in the optical continuum flux over a three month
than 0.15 magnitudes in both GALEX bands). Three out
periodandevenlargerfactorshavebeenfoundforNGC5548
of these 23 objects have been detected in different surveys:
(a factor of ∼7 over a seven years period, Peterson et al.
PG 2130+099 was found in the MIS, while NGC 5548 and
1999). Therefore variability is the main challenge for the
PG 1229+204 were selected from the Guest Investigator
work presented here.
(GI)program.For17objectswealsoassembledoptical/UV
The information that is available for most sources is a
photometryfromtheXMM-NewtonOpticalMonitor(OM):
combination of optical spectroscopy and GALEX photom-
V 5430˚A, B 4500˚A, U 3440˚A, UVW1 2910˚A, UVM2 2310˚A
etry. Most of these data are not simultaneous. In order to
andUVW22120˚A.Fromvisuallyinspectionoftheavailable
avoidasmuchaspossiblethenon-simultaneity,andtocom-
Hubble Space Telescope (HST) data we cannot extend our
pletetheUVGALEXinformation,weuseddataobtainedby
datasetintotheextremeUVbecauseofvariability,i.e.there
the XMM-Newton Optical Monitor (OM). OM photometry
isatleastafactor2onfluxbetweentheHSTspectrumand
was available for 17 out of 29 objects and contemporane-
the selected optical/UV dataset.
ous optical/UV observations (at least 5 simultaneous OM
Summarizing,ofour29sourceswehaveopticalspectra
points) for only 6 sources (see Table 1). These contempora-
inconjunctionwithGALEXphotometryfor26sourcesand
neous optical/UV observations provide reliable constraints
16sourceswhich,inadditiontothesedata,alsohaveXMM-
on the source SEDs and help to reduce the fitting uncer-
OM photometry, and in 8 cases simultaneous optical/UV
tainty (see Appendix B).
photometry. A detailed list of all the observations is given
Theheterogeneityofouroptical/UVdatasetsdictatea
in Table 1.
somewhat different procedure for each of the fitted sources.
Threedifferentprocedureswereused:i)asimultaneousand
non-simultaneous SED were fitted in order to quantify the
non-contemporaneous SED shape; ii) when simultaneous
3 OBSERVED AND INTRINSIC SEDS
A central goal of the present work is to compare the multi-
wavelength SEDs of super-Eddington and sub-Eddington 2 http://irsa.ipac.caltech.edu/applications/DUST/
MNRAS000,1–21(2015)
super-Eddington AGN SED 5
Table1.OpticalandUVobservationsusedtomodeltheSEDofsuper-Eddington(upperblock)andsub-EddingtonAGN(lowerblock).
OpticalSpectrum GALEX XMM-NewtonOM
Object E(B−V)Gal survey date survey date XMMOM date
Mrk142 0.016 SDSSDR10 2003-03-09 AIS 2004-01-22
Mrk335 0.035 SEAMBHCollaboration 2012Oct–2013Feb AIS 2007-04-02 J000619.5+201211 2007-07-10
Mrk382 0.048 SDSSDR10 2001-10-19 AIS 2007-01-21 J075525.3+391110 2011-11-02
Mrk486 0.015 SEAMBHCollaboration 2013Mar–2013Jul
Mrk493 0.025 SDSSDR10 2004-05-16 AIS 2007-05-26
Mrk1044 0.033 SEAMBHCollaboration 2012Oct–2013Feb MIS 2008-10-11
IRAS04416+1215 0.436 SDSSDR10 2006-12-17 MIS 2008-12-17
IRASF12397+3333 0.019 SDSSDR10 2005-03-02 AIS 2004-04-15 J124210.6+331702 2005-06-20
SDSSJ075101.42+291419.1 0.042 SDSSDR10 2002-12-28 AIS 2007-01-18
SDSSJ080101.41+184840.7 0.032 SDSSDR10 2004-11-10
SDSSJ081441.91+212918.5 0.039 SDSSDR10 2004-11-18 AIS 2006-01-27
SDSSJ081456.10+532533.5 0.032 SDSSDR10 2004-10-19 AIS 2004-01-13
SDSSJ093922.89+370943.9 0.014 SDSSDR10 2003-12-23 AIS 2005-08-12 J093922.9+370945 2006-11-01
PG2130+099 0.044 Kaspietal.(2000) 1991Aug–1997Oct MIS 2009-08-22 J213227.8+100819 2003-05-16
PG0844+349 0.037 Kaspietal.(2000) 1991Aug–1997Oct J084742.4+344504 2009-05-03
Mrk110 0.012 SDSSDR10 2001-12-09 AIS 2007-01-23 J092512.9+521711 2004-11-15
Mrk79 0.071 Petersonetal.(1998) 1983–1985 AIS 2007-01-20 J074232.8+494835 2008-04-26
Mrk279 0.016 AGNWatch 1996Dec–1997Jan AIS 2004-01-25 J135303.5+691830 2005-11-19
Mrk290 0.014 SDSSDR10 2002-03-14 AIS 2005-10-22 J153552.3+575409 2006-05-04
Mrk509 0.057 AGNWatch 1988Sep–1993Dec AIS 2007-02-09 J204409.7-104324 2009-11-14
Mrk590 0.037 SDSSDR10 2003-01-08 MIS 2008-10-23
Mrk817 0.007 Petersonetal.(1998) 1983–1985 AIS 2007-02-21
Mrk1511 0.041 SDSSDR10 2007-04-17 AIS 2007-04-17 J153118.1+072729 2012-02-23
NGC5548 0.020 SDSSDR10 2006-05-04 GII 2006-10-30 J141759.5+250813 2013-07-29
NGC7469 0.068 AGNWatch 1996Jun–1996Jul AIS 2007-03-22 J230315.6+085226 2004-11-30
PG1229+204 0.027 SDSSDR10 2008-01-16 GII 2007-05-22 J123203.7+200929 2005-07-09
PG1617+175 0.042 Kaspietal.(2000) 1991Aug–1997Oct AIS 2006-06-14
Fairall9 0.025 AGNWatch 1994May–1995Jan AIS 2007-08-17 J012345.7-584820 2013-12-19
MCG+06-26-012 0.019 SDSSDR10 2005-03-31 AIS 2007-03-22
dataset was not possible, both GALEX and OM photom- et al. 2014a). Published galaxy light profiles were used in
etry in conjunction with the optical spectrum were fitted thesecasesandnormalizedtotheapertureinquestion.This
as long as there is no signs of variability, i.e. all segments correctionisnormallysmall(20-70%)forthe3 SDSSfiber
(cid:48)(cid:48)
of the fitted continuum join smoothly; iii) for those sources andconsiderablylargerforthe6 apertureoftheopticalOM
(cid:48)(cid:48)
without XMM-Newton OM observations we fitted the non- filter.Thesecondmethodmakesuseofempiricalexpressions
simultaneous optical-GALEX SED assuming they were ob- derived by Shen et al. (2011) and Elvis et al. (2012). The
served at a similar flux epoch. Comments on individual uncertainty here is much larger because of the considerable
sources are given in Appendix B. This general procedure range in host properties. Finally, one can use the fact that
works well in all the objects discussed in the present paper the broad Hβ lines show no Baldwin effect (Dietrich et al.
although non-simultaneity is still the main limiting factor. 2002).Thismeansthattheobservedequivalentwidthofthe
line,EW(Hβ),canbeusedtoderivethefractionofthenon-
AGNlightenteringtheapertureat4861˚A.Duetal.(2015)
3.2 Host Galaxy Contribution show the distributions of EW(Hβ) in low Eddington ratio
AGN and in SEAMBHs. The two differ by a considerable
Theoptical/UVemissionmaybecontaminatedbystarlight
amountwithmediansthatare(122±44)˚Aand(89±31)˚A,
from the host galaxy. The relative contribution depends on
respectively.ForthosesourceswithoutHSTobservations,we
the aperture size, AGN luminosity, and the stellar popula-
prefer the use of this method over the expressions given by
tioninthehost.Thuscarefulgalaxysubtractionisnecessary
Shenetal.(2011).Themainconcernishowtotreatthefive
if we are to determine the SED shape. Such corrections are
objects without HST observations.
often made by subtracting galaxy bulge template spectra.
Star formation in the central part must also be considered The subtraction procedure starts with an estimate of
sincethehostofmanyAGNarestarformationgalaxies(e.g. the non-AGN flux at 4861˚A followed by a subtraction of a
Sani et al. 2010). single simple stellar model spectrum which is scaled to this
Therearethreemethodsthatcanbeusedtocorrectthe flux. A nominal 20% uncertainty on this flux was adopted
observed spectra for host-galaxy contamination. The first due to a combination of the host-galaxy modelling and the
is a direct subtraction of the stellar light measured from uncertainties on the emission line fluxes. The SDSS fibers
HST images. Such data are available for 25 of the sources include only the innermost few kpc of the host, hence we
(Petersonetal.1998;Bentzetal.2009;Duetal.2014;Wang chose a quiescent galaxy model from the evolutionary spec-
MNRAS000,1–21(2015)
6 Castell´o-Mor et al.
tral library of Charlot & Bruzual (1991). For 15 out of 29 (SMC) curves, a simple power-law extinction, or a combi-
objects,theinstantaneous-burstmodelwithanageof11Gyr nation of a power law with a flatter curve in the far-UV
andsolarmetallicity(Z=0.02)providedsufficientlygoodfit (Gaskell et al. 2004). Several earlier works (Hopkins et al.
to the stellar spectrum. For the remaining 14 objects, this 2004; Glikman et al. 2012) claimed that the typical bump
templategivesa“fluxexcess”atthelongestobservedwave- at2175˚AoftheMW-likeextinctioncurveisnotobservedin
lengths.Atemplatewith11GyrandZ=0.05providesbetter AGNspectra.Recent,higherqualityspectraandmorecare-
fit to such spectra and was adopted in these cases. We also fullyfittedSEDsbyCapellupoetal.(2015)clearlyshowthis
experimented with adding younger stellar population com- bump in some of the spectra.
ponents but did not find significant improvements over the Sincethewavelengthdependenceoftheextinctionisun-
old population templates. known,apriori,weexperimentedwiththreepossibilities:i)
ThedifferenceonaperturebetweentheSDSSfibres(3 ) theCardellietal.(1989)asaMW-likecurvewithR =3.1,
(cid:48)(cid:48) V
andtheOMbroadbandfilters(12 and35 diameterforthe ii)asimplepower-lawA(λ)=A λ 1,andiii)theSMCex-
(cid:48)(cid:48) (cid:48)(cid:48) 0 −
opticalandUVfilter,respectively)preventusfromusingthe tinction curve as given in Gordon et al. (2003). According
estimated host-galaxy contribution to the optical spectrum to the wavelength dependence of the extinction curve, the
to subtract the star-light from the OM photometric data. intrinsic reddening will be larger for the SMC curve than
For the OM photometric dataset, the host galaxy emission for the MW. The most consistent approach is to add this
was modeled by adopting the published galaxy-light radial extinctionasanadditionalparameterintheSEDmodelling
profiles (Bentz et al. 2009) which were integrated over the analysis.Clearly,ifextinctionisimportant,theintrinsicun-
specific OM aperture to estimate the host-galaxy contribu- derlying continuum shape will depend significantly on the
tion to the total observed flux. form of the reddening curve. As a practical point we note
thattheextremelybroadGALEXbandsdonotallowusto
takeintoaccountspectralfeatureslikethe2175˚Aabsorption
3.3 Emission line Contributions and weaker ISM lines and we only treat the total observed
flux in these bands.
The wavelength range for the photometry are wide enough
that the underlying nuclear continuum might be contami-
natedbybroadandnarrowemissionlines,i.e.Balmerlines,
Civ, Mgii, Feii, as well as the Balmer continuum.
4 SED MODEL FITTING
In order to exclude the underlying continuum, we used
thecompositequasarspectrumofVandenBerketal.(2001) 4.1 Accretion disc models
to estimate the emission line fraction for each photometric
Inthiswork,weusethenumericalcodedescribedinSlone&
waveband.Weassumedonlytheemittedfluxinthespectral
Netzer(2012)tocalculatethinADspectra.Thecalculations
windowoverwhichtheeffectivetransmissionisgreaterthan
10% of the peak effective transmission: FUV 1343–1786˚A, assumeaShakura&Sunyaev(1973)discwithavariablevis-
NUV1771–2831˚A,UVW21805–2454˚A,UVM21970–2675˚A, cosityparameter(choseninthispapertobeα=0.1).Asin
UVW1 2410–3565˚A, U 3030–3890˚A, B 3815–4910˚A and allmodelsofthistype(seeSec.1),thespin-dependentISCO
V 5020–5870˚A. For comparison, at redshift z =0.06, which determines the mass-to-energy conversion efficiency, η. The
focusoftheSlone&Netzer(2012)workistheeffectofdisc
isthemeanofoursample,thecontributionsfromtheemis-
winds on the emitted SEDs. Here we do not consider disc
sionlineregionsateachphotometricwindoware5%,11.9%,
winds because of the lack of far UV spectroscopy required
9.4%, 13.0%, 23.4%, 24.6%, 15.9% and 24.9% for FUV,
to deduce their presence and because we focus on slim disc
NUV, UVW2, UVM2, UVW1, U, B and V, respectively.
where there are additional large uncertainties (see below).
Sincelinevariationsfollowcontinuumvariations(albeitwith
AsinSlone&Netzer(2012),ourcalculationsincludeComp-
somewhatdifferentlags),wedonotexpectlargefluctuations
tonizationoftheemittedradiationateverypointinthedisc
in these fractions.
atmosphere and,for BH spin values of a > 0, full General
(cid:63)
Relativistic corrections. For retrograde discs with a < 0,
(cid:63)
the general relativity effects are not included, which is a
3.4 Intrinsic reddening
fair approximation given the large size of the ISCO (>6r ,
g
Intrinsic extinction in AGN can be significant which can wherer =GM /c2isthegravitationalradiusoftheblack
g BH
make it an important factor when determining the SED of hole). Apart from Comptonization, the calculations do not
the optical/UV continuum. Previous studies (Lusso et al. include any other radiative transfer in the disc atmosphere
2013; Capellupo et al. 2015; Collinson et al. 2015, Mejia- that can make significant changes to the far UV SED (e.g.
Restrepo et al. 2015 (submitted), and references therein) Davis & Laor 2011, Fig. 2).
found a range of properties with most AGN showing lit- Standard,Shakura&Sunyaev(1973)thinaccretiondisc
tledustattenuation,i.e.E(B−V)<0.1,independentofthe models are limited to m˙ ≤ 0.3. Beyond this accretion rate,
reddeninglaw.Suchreddeningdoesnotaffectmuchtheop- the disc geometry becomes thick and many of the approxi-
tical fluxes, but its effect on the UV part of the spectrum mationusedinthemodelno-longerhold(seeLaor&Netzer
is significantly larger. Not accounting for this effect will re- 1989,andreferencestherein).AsexplainedinSection1,the
sultinfittingthewrongaccretiondiscmodelsandthemiss- maindifferencesbetweenthinandslimdiscsarethethicker
calculation of source luminosity and accretion rate. geometry, due to the much larger radiation pressure in the
Extinction curves which are commonly used in AGN disc,andthewaytheradiationescapesthesystem.Thisin-
studies include: Milky Way-like extinction with its known cludes radial advection and perhaps saturation of the emit-
broad bump at ∼2175˚A, Small Magellanic Cloud-type ted luminosity. Unfortunately, detailed SED calculations of
MNRAS000,1–21(2015)
super-Eddington AGN SED 7
slim discs are highly simplified and hardly available. A re- maximum disc temperature, there is very little difference
cent detailed model of this type is discussed in Wang et al. between different spin value. We therefore show, for each
(2014a).Accordingtothisandearliermodels,thelongwave- source only two values, a =−1 and a =0.998.
(cid:63) (cid:63)
lengthpartoftheSEDoriginatesoutsideofthethickpartof The fitting procedure includes the comparison of the
thediscandis,therefore,verysimilarinitsshapetothethin observed SED with various combinations of disc SEDs cov-
ADSED.ForsmallBHmasssystems,significantdifferences ering the range of mass, accretion rate, and the two cho-
betweenthetwoappearonlyatveryshortwavelengths,be- sen spins and assumed reddening. The reddening is taken
yondtheLymanlimit.Thespecialgeometrydictatesstrong into account by changing E(B−V) in steps of 0.004 mag,
anisotropy in such systems, much beyond the standard AD calculating, for each value, a new mass accretion rate and
anisotropyduetoinclination.This,again,ismostnoticeable its range of uncertainty. A simple χ2 procedure was used
atshortwavelengths(seeFig.4inWangetal.2014b).The to find the best-fit combination of reddening and thin AD
observations discussed here do not include the wavelength models. We use at least three line-free windows covering
rangebelowλ=1000˚AandhenceweusethethinADmodel the optical spectroscopic data, and all the available photo-
for the fitting of the SED over this range. Later on, when metric data. The line-free windows are centred on 4205˚A,
we discuss the Lyman continuum emission, we consider the 5100˚Aand6855˚A,withwidthsrangingfrom10to30˚A.For
various possibilities regarding slim discs. those objects with no photometric data, three additional
As explained earlier, theoretical slim disc models are line-free continuum windows were used: 5620˚A, 6205˚A and
still highly simplistic and are not in very good agreement 6860˚A. For the error on each continuum point, we combine
with state-of-the-art numerical calculations like those of thestandarderrorfromthePoisonnoise,anassumed5%er-
Sa¸dowski & Narayan (2015). Given this fundamental un- ror on the flux calibration, and the relative error of 20% on
certainty,wetreatallderivedquantitiesthatdependonthe thecombinationoftheuncertaintiesonthehost-galaxycon-
short wavelength part of the model, e.g. the bolometric lu- tribution and the unknown stellar population. To allow for
minosityandEddingtonratio,asthemostuncertainparam- thelargeelapsedtimebetweentheGALEXandtheoptical
eters for the sub-sample of 16 SEAMBHs described in this observations, which increases the uncertainty due to source
work. variability, we added an uncertainty of 20% to the GALEX
fluxes. Note that this is not meant to take into account the
realvariationsbetweenepochsincethisisdealtwith,inour
4.2 Fitting procedure
special method described in Section 3.1 where we provide
WeusedtheSlone&Netzer(2012)codedescribedearlierto moreinformationaboutthewayweusedthebestSEDthat
calculatealargenumberofthindiscspectrathatincludethe avoids, as much as possible, the variability issue. We refer
entire range of BH mass, accretion rate and spin expected the reader to Appendix B for a detailed description of the
inoursample.Theinputforthefitinclude,foreachsource, fitting results for individual sources.
theblackholemass(M inunitsofM ),andthemassac-
BH
cretionrate(M˙ inunitsofM /yr).We(cid:12)usedtheRM-based
masses and the measured (fro(cid:12)m the 5100˚A continuum) ac-
cretion rates listed in Du et al. (2015, see Table 7). The
uncertainty on the mass is estimated to be a factor of ∼3
5 RESULTS
due to the the uncertainties on the measured time lags, the
measuredFWHM(Hβ),andtheuncertaintiesintheconver- ThesamplesdescribedherewereselectedfromtheRMAGN
sion of observed broad line profiles to a“mean gas veloc- sample with high accretion rate, M˙ ≥3 (super-Eddington)
ity”(thef terminM =f cτFWHM(Hβ)2/G). and with sub-Eddington accretion rate M˙ < 3. In the fol-
BLR BH BLR
The accretion rate is obtained from the luminosity at rest- lowing we use the normalized Eddington ratio, m˙ = ηM˙,
wavelength 5100˚A using the method described in Davis & whichallowaneasycomparisonwithpreviousworks.While
Laor(2011)asdetailedinNetzer&Trakhtenbrot(2014,see not all data are of the same quality, we were able to secure
Eqn.1there).Weassumedafactorthreeuncertaintyonthe photometricandspectroscopicdatathatcoverthe0.2-20µm
accretionratederivedinthiswaywhichisacombinationof range for all sources. For 8 sources (4 sub-Eddington and
flux uncertainty (mostly stellar light subtraction), the un- 4 super-Eddington) we have both simultaneous and non-
known inclination, and the fact that the chosen wavelength simultaneous optical/UV photometry. In the following dis-
(5100˚A) is not on the L ∝ν1/3 part of the SED. cussions we do not distinguish between simultaneous and
ν
Allsourcesfittedinthisworkaretype-IAGNwhichare non-simultaneous SEDs.
assumedtobeobservedclosetoface-on.Therangeofincli- Our best fitted disc SEDs are shown in Figure A1 and
nationisroughly0–60degreeandtherangeincos(θ)0.5–1. allthemodelparameters,includingthevirializedM and
BH
We assumed the anisotropy function proposed by Netzer & M˙,arelistedinTableA1.Twomodelsarelistedpersource,
Trakhtenbrot (2014), corresponding to the minimum and maximum spin param-
eter. The error bars on the intrinsic reddening enclose the
f F cosθ(1+b(ν)cosθ)
f(θ)= 0 ν =f (2) 68% confidence range. Finally, the average spectral proper-
F (face-on) 0 1+b(ν)
ν ties for both groups and for the entire sample are summa-
with b(ν) = 2, f = 1.2×1030 erg/s/Hz, and cosθ = 0.75 rized in Table 2. In the following plots, for every derived
0
for all sources. The remaining parameters are the BH spin thinADmodelparameterwedisplaythemeanbetweenthe
andintrinsicreddening.Weexperimentedwiththefullrange higherandlowervaluescorrespondingtotheminimumand
of spin parameters, from -1 to 0.998. Because of the fitted maximum spin, and the uncertainly which represents the
range of wavelengths, which is far from the frequency of range of possible values.
MNRAS000,1–21(2015)
8 Castell´o-Mor et al.
Table 2.MedianParametersforsuper-Eddington(13),sub-Eddington(16)AGNandtheentiresample(29).Theuncertaintiesreflect
the16th and84th percentiles.Whensaturation(Eqn.1)istakenintoaccounttheparameteristaggedwiththetermsat.Twonumbers
arelistedforeachthinADmodelcorrespondingtotheminimumandthemaximumspin.Theunabsorbedluminosityat2keVisgiven
bythebest-fitpowerlawoverthehardX-rayband(2-10keV).
parameter super-Eddington sub-Eddington all
a(cid:63)=0.998 a(cid:63)= 1 a(cid:63)=0.998 a(cid:63)= 1 a(cid:63)=0.998 a(cid:63)= 1
− − −
reverberation-mappedresutls
logλLλ(5100˚A) 44.00+00..2639 44.00+00..2639 43.68+00..3521 43.68+00..3521 43.71+00..5505 43.71+00..5505
z 0.054+−0.078 0.054+−0.078 0.031+−0.017 0.031+−0.017 0.035+−0.081 0.035+−0.081
0.022 0.022 0.009 0.009 0.009 0.009
logMBH 6.86−+00..3358 6.86−+00..3358 7.97−+00..1731 7.97−+00..1731 7.16−+00..8551 7.16−+00..8551
log ˙ 1.61−+0.98 1.60−+0.98 0.70−+0.35 0.72−+0.37 0.84−+1.24 0.83−+1.23
M −0.77 −0.77 − −0.38 − −0.36 −1.68 −1.69
power-lawmodelling
βUV 2.13+00..3376 2.13+00..3376 2.12+00..3360 2.12+00..3360 2.12+00..3394 2.12+00..3394
E(B V)/PL/ 0.00−+0.06 0.00−+0.06 0.00−+0.00 0.00−+0.00 0.00−+0.00 0.00−+0.00
− −0.00 −0.00 −0.00 −0.00 −0.00 −0.00
thinADmodelling
E(B V) 0.07+0.06 0.06+0.06 0.01+0.03 0.00+0.00 0.05+0.04 0.04+0.06
logM−BH 7.25−+000...204836 7.26−+000...204626 7.88−+000...106310 7.80−+000...405402 7.32−+000...605552 7.36−+000...505946
log ˙ 1.81−+0.46 1.80−+0.58 0.62−+0.98 0.32−+0.66 0.36−+1.72 0.50−+1.62
logMLLyman 46.55−+010...449632 45.57−+010...629226 −44.96−+000...445540 −43.70−+000...359286 45.59−+110...208971 44.48−+111...401031
logLLyman,sat 45.84−+00..1399 45.78−+01..2197 44.96−+00..4550 43.70−+00..3926 45.52−+00..3774 44.48−+11..4131
hν 11.8−+2.7 3.2−+1.0 2.4−+2.7 1.3−+0.5 5.3−+8.5 1.8−+2.3
(cid:104)κ510(cid:105)0˚A 208+−41777.501 61−+175.135 17−+045.35 −30+.711 62+−34376.03 10−+097.50
κ5100˚A,sat 53−+5312 65−+6517 18−+456 3−+71 32−+5119 10−+975
logλLλ(5100˚A)(erg/s) 44.23+00−..3572 43.93+00−..4447 43.64+00−..6333 43.63+00..−6441 43.90+00−..5591 43.83+00−..5515
logλLλ(2500˚A)(erg/s) 44.51+−00..2245 44.48+−00..2285 43.95−+00..2245 43.89+−00..2297 44.15−+00..4463 44.12−+00..4449
LAGN (erg/s) 46.56−+00..4960 45.65−+00..5779 45.07−+00..4494 44.28+−00..3454 45.65−+10..2738 44.84−+10..1702
LAGN,sat (erg/s) 45.89−+00..1492 45.82−+00..2995 45.07−+00..4494 44.28−+00..3454 45.56−+00..3689 44.84+−10..1712
∆∆llooggM˙B‡H† 00..3214−+−+000...152984 00..3113−+−+000...142972 −00..1519−+−+000...321739 −00..0544−+−+000...211954 00..1051−+−+000...363486 00..1193−+−+000...243891
M − −0.33 − −0.32 −1.06 −0.62 − −0.55 −0.46
Luminosities
Ltrous (erg/s) 44.52+00..5306 44.52+00..5306 44.53+00..3477 44.53+00..3477 44.53+00..4484 44.53+00..4484
logLν(5µm)(erg/s) 43.99+−00..5308 43.99+−00..5308 43.86+−00..5329 43.86+−00..5329 43.86−+00..6325 43.86−+00..6325
logLν(2keV)(erg/s) 43.23−+00..1400 43.23−+00..1400 42.85−+00..5470 42.85+−00..5470 43.19+−00..2624 43.19−+00..2624
− − − − − −
† Definedas∆logMBH ≡(logMBH)fit−(logMBH)virial
‡ Definedas∆logM˙ ≡(logM˙)fit−(logM˙)virial
5.1 Accretion Disc SED Severalofthedeclaredgoodfitsstillincludesmalldeviations
ofthemodelfromthelocalcontinuumatsomewavelengths.
This is not surprising given the uncertainties on AD mod-
Amajorgoalofthisprojectistodeterminewhatfractionof
els, especially the radiative transfer in the disc atmosphere
sub-Eddington and super-Eddington accretors in our sam-
that was not treated here, as well as on the choice of the
ple can be fit by the simple optically thick, geometrically
host galaxy template. In general, the simple χ2-fitting pro-
thin AD model based on their long wavelength SEDs. By
cedurecannotdistinguishbetweena =−1anda =0.998,
allowing intrinsic reddening as a free parameter of the thin (cid:63) (cid:63)
given the assumed uncertainties on BH mass, BH accretion
AD model, we can fit all the 29 sources of our sample. For
rate, reddening and the long wavelengths used for the fit-
23 sources, our modelling requires some reddening and for
ting. Observations at shorter wavelengths (i.e. extreme-UV
6 sources, which are all sub-Eddington AGN, the amount
λ<2000˚A) are clearly required to make such a distinction.
of reddening is consistent with zero. We found that 14% of
the RM-selected AGN are consistent with E(B−V)> 0.1 Forconsistencywithpreviousworks,the0.2-1µmSED
in good agreement with the work of Krawczyk et al. (2015) was also fitted using a single reddened power-law model
wheretheSEDwasmodelledbyasinglepowerlaw.Indeed, (Lν ∝ ν−αUV). A satisfactory fit (χ2ν < 2) with AV = 0
the reddening distribution of our entire sample is also con- isfoundfor86%(25/29)ofthesources.Theremainingfour
sistentwiththatpresentedbyLussoetal.(2013,Figure3). objects (which are super-Eddington) were successfully fit-
MNRAS000,1–21(2015)
super-Eddington AGN SED 9
more significant in super-Eddington AGN. Figure 3 com-
αUV
paresthereddeningdistributionofthesuper-Eddingtonand
1.0 0.5 0.0 0.5 1.0 1.5
0.4 − − − thesub-EddingtongroupsassumingaSMCextinctioncurve.
The median reddening for the super-Eddington sources is
(cid:104)E(B−V)(cid:105)=0.07magwithascatterofσ=0.08,whilethe
sub-Eddington AGN sample shows (cid:104)E(B−V)(cid:105)≈0.01 mag
0.3
with σ = 0.01. A Kolmogorov-Smirnov test rules out the
hypothesis that both groups are drawn from the same dis-
tribution of E(B−V) with a probability of > 99.9%. Ex-
perimenting with other extinction curves give basically the
N 0.2
same distribution shown in Figure 3.
0.1 5.3 Uncertainties in black hole mass
In order to evaluate the goodness of our fit, we have anal-
ysedtheblackholepropertiesrecoveredbythethinADfit.
We find that the measured“virialized”mass black hole and
0.0
1.0 1.5 2.0 2.5 3.0 3.5 massaccretionrate,withinthealloweduncertainties,leadto
βUV a suitable disc model for all the 29 AGN. The comparisons
are shown in Figure 4. The median deviation of the black
Figure2.ThedistributionofopticaltoUVspectralindicesinthe holemass(∆logMBH ≡logMBH,fit−logMBH,virial)forthe
super-Eddingtonandsub-Eddingtonobjects.Thespectralindices super-Eddington sources is 0.305dex with a 16th and 84th
aredefinedsuchthatLλ∝λ−βUV andβUV =2−αUV.Colour percentileof0.07and0.5dex,whilethemedianforthesub-
codeasFigure1. Eddington sources is -0.07dex with its percentiles at -0.23
and 0.25dex. There is no systematic shift on the mass ac-
cretion rate which is uniformly distributed. This is because
ted with A > 0. The reddening distribution of the entire
V
the accretion rate in the disc model is directly related to
sample is quite consistent, as in the case of the thin AD
the observed luminosity and the best mass estimate. Note
modelling, with that presented in various earlier studies,
also that the assumed inclination of cos(θ) = 0.75 only in-
e.g. Lusso et al. (2013). The range of slopes is large, from
troduce an additional uncertainty on the fitting results and
α = 0.55 to α = −0.98. The Kolmogorov-Smirnov
UV UV
cannotexplaintheshiftontheblackholemassobservedin
test shows that the optical to UV spectral index corrected
the super-Eddington group.
for intrinsic reddening for SEAMBHs and sub-Eddington
Finally,itisinterestingtonotethatourthinADfitrecover
accretors are fully consistent (see Figure 2), with means
almost perfectly both the black hole mass and the mass ac-
(cid:104)α (cid:105) = −0.14 and −0.10 and a standard deviations of
UV
cretionrate.Thisisincontrastwithwhathasbeenfoundin
σ = 0.40 and 0.33, respectively. A comparison with earlier
the work of Calderone et al. (2013), which claims that the
works on much larger samples (Vanden Berk et al. 2001;
best-fitblackholemassisonaverageafactorof∼6greater
Grupe et al. 2010) shows that our sample has bluer opti-
than the corresponding virialized mass estimation.
cal/UV continua.
5.2 Properties of the obscuring dust 5.4 Ionizing Continuum
IftheadoptedthinADmodeldoesindeedexplaintheemit- An additional consistency check of the fitted SEDs can be
tedSEDofAGNinoursample,thendustinthehostgalax- obtained by studying the disc ionizing continuum which is
ies of approximately 50% of the AGN in our sample is con- directly related to the observed emission lines. The relative
tributingtothereddeningoftheoptical-UVspectrum.Itis intensity of the lines, as well as their equivalent widths are
therefore interesting to test the nature of the dust which is relatedtothemeanenergyoftheionizingphotons,theion-
causingtheextinction.Severalearlierstudies(Hopkinsetal. ization parameter and the covering factor by gas near the
2004; Glikman et al. 2012) claimed that an SMC-type ex- BH. The hydrogen Lyman continuum as the part of the
tinctioncurvebestaccountsforthereddening,inpreference SED responsible for most of the heating and ionization of
to MW-type or Gaskell et al. (2004) reddening, Capellupo the broad emission-line gas can be derived from our disc
etal.(2015)findthatintheirsampleof39z=1.5AGN,there SEDs. In particular, we can calculate both Lyman contin-
werenocaseswheretheSMC-typecurveallowedforabetter uum, L , and the mean energy of an ionizing photon,
Lyman
fitthaneithertheMWorsimplepower-lawextinctioncurve. (cid:104)hν(cid:105), and compare the distribution of these numbers in the
Our observations indicate that for those AGN that require two groups.
intrinsicreddening,thereisnopreferenceforoneparticular A comparison of the Lyman continuum properties of
extinction curve. This is probably because the UV observa- the two groups is shown in Figure 5 where both parame-
tions are based on photometry instead of spectroscopy and tersareplottedagainstthenormalizedEddingtonratio.For
mayalsobetheresultofthelimitedwavelengthrangewhich comparison,themeanenergyofanionizingphotonwithan
is essential to differentiate between the various extinction optical/UV power-law SED of slope α = 1.5 is 2.31 Ryd.
ν
laws. Asexpected,thesuper-Eddingtongroup,withitshigherac-
As noted above, the amount of extinction seems to be cretion rate and smaller BH mass, are predicted to have
MNRAS000,1–21(2015)
10 Castell´o-Mor et al.
1.0 1.0
N N
0.5 0.5
0.0 0.0
0.0 0.5 1.0 0.0 0.5 1.0
E(B−V)power−law E(B−V)thinAD
Figure 3. Intrinsic reddening E(B-V) distribution modelled by a classical SMC-like extinction curve with RV =2.74. The amount of
reddening was derived from model fittings assuming a single power law (left) and a thin AD model (right). The red crosses with error
bars represent the reddening distribution for the entire sample (29 objects) and the green circles with error bars show the distribution
ofE(B V)presentedbyLussoetal.(2013).HistogramcolourcodeasinFig.1.
−
largerL /L andhigher(cid:104)hν(cid:105).Thediagramsshowa
9.0 Lyman AGN
very strong trends with the mass accretion rate. This is in
line with the results of V09, although their sample is in the
8.5 lower-Eddington ratio regime. The average ionization frac-
tion is ∼0.5 for the lowest Eddington ratio M˙ ≤3 (similar
](cid:12)8.0 with V09) and rise to ∼0.9 for the highest M˙ >3.
M
[ Here,andinallotherresultsrelatedtotheLymancon-
)H,fit 7.5 triencutiuomn,fatchteorssofatndX-troaryusemcoivsseiroinn,gafnacdtotrhsedbisocluosmseedtribcelcoowr-,
B
M we must take into account two important factors related to
og( 7.0 the highly simplified assumption of the high energy part of
l
the SED in super-Eddington sources. First, the entire disc
6.5 structure canchange athighaccretionrate,leadingtotwo-
componentdiscliketheoneproposedbyDoneetal.(2012).
This possibility is not investigated here, partly because of
6.0
6.0 6.5 7.0 7.5 8.0 8.5 9.0 the large uncertainties on such models and mostly because
log(MBH,virial) [M ] of the lack of short wavelength observations. Second, sat-
(cid:12)
uration, if exists, affects the far-UV SED much more than
3 thenear-UVandopticalcontinuum(Duetal.2015,andref-
erences therein). In this scenario the bolometric luminosity
2 for sources with M˙ > 20 has only a weak logarithmic de-
pendency on the mass accretion rate (Eqn. 1). Such values
1]− 1 of M˙ are found in most of the super-Eddington sources in
r our sample (11/16, see Table A1). The case of saturation
y
(cid:12) is compared with the thin AD predictions in Figure 5, and
M
[ 0 comparison with the other relevant quantities are shown in
)
˙Mfit Figs. 6 and 7. Indeed it results in lower Lyman continuum
g( 1 bolometriccorrectionfactorsandtoruscoveringfactors(see
lo − below).However,allthesearestillconsiderablyhigherthan
the values estimated for the sub-Eddington sources. Note
2
− thatwedonothaveasimplewayofestimatingthechanges
inthemeanenergyoftheionizingphotonsincasesofsatu-
3 ration and hence do no comment of this additional effects.
− 3 2 1 0 1 2 3
− − − In principle, the higher expected L /L and
log(M˙virial) [M yr−1] Lyman AGN
(cid:12) (cid:104)hν(cid:105) for super-Eddington sources could be investigated by
comparinglineintensitiesandEWsbetweenthetwogroups.
Figure 4. Comparison of the virialized black hole mass (top)
One prediction is that, given similar BLR covering factors,
andthedimensionlessmassaccretionrate(bottom)againstthose
super-Eddington sources will show larger line EWs and
given by our best-fit thin AD model. The straight lines indicate
stronger lines of highly ionized species. While this general
the1:1ratios.Gray-bluecirclesrepresentsuper-EddingtonAGN
andgray-emptysquaressub-EddingtonAGN. issue is beyond the scope of the present work, we note that
such differences in EWs have never been reported. In fact,
several studies show tends in the opposite directions (see
MNRAS000,1–21(2015)