Fe II Emission in AGN: The Role of Total and Gas-Phase Iron Abundance
aa r X i v : . [ a s t r o - ph . C O ] S e p ApJ in press
Preprint typeset using L A TEX style emulateapj v. 11/10/09 Fe ii EMISSION IN AGN: THE ROLE OF TOTAL AND GAS-PHASE IRON ABUNDANCE
Gregory A. Shields , Randi R. Ludwig , Sarah Salviander (Received 2010 June 9; Accepted 2010 August 5) ApJ in press
ABSTRACTActive galactic nuclei (AGN) have Fe ii emission from the broad line region (BLR) that differs greatlyin strength from object to object. We examine the role of the total and gas-phase iron abundancein determining Fe ii strength. Using AGN spectra from the Sloan Digital Sky Survey (SDSS) in theredshift range of 0 . < z < .
35, we measure the Fe/Ne abundance of the narrow line region (NLR)using the [Fe vii ]/[Ne v ] line intensity ratio. We find no significant difference in the abundance of Ferelative to Ne in the NLR as a function of Fe ii /H β . However, the [N ii] / [S ii] ratio increases a bya factor of 2 with increasing Fe ii strength. This indicates a trend in N/S abundance ratio, and byimplication in the overall metallicity of the NLR gas, with increasing Fe ii strength.We propose that the wide range of Fe ii strength in AGN largely results from the selective depletion ofFe into grains in the low ionization portion of the BLR. Photoionization models show that the strengthof the optical Fe ii lines varies almost linearly with gas-phase Fe abundance, while the ultraviolet Fe ii strength varies more weakly. Interstellar depletions of Fe can be as large as two orders of magnitude,sufficient to explain the wide range of optical Fe ii strength in AGN. This picture is consistent withthe similarity of the BLR radius to the dust sublimation radius and with indications of Fe ii emittinggas flowing inwards from the dusty torus. Subject headings: galaxies: active — quasars: general INTRODUCTION
The broad emission-line spectrum of quasars often in-cludes strong Fe ii in the optical and ultraviolet. The dif-ference between the weakest and strongest optical Fe ii emission exceeds a factor of 10, measured as equivalentwidth (EW) or as Fe ii /H β line ratio. For recent discus-sions of Fe ii intensities in AGN, and references to ear-lier work, see Kova˘cevi´c et al. (2010) and Ferland et al.(2009). The Fe ii strength anti-correlates with thestrength of the narrow [O iii] emission line. This trendalong with several associated correlations defines the so-called “Eigenvector 1” (EV1), which characterizes someof the most conspicuous differences among the propertiesof AGN (Boroson & Green 1992, hereinafter BG92). Thequest for physical drivers of EV1 has inspired a numberof studies (e.g., Wills et al. 1999; Marziani et al. 2003).Boroson (2002) and Netzer & Trakhtenbrot (2007) findthat Fe ii increases with Eddington ratio L/L Ed , as orig-inally suggested by BG92. However, the physics under-lying this correlation remains unclear. There is even de-bate as to whether the Fe ii emission is entirely pow-ered by the ionizing continuum of the central source,or comes in some measure from a mechanically heatedregion (Wills, Netzer, & Wills 1985; Collin & Joly 2000;Sigut & Pradhan 2003; Bruhweiler & Verner 2008). Inthis situation, any observational clues to the nature ofEV1 and the great range of Fe ii strength are of value.It is generally assumed that the strength of Fe II emis-sion is driven by physical conditions within the BLR,such as ionizing continuum, BLR density and geometry, Department of Astronomy, University of Texas, Austin, TX78712; [email protected]; [email protected];[email protected] Department of Physics, Southwestern University, George-town, TX 78626 column density, and turbulent velocity. However, a highabundance of Fe has been discussed to help producethe strongest Fe ii observed (Wills, Netzer, & Wills1985; Collin-Souffrin, Hameury, & Joly 1988;Hamann & Ferland 1993). The utility of Fe ii to assessthe Fe abundance in high redshift QSOs has receivedconsiderable interest, in the context of galactic chemicalevolution (Hamann & Ferland 1993; Verner et al. 2003;Baldwin et al. 2004; Netzer & Trakhtenbrot 2007, andreferences therein). Here we assess the importance of theabundance of Fe relative to the α -elements, the overallmetallicity of the nuclear gas, and the depletion of Feinto grains for the strength of Fe ii emission in QSOs. QSO SAMPLE AND MEASUREMENTS
We investigated the influence of differing abundancesin quasars on their optical Fe ii emission strength bystudying the optical emission-line properties of a sampleof QSOs from the Sloan Digital Sky Survey (SDSS) .It is difficult to measure abundances within the BLR di-rectly, because it is a region of high density, line width,and line optical depth. On the assumption that the abun-dance in the NLR and BLR is the same for a given object(see below), we used the narrow emission-line spectrumto assess abundances in the NLR. We considered twokey line ratios: (1) The [Fe vii ]/[Ne v ] intensity ratiogives a measure the Fe/Ne abundance ratio. (2) The [N ii] / [S ii] ratio gives a measure of the N/S abundanceratio, which is in turn a secondary indicator of the over-all metallicity of the gas. Our sample consists of 1571quasars from SDSS Data Release 7 (DR7). These ob-jects were selected in the manner of the “HO3” sampleof Salviander et al. (2007), with the additional require-ment of a signal-to-noise (S/N) ratio greater than 10 in Shields, Ludwig, & Salvianderthe continuum at λ β line and the opti-cal Fe ii blends were measured with the aid of a spectrumfitting program described by Salviander et al. (2007), us-ing a template fitting procedure to establish the flux inFe ii relative to the local continuum. We characterizedthe Fe ii emission strength using the flux ratio of theFe ii β , following BG92. Weused quasars at redshifts 0 . < z < .
35 to ensure cov-erage of both [Ne v ] λ vii ] λ vii ] line, we binned the 1571objects by Fe ii strength, and made five composite spec-tra of “very weak,” “weak,” “medium,” “strong,” and“very strong” Fe ii emission. These composites had 312to 315 objects, within bins bounded by Fe ii /H β valuesof 0, 0.215, 0.372, 0.524, 0.708, and 1.82, respectively.Individual spectra were corrected for Galactic redden-ing using the extinction values A g given by the SDSSpipeline, and normalized to a mean flux density F λ ofunity using all wavelength points in a particular spec-trum. The individual spectra in each group were shiftedin wavelength to the rest frame and re-binned to a com-mon wavelength grid at a spacing of 1.41 ˚A. The adoptedcomposite spectrum was a mean of the rebinned F λ forall contributing spectra at a given wavelength. The com-posite spectra are shown in Figures 1 and 2. The regionof the [Fe vii ] and [Ne v ] line is shown in Figure 2. For adiscussion of issues involving composite spectra of QSOs,see Vanden Berk et al. (2001), and references therein.From these composites, we measured the emission-linefluxes of a number of lines, including [O iii] λ [O ii] λ iii ] λ v ] λ [S ii] λλ , vii ] λ ii /H β val-ues are averages of the values for the individual spectrathat compose each composite. Most lines were measuredusing a Gaussian fit with the IRAF task SPLOT . Thebroad Balmer emission lines (H α , H β ) were measuredusing a Lorentzian profile. Estimated uncertainties are10% for the stronger lines, including continuum place-ment and faithfulness of the fit. For [Fe vii ] and [Ne v ],the uncertainty is as much as 20%, based on noise, con-tinuum uncertainty, and the presence of a strong bluewing on both lines that we excluded from the fit. The [N ii] λ λ α line and relatively difficult tomeasure. Therefore we measured the [N ii] intensity bysubtracting from the H α – [N ii] blend a doublet withthe theoretical 3.0-to-1 intensity ratio, each line havinga Gaussian profile with a central wavelength and widthbased on the redshift and line width of [S ii] λ α linehad a smooth profile with no visible residual intensity orover-subtraction of [N ii] . Error bars were estimated bydetermining [N ii] intensities giving a slight under- orover-subtraction as judged by eye. This gave an uncer-tainty of about ±
12% for each composite.Our results will be discussed in terms of trends of ob-served line ratios. For actual ionic abundances, collision IRAF is distributed by the National Optical Astronomy Ob-servatories, which are operated by the Association of Universitiesfor Research in Astronomy, Inc., under cooperative agreement withthe National Science Foundation.
TABLE 1Emission Line Ratios for Composite Spectra
Emission Intensity RatioLines vlow low med high vhigh [N ii] / [S ii] [N ii] / [O ii] [O ii] / [O iii] [S ii] / [O ii] vii ]/[Ne v ] 0.33 0.32 0.37 0.35 0.32[Ne v ]/[Ne iii ] 0.97 1.10 1.15 1.61 1.81H α /H β [S ii] (6720/4072) 9.2: 10.3: 9.7: 9.9: 13.5: [S ii] (6716/6731) 1.19 1.05 1.07 1.10 1.08Fe ii /H β Note . — Intensity ratio for emission lines measured from com-posite spectra binned by broad Fe ii strength. Values refer to thenarrow emission lines except for the Balmer lines and Fe ii . In-tensities include both lines for the [O ii] and [S ii] doublets butonly the stronger line for [N ii] , [O iii] , [Ne iii ], and [Ne v ]. Colonindicates large uncertainty. See text for discussion. strengths from Berrington et al. (2000) and Osterbrock& Ferland (2006) lead to the relation n ( F e +6 ) /n ( N e +4 ) = 0 . I ( λ /I ( λ [N ii] / [S ii] ratio we have n ( N + ) /n ( S + ) = 6 . I ( λλ , /I ( λλ , RESULTS FOR THE NARROW LINE REGION
Iron
Nussbaumer & Osterbrock (1970) suggested that[Fe vii ]/[Ne v ] should be a good measure of the Fe/Neratio, based on the similarity of the ionization potentials.There is no significant trend in [Fe vii ]/[Ne v ] in ourcomposite spectra. The Fe ii /H β ratio varies by a factorof 8 from the “very low” to “very high” composite.Taking neon to represent the α -elements in general, weconclude that differences in Fe abundance, relative tothe α -elements is not a significant cause of the observedrange of Fe ii emission strength, from a statistical pointof view. In particular, overabundances of Fe/O andFe/Mg of a factor of 2 to 10, as motivated by chemicalevolution models and attempts to fit the Fe ii /Mg ii ra-tio with photoionization models (Wills, Netzer, & Wills1985; Hamann & Ferland 1993), appear to be ruled out.This conclusion is based on the assumption that the[Fe vii ]/[Ne v ] ratio is a faithful measure of Fe/Ne, andthat Fe is not significantly depleted into grains in the[Fe vii ] zone of the NLR. Ferguson et al. (1997a) andNagao et al. (2003) conclude that refractory elements arenot depleted in the coronal line region of the NLR, butthey also find that [Fe vii ] and [Ne v ] do not comefrom the same place in “locally optimally emitting cloud”(LOC) models of the NLR. Here we assume that anyionization correction for Fe +6 /Ne +4 in the NLR doesnot change systematically with the broad line Fe ii in-tensity among our composites. We further assume thatthe abundances in the BLR are similar to those in thee ii Emission in AGN 3NLR. However, it is possible that intense star formationin the nucleus may give chemical enrichment on a spa-tial scale smaller than the NLR (see Hamann et al. 2007;Hamann & Ferland 1999, and references therein).
Ionization
The composite spectra were originally constructed toassess the [Fe vii ] strength in the NLR, but they alsoafford an opportunity to examine other narrow line ra-tios for any systematic dependence on Fe ii strength.One issue is the level of ionization of the gas. Table 1and Figure 3 show that the [O ii] / [O iii] ratio is closelysimilar for the various Fe ii bins. There is a significantincrease in [Ne v ]/[Ne iii ] with increasing Fe ii strength,reflecting a ∼ decrease in the [Ne v ] equivalent widthacross the bins and a much larger decrease in the EW of[Ne iii ]. The narrow He ii λ [O iii] λ ii together with a systematicdecrease in the EW of λ ii and [O iii] is well known(BG92). Although the constancy of [Fe vii ]/[Ne v ] couldresult from offsetting ionization and abundance trends,a straightforward interpretation is that the relative sizeof the zone containing O +2 and Ne +2 decreases with in-creasing Fe ii without a major effect on the Fe +6 /Ne +4 ionization correction. Reddening
The [S ii] / [O ii] ratio shows an increase of about 30%with increasing Fe ii strength. This may be an indica-tion of a modest increase in reddening of the NLR withincreasing Fe ii . The [S ii] I ( λ /I ( λ I (H α ) / I(H β ) intensity ratio. Nitrogen
The [N ii] / [S ii] ratio shows a systematic increase bya factor of 2 from the very low to the very high Fe ii bins. This ratio is insensitive to reddening and electrontemperature. The N + and S + ions have similar ionizationpotentials and occupy similar zones of the nebular ioniza-tion structure. This and the constancy of [O ii] / [O iii] suggests that the trend in [N ii] / [S ii] is not a resultof ionization of the NLR. The trend of increasing [N ii] strength is evident in the [N ii] / [O ii] and [N ii] / [O iii] ratios as well. These results imply a real trend in theN/S and N/O chemical abundance ratios with increas-ing Fe ii strength, amounting to a factor of 2 from the“very low” to the “very high” Fe ii strengths. Nitro-gen is largely a secondary nucleosynthetic product, sothat N/O increases with O/H. The H ii region resultsof van Zee et al. (1998) show N/O increasing almost lin-early with O/H above 12+log O / H = 8 .
5. Assuming thatO/H is in this range in the AGN studied here, then thetrend of N/O with Fe ii implies an increase of a factorof 2 in O/H. Although the chemical evolution of AGNhost galaxies may be complicated (Hamann & Ferland1999; Hamann et al. 2002; Netzer & Trakhtenbrot 2007),it may be reasonable to assume that Fe/H varies roughly with O/H for the modest redshifts considered here. Inthis case, the Fe/H abundance ratio may contributeroughly a factor of 2 to the range in Fe ii strength fromour “very low” to “very high” bins. This conclusionis consistent with the work of Netzer & Trakhtenbrot(2007), who find an increase of Fe ii strength with Ed-dington ratio L/L Ed in a large sample of SDSS quasars.Combining this with indications of an increasing N/Cwith L/L Ed in the BLR (Scott et al. 2004; Hamann et al.2002), they argue that the overall metallicity in QSOs in-creases with Fe ii strength. However, our results indicatethat this makes only a modest contribution to the fullrange of optical Fe ii strength in AGN. Note, however,that our QSO sample is at low redshift, whereas much ofthe interest in iron abundances in QSOs has focused onhigh redshifts. Fe ii STRENGTH AND X-RAY HEATING
It is widely assumed that the Fe ii emission is largelypowered by the soft X-ray portion of the ionizing spec-trum, given that photoionization is indeed the primaryexcitation mechanism. The harder photons in the ioniz-ing continuum create an extensive warm, partially ion-ized zone where Fe ii and other low ions are subject tocollisional excitation and in some cases other excitationmechanisms (see Kwan & Krolik 1981; Ferland et al.2009, and references therein). Does the relative strengthof the X-ray continuum drive the range of Fe ii strengthobserved in AGN? The following considerations suggestthat it does not:1) A number of empirical studies have examined corre-lations of Fe ii strength with the X-ray slope or with theX-ray/optical ratio (Sulentic, Marziani, & Dultzin 2000,and references therein). Lawrence et al. (1997) stud-ied a sample of AGN with extreme values of R Fe II ≡ I ( λ /I (H β ). The strong Fe ii emitters have X-rayproperties, including α ox , similar to the weak Fe ii ob-jects. In particular, the prototype strong Fe ii object,I Zw 1, has α ox = − .
4, an entirely typical value. Com-bining their data with the complete sample of Laor et al.(1997), Lawrence et al. found little correlation between α ox and R Fe II , and only an ambiguous correlation with α x . Indeed, Figure 3 of Lawrence et al. shows a slighttrend in the sense of strong optical Fe ii for weak X-ray luminosity (steep α ox ). They did find a significantcorrelation with the X-ray to IR slope α ix in the senseof stronger infrared for stronger Fe ii . Using compositespectra for X-ray bright and X-ray faint QSOs, Green(1998) found that UV Fe ii was stronger whereas opti-cal Fe ii was weaker for X-ray bright objects. However,the differences were small compared to the full rangeof Fe ii strength among individual AGN. Leighly et al.(2007) find a weak X-ray continuum but strong Fe ii emission in PHL 1811, and discuss other similar exam-ples. None of these results gives support for the ideathat stronger X-ray continuum drives stronger Fe ii forAGN as a class. Ferland & Persson (1989) reached a sim-ilar conclusion regarding the Ca ii emission from AGN,which also comes from the partially ionized zone.2) In order to explore the expected response of theFe ii emission to differences in X-ray luminosity (andother parameters) we have computed a set of models ofthe BLR using version 07.02.00 of the photoionizationcode Cloudy, most recently described by Ferland et al. Shields, Ludwig, & Salviander TABLE 2Photoionization Model Results
Model 1 2 3 4 α ox -1.4 -2.0 -1.4 -1.4Fe Depletion 0.0 0.0 -1.5 -1.5Other Depletion 0.0 0.0 0.0 -1.5 I/I H β ii iii] iv α α iii ] 1.69 1.78 1.54 1.45C iv Note . — Results of Cloudy photoionization models.Logarithmic depletion relative to hydrogen is given foriron and for other refractory elements. See the text fordiscussion. (1998). As a reference model, following Ferland et al.(2009), we used solar abundances, an ionizing flux φ =10 cm − s − , and a gas density N = 10 cm − , giv-ing an ionization parameter U ≡ φ/N c = 10 − . . Theinternal turbulent velocity was u turb = 100 km s − , andthe stopping column density was 10 cm − , in orderto include an extensive partially ionized zone. The ion-izing continuum was the sum of (1) a UV component L ν ∝ ν − . exp( − hν/kT cut ) with T cut = 10 . K to rep-resent the Big Blue Bump, and (2) an X-ray power law L ν ∝ ν − . The continuum had a low frequency expo-nential cutoff below 0.01 Ryd. The models used the fulltreatment of the Fe ii ion (371 levels) and were iteratedto convergence of the diffuse radiation field. The ratio ofthe X-ray to the UV component is controlled by the pa-rameter α ox . For the reference model (Model 1), we used α ox = − .
4, a typical observed value (Lawrence et al.1997). This model has not been optimized to fit a typ-ical AGN emission-line spectrum, but simply serves asa reference point for exploring the effect on Fe ii ofchanging various model parameters. A simultaneous fitto AGN emission-line spectra requires a combination ofphotoionized clouds with a range of physical conditions(Baldwin et al. 1995).Table 2 gives line intensities from the Cloudy mod-els. The H α , C iii ], C iv , and Mg ii intensities are rea-sonable. Ly α is stronger than observed relative to H β ,a familiar problem with photoionization models (e.g.,Kwan & Krolik 1981). The Cloudy output gives Fe ii intensities summed over broad wavelength bands at 1000– 2000, 2000 – 3000, 4000 – 6000, 6000 – 7800, and 7800– 30000 ˚A. Note that the λ ii bump, and the λ ii blends at λ λ ii intensity seen in strong Fe ii ob-jects, a common problem with photoionization models asmentioned above. In Model 2, we used α ox = − . α ox (Lawrence et al. 1997).The Fe ii intensity (relative to H β ) actually increasedslightly in the X-ray weak model. We also computedan alternative pair of models with φ = 10 cm − s − , N = 10 cm − , T cut = 10 . K, and no turbulence. Inthis case, the Fe ii λ λ α ox = − . − .
0. Even in this case, the degree of change is insuffi-cient to give the full observed range of Fe ii strength. Anextreme systematic variation of α ox with Fe ii would berequired, for which the observations give little support. Fe ii STRENGTH AND IRON DEPLETION INTO GRAINS
The above results indicate that changes in the total ele-mental abundance of Fe and the X-ray luminosity do notcause the wide observed range of Fe ii emission strengthamong AGN. We propose instead that Fe ii strength inAGN largely results from differing degrees of depletionof the gas-phase abundance of iron into grains in therelevant portions of the BLR. Gas-phase depletions ofrefractory elements in the interstellar medium can be se-vere. Iron is depleted by up to 2 orders of magnitude inthe interstellar medium and in ionized nebulae (see dis-cussion below). Such a degree of depletion, if present inAGN with weak Fe ii but not those with strong Fe ii , canaccount for much of the observed range of Fe ii strengthin AGN.There has been considerable discussion of refractoryelement depletions in the NLRs of AGN (Ferguson et al.1997b). Gaskell, Shields, & Wampler (1981) consideredthe question of refractory element depletions in the BLRand concluded that depletions of Si, Mg, and Fe as severeas those in the ISM did not occur. This conclusion wasbased mostly on the Mg ii , Si iii , and Si iv lines, andthe intensity of Fe ii in strong Fe ii objects. Here wesuggest that the degree of depletion of iron varies fromobject to object, and can be severe in objects with weakFe ii . In a study of [O i] and Ca ii emission from AGN,Matsuoka, Kawara, & Oyabu (2008) note that depletionof Ca ii might help to reconcile predicted and observedintensities and mention the possibility that depletionsmay affect the Fe ii lines. Ferland & Persson (1989) notethat dust mixed with the BLR gas could help to explainlow observed values of the Ca ii H and K emission lines,relative to the infrared triplet. As discussed below, theradius of the BLR is interestingly close to the dust sub-limation radius; and for typical parameters, grains mayexist in the partially ionized zone of the BLR clouds butnot in the highly ionized surface layers.The present proposal does not address the long-standing question of how to explain the large Fe ii strength observed in many AGN (Wills, Netzer, & Wills1985; Bruhweiler & Verner 2008, and references therein).Rather, it serves to separate the question of Fe ii strengthinto two parts: (1) why is Fe ii so strong in some ob-jects, and (2) why is it so weak in others? We do notattempt to resolve the first question in this paper, ex-cept to note that under the present proposal, the Fe ii physical excitation mechanism is freed from the require-ment of explaining by itself the wide range of observedFe ii strengths.e ii Emission in AGN 5
Photoionization Models
In order to explore the dependence of Fe ii strengthon gas-phase Fe abundance we have computed two ad-ditional Cloudy models (see Table 2). Model 3 is oth-erwise identical to Model 1 but has a depletion of ironby 1.5 dex. The λ λ . and (Fe/H) . , respectively. Inthe N = 10 cm − model described above, a 1.5 dex de-pletion of Fe abundance reduced the λ λ ii intensity varied almost linearly with thegas-phase iron abundance, and the UV Fe ii less strongly.These results are consistent with previous photoioniza-tion studies. Verner et al. (2003) and Baldwin et al.(2004) found that the Fe ii intensity varies approximatelyas (Fe/H) . for the optical lines and as (Fe/H) . forthe UV bump. Some of these models assume differ-ent physical parameters for the BLR from ours, and theVerner et al. (2003) models use an 830 level Fe + modelatom as opposed to 371 levels in our models and thoseof Baldwin et al. (2004).Other refractory elements may be depleted where ironis. In Model 4, the abundances of Al, Mg, Si, Ca, andFe were all depleted by 1.5 dex. Table 2 shows that themultiplets Mg ii λ iii λ iv λ ii multiplets in Model 4is similar to the case in which only Fe was depleted.Different refractory elements have different depletionsin the ISM, and different grain compositions may be moreor less easily destroyed in the AGN environment. Thismay lead to useful diagnostics for grain destruction andgas-phase depletions. Delgado Inglada et al. (2009) sum-marize published depletion factors in planetary nebulae:1/6 to 1/300 for Ca, 1/2 to 1/350 for Al, 1/3 to 1/300 forFe, near solar to 1/10 for Mg, and near solar to 1/20 forSi. Silicon appears to be depleted by a lesser factor thaniron in H ii regions. Garnett et al. (1995) found Si deple-tions of only -0.1 to -0.6 dex in extragalactic H II regions,significantly less than in dense interstellar clouds. If Si ismore easily restored to the gas phase in H II regions, thesame may be true in AGN, allowing objects with weakFe ii to have normal intensities of Si iii and Si iv . Geometrical Considerations
What might cause widely differing degrees of depletionof iron, and possibly other refractory elements, amongAGN? From considerations of BLR covering factor, Ly-man continuum absorption, line widths, and reverbera-tion mapping, Gaskell (2009) argues that the BLR is theinward extension of the dusty torus, with an inflow veloc-ity a substantial fraction of the orbital velocity. Such apicture lends itself to the idea of refractory element deple-tions in the low ionization zone. As material flows inwardthrough the dusty torus, its equilibrium temperaturerises as it experiences a stronger radiation field, eventu-ally reaching the sublimation temperature of ∼ R d = (0 . L , where L is the AGN luminosity in units of 10 erg s − .If this happens before the material reaches the main re-gion of low ionization line emission, then strong emissionin Fe ii and other lines of refractory elements will occur.On the other hand, if the dust survives through most ofthe low ionization zone, then these emission lines will beweak, because of the lack of emitting ions and becauseof attenuation of the ionizing radiation and line emissionby the dust itself.Netzer & Laor (1993) suggested that the survival ofdust outside the sublimation radius leads to extinctionof the ionizing continuum and suppression of line emis-sion, setting a natural limit to radius of the BLR (see alsoLaor 2007). Infrared variability studies give a radius ofthe dusty torus just outside the BLR (Suganuma et al.2006), supporting this picture. Here we suggest thatthe relationship between the sublimation radius and theBLR outer boundary differs from object to object. Crit-ical to this picture is the actual radius of Fe ii emissionrelative to the dust sublimation radius. Reverberationstudies and line widths indicate that the lower ionizationlines often come from larger radii in the BLR (Peterson1997; Sulentic, Marziani, & Dultzin 2000). This may re-sult from some combination of actual ionization strati-fication and line emissivity effects as considered in theLOC model (Baldwin et al. 1995). There have been fewsuccessful reverberation studies of the Fe ii emitting ra-dius in AGN. For the Seyfert galaxy NGC 5548, Maoz(1993) found a similar time lag for the UV Fe ii linesas for Ly α . In a reverberation study of Ark 120, whichhas strong optical Fe ii , Kuehn et al. (2008) found thatthe lag for the optical Fe ii lines was ill-defined but maybe around 300 days, larger than for H β . They estimatea dust sublimation radius in this object of ∼
460 lightdays, and conclude that within the uncertainties “it isplausible that the optical Fe ii emission is produced ator just inside the dust sublimation radius.” In a study ofline profiles of AGN in SDSS, Hu et al. (2008) concludethat the Fe ii emission comes from an inflowing zone inthe outer part of the BLR. Taken together, these studiesare consistent with Fe ii emission from gas entering theBLR from the dusty torus.A refinement of this picture takes account of the atten-uation of the ionizing radiation field with depth withinthe ionized cloud or ionized surface layer of the diskor torus. Reverberation measurements of the BLR ra-dius (often involving the H β line) typically give a ra-dius smaller than the dust sublimation radius by a fac-tor of order 2, based on radii and luminosities given byBentz et al. (2009). Thus, refractory grains should notsurvive at the illuminated face of a BLR cloud. However,the radiation field diminishes with depth in the cloud,allowing dust to survive at large depths. As the dustygas flows from deep in the torus toward the irradiatedsurface, the ambient radiation field intensifies and thegrain temperature rises, reaching the sublimation pointat some depth in the ionization structure. Grain equi-librium temperatures calculated by the Cloudy programsupport this picture. In the reference model (Model 1)described above, the temperature of silicate grains witha radius of 0.094 microns is 680 K at the maximum col-umn depth of 10 cm − . This allows grains to surviveat this depth, even though the incident flux at the cloudface would easily evaporate refractory grains. The grain Shields, Ludwig, & Salviandertemperature rises with decreasing depth, reaching thesublimation temperature of 1500 K at a column den-sity of 10 . cm − . This is well below the “Str¨omgrendepth” of 10 . cm − , where hydrogren is 50% ionized.The grain temperature as calculated by Cloudy wouldbe 2000 K at this latter depth, and 3100 K at the cloudsurface. Reference to Figure 3 of Ferland et al. (2009)shows that in their model (similar to our Model 1), mostof the optical Fe ii emission occurs below the sublima-tion depth of 10 . cm − , whereas most of the ultra-violet Fe ii emission occurs at shallower depths. Thus,evaporation of the grains occurs at a depth giving se-vere reduction of the optical but not the ultraviolet Fe ii emission. Dust will also not affect the emission from thehighly ionized gas above the Str¨omgren depth, such asSi iii] and Si iv ; and weak optical Fe ii could accompanya normal ratio of the ultraviolet Fe ii to Mg ii lines. Theexact sublimation depth depends on grain size and com-position, but the qualitative pattern remains for a rangeof compositions and sizes included in the Cloudy out-put. For different AGN with different parameters, thesublimation point will occur at different depths in theionization structure. Gaskell et al. (2007) have also dis-cussed the affect of attenuation of the AGN continuumin the BLR on the radius of dust sublimation.An alternative geometry for the BLR involvesa radiatively driven wind from the accretion disk(Murray & Chiang 1998). In this picture, one issue iswhether grains can survive in the inflowing material tothe radius where it is expelled in the wind. Standardaccretion disk physics (Peterson 1997) gives for the diskeffective temperature due to internal viscous dissipation T eff = (10 . K)( ˙
M / ˙ M E ) / M − / v / , (3)where ˙ M / ˙ M E is the accretion rate relative to that giv-ing the Eddington luminosity, M is the black hole massin units of 10 M ⊙ , and v is the orbital velocity atthe radius of interest in units of 3000 km s − . Thus, thesublimation temperature for refractory grains is reachedat orbital velocities appropriate for the broad emissionlines. This may be consistent with differing degrees ofgrain evaporation in different objects. (The mid-planetemperature of the disk will be higher.) The above ex-pression qualitatively agrees with the observed trend ofstronger optical Fe ii with increasing Eddington ratio,but the dependence on black hole mass may be prob-lematic. Moreover, it does not give such a natural L . dependence for the BLR radius as does the dust subli-mation model of Netzer & Laor (1993). Note that in thedisk-wind picture, the energy for the line emission stillcomes from photoionization by the central continuum; lo-cally produced energy is insufficient at this shallow depthin the gravitational potential. Sublimation at an inter-mediate depth in the ionization structure, as discussedabove, could also occur in the disk-wind model. Turbulent Velocity and Column Density
Local turbulence substantially affects the Fe ii spec-trum in photoionization models by facilitating continuumand line-line fluorescence. Increasing the turbulence canincrease the Fe ii strength and give better agreement be-tween the predicted shape of the Fe ii blends and obser-vation (Baldwin et al. 2004; Bruhweiler & Verner 2008, and references therein). Bruhweiler & Verner (2008) finda factor of 2 increase in the UV Fe ii strength rel-ative to Mg ii as the turbulence increases from 5 to50 km s − (their Table 2). When we decreased v turb from100 km s − to zero in Model 1, the λ ii band de-creased a factor 10, and the λ ii in AGN. On the other hand, turbulence still appears tobe inadequate to give the full observed range in opticalFe ii . Moreover, Baldwin et al. (2004) argue that sub-stantial turbulence is required to fit the detailed shapeof the UV Fe ii feature, so that there may be limitedfreedom to vary the turbulence. A useful test may be toexamine the observed shape of the UV and optical Fe ii blends as a function of optical Fe ii strength, for com-parison with photoionization models that vary either theturbulence or the gas-phase abundance of Fe.The column density of the emitting clouds also hasan important effect on the Fe ii emission. Ferland et al.(2009) illustrate the increase in Fe ii /H β with increasingcolumn density. They find that the minimum columndensity is ∼ cm − for gravity to overpower radiationpressure and allow infall of clouds as found by Hu et al.(2008). The UV Fe ii lines show little change above thiscolumn density, but the optical Fe ii increases a factor ∼ . ∼ cm − to ∼ cm − . Usingarguments based on virial determinations of the blackhole mass in AGN, Netzer (2008) also concludes that thecolumn densities must substantially exceed ∼ cm − to avoid excessive effects of radiation pressure on theorbital velocities of the BLR clouds. Thus, there may belimited freedom to vary the column density in order toproduce the wide range of optical Fe ii strength observed.The relative behavior of the optical and UV Fe ii bandsmay provide clues to the predominant cause of the rangeof Fe ii strength. In our models, the UV and optical Fe ii both increase with Fe abundance, albeit more weaklyfor the UV blends. However, increasing the microtur-bulence increased the UV Fe ii by a greater factor thatthe optical Fe ii ; and increasing the column density be-yond ∼ cm − mainly increases the optical Fe ii , asnoted above. Shang et al. (2007) give optical and Fe ii strengths for a sample of AGN. Their results show a muchgreater range in the EW of the optical Fe ii bands than inthe UV bands, and a weak anti-correlation between theoptical and UV (see also Wills, Netzer, & Wills 1985).The fact that the optical Fe ii shows a greater rangeof intensity that the UV Fe ii may favor an explana-tion other than microturbulence. One complication iswhether the optical and UV lines originate at substan-tially different radii in the BLR, as suggested by somereverberation and line-width studies (e.g., Maoz 1993;Hu et al. 2008). Ferland et al. (2009) suggest that theobserved optical Fe ii may be strongly affected by radia-tion escaping from the shielded face of the photoionizedclouds. The observational and theoretical situation forFe ii is complex, and further work will be needed to de-vise definitive tests of the role of chemical abundancesand physical conditions. CONCLUSION e ii Emission in AGN 7We have used composite SDSS spectra of AGN to ex-amine the behavior of the narrow emission lines as a func-tion of Fe ii strength. The [Fe vii ] line shows only a weakincrease with increasing Fe ii strength, indicating thatthe iron abundance contributes little to the wide rangeof Fe ii strength in AGN. There is, however, a significantincrease in the N/O abundance ratio with Fe ii strength,which suggests an increase in overall metallicity. Thereis little change in the level of ionization in the NLR asa function of Fe ii strength. This, together with resultsof photoionization models, suggests that differences inthe shape of the ionizing continuum, specifically the softX-ray luminosity, are not the main drivers of the Fe ii strength. We propose that differences in the degree ofdepletion of Fe into grains in the low ionization portionof the BLR are largely responsible for the weakness ofFe ii in some AGN, while it is strong in others. Thispicture is consistent with the approximate coincidence ofthe BLR radius and the dust sublimation radius, withindications that the BLR consists of material flowing in-ward from the dusty torus toward the central black hole,and with the variation of grain temperature with depthin the ionization structure of the BLR gas.The strength of Fe ii emission is a major component ofthe set of correlations known as “Eigenvector 1” (EV1)discussed by BG92. These include weak [O iii] associ-ated with strong Fe ii and narrower widths of the broadH β line. Radio loud AGN tend to have strong [O iii] andweak Fe ii . These trends have been the subject of manystudies, but a good physical understanding of their originremains lacking. BG92 suggested that high column den-sities in the BLR enhance Fe ii while diminishing the ion-izing radiation reaching the NLR. Ludwig et al. (2009),in a spectral principal components analysis of AGN inSDSS, found the interpretation of the eigenvectors to becomplicated. They argued that covering factor of theNLR was the likely cause of the range in [O iii] strength.Ferland et al. (2009) suggest that the higher column den-sities required for infall in more luminous AGN can helpto explain the correlation of Fe ii strength with L/L Ed .The interpretation of Fe ii strength in terms of dust depletion opens many questions for investigation. Cansupport for this picture be found in the line intensitiesof other refractory elements? Does the Fe ii emittingradius bear a different relationship to the dust sublima-tion radius for strong and weak Fe ii emitters? What arethe implications for the infrared emission of AGN? Whatunderlying causes lead to the correlations between Fe ii strength and other properties such as [O iii] strength, ra-dio emission, and Eddington ratio? This paper does notoffer answers to these larger questions, but the explana-tion of Fe ii strength in terms of gas-phase depletionsgives a new context in which to address them.We thank Mark Botorff, Gary Ferland, Martin Gaskell,Richard Green, Fred Hamann, Ari Laor, Hagai Netzer,and Bev Wills for helpful discussions and comments onthe manuscript, and Alyx Stevens for assistance. G.S.acknowledges support from the Jane and Roland Blum-berg Centennial Professorship in Astronomy at the Uni-versity of Texas at Austin. We thank Karl Gebhardtfor the composite spectrum program and Erin Bonningfor the emission-line subtraction program used for thenitrogen line measurements.Funding for the Sloan Digital Sky Survey (SDSS) hasbeen provided by the Alfred P. Sloan Foundation, theParticipating Institutions, the National Aeronautics andSpace Administration, the National Science Foundation,the U.S. Department of Energy, the Japanese Monbuka-gakusho, and the Max Planck Society. The SDSS is man-aged by the Astrophysical Research Consortium (ARC)for the Participating Institutions. The ParticipatingInstitutions are The University of Chicago, Fermilab,the Institute for Advanced Study, the Japan Participa-tion Group, The Johns Hopkins University, the KoreanScientist Group, Los Alamos National Laboratory, theMax-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New MexicoState University, University of Pittsburgh, Universityof Portsmouth, Princeton University, the United StatesNaval Observatory, and the University of Washington. REFERENCESAntonucci, R. R. J., & Miller, J. S. 1985, ApJ, 297, 621Baldwin, J., Ferland, G., Korista, K., & Verner, D. 1995, ApJ,455, L119Baldwin, J. A., Ferland, G. J., Korista, K. T., Hamann, F., &LaCluyz´e, A. 2004, ApJ, 615, 610Bentz, M. C., Peterson, B. M., Netzer, H., Pogge, R. W., &Vestergaard, M. 2009, ApJ, 697, 160Berrington, K. A., Nakazaki, S., Norrington, P. H. 2000, A&AS,142, 313Boroson, T. A. 2002, ApJ, 565, 78Boroson, T. A. & Green, R. F. 1992, ApJS, 80, 109 (BG92)Bruhweiler, F., & Verner, E. 2008, ApJ, 675, 83Collin, S., & Joly, M. 2000, New Astronomy Reviews, 44, 531Collin, S., Hameury, J.-M., & Joly, M. 1988, A&A, 205, 19Delgado Inglada, G., Rodr´ıquez, M., Mampaso, A., & Vironen, K.2009, ApJ, 693, 1335Ferguson, J. W., Korista, K. T., & Ferland, G. J. 1997a, ApJS,110, 287Ferguson, J. W., Korista, K. T., Baldwin, J. A., & Ferland, G. J.1997b, ApJ, 487, 122Ferland, G. J., & Persson, S. E. 1989, ApJ, 347, 656Ferland, G. J., Korista, K.T., Verner, D.A., Ferguson, J.W.,Kingdon, J.B. & Verner, E.M. 1998, PASP, 110, 761 Ferland, G. J., Hu, C., Wang, J.-M., Baldwin, J. A., Porter, R. L.,van Hoof, P. A. M., & Williams, R. J. R. 2009, ApJ, 707, L82Garnett, D. R., Dufour, R. J., Peimbert, M., Torres-Peimbert, S.,Shields, G. A., Skillman, E. D., Terlevich, E., & Terlevich, R. J.1995, ApJ, 449, L77Gaskell, M. 2009, New Astronomy Reviews, 53, 140Gaskell, C. M., Klimek, E. S., & Nazarova, L. S. 2007,arXiv:0711.1025Gaskell, C. M., Shields, G. A., & Wampler 1981, ApJ, 249, 443Green, P. J. 1998, ApJ, 498, 170Hamann, F., & Ferland, G. J. 1993, ApJ, 418, 11Hamann, F., & Ferland, G. 1999, ARA&A, 37, 487Hamann, F., Korista, K. T., Ferland, G. J., Warner, C., &Baldwin, J. 2002, ApJ, 564, 592Hamann, F., Warner, C., Dietrich, M., & Ferland, G. 2007, inThe Central Engine of Active Galactic Nuclei, ASP ConferenceSeries, ed. Luis C. Ho and Jian-Min Wang, Vol. 373, p. 653Hu, C., Wang, J.-M., Ho, L. C., Chen, Y.-M., Zhang, H.-T., Bian,W.-H., & Xue, S.-J. 2008, ApJ, 687, 78Kova˘cevi´c, J., Popovi´c, L. C., & Dimitrijevi´c, M. S. 2010, ApJS,189, 15Kuehn, C. A., Baldwin, J. A., Peterson, B. M., & Korista, K. T.2008, ApJ, 673, 69Kwan, J., & Krolik, J. H. 1981, ApJ, 250, 478
Shields, Ludwig, & Salviander
Laor, A., in The Central Engine of Active Galactic Nuclei, ASPConference Series, ed. Luis C. Ho and Jian-Min Wang, Vol.373, p. 383Laor, A., & Draine, B. T. 1993, ApJ, 402, 441Laor, A., Fiore, F., Elvis, M., Wilkes, B. J., & McDowell, J.C.1997, ApJ, 477, 93Lawrence, A., Elvis, M., Wilkes, B J., McHardy, I. & Brandt, N.1997, MNRAS, 285, 879Leighly, K. M., Halpern, J. P., Jenkins, E. B., & Casebeer, D.2007, ApJS, 173, 1Ludwig, R. R., Wills, B., Greene, J. E., & Robinson, E. L. 2009,ApJ, 706, 995Maoz, D., et al. 1993, ApJ, 404, 576Marziani, P., Zamanov, R. K., Sulentic, J. W., & Calvani, M.2003, MNRAS, 435, 1133Matsuoka, Y., Kawara, K., & Oyabu, S. 2008, ApJ, 673, 62Murray, N., & Chiang, J., 1998, ApJ, 494, 125Nagao, T., Murayama, T., Shioya, Y., & Taniguchi, Y., AJ, 125,1729Netzer, H., 2008, ApJ, 695, 793Netzer, H., & Laor, A. 1993, ApJ, 404, L51Netzer, H., & Trakhtenbrot, B. 2007, ApJ, 654, 754Nussbaumer, H., & Osterbrock, D. E. 1970, ApJ, 161, 811Osterbrock, D. E., & Ferland 2006, ‘Astrophysics of GaseousNebulae and Active Galactic Nuclei,’ 2nd ed., Sausalito,California: University Science Books Peterson, B. 1997, An Introduction to Active Galactic Nuclei(Cambridge: Cambridge Univ. Press)Salviander, S., Shields, G. A., Gebhardt, K., & Bonning, E. W.2007, ApJ, 662, 131Scott, J. E., Kriss, G. A., Brotherton, M., Green, R. F.,Hutchings, J., Shull, J. M., & Wei, Z. 2004, ApJ, 615, 135Shang, Z., Wills, B. J., Wills, D., & Brotherton, M. S., 2007, AJ,134, 294Sigut, T. A. A., & Pradhan, A. K. 2003, ApJS, 145, 15Suganuma, M., et al. 2006, ApJ, 639, 46Sulentic, J. W., Marizani, P., & Dultzin-Hacyan, D. 2000,ARA&A, 38, 521van Zee, L., Salzer, J. J., & Haynes, M. P. 1998, ApJ, 497, L1Vanden Berk, D., et al. 2001 AJ, 122, 549Verner, E., Bruhweiler, F., Verner, D., Johansson, S., & Gull, T.2003, ApJ, 592, L59Wills, B. J., Netzer, H., & Wills, D. 1985, ApJ, 288, 94Wills, B. J., Laor, A., Brotherton, M. S., Wills, D., Wilkes, B. J.,Ferland, G. J., & Shang, Z. 1999, ApJ, 515, L53 e ii Emission in AGN 9
Fig. 1.—
Composite SDSS spectra for the five bins in optical Fe ii strength. Vertical axis gives specific flux F λ . Note the broad Fe ii blends at λ λ Fig. 2.— [Fe vii ] and [Ne v ] narrow emission lines in the composite spectra. See text for discussion. e ii Emission in AGN 11