Origin and Dynamical Support of Ionized Gas in Galaxy Bulges
aa r X i v : . [ a s t r o - ph . GA ] J un T O APPEAR IN
The Astrophysical Journal
Preprint typeset using L A TEX style emulateapj v. 26/01/00
ORIGIN AND DYNAMICAL SUPPORT OF IONIZED GAS IN GALAXY BULGES L UIS
C. H O The Observatories of the Carnegie Institution of Washington, 813 Santa Barbara Street, Pasadena, CA 91101
To appear in The Astrophysical Journal
ABSTRACTWe combine ionized gas ([N II ] λ σ ≈ - - , such that σ g /σ ∗ ≈ . - .
4, with an average value of 0.80. These results are independent of Hubble type(for galaxies from E to Sbc), presence or absence of a bar, or local galaxy environment. For galaxies of type Scand later and that have σ ∗ ∼ <
40 km s - , the gas seems to have a minimum threshold of σ g ≈
30 km s - , such that σ g / σ ∗ always exceeds 1. Within the sample of AGNs, σ g / σ ∗ increases with nuclear luminosity or Eddington ratio,a possible manifestation of AGN feedback associated with accretion disk winds or outflows. This extra source ofnongravitational line broadening should be removed when trying to use σ g to estimate σ ∗ . We show that the massbudget of the narrow-line region can be accounted for by mass loss from evolved stars. The kinematics of thegas, dominated by random motions, largely reflect the velocity field of the hot gas in the bulge. Lastly, we offer asimple explanation for the correlation between line width and line luminosity observed in the narrow-line regionof AGNs. Subject headings: galaxies: active — galaxies: bulges — galaxies: ISM — galaxies: kinematics and dynamics— galaxies: nuclei — galaxies: Seyfert INTRODUCTION
Warm ( ∼ K), ionized gas is pervasive in the central re-gions of nearby galaxies of all morphological types. Easily de-tected through optical emission lines, either by imaging or spec-troscopy, this component of the interstellar medium has servedas a diagnostic of the physical conditions in galactic nuclei,including their source of excitation and chemical abundances.The optical line emission also provides a powerful tracer of thenuclear kinematics of the galaxy, in efforts to probe the massdistribution of the bulge, to detect central dark massive ob-jects, and to study substructure such as kinematically decoupledcores.What is the dynamical state of the nuclear gas? How does itrelate to the dynamics of the stars? Is the gas primarily rotation-ally supported, or do random motions also contribute? Whilethese issues have been addressed in the past, they have beenbased either on investigations of individual objects or small, re-stricted samples. A number of authors have noted, for example,that in the bulges of disk (lenticular and spiral) galaxies, thegas often rotates slower than the stars (Fillmore et al. 1986;Kent 1988; Kormendy & Westpfahl 1989; Bertola et al. 1995;Cinzano et al. 1999; Pignatelli et al. 2001). The same phe-nomenon has been found in some elliptical galaxies (e.g., Cald-well 1984; Caldwell et al. 1986). The physical origin for thiseffect, however, is not entirely clear. An interesting possibil-ity is that the gas experiences nonnegligible pressure support(e.g., Kent 1988; Cinzano & van der Marel 1994; Bertola etal. 1995; Fisher 1997), an idea consistent with the detectionof large central line widths in many bulge-dominated galaxies(e.g., Demoulin-Ulrich et al. 1984; Phillips et al. 1986; Bertolaet al. 1995). Vega Beltrán et al. (2001) compiled gas and stel-lar velocity measurements for a sample of ∼
40 disk galaxies toexamine trends with Hubble type. They conclude that, with afew exceptions, the gas is dynamically cold, confined to a disksupported mainly by rotation. What is the role of nongravitational forces on the gas dy-namics, especially in objects harboring active galactic nuclei(AGNs)? This issue has been investigated most thoroughly inthe context of the narrow-line region (NLR) in radio (Heckmanet al. 1985; Smith et al. 1990; Baum & McCarthy 2000; Noel-Storr et al. 2007) and luminous Seyfert (e.g., Whittle 1992b,1992c; Nelson & Whittle 1996) galaxies. The overall consen-sus from these studies is that the velocity field of the NLR gaslargely traces the gravitational potential of the bulge of the hostgalaxy. Super-virial motions can be found, but only in the mi-nority of objects exhibiting powerful, jetlike radio structures,which presumably impart additional mechanical acceleration tothe gas. Comparing the widths of the optical nebular lines andthe stellar velocity dispersions of a large sample of relativelyluminous Seyfert 2 galaxies, Greene & Ho (2005a) reinforcedthe notion that the stellar potential principally governs the kine-matics of the NLR, with important exceptions arising in caseswhere nuclear activity is especially strong. To date, much lessattention has been paid to more garden-variety AGNs such aslow-luminosity Seyferts and LINERs, or to nuclei powered bymassive stars.The Palomar spectroscopic survey of nearby galaxies (Ho etal. 1995, 1997a, and references therein) offers an excellentopportunity to examine some of the issues mentioned above.In addition to the existing extensive database of emission-linewidths, a catalog of central stellar velocity dispersions is nowavailable (Ho et al. 2009). This allows a detailed comparison ofthe central velocity dispersions of the gas and stars for a large,well-defined sample of galaxies spanning a wide range of Hub-ble types and nuclear spectral classes. This is the goal of thispaper.1 HO F IG . 1.— Distribution of gas ( σ g ) versus stellar ( σ ∗ ) velocity dispersion as a function of Hubble type. Unbarred and barred galaxies are plotted as filled andencircled points, respectively. The dashed line denotes σ g = σ ∗ . A typical error bar is given in the lower-right corner of the first panel. SAMPLE AND DATA
Our analysis draws from the nearly complete, magnitude-limited sample of 486 bright galaxies spectroscopically sur-veyed using the Palomar 5-meter telescope. We focus on thegalaxies that have well-measured widths (FWHM) for [N II ] λ II ]) measurements, 372(91%) have stellar velocity dispersions. If we confine our at-tention to the sources that have FWHM([N II ]) measurementswith a quality rating of “b” or better (see Ho et al. 1997a), ex-cluding a handful objects with only upper limits on their linewidths, 345 have stellar velocity dispersions.The above subset of 345 objects, the focus of this paper, rep- resents fairly the parent population of emission-line nuclei inthe Palomar survey. The 25 E, 70 S0–S0/a, 149 Sa–Sbc, 62 Sc,and 30 Scd–Sm galaxies constitute, respectively, 81%, 93%,96%, 84%, and 48% of the parent population of emission-lineobjects. In terms of the nuclear spectral classes, the coverageis excellent for the AGNs: the present subsample contains 94%of the Seyferts, 93% of the LINERs, and 92% of the transitionnuclei in the parent survey. The high completeness fractionssimply reflect the high AGN detection rate in bulge-dominatedgalaxies (Ho et al. 1997b), for which velocity dispersions areeasier to obtain. By comparison, H II nuclei are 72% completein the current sample—still a very high fraction. THE CORRELATION BETWEEN σ g AND σ ∗ Dependence on Hubble Type
Figure 1 shows the overall correlation between gas ( σ g ) andstellar ( σ ∗ ) velocity dispersions, first for the entire sample, andthen individually for several Hubble type bins. Most of thenarrow emission lines, especially near the core, do not devi-ate strongly from a simple Gaussian (see Ho et al. 1997d), andwe define σ g ≡ FWHM([N II ])/2.35. Barred galaxies, taken tobe those designated as “SB” or “SAB” in the Third ReferenceCatalogue of Bright Galaxies (de Vaucouleurs et al. 1991), are See Ho et al. (1997a) for definition of the spectral classifications.
ONIZED GAS IN BULGES 3 F IG . 2.— Distribution of gas ( σ g ) versus stellar ( σ ∗ ) velocity dispersion as a function of nuclear spectral class. The symbols are coded according to H α luminosity.Objects containing linear radio sources with 6 cm powers greater than 10 . W Hz - are marked with boxes. The dashed line denotes σ g = σ ∗ . A typical error baris given in the lower-right corner of the first panel. marked for emphasis. It is clear that σ g correlates strongly with σ ∗ , albeit with considerable scatter. For the sample as a whole,the Kendall’s τ correlation coefficient (Isobe et al. 1986) is r = 1 .
19, which rejects the null hypothesis of no correlationwith a probability of P null < - . Both σ g and σ ∗ cover ap-proximately the same range of velocities, from ∼
30 to 350km s - . Although the correlation is most evident for the en-tire sample, the same trend replicates for the various Hubbletype bins. It is least well-defined for the elliptical and extremelate-type (Scd–Sm and I0) galaxies because of the small sam- ples and limited dynamic range in velocity dispersions. Thus,to first order, the gas velocities seem to track the random ve-locities of the stars. Quantitatively, however, the gas disper-sions are systematically slightly lower than the stellar disper-sions, on average by a constant offset of ∼ σ g = σ ∗ line by ∆ σ ≡ log σ g - log σ ∗ , h ∆ σ i = - . ± . h ∆ σ i = - . ± .
014 and - . ± . σ ∗ ∼ <
40 km s - , nearly allthe objects have ∆ σ >
0. This subgroup of high- ∆ σ objects The lower limit of the velocities, however, is strongly influenced by the spectral resolution of the Palomar survey, which is σ ≈
40 km s - in the red and ∼ - in the blue (Ho et al. 1997a). Published stellar velocity dispersions often are also not very reliable for low-dispersion galaxies. Thus, we consider the regiondefined by σ g ∼ <
40 km s - and σ ∗ ∼ <
40 km s - to be quite uncertain. HO F IG . 3.— Residuals of the σ g - σ ∗ correlation, ∆ σ ≡ log σ g - log σ ∗ , asa function of H α luminosity L H α , for ( a ) all AGNs (Seyferts, LINERs, andtransition objects), including the sample of luminous Seyferts from Nelson &Whittle (1995), and ( b ) H II nuclei. Objects with linear radio sources and P > . W Hz - are marked with boxes. The dashed line denotes σ g = σ ∗ . The solid line in panel ( a ) shows the best-fit regression line. A typicalerror bar is given in the lower-right corner of the panel ( b ). cannot be attributed entirely to resolution effects because onlya small number of [N II ] profiles are unresolved and includingthem does not alter the distribution of data points. Rather, itappears that galaxy centers reach a “floor” at σ g ≈
30 km s - .In detail, the different Hubble types also have similar dis-tributions: h ∆ σ i = - . ± . - . ± . - . ± . - . ± .
026 for E, S0–S0/a, Sa–Sbc, and Sc,respectively. Ellipticals appear to have somewhat larger off-sets than the disk galaxies, but the means of the different sam-ples are not statistically different according to the Student’s t test. By contrast, very late-type objects, in our sample dom-inated by Scd–Sm galaxies, clearly stand apart with h ∆ σ i = - . ± . Dependence on Local Environment
Ho et al. (1997a) give two parameters to gauge the tidal in-fluence of nearby neighbors: (1) ρ gal , the density of galaxiesbrighter than M B = -
16 mag in the object’s local vicinity, and(2) θ p , the projected angular separation to the nearest neighborwithin a magnitude difference of ± ±
500 km s - . We find that ∆ σ shows no correlationwith either ρ gal or θ p .3.3. Dependence on AGN Properties
We next evaluate the dependence of the σ g - σ ∗ correlationon nuclear spectral classification (Fig. 2). The four classes of emission-line objects appear broadly similar, but in detailthere are some subtle, important differences. Taken at facevalue, the Seyfert sample shows the least degree of offsetfrom the σ g = σ ∗ line. Its distribution of ∆ σ , with an av-erage value of - . ± . h ∆ σ i = - . ± . t test. LINERs, with h ∆ σ i = - . ± . h ∆ σ i = - . ± . II nuclei have an overall h ∆ σ i = - . ± . II class, seem to comprise two distinctpopulations divided roughly at σ ∗ ≈
40 km s - . Below thisvalue, our sample of H II nuclei has h ∆ σ i = + . ± . h ∆ σ i = - . ± . and linearly extended tend to have emission lines that aresystematically broader than the stellar velocities. They interpretthis to mean that the outflowing radio jets provide a secondary,nongravitational source of acceleration to the NLR gas. Whittle(1992b, 1992c) chose P = 10 . W Hz - as the threshold be-tween strong and weak radio sources. We have adopted this cri-terion , and with the radio data for the Palomar low-luminosityAGNs presented in Ho & Ulvestad (2001) and elsewhere (Ho2008, and references therein), highlighted in Figure 2 the ob-jects with luminous, jetlike radio sources. Two of the objects inthis subset—NGC 1068 and NGC 3079—are clearly systemati-cally offset above the rest. But another six that satisfy the sameradio criteria do not look extraordinary whatsoever, and not allhigh- ∆ σ objects have strong radio jets. The same trends holdfor luminous radio sources among the LINERs and transitionobjects. As a whole, strong radio jets are neither necessary norsufficient to produce super-virial velocities in the NLR.Closer inspection of Figure 2 reveals a more interesting effectwithin the AGN samples. When the objects are flagged accord-ing to H α luminosity, those with high luminosities—here, forillustrative purposes, chosen somewhat arbitrarily to be thoseabove L H α = 10 erg s - —tend to lie systematically displacedtoward higher ∆ σ compared to objects with lower luminosities.(We use the narrow component of the H α luminosity, correctedfor Galactic and internal extinction.) This is most clearly seenfor the Seyfert sample. It is less obvious for the LINERs andtransition objects because of the paucity of luminous sources inthese classes (Ho et al. 2003), but the same trend is noticeablenonetheless. Within the Seyfert sample, h ∆ σ i = + . ± . L H α ≥ erg s - and h ∆ σ i = - . ± . L H α < erg s - . The difference betweenthe two means is significant at the level of 96%.To explore the luminosity effect further, we augmented thePalomar sample with 52 Seyfert nuclei from the work of Nelson& Whittle (1995) that were not already in the Palomar survey.These additional objects provide better coverage at the high-luminosity end. H α luminosities were converted from H β lumi-nosities given in Nelson & Whittle (1995) and Whittle (1992a),assuming H α /H β = 3.1 (Gaskell & Ferland 1984) and adjustedto a Hubble constant of H = 75 km s - Mpc - . Figure 3 a plots ∆ σ against L H α for the combined sample of AGNs, again flag-ging the luminous, extended radio sources. Note the nonzero Whittle (1992b, 1992c) assumes a Hubble constant of H = 50 km s - Mpc - and a reference wavelength of 20 cm, whereas we use H = 75 km s - Mpc - and areference wavelength of 6 cm. The corresponding threshold, for a spectrum f ν ∝ ν - . , is P = 10 . W Hz - . ONIZED GAS IN BULGES 5 F IG . 4.— Residuals of the σ g - σ ∗ correlation for the AGNs (Seyferts, LIN-ERs, and transition objects in the Palomar survey, plus the luminous Seyfertsfrom Nelson & Whittle 1995), ∆ σ , as a function of P , the 6 cm power. Thesymbols for the different sources are given in the figure legend. Objects withlinear radio sources and P > . W Hz - are marked with boxes. Thedashed line denotes σ g = σ ∗ . The solid line gives the best-fit regression line.A typical error bar is shown in the lower-right corner. slope. A correlation can be seen for the Seyferts alone, as wellas for the combined sample of LINERs and transition nuclei.Treating log L H α as the independent variable, an ordinary least-squares regression fit to all the points gives ∆ σ = (0 . ± . L H α - (1 . ± . , (1)with an rms scatter of 0.17 dex. The scatter reduces to 0.16dex with the luminous radio sources excluded. The correla-tion has high statistical significance: the generalized Kendall’s τ test yields r = 0 .
47 and P null < - . Omitting the flaggedradio sources has little effect on the results. By contrast, thesample of H II nuclei (Fig. 3 b ) does not show this effect, eventhough the H II nuclei and AGNs span essentially the samerange in H α luminosities. The different behavior of H II nu-clei compared to AGNs cannot be attributed to differences inmorphological types. Roughly half of the H II nuclei are hostedin bulge-dominated (S0–Sbc; T = - b ).In Figure 4, we substitute the H α luminosity with the 6 cmradio power. Radio data for the Palomar Seyferts come fromHo & Ulvestad (2001), and those for the supplementary sampleof luminous objects come from Nelson & Whittle (1995) andWhittle (1992a). Data for the LINERs and transition objects aretaken from a variety of sources, as summarized in Ho (2008).Again, we see a positive, significant trend, although formally itis weaker than that for L H α . In this case the Kendall’s τ cor-relation coefficient is r = 0 .
29, with P null = 6 × - . A linearregression fit, calculated using the method of Schmitt (1985),which properly treats the few censored radio data points, gives ∆ σ = (0 . ± . P - (0 . ± . . (2)The tendency for ∆ σ to increase with optical and radio powersuggests that the AGN luminosity, over a wide range of values,injects an additional, systematic contribution to the line broad-ening.Lastly, Figure 5 illustrates that an even stronger correlationexists between ∆ σ and the Eddington ratio, which is propor-tional to the mass accretion rate. As in Greene & Ho (2007),we estimate the bolometric luminosity from the H α luminosity,converting L H α to L using the line-continuum correlation ofGreene & Ho (2005b) and then adopting L bol = 9 L . F IG . 5.— Residuals of the σ g - σ ∗ correlation for the AGNs (Seyferts, LIN-ERs, and transition objects in the Palomar survey, plus the luminous Seyfertsfrom Nelson & Whittle 1995), ∆ σ , as a function of Eddington ratio, L bol / L Edd .The symbols for the different sources are given in the figure legend. Objectswith linear radio sources and P > . W Hz - are marked with boxes.The dashed line denotes σ g = σ ∗ . The solid line gives the best-fit regressionline; the dotted line shows the relation found by Greene & Ho (2005a; Equa-tion 4). A typical error bar is shown in the upper-right corner. We use the M BH - σ ∗ relation of Tremaine et al. (2002) to ob-tain the black hole mass and hence the Eddington luminosity, L Edd = 1 . × (cid:0) M BH / M ⊙ (cid:1) erg s - . The Kendall’s τ corre-lation coefficient now rises to r = 0 .
74, with P null < - . Anordinary least-squares regression fit with log L bol / L Edd as theindependent variable gives ∆ σ = (0 . ± . L bol / L Edd + (0 . ± . . (3)The correlation is surprisingly tight, with an rms scatter of only0.15 dex. If we omit the sources with luminous radio jets, thescatter reduces even further to 0.14 dex. ∆ σ seems to corre-late somewhat better with L bol / L Edd than with L H α , althoughformally, according to the t and F tests, neither the mean northe variance of the residuals about the best-fit lines differs at astatistically significant level.3.4. Geometry and Kinematics of the Ionized Gas
The gas line widths presented in this study were extractedfrom a single 2 ′′ × ′′ aperture centered on the nucleus, whichcorresponds to a region 200 pc ×
400 pc for a typical distance of20 Mpc (Ho et al. 1997a). Without spatially resolved data, it isimpossible to know a priori in any individual object whether theline width represents true velocity dispersion or instead arisesfrom spatial smearing of a steep inner rotational gradient. Nev-ertheless, we can appeal to general arguments to test the sce-nario of rotational broadening. If the nuclear gas is confined toa rotating disk aligned with the large-scale galactic plane, onemight naively expect the line width to correlate with the galaxyinclination angle. This effect is not seen in the Palomar survey(Ho 1996; Ho et al. 2003). However, as discussed by Whittle(1992b), this simple test is likely to be inconclusive because theline widths are primarily governed by virial velocities and AGNproperties rather than by projection effects.Instead, Whittle (1992b), as did Heckman et al. (1989), ex-amined the correlation between inclination angle and the ratio ∆ V rot /FWHM, where ∆ V rot is the observed rotation amplitudeof the large-scale disk and FWHM is the full-width at half max-imum of the central line profile. If the central nebular emission HO F IG . 6.— Distribution of the ratio ∆ V rot /FWHM([N II ]) versus sin i as afunction of Hubble type. The dashed line denotes an idealized model whereinFWHM([N II ]) arises entirely from rotation in the galactic plane. Actual fitsto the data, after excluding the luminous, linear radio sources (marked withsquares), are given as solid lines; the regressions are calculated using an ordi-nary least-squares fit with log sin i as the independent variable. arises from a coplanar nuclear disk, FWHM, like ∆ V rot , is a pro-jected quantity, and ∆ V rot /FWHM will be independent of incli-nation. In the extreme alternative scenario where FWHM repre-sents purely random velocity dispersion, ∆ V rot /FWHM shouldcorrelate positively with inclination. The results of this exerciseare shown in Figure 6 for all the non-elliptical galaxies, usingdata for ∆ V rot (derived from integrated H I line widths) and in-clination angle i from Ho et al. (1997a). In each panel, thecoplanar rotation model is marked with the horizontal dashedline. The actual regression fit to the data (after excluding theflagged radio sources) is shown as a solid line. With the excep-tion of the small number of very late-type (Scd–Sm) spirals, thevast majority of the sample strongly disagree with the coplanarrotation model.As a final check, we also plotted the residuals ∆ σ from Fig-ure 1 against sin i . We find no correlation, as to be expected if σ g and σ ∗ are both inclination-independent. DISCUSSION
Estimating σ ∗ from σ g We combine new central stellar velocity dispersions (Ho etal. 2009) with published emission-line width measurements(Ho et al. 1997a) to examine the relationship between the kine-matics of the ionized gas and stars in the central regions ofgalaxies. Unlike most studies that primarily focus on the [O
III ] λ II ] λ II ] has a lower ionization potential and alower critical density than [O III ], making it a better tracer overa larger radial extent of the bulge. [N II ] also provides a more re-liable probe of the gravitational potential because it is less sus-ceptible to contamination by outflows or other radial motions. Like most narrow emission lines in AGNs, [N II ] does exhibitline asymmetries indicative of radial flows (Ho 1996; Ho et al.2003), but they seem to be less prevalent than in high-ionizationtransitions such as [O III ] (Heckman et al. 1981; Vrtilek & Car-leton 1985; Whittle 1985). Studies of AGN line profiles that in-clude transitions from a wide range of ionization states (Busko& Steiner 1992; Greene & Ho 2005a) find that [S II ] λλ III ] (but seeRice et al. 2006). Since the profile of [S II ] empirically matcheswell the profile of [N II ] (e.g., Ho et al. 1997d), it follows that[N II ] is also less asymmetric than [O III ]. Moreover, the aver-age line centroid of [N II ] is consistent with the systemic ve-locity of the host galaxy, whereas high-ionization lines such as[O III ] are often blueshifted (Boroson 2005). Previous work thatmakes use of [N II ] to probe the kinematics of the NLR includethose of Phillips et al. (1986), Verdoes Kleijn et al. (2006),Zhou et al. (2006), Chen et al. (2008), and Walsh et al. (2008).Our principal finding is that, within the central few hundredparsecs of galaxy bulges, the velocity dispersion of the ionizedgas is comparable to, but typically somewhat lower than, thevelocity dispersion of the stars. The rough correspondence be-tween σ g and σ ∗ indicates that the NLR is in dynamical equi-librium with the stellar potential. Defining ∆ σ = log σ g - log σ ∗ , ∆ σ ranges from - .
15 to + .
08, with an average value of - . h σ g /σ ∗ i = 0 .
80. In detail, ∆ σ shows, at best, amild variation with galaxy morphology but no discernible de-pendence on the presence of a large-scale bar or local galaxyenvironment. By contrast, the more luminous AGNs studiedby Whittle (1992b) seem to exhibit broader lines in barred andinteracting galaxies.As an aside, we note that the late-type spirals in our sam-ple show an intriguing property: below σ ∗ ≈
40 km s - , ∆ σ is always positive. It seems that in these late-type galaxies withsmall bulges (possibly all “pseudo-bulges”; Kormendy & Ken-nicutt 2004), the velocity dispersion of the warm ( ∼ K)ionized gas has a floor of ∼
30 km s - . This might represent aminimum threshold of interstellar turbulence in the central re-gions of present-day disk-dominated galaxies, maintained, per-haps, through stellar feedback associated with their modest starformation rates (Ho et al. 1997c). The dispersion thresholdwe find is qualitatively similar to the magnitude of non-orderedmotions detected in the central regions of low-surface bright-ness galaxies (e.g., Kuzio de Naray et al. 2008; Pizzella et al.2008a, 2008b).The availability of a uniform set of nuclear spectroscopicclassifications for the Palomar galaxies has prompted us to re-examine the kinematics of the NLR, their relationship with thestellar potential of the bulge of the host galaxy, and the possi-ble role of nongravitational perturbations by AGNs. Previoustreatment of this subject has concentrated almost exclusivelyon radio galaxies (Smith et al. 1990; Verdoes Kleijn et al.2006; Noel-Storr et al. 2007) and relatively luminous Seyfertnuclei (Véron 1981; Terlevich et al. 1990; Veilleux 1991; Whit-tle 1992b, 1992c; Nelson & Whittle 1996; Jiménez-Benito etal. 2000; Botte et al. 2005; Greene & Ho 2005a; Bian etal. 2006; Zhou et al. 2006; Chen et al. 2008; Dasyra et al.2008). The reviews by Wilson & Heckman (1985) and Whittle(1993) included some preliminary results on LINERs, althoughthey were based on rather fragmentary data. They concludedthat the NLR kinematics of LINERs, as in Seyferts, are largelyvirial. In their survey of ionized gas in E and S0 galaxies, mostof which contain LINER nuclei, Phillips et al. (1986) comparedONIZED GAS IN BULGES 7the FWHM of [N II ] λ σ g ∼ < σ ∗ .Our analysis strongly reinforces the notion that gravity dom-inates the kinematics of the NLR, not only in Seyferts, but in-deed in all emission-line nuclei found in nearby galaxies, in-cluding low-ionization, low-luminosity AGNs such as LINERsand transition objects, and even H II nuclei. The gas veloc-ity dispersions are comparable to, but typically somewhat lessthan, the stellar velocity dispersions measured over a phys-ical scale of typically 200 -
400 pc. The velocity discrep-ancy observed in H II nuclei and AGNs is about the same, h σ g / σ ∗ i ≈ .
8, with a surprisingly small standard deviation ofonly ∼ . σ g and σ ∗ varies mildly but systematically with the level of AGN ac-tivity. The effect can be seen with nuclear activity as gaugedthrough optical (H α ) luminosity (Fig. 3 a ), radio luminosity(Fig. 4), and especially Eddington ratio (Fig. 5). The weak-est AGNs in the Palomar sample (LINERs and transition nucleiwith L H α ≈ × erg s - , P ∼ < W Hz - , and L bol / L Edd ≈ × - ) are characterized by σ g / σ ∗ ≈ . - .
8, while lumi-nous sources (Seyferts with L H α ≈ × erg s - , P ≈ W Hz - , and L bol / L Edd ≈ × - ) can attain σ g / σ ∗ ≈ . - . II nuclei, even those spanning a similar rangein H α luminosity and Hubble type as the AGNs, do not fol-low these trends (Fig. 3 b ). Evidently the physical nature of theactivity—stellar or nonstellar—matters.It is worth noting that Whittle (1992b, 1992c) and Nelson &Whittle (1996) find h σ g / σ ∗ i ≈ . h σ g / σ ∗ i ≈ . σ g / σ ∗ varies slightly but systematically with nuclear spec-tral classification: transition objects have the lowest values andSeyferts the highest, with LINERs in between. This trend re-flects none other than the dependence of σ g / σ ∗ on luminosityor Eddington ratio. Ho (2008, 2009) proposes that the massaccretion rate, which scales with the Eddington ratio, is the pri-mary physical driver responsible for variations in spectral clas-sification. The accretion rate increases systematically along thesequence transition objects → LINERs → Seyferts.If the size of the NLR scales with luminosity, Laor (2003)suggests that the NLR in low-luminosity AGNs may be suffi-ciently compact that its kinematics may be influenced more bythe gravitational potential of the central black hole than by thatof the surrounding bulge. Spatially resolved observations of theNLR in some LINERs at
Hubble Space Telescope resolution in-deed do find that the line widths are largest interior to the blackhole’s gravitational sphere of influence, gradually merging atlarger radii with the stellar dispersion of the bulge (Walsh et al. 2008). If this effect were dominant, however, σ g / σ ∗ would risewith decreasing luminosity, exactly the opposite of what is seenin our sample. Evidently on a scale of several hundred parsecsthe integrated kinematics of the NLR are controlled primarilyby the gravitational potential of the stars and by AGN feedback.The discovery of a tight relation between central black holemass and bulge stellar velocity dispersion (the M BH - σ ∗ relation:Gebhardt et al. 2000; Ferrarese & Merritt 2000) has promptedrenewed interest in efficient methods to estimate σ ∗ , especiallyin objects such as luminous AGNs where this parameter is chal-lenging to measure directly (Greene & Ho 2006). In view ofthe approximate equality between σ g and σ ∗ and the ease withwhich [O III ] λ σ ∗ = σ g ([O III ]). Numerous stud-ies have since adopted this strategy to investigate the M BH - σ ∗ relation in AGNs (e.g., Wang & Lu 2001; Grupe & Mathur2004) and its possible evolution with redshift (e.g., Shields etal. 2003; Salviander et al. 2007). However, the large observedscatter of the σ g - σ ∗ correlation (Boroson 2003), not to mentionthe kinematic anomalies long known to afflict the [O III ] line(Boroson 2005), gave cause for concern. To mitigate these un-certainties, Greene & Ho (2005a) recommended that only thecore of the [O
III ] line be used, and, if possible, that [O
III ]be abandoned altogether in favor of lower-ionization transitionssuch as [O II ] λ II ] λλ II ], another low-ionization line, also traces the stellar velocity dispersion of thebulge. The scatter is modest, ∼ . M BH ∝ σ ∗ (Tremaine et al. 2002), a 0.2 dex uncertaintyon σ ∗ translates into a 0.8 dex (or factor of ∼
6) uncertaintyon M BH . Importantly, we find that the velocity dispersion ofthe ionized gas does not identically match the velocity disper-sion of the stars, but rather, on average, σ g / σ ∗ ≈ .
8. Closerexamination reveals that the residuals of the σ g - σ ∗ relation infact depend on AGN properties. The principal “second pa-rameter” appears to be either the Eddington ratio ( L bol / L Edd ) orsome measure of the AGN luminosity (we use L H α ). The resid-uals seem to correlate somewhat better with L bol / L Edd than with L H α , although from our data alone we cannot truly tell whichvariable is more fundamental. Since the optical NLR luminos-ity correlates strongly with radio continuum luminosity (e.g.,Ho & Peng 2001; Ulvestad & Ho 2001; Nagar et al. 2005),it is hardly surprising that ∆ σ correlates with radio power too,although the scatter is larger and the statistical significance isinferior compared to either L H α or L bol / L Edd . The dependenceof the residuals of the σ g - σ ∗ relation on L bol / L Edd was first no-ticed by Greene & Ho (2005a; later confirmed by Bian et al.2006) for a large sample of more luminous AGNs selected fromSDSS, and, within the uncertainties, the Palomar sources seemto follow roughly the low-luminosity extrapolation of the SDSSsample (Fig. 5). Curiously, Greene & Ho did not see any cor-relation between ∆ σ and AGN optical ([O III ]) luminosity orradio power. It is possible that the large linear scales probedby SDSS (the 3 ′′ fibers correspond to 5.4 kpc at z = 0 .
1) intro-duce substantial uncertainties into measurements of the bulgeand NLR.The dependence of ∆ σ on AGN properties can be exploitedto sharpen the σ g - σ ∗ correlation as a tool to predict σ ∗ whenthe latter is, as is often the case, difficult or impractical to mea-sure directly in AGNs. Even when a low-ionization line or onlythe core of the [O III ] line is being used, as recommended by HOGreene & Ho (2005a), it is important to realize that σ g / σ ∗ is not identically equal to unity. Rather, σ g / σ ∗ varies from ∼ . ∼ . σ ∗ . Moreover, significant scatter still remains ( ∼ .
15 dex),even after correcting for the second parameter, most of whichmay be intrinsic and irreducible owing to the complexities ofthe NLR (Rice et al. 2006).4.2.
Evidence for AGN Feedback
From a physical point of view, what is the origin of the ∆ σ - L H α or ∆ σ - L bol / L Edd correlations? As discussed further below,if the gas derives principally from mass loss from bulge stars,its kinematics should generally track the kinematics of the stars.But because the gas is collisional and experiences hydrodynam-ical drag against the surrounding hot medium, we expect it tobe kinematically slightly colder than the stars. In the absence ofadditional energy input from other sources, we anticipate σ g / σ ∗ ∼ <
1, as observed. As additional energy is injected into the sys-tem, for example from activation of the central black hole, thegas gains energy, to the point that σ g approaches or even over-takes σ ∗ . Precisely how this is accomplished is unclear. DoAGNs impart mostly mechanical energy from outflows/windsor more highly collimated fast jets? Does radiation pressurematter? The empirical trends presented in this paper offer someinsights into these issues, which ultimately have bearing on cur-rent interests in AGN feedback in galaxies.The fact that the emission-line widths become systematicallybroader with increasing AGN activity—whatever the principalsource of energy—implies that the gas is primarily acceleratedrather than directly heated. The H α luminosity, and presumablyalso the bolometric luminosity, strongly correlates with ∆ σ , butit seems doubtful that the radiation density itself plays a centralrole because H II nuclei with the same range in luminosity andsampled over a similar volume do not show the same effect.Compared to nuclear star formation, AGNs have a characteris-tically harder ionizing spectrum and a more centrally concen-trated radiation field. Whether these factors result in more effi-cient radiative acceleration of the ionized gas in AGNs needs tobe further investigated.The radio-emitting plasma is an obvious suspect—that is, ascaled-down, but qualitatively similar agent as that responsiblefor super-virial acceleration in the most extreme, jet-dominatedsources. This is a tenable hypothesis because nearly all Seyferts(e.g., Kukula et al. 1995; Ho & Ulvestad 2001; Ulvestad & Ho2001) and LINERs (Nagar et al. 2005) and a sizable fractionof transition objects (Filho et al. 2000, 2002) contain nuclearradio sources, many with morphological structures resemblingjets. The radio luminosity function of low-luminosity AGNsvaries continuously over at least 4 orders of magnitude in ra-dio power, from P ≈ to 10 W Hz - (Ulvestad &Ho 2001; Filho et al. 2006). There is no characteristic powerthat might demarcate a luminosity threshold above which ra-dio sources become effective in perturbing the NLR. Althoughthe degree of radio-loudness (expressed as the relative fractionof the radiative output emerging in the radio band) actually de-creases with increasing Eddington ratio (Ho 2002; Terashima& Wilson 2003; Greene et al. 2006), Ho (2008) notes that ex-tended, jetlike structures are more prevalent in high- L bol / L Edd sources (Seyferts) than in low- L bol / L Edd sources (LINERs). Theradio cores in LINERs, despite being energetically more dom-inant, may be less effective at stirring up the NLR because of their compact structure. The extended, linear features inSeyferts, on the other hand, can impact a larger pool of thecircumnuclear material. Jet-induced interactions have been in-voked to explain the morphology and kinematics of the NLR ina number of well-studied Seyfert galaxies (e.g., Mrk 78: Whit-tle & Wilson 2004; NGC 1068: Das et al. 2006; NGC 4151:Mundell et al. 2003). Despite the natural appeal of radio jets,however, our analysis shows that radio power correlates morepoorly with ∆ σ than H α luminosity or Eddington ratio.We speculate that the dominant source of energy input comesfrom accretion disk winds, a common feature in AGNs (Cren-shaw et al. 2003). Disk winds or outflows presumably agi-tate the NLR through shocks, provided that they do not violateother constraints that limit the role of shocks in the excitationof the gas (Ho 2008). The ability of an accretion disk to gen-erate a wind may depend critically on L bol / L Edd (Proga 2007),and systems such as LINERs with ultra-low L bol / L Edd seem tolack outflows altogether (Ho 2008). Consistent with Greene& Ho (2005a), we propose that the Eddington ratio primarilyaccounts for the secondary line broadening in the NLR. Theapparent interchangeability between L bol / L Edd and L H α is prob-ably an artifact of the small dynamic range in black hole massfor the majority of the Palomar sample, coupled with the sizableintrinsic scatter in the σ g - σ ∗ relation.4.3. Line Width-Luminosity Correlation in AGNs
Phillips et al. (1983) first noticed that the luminosity of the[O
III ] line in Seyfert galaxies loosely correlates with its linewidth. This result has since been extended to much larger sam-ples of Seyferts by Whittle (1985, 1992b) and Gu et al. (2006),as well as to lower-luminosity AGNs in the Palomar survey byHo et al. (2003; using H α for line luminosity and [N II ] forline width). The form of the correlation is L NLR ∝ FWHM a ,where, depending on the study and sample, the slope can takea wide range of values, from a ≈ ∼
4. The physical origin of the line width-luminosity rela-tion has been unclear. Whittle (1992b) suggests that the linewidth-luminosity relation is actually a secondary correlation,reflecting on the one hand the more primary link between bulgemass and NLR velocity, and on the other various possible inter-dependences among bulge mass, NLR luminosity, radio power,and line width.In light of the results of this study and other recent develop-ments, we offer a simple explanation. We have confirmed, ashas long been suspected, that the widths of the narrow emis-sion lines largely reflect the gravitational potential of the bulge,such that FWHM ∝ σ ∗ . The bulge, meanwhile, directly links tothe mass of the central black hole via the fundamental relation M BH ∝ σ ∗ (Tremaine et al. 2002). Now, a given M BH radiatesup its Eddington luminosity, which in turn is related to the NLRluminosity by a bolometric correction and the Eddington ratio(accretion rate) of the system. Hence, AGNs should populate adistribution whose upper envelope follows a ridgeline definedby L NLR ∝ FWHM . At a given FWHM, L NLR has a maximumbut no hard minimum (other than that dictated by detection sen-sitivity). Indeed, this is what is seen. Ho et al. (2003) presentedthe line width-luminosity correlation separately for each of thethree classes of AGNs in the Palomar survey, and it is clear thatthe zero point of the correlation shifts with the average lumi-nosity of each class.4.4.
Source of the Gas and Its Dynamical Support
ONIZED GAS IN BULGES 9As discussed in Ho (2008, 2009), the gas that ultimately fu-els AGNs in nearby galaxies can be readily supplied internallyby mass loss from evolved stars (red giants and planetary neb-ulae) in the bulge. The same argument, the gist of which wasalready advanced by Minkowski & Osterbrock (1959) to ex-plain the origin of nebular emission seen in ellipticals, can beextended to include the entire gas reservoir that makes up theNLR in low-luminosity AGNs. Assuming T e = 10 K, n e = 100cm - , H α /H β = 3.1, and an effective recombination rate for H α of α effH α = 9 . × - cm s - (Osterbrock 1989), M NLR = 2 . × (cid:18)
100 cm - n e (cid:19) (cid:18) L H α erg s - (cid:19) M ⊙ . (4)From the population statistics of the Palomar survey assembledin Ho et al. (2003), LINERs and transition objects have me-dian L H α ≈ × erg s - and n e ≈
200 cm - , and hence M NLR ≈ × M ⊙ . The corresponding values for Seyferts are L H α ≈ × erg s - , n e ≈
400 cm - , and M NLR ≈ × M ⊙ . Strictly speaking, these values only pertain to the central200 pc ×
400 pc region sampled by the spectroscopic aper-ture, but given the high degree of central concentration of theline-emitting gas (Ho 2008), they probably underestimate theintegrated values for the entire NLR by no more than a fac-tor of ∼
2. These mass estimates can be compared with theamount of material shed from evolved stars. For a Salpeter stel-lar initial mass function with a lower-mass cutoff of 0.1 M ⊙ , anupper-mass cutoff of 100 M ⊙ , solar metallicities, and an age of15 Gyr (Padovani & Matteucci 1993), ˙ M ∗ ≈ × - (cid:18) L ∗ L ⊙ , V (cid:19) M ⊙ yr - . (5)With a median Hubble type of Sa, the host galaxies of the Palo-mar AGNs have sizable bulges and correspondingly healthy gassupply rates. For a median bulge luminosity of M B ≈ - . B - V = 0 . ˙ M ∗ ≈ . M ⊙ yr - . These values pertain to the entirebulge. Within the spectroscopic aperture the nuclear stellar con-tinuum luminosities (Ho et al. 1997a) are approximately a fac-tor of 10 lower than the integrated luminosities, which implies ˙ M ∗ ≈ . M ⊙ yr - . Thus, stellar mass loss can sustain the gasreservoir in the NLR if the stellar debris survives for periodslonger than ∼ × - yr before it dissipates and mergeswith the surrounding hot interstellar medium (Mathews 1990;Parriott & Bregman 2008).If the ionized gas is mainly produced internally through stel-lar mass loss in the bulge , what is the expected kinemati-cal signature of this material? The supersonic motion of theejecta through the ambient medium generates a strong shockthat thermalizes the gas to the kinetic temperature of the stars, ∼ - K, on a timescale of ∼ - yr (Sanders 1981;Mathews 1990). Not all of the gas gets heated, however. Nu-merical simulations of mass loss from evolved stars in ellipti-cal galaxies, after accounting for radiative cooling, predict thatsome fraction of the gas, perhaps ∼ During the early stages of theirevolution, the mass loss envelopes orbit roughly cospatiallywith their parent stars and thus closely track their natal veloci-ties. As bulge stars do not have a strong rotational componentto their velocities, neither should the gas shed by them. Afternormalizing the gas line widths by the observed (projected) ro-tational velocity of the large-scale (H I ) disk (Fig. 6), the signif-icant, positive trend with galaxy inclination angle that remainsimplies that the ionized gas does not reside in a plane alignedwith the large-scale galaxy. Moreover, the ejecta quickly losememory of their birth sites as the material experiences drag de-celeration against the surrounding hot medium. As discussedby Mathews (1990), the warm, optical-emitting clouds cannotbe self-gravitating. Instead, they must be pressure-confinedby the ambient hot medium, and hence their kinematics willlargely imprint the local velocity of the hot gas. Since the hotgas is in virial equilibrium with the bulge stars, to first orderwe anticipate σ g ≈ σ ∗ . In detail, our analysis finds σ g ≈ . σ ∗ .As the referee points out, supersonic turbulence in the hot gasrapidly dissipates through shocks. It is thus natural for σ g tobe slightly subvirial because the turbulence—the source of therandom motions—is subsonic. When an AGN is present, thehot gas may be more agitated.The scenario presented above provides a useful frameworkfor interpreting a number of trends gleaned from spatially re-solved observations of the central regions of galaxies. HubbleSpace Telescope images frequently reveal dust lanes, spiral pat-terns, and a variety of other small-scale fine features in galaxycenters, but rarely do the structures ever take the form of thin,well-defined disks (Phillips et al. 1996; Carollo et al. 1997;Malkan et al. 1998; Pogge et al. 2000; Ho et al. 2002; Martiniet al. 2003; Simões Lopes et al. 2007). Since the dust generallytraces the nebular gas (e.g., Ravindranath et al. 2001; Falcón-Barroso et al. 2006), we can infer that the ionized gas more orless follows the same topology as the dust. The complex mor-phology of the gas strongly hints that its velocity field must besimilarly chaotic. Where available, observations support thispicture. For example, a number of authors have noted that thecentral rotation curves of the ionized gas often rise more slowlythan the circular velocities predicted from the luminosity pro-file of the stars (Caldwell 1984; Caldwell et al. 1986; Fillmoreet al. 1986; Kent 1988; Kormendy & Westpfahl 1989; Cinzano& van der Marel 1994; Bertola et al. 1995; Fisher 1997; Cin-zano et al. 1999; Pignatelli et al. 2001), suggesting that randommotions contribute appreciably to the dynamical support of thegas. On scales probed by the
Hubble Space Telescope , very fewobjects exhibit clean signatures of dynamically cold disks un-dergoing circular rotation (Sarzi et al. 2001; Ho et al. 2002;Atkinson et al. 2005; Noel-Storr et al. 2007). Although thestatistics are still quite limited, evidence is emerging that thedegree of kinematic disturbance may be directly linked to thelevel of AGN activity (Verdoes Kleijn et al. 2006; Dumas et al.2007; Walsh et al. 2008; § 4.2).We end with the following note. The above discussion pre-supposes that the optical line emission comes exclusively from We do not mean to suggest that external acquisition of gas does not take place. Indeed, an external origin has been invoked to explain the distribution of kinematicaxis misalignments between the gas and stars (Bertola et al. 1992; Caon et al. 2000; Sarzi et al. 2006), as well as certain H I properties (Ho 2007), in early-typegalaxies. The optical line emission in ellipticals and bulges in principle can be excited by a variety of mechanisms, including photoionization by young or old hot stars,photoionization by a central AGN, shocks, self-irradiation from cooling condensations, and even cosmic ray heating by the radio-emitting plasma. With the possibleexception of the weakest emission-line objects, however, Ho (2008) argues that the majority of the low-luminosity AGNs in the Palomar survey—essentially allnearby galaxies with sizable bulges—are powered by nonstellar photoionization.
Chandra and
XMM-Newton observations of ellipticals andbulges show that the diffuse X-ray-emitting medium typicallyhas temperatures of kT ≈ . - n ≈ . - . - (see Ho 2009 and references therein). Assum-ing kT = 0 . n = 0 . - , the cooling time is merely ∼ × yr. This simple picture of galactic-scale cooling flowsencounters a number of long-standing difficulties familiar in thecontext of galaxy clusters, and their resolution appeals to sim-ilar remedies (Mathews & Brighenti 2003). A key ingredientto counter the overcooling problem invokes some form of en-ergy feedback. It is tempting to associate the evidence for AGNfeedback discussed in this paper (§ 4.2) with such a mecha-nism. The kinematics that ensue from the cooling flow scenarioare likely to be quite complicated, but in general we also ex-pect the cooling condensations to be in pressure balance withthe surrounding hot medium and hence to reflect its kinematics.Depending on the shape of the potential, the cooling flow fila-ments can settle in different planes at different radii (Tohline etal. 1982). This may account for the variety of dust structuresoften seen in the central regions of early-type galaxies (e.g.,Lauer et al. 2005). SUMMARY
We use a new catalog of stellar velocity dispersions to studythe relationship between the kinematics of the ionized gas andthe stars in the central few hundred parsecs of a large sample ofnearby galaxies. The gas dispersions are based on the width ofthe [N II ] λ σ g and σ ∗ are available for 345 galaxies spanning a wide range in Hubbletype and level of nuclear activity. The principal results of ouranalysis can be summarized as follows.1. The velocity dispersion of the ionized gas strongly cor-relates with, and is roughly comparable to, the velocitydispersion of the stars over the velocity range ∼ - - , across all Hubble types. This confirms the no-tion that the widths of the narrow emission lines primar-ily reflect the virial velocity of the gravitational potentialof the stellar bulge. In detail, σ g /σ ∗ ranges from ∼ . σ g /σ ∗ shows no strong dependence on Hubbletype, for galaxies ranging from E to Sbc. Among galax-ies of type Sc and later, systems with σ ∗ ∼ <
40 km s - , all spectroscopically classified as H II nuclei, σ g / σ ∗ ∼ >
1. The ionized gas maintains a minimum “floor” disper-sion of σ g ≈
30 km s - , which might signify a lowerthreshold level of interstellar turbulence generated byenergy feedback from massive stars. The σ g - σ ∗ relationdoes not depend on the presence of a large-scale bar oron the local environment of the galaxy.3. The residuals of the σ g - σ ∗ relation for AGNs (Seyferts,LINERs, and transition objects) correlate strongly withthe nuclear optical (H α ) luminosity, and perhaps evenstronger with the Eddington ratio. Radio power playsa less essential role. We argue that energy injected byaccretion disk winds or outflows, not radio jets, providea source of AGN feedback that accelerates the NLR gasto speeds above the virial velocities of the stars.4. We confirm that the line widths of narrow emission linescan be used to obtain reasonably reliable estimates of σ ∗ , but, for maximum accuracy, preference should begiven to low-ionization transitions such as [N II ] andcare should be taken to correct for the secondary de-pendence of σ g on AGN luminosity or Eddington ratio.5. We show that the mass budget of the ionized gas canbe accounted for by mass loss from evolved stars. Theoptical line emission arises from photoionization of thejust-shed, mass loss envelopes prior to their being ther-malized, in conjunction with an additional contributionfrom filaments recondensing from the hot medium. Weargue that the line widths are dominated by random mo-tions, reflecting the kinematics of the hot gas that pres-sure confines the line-emitting clouds.6. Lastly, we suggest that the line width-luminosity corre-lation in AGNs actually defines the upper envelope ofa distribution of points governed by the σ g - σ ∗ relation,the M BH - σ ∗ relation, and the Eddington limit.I am grateful to Joel Bregman, Jenny Greene, Janice Lee, andMark Whittle for helpful correspondence. I thank Bill Math-ews, the referee, for constructive criticisms that helped to clar-ify the discussion in §4.4. This work was supported by theCarnegie Institution of Washington and by NASA grants fromthe Space Telescope Science Institute, which is operated by theAssociation of Universities for Research in Astronomy, Inc., forNASA, under contract NAS5-26555. REFERENCESAtkinson, J. W., et al. 2005, MNRAS, 359, 504Baum, S. A., & McCarthy, P. J. 2000, AJ, 119, 2634Bertola, F., Buson, L. M., & Zeilinger, W. W. 1992, ApJ, 401, L79Bertola, F., Cinzano, P., Corsini, E. M., Rix, H.-W., & Zeilinger, W. W. 1995,ApJ, 448, L13Bian, W., Gu, Q., Zhao, Y., Chao, L., & Cui, Q. 2006, MNRAS, 372, 876Boroson, T. A. 2003, ApJ, 585, 647——. 2005, AJ, 130, 381Botte, V., Ciroi, S., Di Mille, F., Rafanelli, P., & Romano, A. 2005, MNRAS,356, 789Busko, I. C., & Steiner, J. E. 1992, MNRAS, 258, 306Caldwell, N. 1984, PASP, 96, 287Caldwell, N., Kirshner, R. P., & Richstone, D. O. 1986, ApJ, 305, 136Caon, N., Macchetto, D., & Pastoriza, M. 2000, ApJS, 127, 39Carollo, C. M., Stiavelli, M., de Zeeuw, P. T., & Mack, J. 1997, AJ, 114, 2366Chen, X.-Y., Hao, C.-N., & Wang, J. 2008, ChJAA, 8, 25 Cinzano, P., Rix, H.-W., Sarzi, M., Corsini, E. M., Zeilinger, W. W., & Bertola,F. 1999, MNRAS, 307, 433Cinzano, P., & van der Marel, R. P. 1994, MNRAS, 270, 325Crenshaw, D. M., Kraemer, S. B., & George, I. M. 2003, ARA&A, 41, 117Das, V., Crenshaw, D. M., Deo, R. P., & Kraemer, S. B. 2006, AJ, 132, 620Dasyra, K. M., et al. 2008, ApJ, 674, L9Demoulin-Ulrich, M.-H., Butcher, H. R., Boksenberg, A. 1984, ApJ, 285, 527de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Jr., Buta, R. J., Paturel,G., & Fouqué, R. 1991, Third Reference Catalogue of Bright Galaxies (NewYork: Springer)Dumas, G., Mundell, C. G., Emsellem, E., & Nagar, N. M. 2007, MNRAS, 379,1249Falcón-Barroso, J., et al. 2006, MNRAS, 369, 529Ferrarese, L., & Merritt, D. 2000, ApJ, 539, L9Filho, M. E., Barthel, P. D., & Ho, L. C. 2000, ApJS, 129, 93——. 2002, ApJS, 142, 223——. 2006, A&A, 451, 71