Type 1 Active Galactic Nucleus Fraction in SDSS/FIRST Survey
aa r X i v : . [ a s t r o - ph . C O ] F e b Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 31 October 2018 (MN L A TEX style file v2.2)
Type 1 Active Galactic Nucleus Fraction in SDSS / FIRST Survey
Yu Lu, Ting-Gui Wang, Xiao-Bo Dong, Hong-Yan Zhou CAS Key Laboratory for Research in Galaxies and Cosmology, University of Science and Technology of China, Center for Astrophysics, University of Science and Technology of China, Hefei, Anhui, 230026, P.R.ChinaHefei, Anhui, 230026, China
31 October 2018
ABSTRACT
In the unification scheme, narrow lined (type 2) active galactic nuclei (AGN) are intrinsicallysimilar to broad lined (type 1) AGN with the exception that the line of sight to the broademission line region and accretion disk is blocked by a dusty torus. The fraction of type 1 AGNmeasures the average covering factor of the torus. In this paper, we explore the dependenceof this fraction on nuclear properties for a sample of low redshift ( z .
35) radio strong( P . > W Hz − ) AGN selected by matching the spectroscopic catalog of Sloan DigitalSky Survey and the radio source catalog of Faint Image of Radio Sky at Twenty cm. Aftercorrecting for several selection e ff ects, we find that : (1) type 1 fraction f keeps at a constantof ∼
20% in the [O iii ] luminosity range of 40 . < log( L [OIII] / erg s − ) < .
5. This result issignificantly di ff erent from previous studies, and the di ff erence can be explained by extinctioncorrection and di ff erent treatment of selection e ff ects. (2) f rises with black hole mass from ∼
20% below 10 M ⊙ to 30% above that. This coincides with the decrease of the fraction ofhighly-inclined disk galaxies with black hole mass, implying a population of Seyfert galaxiesseen as Type-2 due to galaxy-scale obscuration in disk when the host galaxy type transferfrom bulge-dominant to disk-dominant. (3) f is independent of the Eddington ratio for itsvalue between 0.01 and 1; (4) f ascends from 15% to 30% in the radio power range of23 < log( P . / WHz − ) <
24, then remain a constant at ∼
30% up to 10 W Hz − . Key words: quasars: general—galaxies
Active Galactic Nuclei (AGN) are traditionally divided into 2 sub-classes: type 2 and type 1, according to the absence or presenceof the broad emission lines. Their di ff erent observed properties canbe explained largely via anisotropic obscuration, likely by a dustytorus on scales of several to tens parsecs, of otherwise the sametype of objects (Antonucci 1993; Tran 1995; Nenkova et al. 2002).Because of its large physical scale, Narrow emission Line Region(NLR) cannot be (completely) obscured by such torus. Therefore,an AGN will appear as a type 2 source, when our line of sight tothe Broad emission Line Region (BLR) and the accretion disk isblocked by the dusty torus. Scattered broad lines have been detectedin the polarized light in almost half of type-2 AGN, which stronglysupports the unification scheme (Antonucci & Miller 1985; Miller& Goodrich 1990; Moran et al. 2000).An important parameter in this model is the opening angle ofthe torus. The fraction of type 1 AGN (hereafter f ) is a measureof the average opening angle of torus in such a unification scheme.Previous studies have shown a type 2 to type 1 ratio of 3:1 for localSeyfert galaxies (e.g., Maia et al. 2003; c.f, Ho et al. 1997). How- ⋆ Email: [email protected] ever, there is no reason that the torus opening angle should be thesame for every AGN, rather it may depend on the black hole mass,accretion rate, luminosity or other intrinsic parameters. Exploringthese parameters dependence will be an important extension to thesimple unification scheme, and yield the insight into the origin ofthe dust torus as well.The luminosities of narrow emission lines, such as [O iii ], mid-infrared light and hard X-rays have been considered as isotropicproperties (e.g. Meisenheimer et al.2001; Kau ff mann et al. 2003;Wang et al. 2006). Thus, we can examine the luminosity depen-dence of f using those luminosities. It was found that f increaseswith [O iii ] luminosity in a large sample of Seyfert galaxies fromSloan Digital Sky Survey (SDSS, York et al 2000) Data Release 2(DR2) (Simpson 2005; Hao et al. 2005). These results were inter-preted in the context of “receding torus” (Lawrence 1991; Hill et al.1996). Similar results have been reported from statistical studies ofhard X-ray selected AGN (Ueda et al. 2003; Hasinger et al. 2004;Gilli et al. 2007; Hasinger et al. 2008; c.f, Wang et al. 2006).Radio loud AGN, including radio quasars and radio galaxies,can also be unified in such scheme (see Urry & Padovani 1995 fora review). As in the case for Seyfert galaxies, broad permitted lineswere detected in the polarized light of narrow line radio galaxies(e.g., Antonuuci, Hurt & Kinney 1994; Tran, Cohen & Goodrich c (cid:13) Lu et al. o − o to mark the division between the radioquasar and galaxy. Within this scheme, by using 172 3CR radiosources, Lawrence (1991) found that the type 2 fractions are anti-correlated with radio luminosity in the range of L > W Hz − . Falcke et al. (1995) proposed that for radio-loud sources,jets may clear a path through the dust, and cause the obscurationalong the jet’s periphery, and give rise to the anti-correlation be-tween obscuration and radio luminosity. These mentioned aboveonly focused on the most powerful radio AGN.In this paper, we will extend those analysis to radiomoderately-strong AGN, and examine in more details f as a func-tion of radio power, [O iii ] luminosity, black hole mass and accre-tion rate using the large sample of AGN in the SDSS spectroscopiccatalogues of galaxies and quasars that have been detected in FaintImage of Radio Sources at Twenty cm ( FIRST, White et al.1997).A combination of large sky coverage, high completeness to a rela-tively deep magnitude for galaxies and quasars, and moderate spec-tral resolution makes the sample ideal for such a study.Based on the SDSS spectroscopic data set, we culled 711type 2 objects and 286 type 1 objects having FIRST luminosity P . > W Hz − and z .
35. After correcting several se-lection e ff ects, we obtained f as a function of nuclear parameters.The paper is arranged as follows. The sample is described in thenext section. The selection function is estimated in § §
4. Finally, we will discuss our result in § H =
71 km s − Mpc − , Ω m = .
27, and Ω λ = .
73 (Spergel et al.2003).
Starting with the spectroscopic samples of quasars and galaxiesin the SDSS data release four (DR4), we construct the low red-shift sample of radio detected galaxies and AGN. A redshift cut z .
35 is applied so that H α falls in the SDSS spectral cover-age. To simplify the estimation of selection e ff ects, we consideronly the objects targeted as main galaxies (Petrosian magnitude r .
77 and the target mask as “TARGET GALAXY”) or lowredshift quasars (psf magnitude i . / SKIRT”), or FIRST counterparts (unre-solved objects with FIRST counterparts within 2 ′′ and psf magni-tude i .
0, masked as “TARGET QSOs FIRST CAP / SKIRT”).This low- z galaxy and quasar sample (405,904 SDSS spectra intotal) are then cross-correlated with the FIRST source catalog(Becker et al. 2003) to form the radio detected galaxy and AGNsample with a procedure described in Lu et al. (2007).We use positional coincidence to select radio point sources,and visually inspect the FIRST images for all the candidates withextended radio morphology (refer to Lu et al. 2007 for details).Briefly, we take 2 . ′′ ff of position o ff set for SDSS-FIRSTpositional coincidence, which is a trade-o ff between the complete-ness and random contamination. Then, we visually inspected thecutouts of FIRST image to select the extended lobe(s) apart fromSDSS nucleus, using a degraded image of 3CR radio sources as thereference for physical association. For z > . ′ × ′ , which corresponds to a physical size of500 kpc at the redshift z = z . z < .
35) objects fromSDSS-FIRST matching process.Next, we apply a radio luminosity cuto ff to the sample. Ac-cording to Yun et al. (2001) and Hopkins et al. (2003), star forma-tion (SF) galaxies rarely have radio powers larger than P . = W Hz − . Best et al. (2005) found that the radio luminosityfunction at 1.4 GHz of SF galaxies intersect with radio loud AGNat ∼ W Hz − . Beyond this value, the luminosity function ofSF galaxies drops dramatically, while that of radio AGN decreasesmildly. Although at ∼ W Hz − , SF galaxies account for nearly10% of the population, they will be rejected on the emission line-ratio diagrams (hereafter, BPT diagram; Baldwin, Phillips & Ter-levich 1981). So we will limit our analysis to the sources with radioluminosities above 10 W Hz − .A k -correction to the radio luminosity is applied by assuminga radio spectral index α = . f ν ∝ ν − α ). After retaining only thehighest S / N spectrum for the duplicated observations, we obtain7,810 SDSS radio loud sources, 276 from SDSS quasar sample and7,534 from SDSS galaxy sample, at redshift z .
35 and with P . > W Hz − .Our criteria are similar to Best et al. (2005), who cross-correlated SDSS-FIRST “compact” sources with 3 ′′ and selectedthe “extended” radio counterparts within 30 ′′ by assuming that ra-dio sources are “double lobe” or “core-lobe”, but extend to a fainterlimit and also include larger radio sources. These authors extracted2,215 radio AGN and 497 SF galaxies brighter than 5 mJy (corre-sponding to 10 W Hz − at z ∼ z < . D n (4000) versus L . / M ∗ diagnosticplane. In order to classify and obtain the intrinsic properties of the radiogalaxies and AGN, precise measurements of emission line fluxesare necessary. We use the measured parameters from the Value-added Extra-GAlactic Catalog developed and maintained by Centerfor Astrophysics, University of Science and Technology of China(USTC-VEGAC; X.-B. Dong et al. in preparation). We will de-scribe briefly the steps relevant here and leave details to the re-ferred paper. We correct the SDSS spectra for the Galactic extinc-tion (Schlegel, Finkbeiner & Davis 1998) using the extinction curveof Fitzpatrick (1999). The spectra are then brought into their restframe using the redshift provided by SDSS pipeline. The contin-uum subtraction and emission line measurements are done sepa-rately for the di ff erent type of objects as follows:The continuum subtraction is done according to their relativecontribution of star-light and AGN continuum. For the narrow lineobjects, the continuum is dominated by star-light, thus, can be mod-eled with Independent Component templates (IC templates) fol-lowing the procedures described in Lu et al. (2006). In brief, we IC templates were derived from Independent Component Analysis(ICA), developed by Lu et al. (2006). Using this technique, Lu et al. (2006)compressed the synthetic galaxy spectral library to six nonnegative inde-c (cid:13) , 000–000 ype 1 Active Galactic Nuclei Fraction fit galaxy spectra with the templates derived by applying Essem-bling Learning for Independent Component Analysis (EL-ICA) tothe simple stellar population library (Bruzual & Charlot 2003). Thetemplates were then broadened and shifted to match the stellar ve-locity dispersion of the galaxy. In this way, stellar absorption linesare reasonably modeled to ensure the reliable measurement of weakemission lines. At the same time, stellar velocity dispersion, a cor-rection to the redshift as well as an average internal extinction tothe stellar light is obtained.For nucleus dominated type 1 AGN where Fe ii multiplets andother broad emission lines are highly blended, we fit simultane-ously the nuclear continuum, the Fe ii multiplets and emission lines(see Dong et al. 2008 for details). The nuclear continuum is approx-imated by a broken power-law. Fe ii emission, both broad and nar-row, is modeled using the Fe ii templates provided by Veron-Cettyet al. (2004). For those Seyfert 1s with significant contribution ofstarlight as measured by the equivalent widths (EWs) of the Ca ii k λ i λλ , ii emission were carried out following the proce-dure as described in detail in Zhou et al. (2006).After subtracting the continuum, we fit emission lines withmulti-gaussian model using the code described in detail in Donget al. (2005, 2008). Briefly, each line is fitted with one or moreGaussians as statistically justified (mostly with 1–2 Gaussians);the line parameters are determined by minimizing χ . The [O iii ] λλ , ii ] λλ , ii ] λλ , iii ] doublet and [N ii ] doublet are fixed to the theoreticalvalues. Usually, H α and [N ii ] doublets are highly blended and thushard to be isolated; in such cases, we fit them assuming they havethe same profile as [S ii ] doublets, which is empirically justified(e.g., Filippenko & Sargent 1988; Ho et al. 1997; Zhou et al. 2006).For the possible broad H α and H β lines, we use multiple Gaussiansto fit them, as many as they could be statistically justified. If a broademission line is detected with S / N >
5, we regard it as genuine. Ifthe broad H β line is too weak to achieve a reliable fit, we then re-fitit assuming that it has the same profile and redshift as the broad H α line. Examples of emission line modeling in H α + [N ii ] andH β + [O iii ] regions are illustrated in Figure 1. The two upper pan-els shown the objects which was classified by SDSS as galaxies,and the two lower ones shown the SDSS classified quasars. The to-tal emission line flux is estimated by adding di ff erent components,while the line width is measured in the combined model. Based on the emission line parameters measured in the last section,we classify optical spectra into broad lined and narrow lined AGN,and SF galaxies, respectively. Detail criteria for each class are asfollows:Broad line AGN are culled according to the following crite-ria (cf. Zhang et al. 2008, X.-B. Dong et al. in preparation): (1)The broad H α component has been detected with S / N >
5; (2) Theheight of broad H α is more than twice of root mean square (RMS) pendent components (ICs), which were proved to be good templates formodeling most normal galaxy spectra. Figure 1.
Emission line fitting in the H α and H β regions. The black linesrepresent the observed “continuum free” emission lines and noise. Theblue lines represent for the fitted broad components and the green lines forbroad + narrow fitting. Figure 2.
The BPT diagram for narrow emission line galaxies with radioluminosity P . > W Hz − and z .
35. The solid lines sepa-rate Seyfert from SF galaxies, and the dashed lines divide the “composite”galaxies from SF galaxies. The blue line represents for Seyfert / LINER sep-aration. of the continuum-subtracted spectrum in the neighbor emission-line free region to account for potential systematic errors broughtin the continuum subtraction; (3) The equivalent width of broadH α line EW (H α ) > c (cid:13) , 000–000 Lu et al.
Figure 3.
The distribution of H δ absorption line equivalent width (leftpanel) and strength of 4000Å break (right panel) for Seyfert 2 galaxies(dashed red line), SF galaxies (dashed green line), LINERs (dashed purpleline), and weak lined objects (solid black line). et al. (2006). First, we divide narrow line radio galaxies into star-forming galaxies, composite galaxies and ’pure’ AGN on [N ii ] / H α versus [O iii ] / H β diagram. Second, ’pure’ AGN are further sepa-rated into ’LINER’ or Seyfert according to their locations on the[O iii ] / H β versus [S ii ] / H α or [O i ] / H α diagram (see Figure 2). For137 objects with strong detections in [O iii ] (with S / N >
10) butnone in H β (S / N < β by using the averageH α / H β of 6 .
08. Most non-detections are caused by low signal tonoise of the spectrum around H β region or low H β equivalent width,but not due to large extinctions. Their 3 σ upper limits are consis-tent with the average value of H α / H β ∼ . β flux (with S / N < α/ H β ) BL >
10, inanalogy with type 1.8 / β flux; (2) the sample becomes very incompleteat large Balmer decrements due to their large obscuration.Finally, 286 broad lined objects, including 248 from SDSSquasar sample and 38 from the main galaxy sample, will be used forfurther analysis. We note that 55 of these Type 1 AGN (46 quasarsand 9 galaxies) are Narrow-line Seyfert 1 galaxies (NLS1’s), ac-cording to the criteria of Zhou et al. (2006), with broad componentof H α or H β > σ confidence level and FWHM(H α ) ,
200 kms − . 35 of them are already included in the NLS1 sample of Zhouet al. (2006).Among the 3,297 galaxies with detectable emission lines,1,771 ones can be un-ambiguously classified, yielding 711 Seyfert2 galaxies, 342 LINERS, 216 star-forming galaxies and 502 “com-posite” galaxies. The remaining 1,526 weak emission line galaxieseither do not have su ffi cient number of measured line ratios to al-low a meaningful diagnostic on the BPT diagram (1510 objects), orthe narrow line decrement H α / H β >
15 (16 objects). Most of thoseweak lined objects are likely to be LINERs rather than Seyfert 2galaxies because the distributions of their H δ absorption line equiv-alent width and strength of 4000Å break are more similar to thatof LINERs rather than Seyfert galaxies (see Figure 3). Therefore,excluding those weak lined objects will not seriously a ff ect our es-timate of type 1 fraction.The final sample consists of 711 type 2 AGN and 286 type 1AGN, selected from original 7810 radio galaxies and quasars with z .
35 and P . > W Hz − . Their redshift distribution areshown in Figure 4. Their locations on the radio luminosity versusthe narrow H α luminosity L H α are shown on Figure 5. For compar-ison, we also add the empirical relation for SF galaxies (Yun et al. Figure 4.
The redshift distribution of type 1s (blue line), and type 2s (redline).
Figure 5.
Radio luminosity P . versus narrow H α luminosity L H α . Thepoints represents for all SDSS selected radio AGN, regardless their radioluminosity. The solid line indicates P . − L H α relation in SF galaxies.Beneath this line, the radio luminosity P . will be barely dominated bystar formation. The dashed line indicates the 10 W Hz − threshold. For the Seyfert 2 galaxies, the black hole mass ( M • ) is estimatedfrom the stellar velocity dispersion using the M • − σ ∗ relation ob-tained from local host spheroids (Tremaine et al. 2002), includ-ing elliptical galaxies and the bulges of disk galaxies. The intrinsicscatter of this relation is estimated to be about 0.3 dex. The stellarvelocity dispersion has been obtained by fitting the SDSS spectrumwith the IC templates (refer to § . ′′ of the SDSS fibre may contain only afraction of galaxy spheroid or have a substantial contribution fromthe galactic disk. In the former case, the correction is usually smallfor most of our galaxies according to the formula of Jorgensen etal. (1995). In the latter case, the measured σ ∗ may over-estimatethe stellar velocity dispersion of the bulge component in high in-clination disk galaxies, or under-estimate the true value in low in-clination disk galaxies. The correction depends on the size of thedisk and bulge, the bulge disk ratio, the bulge velocity dispersionto the circular velocity of the disk, the distance of the galaxy, aswell as the inclination of the disk. There is no simple formula forthis correction. Fortunately, most radio selected AGN have largeblack hole masses, thus reside in the galaxies with large bulges. Inthese galaxies, the disk-light contamination within SDSS fibre maynot be very severe. In addition, the rotational velocity is propor-tional to the central stellar velocity dispersion of these large galax-ies (Courteau et al. 2008). Therefore, this correction may be lesssevere to our sample. c (cid:13) , 000–000 ype 1 Active Galactic Nuclei Fraction Figure 6.
Black hole mass log( M • ) versus the absolute optical magnitude M r for the radio galaxies. Left panel displays the Seyfert 2 galaxies. Theorange points represent for the high inclination disk galaxies with axial ra-tio b / a < . σ ∗ / err >
10, the green points for low the inclination diskgalaxies with b / a > . σ ∗ / err >
10, and black points for ellipticalgalaxies. The solid line shows the mean M r value in the log( M • ) bins. Thedotted line show the M r - M • fitting of inactive galaxies suggested by Mclure& Dunlop (2002), and the dashed line is referred from Faber-Jackson rela-tion of Desroches et al. (2007). The right panel displays the M r - M • relationof the absorption line galaxies for comparison. The entire objects in thesetwo panels have radio luminosity log( P . / W Hz − ) > In order to check whether the latter e ff ect is important forour sample, we divide the galaxies into the disk and bulge domi-nated by using the profile parameters provided by SDSS pipeline.We consider it a disk galaxy if the likelihood for the exponentialprofile fit is larger than de Vacouleurs model fit by 0.2 dex, or in-verse concentration index petroR / petroR > .
33. Note thatthe concentration index criterion C = .
33 from Shimasaku etal. (2001) would induce a 15 ∼
20% contamination from oppo-site types of both disk galaxies and elliptical galaxies. But we con-sider this contamination acceptable. We separate the disk galaxiesinto high and low inclination groups according to their axial ratios(b / a; b / a < . b / a > . b / a < .
3) objectsare considered as high inclination disks, regardless of their radialprofile. Figure 6 shows their distribution on the M r versus M • di-agram; where M r is the k -corrected absolute Petrosian magnitudeof the host galaxy in the galaxy rest frame by interpolating the fiveSDSS apparent magnitudes with Spline function. For comparison,we plot the Seyfert 2 galaxies of our sample in the left panel andplot the absorption line dominated radio galaxies in the right panel.We find: (1) absorption-line dominated radio galaxies are concen-trated on the high luminosity regime, while the spectroscopicallytype 2 AGN spread more to the lower luminosities. (2) M r vs M • istilted for type 2 AGN. It departs significantly from what is definedby the absorption line galaxies or the early type galaxies at low M • ,and towards higher luminosity. This can be attributed most likely tothe disk light contribution. (3) the high inclination galaxies (orangepoints) show systematically lower luminosities than that of low in-clination galaxies (green points) at a given M • (i.e. σ ∗ ) despite oftheir overlap on the plot, and they locate more closely to that ofabsorption line galaxies. Shao et al. (2007) showed that ’edge-on’galaxies are on average 0.8-0.9 fainter in magnitude than that of’face-on’ galaxies in the r -band due to the dust extinction. Thisseems su ffi cient to explain the o ff set between low inclination andhigh inclination disk galaxies observed here without invoking anyadditional e ff ect of a disk component on the σ ∗ .For type 1 AGN, the virial black hole mass can be estimatedusing the empirical relation between the size of BLR and contin-uum luminosity, and the emission line width (e.g., Wandel et al.1999; Kaspi et al. 2005; Peterson & Bentz 2006;). We use the broad Figure 7.
The [O iii ] luminosity L [OIII] versus broad H α luminosity L H α, Br . They are all extinction corrected on the basis of their Balmer decrementH α / H β . H α luminosity as a surrogate for the continuum luminosity (Wang& Zhang 2003), to account for the potential contamination fromthe stellar-light in the weak BLR source and from nonthermal jetemission in some cases, using the formula of Greene& Ho (2007): M • = (3 . + . − . ) × L H α ergs − ! . + − . (1) × (cid:18) FWHM H α kms − (cid:19) . + − . M ⊙ . In section 3.3, this relation will be used also in the simulation ofselection e ff ect for the potentially overlooked broad emission linesin the observed type 2 AGN. The typical uncertainty in this relationis estimated to be around 0.5 dex (Vestergaard & Peterson 2006).We will use reddening corrected [O iii ] luminosity as an indi-cator for the nuclear power of type 2 AGN (Mulchaey et al. 1994;Dietrich et al. 2002; Kau ff mann et al. 2003; Haas et al. 2007; Net-zer et al. 2006; See Reyes et al. 2008 for an extended discussion)for two reasons. First, [O iii ] emission is less contaminated by star-formation process, especially in these massive metal rich galaxies,than other lines such as H α and [O ii ]. Second, Balmer decrementcan be used as an indicator for the global extinction to the narrowemission lines, thus the intrinsic luminosity of [O iii ] can be recov-ered based on the extinction corrected line flux.The extinction correction to [O iii ] luminosity is still a com-plex issue. Some authors argued that the some of the Balmer linesmay be totally blocked according to polarization observation (diSerego Alighieri et al. 1997). Because the ionization potential ofO + is much higher than that of hydrogen, [O iii ] emission regionmay be even smaller than that of Balmer lines, and subject moreseverely to the dust extinction. Therefore, extinction derived fromnarrow line Balmer decrements may still underestimate the [O iii ]extinction. However, most of the former arguments are based onthe directly observed [O iii ] luminosity rather than on the extinc-tion corrected [O iii ] luminosity. We consider applying correctionshould be better than applying no correction.We calibrate the relation between the bolometric luminosityand [O iii ] luminosity with the broad line radio AGN and then ap-plied it to the type 2 objects as follows. First, we establish a relationbetween the extinction corrected [O iii ] luminosity and the broadH α luminosity for the type 1 radio AGN. The extinctions are esti-mated based on the Balmer decrements of narrow and broad lines,respectively, assuming an intrinsic H α/ H β = .
1. [O iii ] and broadH α luminosities are fairly well correlated (see Figure 7).A least-square logarithmic-linear fit yieldslog( L H α, Br ) = (1 . ± . + (0 . ± . × log( L [OIII] ) (2) c (cid:13) , 000–000 Lu et al.
Figure 8.
The black hole mass M • distribution of type 1s (estimated fromreverberation-mapping, blue line) and that of type 2s (estimated from bulge- M • correlation, red line). The inset small diagram shows M • distributionwithin restricted L / L Edd and L regime, refer to Figure 12 and § Figure 9.
Eddington ratio log( ℓ ) = log( L / L Edd ) distribution for type 1s(blue line) and type 2s (red line). , with a scatter in log( L H α, Br ) of 0.39. Note that the slope is closeto one, suggesting a linear relation between the two luminosities,which is consistent with Zhang et al. (2008) for radio quiet AGN.Second, we estimate the continuum luminosity at 5100Å fromthe broad H α luminosity using the relation obtained by Greene &Ho (2007), for the nuclear light dominated AGN. Finally, bolomet-ric luminosity is estimated as L = λ L λ ) . With the bolometricluminosity and the black hole mass, it is straight forward to calcu-late the Eddington ratio ℓ = L bol / L Edd .The distributions of the black hole mass for the type 1 andtype 2 AGN are displayed in Figure 8. On average, the type 1 AGNhave slightly smaller black hole mass than that of type 2 AGN.
Figure 10.
The extinction corrected [O iii ] luminosity distribution for type1s (blue line) and type 2s (red line). The small diagram shows L [OIII] distri-bution within restricted log( L / L Edd ) and M • regime, refer to Figure 12 and § The medians are 1.0 × M ⊙ and 1.4 × M ⊙ for the type 1 andtype 2 AGN, respectively. However, the type 1 AGN are on aver-age 1.4 times more luminous than type 2 AGN in the extinctioncorrected [O iii ] luminosity (Figure 10). Jackson & Browne (1990)found that radio quasars are a factor 10 more luminous in the [O iii ]line luminosities than that of radio galaxies, and they interpretedtheir results as a larger extinction to narrow line region in radiogalaxies than that in quasars. The di ff erence found here is muchsmaller because we have corrected the intrinsic reddening usingBalmer decrement while Jackson & Browne have not. Indeed, theaverage [O iii ] luminosity of the type 1 AGN is more luminous by afactor of 6 than type 2 AGN before the extinction correction. Heck-man et al. (1992; see also Meisenheimer 2001) found that the mid-to far-IR emission ( λ ∼ − µ m in AGN rest frame) is 4 timesstronger on average in quasars than that in NLRG for a 178MHz3CR sample (including 42 quasars and 75 NLRG with z > . § ℓ ) = log( L bol / L Edd ) distribution wasdisplayed in figure 9. We found that the distribution of the type1 AGN peaked at log( ℓ ) ∼ − . ℓ ∼ .
13) in the range of − log( ℓ )
0, while the distribution of the type 2 objects iswider and skewed to a lower log( ℓ ) although it is also peaked nearlog( ℓ ) ∼ − .
9. The more extended tail towards small log( ℓ ) in thetype 2 AGN may be attributed to a selection e ff ect because thebroad line components are more di ffi cult to be detected at a lowerlog( ℓ ). Note that scatters in the estimation of black hole mass andbolometric luminosity will cause a distortion in log( ℓ ) distribution.Fortunately, the scatter (0.5 dex) of the black hole estimation fortype 1 using the line width and continuum luminosity is compara-ble to the combination of the scatter (0.3 dex) in M • − σ ∗ relationand the scatter (0.4 dex) in the bolometric luminosity estimationusing [O iii ] luminosity for type 2. Thus, the scatter e ff ect on type1 and type 2 distributions is similar. We will discuss this further inSection 4. Selection e ff ects are introduced during the spectroscopic target se-lection and the definition of narrow and broad line AGN sample. Inthis section we will quantify these selection e ff ects as a function ofnuclear properties for both the type 1 and type 2 AGN. We denotethe number density of AGN in unit nuclear luminosity ( L ) and unit M • intervals as φ ( L , M • ). The expected number of AGNs in the nu-clear luminosity bin L − L + ∆ L and black hole bin M • − M • + ∆ M • can be written as, ∆ N ( L , M • ) = φ ( L , M • ) Z L smin ( z ) p ( L s | M • , L ) S ( M • , L , L s ) (3) × dL s V max ( L s ) ∆ M • ∆ L where L s is the luminosity of the band that is used in defining themagnitude limit of the sample, e.g., the r -band luminosity of thehost galaxy dominated targets and i -band luminosity for the nucleardominated targets. p ( L s | M • , L ) is the conditional probability that anAGN has L s at the given M • and L . Selection function S ( M • , L , L s )is the probability to classify the spectrum as type 1 or type 2; V max is the comoving volume corresponding to the maximum redshiftthat an object with a luminosity L s could be detected within the c (cid:13) , 000–000 ype 1 Active Galactic Nuclei Fraction magnitude limit. We have assumed that there is no cosmologicalevolution for both types of AGN within z < .
35, and we will checkthe assumption in § § φ ( L , M • ) = N X i = V max , i h ( L , L + ∆ L ; M • , M • + ∆ M • ) R ∞ L slim p ( L s | M • , L ) S ( M • , L , L s ) dL s (4)for N AGN in the sample, and h ( L , L + ∆ L ; M • , M • + ∆ M • ) is therectangular function. Due to the additive nature of Eq 4, AGN se-lected via exclusive rules can be added up simply.With φ ( L , M • ), we can obtain the density of AGN in the unitinterval of Eddington ratio ℓ = L bol / L Edd and M • : ϕ ( ℓ, M • ) = Z φ ( L , M • ) δ ( ℓ − ℓ ( L ))( ∂ℓ ( L ) /∂ L ) M • dL (5)The density at radio luminosity bin could be wrote as ψ ( P . ) = Z φ ( L , M • ) p ( P . | L , M • ) dLdM • (6), where p ( P . | L , M • ) is the conditional probability that an AGNhas a radio luminosity of P . at a given M • and L . V max Di ff erent magnitude limits have been used in the selection of dif-ferent type spectroscopic targets. The apparent magnitude limitsfor main quasars and FIRST counterparts are i psf . r Petrosian .
77 in r band.We calculate V optmax for each object according to its relevant opticalmagnitude limit and the corresponding optical luminosity . Sinceour objects were detected by FIRST, which has a detection limit of f . > V radiomax can be estimated from the radio power andthe radio flux limit for each object. Furthermore, we adopt a cuto ff in the redshift z .
35, so the maximum volume is V . . The V max is the smallest one among the above three values.For objects uniformly distributed in the unverse, the average < V / V max > will be around 0.5. We calculated this value for both type1 and type 2 AGN. We find < V / V max > = .
55 for the type 1 AGN,and < V / V max > = .
54 for the type 2 AGN. This may be taken asan evidence that the number density of AGN increases mildly withincrease redshift, and we will discuss the evolution e ff ect in § Ignoring the contribution of emission lines to the optical magni-tude, the type 2 AGNs are selected according to their host galax-ies. In order to quantify the conditional probability p ( L s | M • , L ) oftype 2 AGN for a given black hole mass and nuclear luminos-ity (or [O iii ] luminosity), we need to establish a relation amongthe host galaxy magnitude L s , the [O iii ] luminosity L [OIII] , andthe black hole mass M • . The strong correlation between the massof the massive black hole and the bulge luminosity of its host Because we are interested in the ratio of type 1 and type 2 AGNs, ratherthan their comoving density, we do not take the survey area into considera-tion. galaxy was established for the bulge dominated quiescent galax-ies in the local universe (e.g., H¨aring & Rix 2004) and for the ac-tive galaxies and quasars (e.g., Peng et al. 2006). However, thereis still controversial whether this relation depends on the level ofnuclear activity (McLure & Dunlop 2002). To check this, we di-vide the type 2 AGN into high (log( L [OIII] / erg s − ) >
42) and low(log( L [OIII] / erg s − ) <
42) [O iii ] luminosity groups, and examinetheir distributions on M r − M • diagram. We find that at the same M • , the [O iii ] luminous galaxies are brighter than the low L [OIII] counterparts for only 0.1-0.2 mag on average, which can be ac-counted for by their di ff erent emission line luminosities. Therefore,we will assume that the host’s luminosity is independence of the nu-clear luminosity at a giving black hole mass. Therefore, we write p ( L s | M • , L ) as p ( L s | M • ), i.e., independence of the nuclear luminos-ity L .The galaxies are selected based on their Petrosian magnitudes,i.e., L s should be the total magnitude of the galaxy. In order toquantify p ( L s | M • ), we need to characterize the relation between theblack hole mass and the total luminosity of its host galaxy, and thescatter of this relation as well. For the known Seyfert 2 galaxies,their distributions are already shown in Figure 6. We assume theincluding of the missed Seyfert galaxies does not change this re-lation. Because the disk contribution increases as the black holemass decreases, the relation between the galaxy luminosity andblack hole mass is flatter than the bulge-black hole relation shownin Figure 6. For comparison, we also show the M • − M (bulge) rela-tion for the inactive galaxies: log( M • ) = − . ± . M R (bulge) − . ± .
04) (McLure & Dunlop 2002), and the Faber-Jackson re-lation of Desroches et al. (2007) for the normal elliptical galaxies.Apparently, M r − log( M • ) relation of our sample is much flatter thanthe latter relations. To describe quantitatively the distribution of the M r as a function of M • , we calculate the average value and thesecond momentum of M r over each M • bin, and approximately es-timate the p ( L s | M • ) with a gaussian distribution around their meanvalue.We assume the type 2 galaxies are relative easy to identify,so S ( M • , L , L s ) = ff ect our study of the type 1 to type 2 ratios, be-cause the detection of a type 1 AGN will require an even higherEddington ratio, and our analysis is limited to the parameter spacethat a substantial fraction of both the type 1 and type 2 can be de-tected (see Figure 12). In the second case, we will drop some of thetype 2 AGN in the late type of galaxies, which tend to have lowerblack hole masses. We will discuss this in § The probability of identifying a type 1 object depends on the signalto noise ratio of the spectrum and broad line parameters, its pro-file and intensity, in a rather complicated manner. To quantify sucha selection e ff ect, we generate a large number of spectra coveringthe H α blending and H β regimes using Monte-Carlo simulationssimilar to what has been done by Hao et al. (2005) but taking addi-tional parameters, like the intrinsic reddening and black hole mass,into consideration. These spectra are modeled in exactly the same c (cid:13) , 000–000 Lu et al. way as we did for the real data to obtain the emission line param-eters, and the type 1 objects are selected with the same criteria asdescribed in § ff erent physical parameter regime.A simulated emission line spectrum is the sum of three com-ponents: the narrow line spectrum ( f narrow ), broad line spectrum( f broad ) and the noise spectrum ( noise ). In order to mimic the di-verse narrow line spectrum and to avoid complicated noise model,we use the observed narrow line spectrum and noise spectrum. Inorder to explore the physical parameters as broad as possible, wetake the narrow line plus noise spectrum from both the type 1 andtype 2 AGN.The narrow line plus the noise spectrum ( f narrow + noise ) isobtained by subtracting the power-law continuum and FeII mod-els for the nuclei dominated objects, or by subtracting the stellarcontinuum model for the host galaxy dominated objects from theobserved spectrum. It should be noted that this treatment of noisespectrum is in-exact for the Seyfert 2 galaxies, because the additionof a broad line component and its corresponding nuclear continuumwould also increase the noise. But its e ff ect is likely to be small be-cause a broad line with a height of ten percent of the continuumflux would be easily detectable in most spectra.For the broad line spectrum ( f broad ), we estimated them withthe parameters of line profile and line flux, which in turn dependson the intrinsic broad-line luminosity and the dust extinction to theBLR. We denotes the broad line component as, f broad = A ( λ ) K f sbroad (7)where A ( λ ) and K are the internal extinction and the scaling factor,respectively, and f sbroad is ’standard’ broad line spectrum.For the type 1 objects, f sbroad is the best fitted model for bothbroad H β and H α after correcting for the internal reddening. Forthe Seyfert 2 objects, we assume that the broad line profile canbe approximated with a single gaussian, and the line width can beestimated from the empirical relation in Eq.1, with the black holemass estimated from M − σ ∗ relation, and line-flux obtained fromthe observed [O iii ] luminosity with Eq.2. A ( λ ) comes from a set of eight extinction values, which cor-responds to a uniformly distributed H α / H β values between 3-10,assuming the intrinsic H α / H β = A ( λ ), a set of ten K n are created randomly following the H α luminosity distributionat the observed [O iii ] luminosity, i.e., p ( L H α | L [OIII] ) (see Figure 7).To summarize, for each AGN, we build a set of simulatedspectra that has identical narrow line + noise spectrum as the ob-served spectrum but with a series of manually-built broad line com-ponents. Assuming the nuclei luminosity scaled with broad line, theapparent magnitude and volume limit will change when we vary f broad component of Seyfert galaxies. We retained only those K n and A ( λ ) that makes the optical flux for the simulated spectrumwithin the magnitude limit ( r < .
77 for the host dominated ob-jects and i < . S ( M • , L , E B − V , L s ). With this selection function, we can estimateapproximately the distribution of E B − V from the observed one byassuming that the real E B − V distribution does not depend on theblack hole mass and optical luminosity. Under this assumption, wecan correct the observed E B − V distribution using the above simu- Figure 11.
Left panel: average correction factor N id ( E B − V ) / N sim ( E B − V )along Balmer decrement (H α / H β ) BL . Right panel: the observed (dashedline) and corrected (solid line) Balmer decrement distribution. lated spectrum p ( E B − V ) = p obs ( E B − V ) N id ( E B − V ) / N sim ( E B − V ) . (8)where the p obs ( E B − V ) and p ( E B − V ) are the observed and real distri-bution of E B − V ; N id ( E B − V ) and N sim ( E B − V ) are the number of spec-tra identified as Seyfert 1 galaxies and total number of the simulatedspectra that meet the SDSS targeting criteria. For this correction,we have assumed that E B − V distribution is independent of blackhole mass and nuclear luminosity. The assumption has not beenfully tested, however, it should not dramatically a ff ect our result asfar as the correction remains small or the dependence is weak. Theobserved and corrected H α / H β is displayed in Figure 11. In com-parison with the observed one, which peaks at 3.1, and is identicalto the value reported by Dong et al. (2008), the corrected H α / H β distribution shows much more objects at higher extinctions. Inte-grating over the distribution of E B − V , we yield: S ( M • , L , L s ) = Z S ( M • , L , E B − V , L s ) p ( E B − V ) dE ( B − V ) (9)Because the correction increases fast with the degree of extinction,one must keep in mind that the true number may become less reli-able at higher extinctions. For this reason, we will limit our analysisonly to E B − V < . α / H β < p ( L s | M • , L ) = δ ( L s − L ) and S ( M • , L , L s ) is independent of L s . For the broad line objectsselected from the galaxy sample, there is a distribution of L s for agiven nuclear luminosity and black hole mass due to the significantcontribution from the host galaxy. Using the relation between thehost galaxy and the black hole for the type 2 Seyfert galaxies (seeFigure 6) and the nuclear luminosity from L H α, Br − L relation(refer to Greene & Ho 2007), we write p ( L s | M • , L ) = πσ σ Z dL host exp − ( L host − L host ( M • )) σ ! (10) × exp − ( L s − L host − L r ( L )) σ ! where L host ( M • ) is the average r -band host luminosity in theAGN rest frame for a black hole mass M • , and σ is the scatterof this relation; L r ( L ) is the average r -band nuclear luminosity fora given L which is estimated via the broad H α luminosity. As wealready seen, σ is much smaller than σ , so we can approximatethe second part of the integrand as a δ function.We show contours of the average selection function on theblack hole mass versus nuclear luminosity plane in Figure 12. Theselection function strongly depends on the BH mass and nuclear lu-minosity [O iii ]: the incompleteness increases towards larger black c (cid:13) , 000–000 ype 1 Active Galactic Nuclei Fraction Figure 12.
The contour of selection function and the restricted regime oflog( M • / M ⊙ ) > . , log( L [OIII] / erg s − ) > . L bol / L Edd ) > − . hole mass and lower luminosities. At [O iii ] luminosity below 10 . erg s − , the selection function is below 0.2, even at black hole massas low as 10 M ⊙ . The dearth of type 1 AGN below this limit canbe well attributed to the selection e ff ects. With all the above corrections, we will analyze the fraction of thetype 1 AGN as a function of the basic parameters, such as blackhole mass, accretion rate, nuclear luminosity and radio power. Aswe can see in the Figure 12, due to selection e ff ect, the type 1AGN can be detected to a reasonable fraction only in a very lim-ited parameter space on either M • versus log( L / L Edd ) plane or L [OIII] versus M • plane. In order to avoid potentially large uncer-tainties introduced with our correction of selection e ff ects, we willlimit our analysis to the parameter regimes of log( M • / M ⊙ ) > . L [OIII] / erg s − ) > . L bol / L Edd ) > − .
1, that a sub-stantial fraction of Type 1 AGN are detected. Using the selectionfunction S ( M • , L , L s ), V max and p ( L s | M • , L ) in the last section, wecan calculate φ ( M • , L ) for broad and narrow line AGN. L [OIII] The regime of M • and L [OIII] used in this analysis is illustrated in theupper panel of Figure 12 with blue lines. In this regime, log( L [OIII] )distribution of type 1s and type 2s are very similar with meanslog( L [OIII] / erg s − ) of 42.47 and 42.40 for type 1 and type 2 AGN,respectively (see inserted small diagram in Figure 10). We inte-grate φ ( M • , L [OIII] ) over either M • in the corresponding range to get Figure 13.
The dependence of type 1 fraction on L [OIII] for observed den-sity (dashed line and diamonds), bias corrected density(solid line and dia-monds), and maximum redshift evolution corrected density (green line andtriangles). The evolution correction is based on extrapolation of quasar lu-minosity function of Richards et al. (2005). φ ( L [OIII] ) for both type 1 and type 2 AGN. It is straight forward tocalculate the type 1 fraction by dividing the bias-corrected type 1objects density by the total AGN density in each bin. The resultsare show in Figure 13.The observed type 1 fraction increases with L [OIII] , which issimilar to, but with a flatter slope than those given by Simpson(2005, hereafter S05) and Hao et al. (2005, hereafter H05) for ra-dio quiet AGN. After correcting for selection e ff ects, f keeps atnearly a constant value of 20% over the [O iii ] luminosity range40 . < log( L [OIII] / erg s − ) < .
5. The least χ fit for a constant f yields f ∼ .
1% which is acceptable at a probability of 0.92.This result is di ff erent from S05 and H05, who found that the frac-tion of Type 1 AGN increases significantly with the nuclear lu-minosity for mainly radio quiet AGN. The di ff erence can not beconsidered (solely) as a di ff erence between radio-loud and radioquiet AGN, but be attributed to several di ff erent treatments. First,S05 and H05 does not corrected [O iii ] luminosity for the inter-nal extinction, while such a correction is included in this work. Itis known that type 2 Seyfert galaxies show systematically largerBalmer decrements of narrow lines than Type 1 Seyfert galaxies(Cohen 1983; Gaskell 1984; Rhee & Larkin 2005). This extinc-tion would reduce the [O iii ] luminosities of Seyfert 2 galaxies withrespect to the Seyfert 1 galaxies systematically. In fact, before ex-tinction correction, the type 1 fraction in our sample would alsoincrease dramatically from ∼
10% at log( L [OIII] / erg s − ) ∼
41 to ∼
90% at log( L [OIII] / erg s − ) ∼
43. Second, Type 2 AGN in the S05includes also LINERs and composite type AGN. While it is stillcontroversy whether composite type 2 AGN are of similar natureas Seyfert galaxies, those spectroscopic LINERs are certainly dif-ferent from Seyfert galaxies (Kau ff mann et al. 2003; Kewley et al.2006; Heckmann et al. 2004). Also we note that most compositegalaxies are located in relative lower black hole mass and LINERSon lower accretion rate regime, while broad lined AGN seldom fallin these regions (refer to Figure 12). Third, as seen in Figure 12,the di ff erence can be attributed at least partly to that we focus ouranalysis to the limited parameter regime. Outside the regime, type2 AGN are detected in large amount but almost no type 1 object hasbeen detected on the regime with large black hole mass and small[O iii ] luminosity. Finally, S05 did not consider selection e ff ect at allwhile H05 used a di ff erent selection function. However, as seen inthe Figure 12, this changes the [O iii ] luminosity dependence onlymoderate. This is understandable because we limit our analysis tothe parameter regime, where the correction is only modest.Hitherto we ignored the redshift evolution, but V / V max testdoes show mild positive evolution with z . We estimate the maximalimpact of the evolution on our result by using the double power-law c (cid:13) , 000–000 Lu et al.
Figure 14.
The dependence of type 1 fraction on Eddington ratio log( ℓ ) = log( L / L Edd ) for observed density (dashed line) and bias corrected density(solid line).
Figure 15.
The dependence of type 1 fraction on M • for the bias correcteddensity ( f cor1 , solid black line and diamonds), the observed density ( f obs1 ,dashed black line and diamonds), and simulated type 1 density ( f sim1 , greenline and triangles, based on the convolved type 2 density). Each bin containsthe same number of sources. form quasar luminosity function Φ (L , z) for z < . { L , z } bin, we can apply correction of Φ (L , z = / Φ (L , z) to both type 1sand type 2s. The result is shown in figure 13 (green line and trian-gles). The redshift evolution pulls down the type 1 fraction by ∼ ff ect does not significantly alter our result. M • andEddington ratio The regimes of M • and ℓ = L bol / L Edd used in this analysis areshown in the bottom panel of Figure 12. In this regime, the average M • for type 1s ( ∼ . M ⊙ ) and type 2s ( ∼ . M ⊙ ) are indistin-guishable (see small diagram in Figure 8). Within the regime, weintegrate φ ( L , M • ) over either L in the corresponding range to get φ ( M • ). Similarly, by integrating of ϕ ( M • , ℓ ) (see Eq.5) over M • , wecan obtain ϕ ( ℓ ) for both type 1 and type 2 AGN. Note that our anal-ysis is restricted to the parameter range illustrated in Figure 12(bluelines), where a substantial fraction of both type 1 and type 2 AGNare detected. Thus we can calculate the type 1 fraction straightly bydividing the bias-corrected type 1s density by the total AGN densityin each log( ℓ ) bin and log( M • ) bin. The results are show in Figure14 and Figure 15.From Figure 14, we find that the type 1 fraction keeps nearly aconstant value of 25% when Eddington ratio changes from log( ℓ ) ≃− . ℓ ) ≃ . . < log( M • / M ⊙ ) < .
5. We also note that the correction to selectione ff ects is only modest in whole range of Eddington ratio, and doesnot significantly a ff ect our conclusion as shown in the Figure 12.In fact, before the correction for selection e ff ects, type 1 fractionis more consistent with a constant value over − . < log( ℓ ) < .
0. At lower Eddington ratios, type 1 fraction becomes very low, andthe transition occurs at log( ℓ ) ∼ −
2. The transition is likely real,rather than caused by under-estimate of the selection function. Asseen in Figure 12, most detected objects in this region are massiveobjects, and the value of the selection function is between 0.3 and0.5. Lack broad emission line of those objects maybe represent atrue di ff erence in the BLR properties or over-luminous of [O iii ]in comparison with its true nuclear luminosity in this parameterregion.As mentioned in § ff ect our results aboutthe dependence of type 1 fraction on black hole mass. Black holemasses for type 2 are estimated using the M • − σ ∗ relation. This re-lation has an intrinsic scatter of only 0.3 dex, while the Type 1 BHmasses are estimated using the Virial estimator, which may a ff ectedby the inclination of the AGN and accretion rate, and has a typicaluncertainty of a factor of 0.5 dex. Because this larger scatter, a tailto high masses and low mass would be greater for the Type 1s thanthe Type 2s when cut-o ff (such as L [OIII] ) is applied. When a ratioof Type 1 over Type 2 is taken, this would then give an increasingType-1 fraction at high and low black hole masses as observed.In order to correct this bias, we convolve the type 2 log( M • )distribution with a Gaussian of σ = .
4, so that type 1s and type2s have the same scatter. We restrict to objects in { L [OIII] , M • } re-gion, as shown in upper panel of Figure 12. Based on convolvedtype 2 density ρ convtype2 , the simulated type 1 density fraction f sim1 = ρ obstype / ( ρ obstype1 + ρ convtype2 ) was displayed in Figure 15, represented withgreen line and triangles. While the observed type 1 density frac-tion f obs1 = ρ obstype1 / ( ρ obstype1 + ρ obstype2 ) was represented with dashed lineand diamonds. Therefore, in every log( M • ) bin, we could correctthe selection e ff ect from di ff erent M • measurement uncertainty fora factor of f sim1 / f obs1 . Combining it with selection function and de-tection probability, we plot the type 1 fraction f cor1 = φ / ( φ + φ )dependence on log( M • ) in figure 15 with solid line. The least χ fit for a constant f yields χ = .
54 for 5 degrees of freedom,which rules out a constant f at a confidence level of ∼ ∼
30% in higherBH mass region (log( M • / M ⊙ ) > M • region. f also rises in the BH mass bin belowlog( M • / M ⊙ ) < .
6. This may be due to the loss of type 2 AGN inthis sample. Low mass black holes are usually hosted in the diskgalaxies with relative small bugle, such galaxies are more likely tohave a star forming disk. Due to relative large aperture of SDSS fi-bre, SDSS spectrum will encompass also the star forming disk. Thiswill shift some type 2 Seyfert galaxies into the composite-type. Be-cause we consider only Seyfert 2 type spectra, those type 2 AGNwill be missed in our Seyfert 2 sample (refer to the “compositegalaxies” in Figure 12, represented with green points). In order tocheck the fraction of such objects, we examine the distribution onthe BPT diagram of type 1 AGN, which are selected based solely onthe presence of broad lines. We find that only 4.05%(9) type 1s withlog( M • / M ⊙ ) > . M • / M ⊙ ) < . We plotted the distribution of radio luminosity P . in the leftpanel of Figure 16. It is clear that more type 2 AGN distribute nearthe lower P . limit 10 W Hz − than type 1 objects, and fewer c (cid:13) , 000–000 ype 1 Active Galactic Nuclei Fraction Figure 16.
Left panel: the radio power P . distribution for type 1s (blueline) and type 2s (red line). The small diagram shows P . distributionwithin restricted log( L / L Edd ) and log( M • ) regime, refer to Figure 12. Rightpanel: the dependence of type 1 fraction on P . for observed density(dashed line) and bias corrected density (solid line). Each bin in right panelcontains the same number of sources. type 2 objects in the high radio luminosity region P . > W Hz − . After limiting to radio AGN in the parameter regime de-fined in Figure 12, the overall distribution is also quite similar. Thetype 1 fraction increases from 15% at log( P . / W Hz − ) = P . / W Hz − ) =
24 and then flattened tolog( P . / W Hz − ) =
26 (Figure 16). Note that the trend is verysimilar before and after correction for selection e ff ects. χ -test rulesout the possibility of a constant f at a confidence level of 99.5%.However, when the lowest radio power bin is removed, the trend inthe rest 5 bins is consistent with a constant f at a probability of50%. Lawrence et al. (1991) found that type 2 fraction decreaseswith radio power for a lower frequency selected 3CR radio sample,which consists of mainly powerful radio sources. Since our sam-ple includes only a smaller number of powerful ( P . > WHz − ) radio sources, thus is not adequate to address whether f risesat high radio luminosity.Comparing to lower frequency selected samples, our sample ismore likely a ff ected by radio selection e ff ects. It is general consid-ered that radio emission from an AGN consists of two components,an isotropic extended lobe component and a beamed jet component(e.g., Urry & Padovani 1995). When the AGN is viewed along theradio jet, the jet component with a relative flat radio spectrum isboosted relativistically, while the lobe emission with a steep spec-trum remains the same, i.e., high frequency radio flux is boostedalong the jet direction more than low frequency radio flux. As such,a high frequency survey is more sensitive to the core-dominated orcompact, flat-spectrum sources than to the lobe-dominated or dif-fused, steep-spectrum sources. Because it is general believed thatthe jet emerges along the symmetric axis, the sample will be biasedto the face-on system, whose optical spectrum is type 1 in unifiedscheme.The beaming e ff ect can be checked with radio morphology. Aradio AGN will appear as a double-lobe source when it is observedside-away, and as core-jet or core dominated structure viewedalong the jet direction due to beaming e ff ect of the relativistic jet.Therefore, the core dominance parameter, defined as C = P c / P t ( P t = P c + P l ), as a rough indicator of the system inclination. Ifradio jet aligns with the symmetric axis of the dusty torus, onewould expect that type 1 fraction increases with C . However, onlyabout 12% sources in our sample have resolved radio structures.The number is still too small to allow to reach a firm conclusion onthis. We have quantified several selection e ff ects on the AGN classi-fication and spectral targeting for di ff erent combinations of hostgalaxy properties, nuclear properties and dust extinction from spec-troscopic samples of radio loud SDSS galaxies and quasars usingMonte-Carlo simulations. Type 1 AGN have much strong selectione ff ect than type 2 AGN. Type 1 radio AGN can be studied onlyin a very limited parameter space on the black hole mass versusEddington ratio or luminosity diagram. After correction for theseselection e ff ects, we find that in the limited parameter region: (1)type 1 fraction is nearly independent of extinction-corrected [O iii ]luminosity and Eddington ratio at log( L [OIII] / erg s − ) > . ℓ ) > − .
5; at very low Eddington ratio, there are almost no type1 AGN, which can not be accounted for solely by selection e ff ects,indicating a transition in the broad line strength around log( ℓ ) ∼ − to 10 W Hz − , then keep flat until 10 W Hz − . (3) The type 1 fraction is ∼
30% in M • > M ⊙ region, ∼
10% higher than that in lower M • region.Our result on the luminosity dependence is quite di ff erentfrom previous studies for radio loud quasar and galaxies (Lawrence1991; Hill et al. 1996; Reyes et al. 2008), or mainly radio quietAGN from SDSS (Simpson 2005, Hao et al. 2005). The di ff erencecan be mainly attributed to three factors. First, we use an extinc-tion corrected [O iii ] luminosity while previous authors did NOT.The average Balmer decrements of narrow components for Seyfert2 galaxies ( ∼ .
08) is significantly higher than for Seyfert 1 galax-ies ( ∼ . ff erence in[O iii ] luminosity for two type of Seyfert galaxies. Second, we re-stricted our analysis to the parameter regimes, where a substantialfraction of objects can be detected according to selection function,this reduces the uncertainty caused by the correction of selectione ff ects. The final parameter regime used here is similar to the onethat found by Hopkins et al. (2009) based on a di ff erent approach.This cuts down the objects with very low Eddington ratios and highblack hole mass, thus raises the type 1 fraction in the low luminos-ity end. Finally, our AGN classification is based on more recentwork of Kewley et al. (2006), which gives a better separation ofLINERs and Seyfert galaxies. The new criteria removes quite somemore LINERs than the previous criteria. Because most of theseLINERs have lower [O iii ] luminosity (see Figure 12), this raisessignificantly the fraction of type 1 in the low luminosity.Although the X-ray observed AGN sample do not su ff er fromabove selection e ff ects, however, the fraction of type 1 object basedon the optical classification into broad and narrow lined objectssu ff ers from the aforementioned selection e ff ects. Thus it is proba-bly the same reasons caused a luminosity-dependence of obscuredtype-1 (Hasinger et al.2008; Gilli et al. 2007; Fiore et al. 2008).Rowan-Robinson et al. (2009) using infrared to X-ray spectral en-ergy distribution as a classification for obscured and un-obscuredAGN, and they reached a similar conclusion as this paper.Two very di ff erent phonometrical models have been proposedfor the unification of broad and narrow line radio AGN. The reced-ing torus model was first proposed by Lawrence (1991) to explainthe decrease of the fraction of narrow line objects with the opti-cal luminosity in radio loud AGN, and was extended to radio quietobjects later (Hill et al. 1996; Simpson 2005; Hao et al. 2005; Sug-anuma et al. 2006). The basic idea is that the opening angle of thedust torus is larger for more luminous AGN because dust sublima-tion radius increases with nuclear luminosity while the height of thetorus is assumed to be independent of the bolometric luminosity. c (cid:13) , 000–000 Lu et al.
Thus the broad-line region can be seen over a larger opening anglein more luminous objects. On the other hand, Grimes et al. (2004)showed that observations are consistent with dual population radiosources, ’starved’ low luminosity AGN, which do not have a broademission line region, and ’Eddington-tuned’ high luminosity AGN,without invoking a receding model. In our analysis, we excludedthese very low Eddington ratio objects . Our results show that thetorus opening angle does not change with the nuclear luminosity,thus do not support ’receding torus’ model, but are consistent withthe two population models.Very little has been known about the dependence of the type 1fraction on black hole mass or accretion rate. Using the ratio of in-frared to bolometric luminosity as an indicator of subtending angleof torus, Cao (2005) found that the opening angle of torus increaseswith the central black hole for a sample of Palomer-Green quasars,but does not correlate with its Eddington ratio. On the other hand,Zhou & Wang (2005) argued that the subtending angle of torus de-creases with increasing accretion rate based on the equivalent widthof narrow FeK α line. It should be noted that the origin of the nar-row FeK α is not clear and the equivalent width is also sensitiveto the column density of the absorbing material. Our results agreewith Cao et al. (2005) in the region of log( M • / M ⊙ ) > . ffi cult to compare our results with theoretical models ofdusty torus. Most dynamic-based dusty torus models are not ableto predict quantitatively how the opening angle of torus changeswith the black hole mass or Eddington ratio, although some relationmay be expected. In the dusty cloud model, the physical processthat maintains a thick torus is still not clear (Krolik & Begelman1988; Zier & Biermann 2002). Beckert & Duschl (2004) proposedthat a geometrically thick torus can be sustained at a high accretionrate by the balance of energy dissipation due to cloud-cloud colli-sion and heating due to accretion. They found a torus height to ra-dius ratio H / R ∼ p G ˙ M / ( c s M ( R )), where ˙ M is the accretion rate, M ( R ) the mass within radius R , and c s sound speed. Their modelhas a clear prediction that subtending angle of the torus increaseswith the accretion rate. Our results do not support this. Radiationpressure support is suggested by Pier & Krolik (1992), and the in-frared radiation and local heating pressure can be very e ff ective insupporting a smooth distributed torus. However, an equilibrium so-lution can be found only in a relative narrow Eddington ratio range(Krolik 2007; Shi & Krolik 2008). It has still to be demonstratedthat models taken into consideration of more physical processes,such as instability, would finally lead to a prediction of a constanttorus opening angle over a fairly broad Eddington ratios.In an alternative model of disk outflow, the entrained dustycold clouds are responsible for the obscuration (e.g., Konigl &Kartje 1994). Dopita et al. (1998) suggested that the torus and theaccretion disk may interacte by accretion-outflow feedback pro-cess. Elitzur & Shlosman (2006) argued that a torus is present onlyat nuclear luminosity greater than 10 erg s − , and its coveringfactor decreases with increase luminosity. However, its argument isbased on a strong assumption that the size of dusty clouds are scaledonly with their launching radius. On the other hand, numerical sim-ulations showed that luminous and high accretion rate ( L bol / L Edd )AGN are likely strongly a ff ected by obscuration in the disk windmodels (Schurch, Done & Proga 2009).It should be noted that there are obscuration sources other thandusty torus. The large-scale galactic dust lane and the nuclear star- Most of low Eddington ratio objects (with L / L Edd < .
01) are LINERs(see Figure 12).
Figure 17.
The fraction of high inclination ((b / a) exp < .
5) type 2s amongdisk host galaxies. Left panel displayed its dependence on log( M • ), and theright panel displayed its dependence on log( P . ). Each bin contains thesame number of sources. burst region are certainly responsible for obscuration in some ofobserved type-2 AGN. As we have noticed that a fraction of type-2objects in the sample resides in relative edge-on disk galaxies. InFig 17, we displayed the fraction of high inclination (b / a < . f with M • found in Fig15: a ∼
10% step function in M • at around 10 M ⊙ . Also, the de-crease of the edge-on disks fraction in right panel of Fig 17 may berelated to the large increase of f in the lowest radio power bin inFig 16. These results can be explained by the the presence of a pop-ulation of Seyfert galaxies, seen as Type-2 due to galaxy-scale ob-scuration. Therefore, the increase of type 1 fraction with black holemass and radio luminosity may be understood as that host galaxiesmoves gradually from disk to early-type galaxies, the extinction bythe host galaxy decreases. This result o ff er a very di ff erent interpre-tation from the rest of the paper discusses. We thank the anonymous referee for useful suggestions to improvethe paper. This work was supported by the Chinese NSF throughNSF-10973013 and 973 program 2007CB815403. This paper hasmade use of the data from the SDSS and FIRST. Funding for thecreation and distribution of the SDSS Archive has been providedby the Alfred P. Sloan Foundation, the Participating Institutions,the National Aeronautics and Space Administration, the NationalScience Foundation, the US Department of Energy, the JapaneseMonbukagakusho and the Max Planck Society. FIRST is fundedby the National Astronomy Observatory (NRAO), and is a researchfacility of the US National Science foundation and uses the NRAOVery Large Array.
REFERENCES
Antonucci, R. R. J., & Miller, J. S. 1985,
ApJ , 297, 621Antonucci, R. 1993,
ARA & A , 31, 473Antonucci, R., Hurt, T., & Kinney, A. 1994, Nature , 371, 313Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981,
PASP , 93,5Barthel, P. D. 1989,
ApJ , 336, 606Becker, R. H., Helfand, D. J., White, R. L., Gregg, M. D., &Laurent-Muehleisen, S. A. 2003, VizieR Online Data Catalog,8071, 0Beckert, T., & Duschl, W. J. 2004, A & A , 426, 445 c (cid:13) , 000–000 ype 1 Active Galactic Nuclei Fraction Best, P. N., Kau ff mann, G., Heckman, T. M., & Ivezi´c, ˇZ. 2005, MNRAS , 362, 9Bruzual, G., & Charlot, S. 2003,
MNRAS , 344, 1000Cao, X. 2005,
ApJ , 619, 86Cohen, R. D. 1983,
ApJ , 273, 489Courteau, S., McDonald, M., & Widrow, L. M. 2008, IAU Sym-posium, 245, 227Desroches, L.-B., Quataert, E., Ma, C.-P., & West, A. A. 2007,
MNRAS , 377, 402Dietrich, M., Hamann, F., Shields, J. C., Constantin, A., Vester-gaard, M., Cha ff ee, F., Foltz, C. B., & Junkkarinen, V. T. 2002, ApJ , 581, 912Dong, X.-B., Zhou, H.-Y., Wang, T.-G., Wang, J.-X., Li, C., &Zhou, Y.-Y. 2005,
ApJ , 620, 629Dong, X., Wang, T., Wang, J., Yuan, W., Zhou, H., Dai, H., &Zhang, K. 2008,
MNRAS , 383, 581Dopita, M. A., Heisler, C., Lumsden, S., & Bailey, J. 1998,
ApJ ,498, 570Elitzur, M., & Shlosman, I. 2006,
ApJ , 648, L101Falcke, H., Gopal-Krishna, & Biermann, P. L. 1995, A & A , 298,395Filippenko, A. V., & Sargent, W. L. W. 1988, ApJ , 324, 134Fiore, F., et al. 2008,
ApJ , 672, 94Fitzpatrick, E. L. 1999,
PASP , 111, 63Gaskell, C. M. 1984,
Astrophys. Lett. , 24, 43Gilli, R., Comastri, A., & Hasinger, G. 2007, A & A , 463, 79Greene, J. E., & Ho, L. C. 2007, ApJ , 670, 92Grimes, J. A., Rawlings, S., & Willott, C. J. 2004,
MNRAS , 349,503Haas, M., Siebenmorgen, R., Pantin, E., Horst, H., Smette, A.,K¨aufl, H.-U., Lagage, P.-O., & Chini, R. 2007, A & A , 473, 369Hao, L., et al. 2005, AJ , 129, 1795H¨aring, N., & Rix, H.-W. 2004, ApJ , 604, L89Hasinger, G. 2004, Nuclear Physics B Proceedings Supplements,132, 86Hasinger, G. 2008, A & A , 490, 905Heckman, T. M., Chambers, K. C., & Postman, M. 1992, ApJ ,391, 39Heckman, T. M., Kau ff mann, G., Brinchmann, J., Charlot, S.,Tremonti, C., & White, S. D. M. 2004, ApJ , 613, 109Hill, G. J., Goodrich, R. W., & Depoy, D. L. 1996,
ApJ , 462, 163Ho, L. C., Filippenko, A. V., Sargent, W. L. W., & Peng, C. Y.1997,
ApJS , 112, 391Hopkins, A. M., et al. 2003,
ApJ , 599, 971Hopkins, P. F., Hickox, R., Quataert, E., & Hernquist, L. 2009,arXiv:0901.2936Jackson, C. A., & Browne, I. W. 1990,
Nature , 343, 43Jorgensen, I., Franx, M., & Kjaergaard, P. 1995,
MNRAS , 276,1341Kaspi, S., Maoz, D., Netzer, H., Peterson, B. M., Vestergaard, M.,& Jannuzi, B. T. 2005,
ApJ , 629, 61Kau ff mann, G., et al. 2003, MNRAS , 346, 1055Kewley, L. J., Geller, M. J., Jansen, R. A., & Dopita, M. A. 2002, AJ , 124, 3135Kewley, L. J., Groves, B., Kau ff mann, G., & Heckman, T. 2006, MNRAS , 372, 961Konigl, A., & Kartje, J. F. 1994,
ApJ , 434, 446Krolik, J. H., & Begelman, M. C. 1988,
ApJ , 329, 702Krolik, J. H. 2007,
ApJ , 661, 52Lawrence, A. 1991,
MNRAS , 252, 586Le Borgne, J.-F., et al. 2003, A & A , 402, 433 Lu, H., Zhou, H., Wang, J., Wang, T., Dong, X., Zhuang, Z., & Li,C. 2006, AJ , 131, 790Lu, Y., Wang, T., Zhou, H., & Wu, J. 2007, AJ , 133, 1615Maia, M. A. G., Machado, R. S., & Willmer, C. N. A. 2003, AJ ,126, 1750McLure, R. J., & Dunlop, J. S. 2002, MNRAS , 331, 795Meisenheimer, K., Haas, M., M¨uller, S. A. H., Chini, R., Klaas,U., & Lemke, D. 2001, A & A , 372, 719Miller, J. S., & Goodrich, R. W. 1990, ApJ , 355, 456Moran, E. C., Barth, A. J., Kay, L. E., & Filippenko, A. V. 2000,
ApJ , 540, L73Mulchaey, J. S., Koratkar, A., Ward, M. J., Wilson, A. S., Whittle,M., Antonucci, R. R. J., Kinney, A. L., & Hurt, T. 1994,
ApJ ,436, 586Nenkova, M., Ivezi´c, ˇZ., & Elitzur, M. 2002,
ApJ , 570, L9Netzer, H., Mainieri, V., Rosati, P., & Trakhtenbrot, B. 2006, A & A ,453, 525Peng, C. Y., Impey, C. D., Ho, L. C., Barton, E. J., & Rix, H.-W.2006, ApJ , 640, 114Peterson, B. M., & Bentz, M. C. 2006, New Astronomy Review,50, 796Pier, E. A., & Krolik, J. H. 1992,
ApJ , 399, L23Reyes, R., et al. 2008, AJ , 136, 2373Rhee J. H., Larkin J. E., 2005, ApJ, 620, 151Richards, G. T., et al. 2005, MNRAS , 360, 839Rowan-Robinson, M., Valtchanov, I., & Nandra, K. 2009,
MN-RAS , 941Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998,
ApJ , 500,525Schurch, N. J., Done, C., & Proga, D. 2009,
ApJ , 694, 1Shao, Z., Xiao, Q., Shen, S., Mo, H. J., Xia, X., & Deng, Z. 2007,
ApJ , 659, 1159Shi, J., & Krolik, J. H. 2008,
ApJ , 679, 1018Simpson, C. 2005,
MNRAS , 360, 565Spergel, D. N., et al. 2003,
ApJS , 148, 175Suganuma, M., et al. 2006,
ApJ , 639, 46Tran, H. D. 1995,
ApJ , 440, 597Tran, H. D., Cohen, M. H., & Goodrich, R. W. 1995, AJ , 110,2597Tran, H. D., Cohen, M. H., Ogle, P. M., Goodrich, R. W., & diSerego Alighieri, S. 1998, ApJ , 500, 660Tremaine, S., et al. 2002,
ApJ , 574, 740Ueda, Y., Akiyama, M., Ohta, K., & Miyaji, T. 2003,
ApJ , 598,886Urry, C. M., & Padovani, P. 1995,
PASP , 107, 803V´eron-Cetty, M.-P., Joly, M., & V´eron, P. 2004, A & A , 417, 515Vestergaard, M., & Peterson, B. M. 2006, ApJ , 641, 689Wandel, A., Peterson, B. M., & Malkan, M. A. 1999,
ApJ , 526,579Wang, T.-G., & Zhang, X.-G. 2003,
MNRAS , 340, 793Wang, J. X., & Jiang, P. 2006,
ApJ , 646, L103White, R. L., Becker, R. H., Helfand, D. J., & Gregg, M. D. 1997,
ApJ , 475, 479York, D. G., et al. 2000, AJ , 120, 1579Young, S., Hough, J. H., Efstathiou, A., Wills, B. J., Bailey, J. A.,Ward, M. J., & Axon, D. J. 1996, MNRAS , 281, 1206Yun, M. S., Reddy, N. A., & Condon, J. J. 2001,
ApJ , 554, 803Zhang, K., Wang, T., Dong, X., & Lu, H. 2008,
ApJ , 685, L109Zhou, X.-L., & Wang, J.-M. 2005,
ApJ , 618, L83Zhou, H., Wang, T., Yuan, W., Lu, H., Dong, X., Wang, J., & Lu,Y. 2006,
ApJS , 166, 128Zier, C., & Biermann, P. L. 2002, A & A , 396, 91 c (cid:13)000