Calibrating Emission Lines as Quasar Bolometers

1.9 MB·

by  Brian Punsly

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed26January2011 (MNLATEXstylefilev2.2)
Calibrating Emission Lines as Quasar Bolometers
Brian Punsly and Shaohua Zhang
1 4014 Emerald Street No.116, Torrance CA, USA 90503 and
1 ICRANet, Piazza della Repubblica 10 Pescara 65100, Italy,
0 E-mail: [email protected] or [email protected]
2
CAS Key Laboratory for Research inGalaxies and Cosmology,
n
Department of Astronomy, Universityof Sciencesand Technology of China,
a
J Hefei, Anhui, 230026, China

5
2
26January2011

]
O
C ABSTRACT
Historically, emission lines have been considered a valuable tool for estimating the
.
h bolometric thermal luminosity of the accretion flow in AGN, Lbol. We study the re-
p
liability of this method by comparing line strengths to the optical/UV continuum
-
o luminosity of SDSS DR7 radioquiet quasarswith 0.4<z <0.8.We find formulae for
r Lbol as a function of single line strengths for the broadcomponents of Hβ and Mg II,
st as well as the narrow lines of [O III] and [O II]. We determine the standard errors of
a theformulaethatarefittedtothedata.Ournewestimatorsareshowntobe moreac-
[ curatethanarchivallinestrengthestimationsintheliterature.Itisdemonstratedthat

1 the broad lines are superior estimators of the continuum luminosity (and Lbol) with
v Hβ being the mostreliable.The fidelityoftheeachofthe estimatorsisdeterminedin
3 the context of the SDSS DR7 radio loud quasars as an illustrative application of our
3 results. In general, individual researchers can use our results as a tool to help decide
8 if a particular line strength provides an adequate estimate of Lbol for their purposes.
4 Finally, it is shown that considering all four line strength, simultaneously, can yield
1. information on both Lbol and the radio jet power.
0
Key words: black hole physics — galaxies: jets—galaxies: active — accretion, ac-
1
cretion disks
1
:
v
i
X
r Theprimarymeasureofthestrengthofanactivegalac- when dusty molecular gas (e.g., the ”dusty torus”) ob-
a tic nucleus (AGN) is its bolometric luminosity, L , the scures the thermal emission produced by the quasar accre-
bol
broadband thermal luminosity from IR to X-ray.The char- tion flow from our view. The latter case is believed to be
acteristic signature of the thermal component is the ”big representative of NLRGs (narrow line radio galaxies) and
blue bump” a large blue/UV excess in the spectral con- Seyfert2galaxies (Antonucci1993).Historically, broad line
tinuum (Sunand Malkan 1989). Typically, one does not strengthshavebeenusedasasurrogateforcontinuumlumi-
have complete frequency coverage of the broadband con- nosity in blazars (Cao and Jiang 2001; Celotti and Fabian
tinuum and L can only be estimated. The most com- 1993;Celotti et al1997;Ghisellini et al2011;Gu et al2009;
bol
mon estimators involve single point rest frame optical or Maraschi and Tavecchio 2003; Wanget al 2004). For NL-
UV luminosity as an approximation to the big blue bump, RGs, the broad line region is not visible by definition and
Kaspi et al (2000), or the luminosity of the optical contin- narrow emission line strengths have been used as a surro-
uum as in Miller et al (1992). However, there are many in- gate for the continuum luminosity (Rawlings and Saunders
stances in which the optical/UV continuum is not directly 1991; Willott et al 1999).
observable or is contaminated with other sources of flux.
Thus, astrophysicists need a surrogate for the continuum TheserelationshipsbetweenlinestrengthandLbol have
luminosity. The two most prevalent situations that are en- never been carefully calibrated or scrutinized for their ac-
counteredare firstly,blazars in which thesynchrotron opti- curacy. This Letter intends to do both with a carefully
cal/UVemissionfromtherelativisticjetisDopplerboosted selected, large sample of SDSS DR7 radio quiet quasars
toabrightnessthateitherswampsthequasarthermalemis- (0.4<z<0.8) inwhichtwoprominentnarrowlines([OIII]
sion or contributes an unknown fraction of the total ob- λ5007 and [OII] λ3727) and two broad lines Hβ λ4861 and
served optical/UV flux. The other common occurrence is Mg II λ2798 are observable for all objects. This allows us
tocompare thevariousderived estimators in the contextof
2 Brian Punsly and Shaohua Zhang
Bolometric Luminosity as a Function of Hβ Broad Line Strength Bolometric Luminosity as a Function of Mg II Broad Line Strength
47.00 47.50
R² = 0.8123
47.00
46.50 R² = 0.6586
LOG(L) (ergs/sec)bol 4456..5000 LOG(L) (ergs/sec)bol 444566...505000
45.00
SDSS DR7 45.00 SDSS DR7
Wang et al (2004) Wang et al (2004)
44.50 44.50
41.50 42.00 42.50 43.00 43.50 44.00 44.50 45.00 41.50 42.00 42.50 43.00 43.50 44.00 44.50
LOG(LHβ) (ergs/sec) LOG(LMg II) (ergs/sec)
Bolometric Luminosity as a Function of O III Line Strength Bolometric Luminosity as a Function of O II Line Strength
47.50 47.00
R² = 0.3481
47.00 46.50
G(L) (ergs/sec)bol 4466..0500 G(L) (ergs/sec)bol 4456..5000
O 45.50 O 45.00
L L SDSS DR7
Willott et al (1999)
45.00 44.50
SDSS DR 7
R² = 0.1606
Rawlings and Sanders (1991)/Willott et al (1999)
44.50 44.00
41.00 41.50 42.00 42.50 43.00 43.50 44.00 40.00 40.50 41.00 41.50 42.00 42.50 43.00
LOG(LOIII) (ergs/sec) LOG(LOII) (ergs/sec)
Figure 1. Scatter plots of the logarithm of the line strengths as a function of the logarithm of Lbol = 15Lcont, where Lcont is the
optical/UVcontinuumluminosityfrom5100˚Ato3000˚A.ThefittotheSDSSDR7dataisgivenbytheblackline.Theaccuracyofthe
linear representation of the data is given by the coefficient of determination, R2, that is displayed on each plot. An archival estimator
from the literature is shown as a red line. Notice that the broad component of Hβ is an excellent representation of Lcont at the top
left.Thenextbestestimatorisbasedonthebroadcomponent ofMgIIinupperrighthandframe.The[OIII]λ5007basedestimatoris
clearlysuperiorto[OII]λ3727basedestimatorinthebottom row.
a single sample of objects, which removes the uncertainties 1969). So any FIRST flux density detection or upper limit
associated with sample selection biases if a different set of that implied R < 10 was considered radio quiet with 1.4
objects is used for the calibrations of each individual line GHz flux density used instead of 5 GHz flux density. We
strengthbasedestimator.Insection4,weconsiderthesees- also eliminated the low ionization broad absorption line
timators in the context of the radio loud subsample of the quasars from our sample since they are known to have
SDSSDR7quasars (0.4<z<0.8). anomalouslyweak[OIII]λ5007and[OII]λ3727emissionlines
(Boroson and Green 1992; Zhanget al 2010). In the end,
therewere6904radioquietsourcesremaininginoursample.
The spectral data of Hβ and Mg II regime were sep-
1 SAMPLE SELECTION
arately reduced using the procedures of Dong et al (2008);
In order to determine the dependence of line luminosity on Wanget al (2009) which see forfurtherdetails, andwe will
the thermal continuum in a quasar, we constructed a sam- onlybrieflyoutline ithere. TomeasuretheHβ lineand the
ple of SDSS DR7 radio quiet quasars with 0.4 < z < 0.8 [OIII]λ5007line,asinglepower-lawfittotheopticalcontin-
and spectra with a median S/N > 7, this yielded 10069 uum in the restframe wavelength range 4200−5600˚A was
AGNs.LongslitspectroscopyofradioloudAGNoftenshow obtainedtakingintoaccountcontributionsfrombothbroad
strong regions of narrow line emission on scales as large as and narrow Fe II multiplets that were modeled using the I
∼ 100 kpc that tend to be aligned with the jet direction Zw1 templates providedbyV´eron-Cetty et al. (2004).The
(Inksip et al 2002; Best et al 2000; McCarthy et al 1995). Hβ emission lines are modeled as multiple Gaussians: at
The magnitude of this contribution to the narrow line lu- most four broad Gaussians and one narrow Gaussian with
minosityislargelyunknown(Willott et al1999).Therefore, FWHM<900kms−1 forHβ,andoneortwoGaussiansfor
we segregated out theradio loud quasars becausethereis a each[OIII]λ4959,5007.TomeasuretheMgIIline,wefitthe
concernthatjetpropagationcanenhancethelinestrengths. continuumfromtheseveralcontinuumwindowsintherest-
The SDSS DR7 data was cross-referenced to the FIRST frame wavelength range 2200-3500˚A, after subtracting the
data base. The radio loudness, R, is usually defined as a estimated UVFeIIcontributionsbytheFeIItemplategen-
5GHzfluxdensity10timeslargerthanthe4400˚Afluxden- erated by Tsuzuki et al (2006) and the Balmer continuum
sity, R = S5GHz/S4400˚A < 10 (Kellermann & Pauliny-Toth using the method by Dietrich et al (2002). Each of the two
Calibrating Emission Lines as Quasar Bolometers 3
MgIIdoubletlinesismodeledwithonebroadfive-parameter linestrength based estimates in thepast.A 5100 ˚A estima-
Gauss-Hermite series component and one single Gaussian tor based on the composite of Punsly and Tingay (2006),
narrow component. Furthermore, the broad components of L ≈7.4λL (5100˚A),iscloser,butstillconsiderablylarger
bol λ
the doublet lines are set to have the same profile. The nar- thanthenormalizationofthearchivallinestrengthbasedes-
rowcomponentsaresettoFWHM<900kms−1andflux< timates. The Miller et al (1992) estimator (which is about
10% of the total Mg II flux. Finally, we fit the spectrum in 65% of the Kaspi et al (2000) value) seems to match the
the [OII] regime using one Gaussian for [OII] emission and normalization of theline strengthbased estimates thebest.
a single power-law for optical continuum in the restframe However, the poor archival fits are not just a consequence
wavelength range 3600-3800˚A. This process resulted in five of thenormalization, buttheslope of thefitsas well. If the
pieces of relevant information for every quasar: normalization were adjusted, the Wang et al (2004) fit for
Hβ would be fairly accurate.
• The optical/UV continuum luminosity from 5100 ˚A to
There is no unique criteria for the quality of the esti-
3000 ˚A, Lcont mator and this must be determined by each researcher per
• The luminosity of the broad component of Hβ λ4861,
their requirements. We choose a ”factor of two” as a figure
L
H•βTheluminosityofthebroadcomponentofMgIIλ2798, of merit (i.e. 0.5 < Lestimated/Lcont < 2, where Lestimated
is theestimated continuum luminosity) for illustrative pur-
LMgII poses. The Hβ based estimate is excellent with 97.4% of
• Theluminosity of thenarrow line [OIII]λ5007, LOIII the estimates accurate to within a factor of two. The other
• Theluminosity of thenarrow line [OII]λ3727, LOII estimators are accurate to within a factor of two in thefol-
lowingorder,MgII,[OIII],[OII];90.1%,80.3% and73.9%
of the time, respectively. The results are presented graphi-
callyinFigure2astheprobabilitydistribution(histogram)
2 LINE STRENGTH FITS TO CONTINUUM
oftheratioofestimated luminositytomeasuredcontinuum
LUMINOSITY
luminosity, Lestimated/Lcont.
The blue continuum luminosity is the most basic signature
of the thermal emission from the quasar, so it is the most
commonly used quantity for estimating L . Perhaps the
bol
3 APPLICATION TO RADIO LOUD QUASARS
most popular bolometric correction is the simple one pro-
posed by Kaspi et al (2000), Lbol ≈9λLλ(5100˚A). Clearly, Thereare2461quasarsintheSDSSDR7sample,0.4<z <
using a portion of the optical/UV continuum is more accu- 0.8,withaFIRSTdetectionandradioloudness,R>10.We
rate than a single point and we have that at our disposal, usethissampleas atest-bedfor theapplication of theesti-
Lcont.Forexample,asinglepointcouldlieinanoisyendof mators that were derived in the last section. We can check
theSDSSspectrum.Theaveragespectralindexinoursam- our calibrations of the bolometric correction by applying
ple of 6904 quasars, from 5100 ˚A to 3000 ˚A, is αν = 0.54, Equations(2)to(5)toeachlineindividuallyandcomparing
where αν is defined in terms of the spectral luminosity as theresultantLestimated toLcont.Theresultsarepresentedin
Lν ∼ ν−α. This spectral slope implies that the Kaspi et al Figure 3 as the probability distribution (histogram) of the
(2000) bolometric correction can be expressed as ratio of estimated luminosity to measured continuum lumi-
Lbol ≈15Lcont . (1) nosity, Lestimated/Lcont. Our ”factor of two” figure of merit
(i.e.0.5<Lestimated/Lcont <2)indicatesthattheHβ based
Equation(1)isrelativelyinsensitivetothechoiceofα,with estimate is still excellent with 94.9% of the estimates ac-
onlyafewpercentchangeintheconstantofproportionality curate to within a factor of two. The other estimators are
as α varies from 0.45 to 0.7. accuratetowithinafactoroftwointhefollowingorder,Mg
Figure 1 shows how well the line strengths represent II, [O III], [O II]; 86.2%, 75.1% and 67.5% of the time, re-
the continuum luminosity with the factor of 15 bolometric spectively.Figure3showsthatestimatorsbasedonthe[OII]
correctionfromequation(1).Thelinearfitsinlog-logspace and[OIII]emission linestrengthshaveapropensitytoover
aregivenbyequations(2)-(5)withthestandarderrorsfor estimate the continuum luminosity in radio loud quasars.
theintercept and theslope This is in contrast to the radio quiet quasars in Figure 2,
where the more errant estimates are equally likely to occur
log(L )=12.32±0.20+(0.78±0.01)log[L ], (2)
bol Hβ above or below ”1”. In other words, [O III] and [O II] are
log(Lbol)=16.76±0.26+(0.68±0.01)log[LMgII],(3) often much stronger than expected from the continuum lu-
log(Lbol)=26.50±0.32+(0.46±0.01)log[LOIII],(4) minosity alone. The implication is that the excess narrow
log(Lbol)=33.96±0.33+(0.29±0.01)log[LOII]. (5) linestrengthisproducedbytheaforementioned excitedgas
that is often observed aligned with the radio jets and is
The coefficients of determination, R2, are shown on the thereforelikelytobeenergizedbythejet(Inksip et al2002;
plots. The Hβ fit is the best, followed closely by Mg II Best et al 2000; McCarthy et al 1995). We also performed
with the narrow line fits considerably worse. Figure 1 con- thesame exercise with a cutoff at R>20. The results were
firms the implications of Simpson (1998) that LOIII rep- verysimilartothoseabove,indicatingthatthedefinitionof
resents Lcont more reliably than LOII does. We also plot radio loudness is a negligible factor in this analysis.
archival estimators of L from the literature to show the We give an example of how the line strengths can be
bol
improvement obtained by these more rigorous calibrations used to analyze radio loud AGN, the blazar 3C 216. We
of the data. It appears that the Kaspi et al (2000) normal- retrieve the line strengths from Lawrence et al. (1996). For
ization is considerably higher than what has been used for Hβ,MgII,[OIII]and[OII],fromequations(2)-(5),these
4 Brian Punsly and Shaohua Zhang
Accuracy of the Hβ Es(cid:129)mator Applied to Radio Quiet Quasars Accuracy of the Mg II Es(cid:127)mator Applied to Radio Quiet Quasars
0.3 0.3
0.25 0.25
Rela(cid:129)ve Frequency0.001..125 Rela(cid:127)ve Frequency000.1..125 Mg II
0.05 0.05
0 0
0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1
Les(cid:129)mator/Lcon(cid:129)nuum Les(cid:127)mator/Lcon(cid:127)nuum
Accuracy of the O III Es(cid:127)mator Applied to Radio Quiet Quasars Accuracy of the O II Es(cid:127)mator Applied to Radio Quiet Quasars
0.3 0.3
0.25 0.25
ncy 0.2 ncy 0.2
ue ue
q q
e Fre0.15 O III e Fre0.15 O II
(cid:127)v (cid:127)v
Rela 0.1 Rela 0.1
0.05 0.05
0 0
0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 4.1 4.3 4.5 4.7 4.9 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 4.1 4.3 4.5 4.7 4.9
Les(cid:127)mator/Lcon(cid:127)nuum Les(cid:127)mator/Lcon(cid:127)nuum
Figure 2. The histograms represent the distributions of the ratio of the estimated continuum flux to the actual continuum flux,
Lestimated/Lcont foreach oftheemissionlinesinthe context ofthe SDSSDR7radioquietsubsample.Thesmallestdispersionabout 1
(the most accurate estimator) is Hβ in the upper left hand frame. The superior Hβ fit is shown in black in the other three frames for
thesakeofcomparison.
line strengths yield the following estimates for Lbol, 8.12× studyingthejetpower,environmentandaccretion powerin
1044ergs/sec,1.23×1045ergs/sec,7.24×1045ergs/sec,1.32× AGN.
1046 ergs/sec, respectively. 3C 216 is argued to be the one TheresultsofthisLetterareapplicabletoblazars, but
ofthemostkineticallydominatedknownquasars,withajet perhapsnottoNLRGsandSeyfert2galaxies.Thereiscom-
power(kineticluminosity)toL ratio>10(Punsly2007). pelling evidence that sometimes the same gas that can at-
bol
The broad line estimates are in good agreement, but the tenuate the nucleus in these sources can also attenuate the
narrow line estimates are much larger. The results of this narrow line emission (Mulchaey et al 1994; Kraemer et al
paper indicate that the broad lines are therefore excellent 2010). In general, without additional information, one does
estimatorsofthecontinuumluminosityandthenarrowlines not know if the narrow line region in a particular Seyfert 2
are dominated by a jet contribution. galaxyorNLRGisattenuatedornot,thusthe[OIII]and[O
II]basedestimatorsarenotreliableinisolation.Weshowed
insection4thatthenarrowlineestimatorsareactuallyuse-
ful in theblazar context,where attenuation is not an issue.
4 CONCLUSION In general, narrow line based estimators are most useful if
complementedbyotherinformationsuchasbroadlinelumi-
We have derived line strength based estimators for L for
bol nosity or theactual continuumluminosity. For NLRGs and
radio quiet quasars from the broad components of Hβ, Mg
Seyfert 2 galaxies, the IR continuum luminosity is a valu-
II,and thenarrow lines, [O III]and [O II] in Equations (2)
ablecomplement tothenarrowlinestrengthssinceit isnot
-(5).Thestrength of thebroad component of Hβ is asupe-
believed to be attenuated (Fernandeset al 2010; Ogle et al
rior estimator of Lcont for either radio quiet or radio loud 2006).
quasars since it is accurate towithin afactor of two, 97.4%
and94.9% oftime,respectively.Thenextbestlinestrength
based estimators are in order of accuracy, Mg II, [O III]
REFERENCES
and [O II]. We applied our results to radio loud quasars
and found strong evidence that the narrow line based esti- Antonucci, R.J. 1993, Annu. Rev. Astron. Astrophys. 31
matesareoftenskewedbywhatislikelyastrongjetinduced 473
contribution. It was also demonstrated that using all four Best,P.,Rottgering,H.,Longair,M.2000, MNRAS3111
line strengths in tandem can be a useful diagnostic tool for Boroson, T. and Green, R.1992, ApJS80 109
Calibrating Emission Lines as Quasar Bolometers 5
Accuracy of the Hβ Es(cid:129)mator Applied to Radio Loud Quasars Accuracy of the Mg II Es(cid:127)mator Applied to Radio Loud Quasars
0.25 0.25
0.2 0.2
Rela(cid:129)ve Frequency0.01.15 Rela(cid:127)ve Frequency00.1.15 Mg II
0.05 0.05
0 0
0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9
Les(cid:129)mator/Lcon(cid:129)nuum Les(cid:127)mator/Lcon(cid:127)nuum
Accuracy of the O III Es(cid:127)mator Applied to Radio Loud Quasars Accuracy of the O II Es(cid:127)mator Applied to Radio Loud Quasars
0.25 0.25
0.2 0.2
ncy ncy
ue0.15 ue0.15
q q
e Fre O III e Fre O II
Rela(cid:127)v 0.1 Rela(cid:127)v 0.1
0.05 0.05
0 0
0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 4.1 4.3 4.5 4.7 4.9 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 4.1 4.3 4.5 4.7 4.9
Les(cid:127)mator/Lcon(cid:127)nuum Les(cid:127)mator/Lcon(cid:127)nuum
Figure 3. The histograms represent the distributions of the ratio of the estimated continuum flux to the actual continuum flux,
Lestimated/Lcont foreach of the emissionlines inthe context of the SDSS DR7radioloud subsample. Thesmallestdispersionabout 1
(themostaccurateestimator)isproducedbytheHβbasedestimateintheupperlefthandframe.ThesuperiorHβ fitisshowninblack
intheotherthreeframesforthesakeofcomparison.
Cao, X.and Jiang, D.2001, MNRAS320 347 Punsly,B. 2007, MNRAS 374 10
Celotti, A.and Fabian, A.1993, MNRAS264 228 Punsly,B., Tingay, S. 2006, ApJL 651 L17
Celotti,A.,PadovaniandGhisellini,G.1997,MNRAS286 Rawlings, S., Saunders, R.1991, Nature349, 138
415 Simpson, C. 1998, MNRASL297 39
Dietrich, M., Appenzeller, I., Vestergaard, M., & Wagner, Sun,W.H., Malkan, M. 1989, ApJ 346 68
S.J. 2002, ApJ, 564, 581 Tsuzuki, Y., Kawara, K., Yoshii, Y., Oyabu, S., Tanab´e,
Dong, X., Wang, T., Wang, J., Yuan, W., Zhou, H., Dai, T., & Matsuoka, Y. 2006, ApJ, 650, 57
H., & Zhang, K. 2008, MNRAS,383, 581 V´eron-Cetty,M.-P.,Joly,M.,&V´eron,P.2004,AAP,417,
Fernandes, C. 2010, MNRASin press arXiv:1010.0691v1 515
Ghisellini, G., Tavecchio, F., Foschini, L., Wang, J.-M., Luo, B, Ho, L. 2004, ApJL 615 9
Ghirlanda, G. 2011, submitted to MNRAS Wang, J.-G., et al. 2009, ApJ,707, 1334
http://xxx.lanl.gov/abs/1012.0308 Willott,C.,Rawlings,S.,Blundell,K.,Lacy,M.1999,MN-
Gu, M., Cao, X. and Jiang, D. 2009, MNRAS396 984 RAS309 1017
Inskip,K. et al 2002, MNRAS337 1318 Zhang, S. et al 2010 ApJ714 367
Kaspi, S.et al 2000, ApJ533 631
Kellermann,K.I.,&Pauliny-Toth,I.I.K.1969ApJ,155,
L71
Kraemer, S. et al 2010, to appear in ApJ
http://arxiv.org/abs/1011.5993v1
Lawrence, C. et al 1996, ApJS 107 541
Maraschi, L. and Tavecchio F., 2003, ApJ593 667
McCarthy,P.,Spinrad,H.,vanBruegel,W.1995,ApJS99
27
Miller, P., Rawlings, S., Saunders, R. and Eales, S. 1992,
MNRAS254 93
Mulchaey, J. et al 1994, ApJ436 586
Ogle,P.,Whysong,D.andAntonucci,R.2006ApJ647161