Host galaxies and black hole masses of low- and high-luminosity radio-loud active nuclei

Abstract

We investigate the host galaxy luminosities of BL Lac objects (BLLs) and radio-loud quasars (RLQs) at z < 0.5, imaged with the Hubble Space Telescope (HST). From a homogeneous treatment of the data, we construct the host-galaxy luminosity functions (HGLFs) and find that RLQ hosts are ∼0.5 mag brighter than those of BLLs: 〈M_R〉_RLQ=−24.0, 〈M_R〉_BLL=−23.5. For both classes, the HGLFs exhibit a remarkably different distribution with respect to that of normal (inactive) ellipticals, with clear preference for more luminous galaxies to show nuclear activity. We make use of the black hole mass–bulge luminosity (M_BH–L_bulge) relation, derived for nearby inactive ellipticals, to estimate the central black hole mass in our sample of radio-loud active galaxies. In spite of a ∼2 mag difference of intrinsic nuclear luminosity, BLLs and RLQs have black holes (BHs) of similar mass (〈M_BH/M_⊙〉_BLL = 5.6 × 10⁸, 〈M_BH/M_⊙〉_RLQ = 1.0 × 10⁹). This implies that the two types of objects are radiating at very different rates with respect to their Eddington luminosity.

black hole physics, galaxies: active, BL Lacertae objects: general, galaxies: photometry, quasars: general

Introduction

There is a general consensus about the existence of supermassive black holes (SBHs) at the centre of normal galaxies, as well as in the nuclei of active galaxies and quasars (see e.g. the recent review of Ferrarese 2002). A large body of data — in particular, based on high-resolution Hubble Space Telescope (HST) observations — is now available to strongly support the presence of such massive black holes (BHs).

SBHs play an important role in the formation and evolution of massive galaxies, and are also a key component in the development of nuclear activity. In spite of this apparently ubiquitous presence of SBH in galaxies, our understanding of how the galaxies and their central BHs are linked in the formation process of the structures remains unclear, but several possible explanations have been proposed (e.g. Silk & Rees 1998; Haehnelt & Kauffmann 2000; Adams, Graff & Richstone 2001; Burkert & Silk 2001; Merritt & Ferrarese 2001a; Balberg & Shapiro 2002).

From the observational point of view, it was shown that the BH mass is correlated with the properties of the bulge component of the host galaxy, which translates into the relationships between the black hole mass (M_BH) and the bulge luminosity (Kormendy & Richstone 1995; Magorrian et al. 1998; Richstone et al. 1998; Kormendy & Gebhardt 2001), and between M_BH and the velocity dispersion σ of the host galaxy (Ferrarese & Merritt 2000; Gebhardt et al. 2000; Merritt & Ferrarese 2001b). Any theory of SBH and galaxy formation must therefore take into account and explain such observed empirical relations (e.g. Silk & Rees 1998; Haehnelt & Kauffmann 2000; Ciotti & van Albada 2001). On the other hand, although these relations have a significant scatter (∼0.4 in log M_BH), they offer a new tool for the evaluation of BH masses in various types of active galactic nuclei (AGNs), if a reliable measurement either of the host galaxy luminosity or of σ is made. While for AGNs with strong emission lines [such as quasi-stellar object (QSO) and Seyfert galaxies] the standard methods (e.g. reverberation mapping), under virial assumptions of the emitting regions, can be used to derive M_BH (see e.g. Wandel, Peterson & Malkan 1999; Kaspi et al. 2000), the above relations may be the only way to estimate M_BH for active galaxies that lack emission lines or that are too far away [such as BL Lac objects (BLLs) and many nearby radio galaxies] to resolve the region of influence of the BH with present-day instrumentation. The two different approaches lead to consistent estimates of BH masses within the assumed uncertainties of the two methods (McLure & Dunlop 2002).

Given the difficulty of obtaining σ from spectroscopy of the galaxies that host active nuclei, it was possible to use σ to evaluate M_BH only for few AGNs (Ferrarese et al. 2001; Barth, Ho & Sargent 2002, 2003; Falomo, Kotilainen & Treves 2002; Falomo et al. 2003). In contrast, the galaxy luminosity is much easier to measure for active galaxies and can therefore be used to determine M_BH for larger data sets.

In this paper we use the black hole mass–bulge luminosity (M_BH–L_bulge) relation to investigate and compare the BH mass distributions of a sample of low- and high-luminosity radio-loud AGNs [BLLs and radio-loud quasars (RLQs) respectively]. Both classes are found to reside in massive giant ellipticals (Urry et al. 2000; Dunlop et al. 2003), which makes them rather homogeneous for such kinds of analysis. To ensure uniformity of the results, we have considered only objects at z < 0.5 and that have been imaged by HST. This also allows us to constrain their host properties better. In Section 2, we describe our samples of BLLs and RLQs and compare their host galaxy luminosity functions. In Section 3 we discuss the M_BH–L_bulge relation and derive the central black hole mass for each object. Finally, we discuss our findings in Section 4, comparing them with recent results on radio galaxies. In our analysis, H₀= 50 km s⁻¹ Mpc⁻¹ and Ω₀= 0 were used.

Luminosity of the Host Galaxies

We have collected host galaxy data for BLLs and RLQs at z < 0.5, imaged by HST with the Wide Field Planetary Camera 2 (WFPC2), and have constructed a homogeneous data set of the luminosities of the host galaxies. This yields a sample of 57 BLLs and 18 RLQs that represent, respectively, low- and high-luminosity radio-loud active galaxies.

Because most of the observations were obtained in the F702W and F675W filters, we converted all HST magnitudes into R Cousins band (Holtzman et al. 1995). In the few cases where the filters F555W and F606W were used, we applied a colour correction V−R= 0.61 for the elliptical host galaxies (Fukugita, Shimasaku & Ichikawa 1995). Absolute magnitudes have been k-corrected, following the prescriptions of Poggianti (1997), and corrected for galactic reddening, using the Bell Lab Survey of neutral hydrogen N_H (Stark et al. 1992) with the conversion log N_H/E(B−V) = 21.83 cm⁻² mag⁻¹ (Shull & Van Steenberg 1985), assuming a total-to-selective extinction A_R= 2.3E(B−V) (Cardelli, Clayton & Mathis 1989). Because the objects are distributed over a significant redshift interval, we have also applied a correction to set the host galaxy luminosity to the present epoch, assuming a passive stellar evolution for massive ellipticals (Bressan, Chiosi & Fagotto 1994). This correction (Δm∼−0.2) allows us to use the M_BH–L_bulge relations appropriately, which refer to local galaxies. In the following, M_R represents the host galaxy absolute magnitude, including all correction terms specified above.

The BL Lac object sample

The HST snapshot image survey of BLLs (Urry et al. 2000; Scarpa et al. 2000) has provided a homogeneous set of 110 short-exposure, high-resolution images through the F702W filter. From this, we have extracted all resolved objects at z < 0.5, yielding 57 sources with redshift between 0.027 and 0.495 (〈z〉= 0.20 ± 0.11). The host galaxy morphology of these objects is always well described by an elliptical model (Scarpa et al. 2000). The absolute M_R magnitude for each object is reported in Table 1. The host galaxy average luminosity is 〈M_R〉=−23.49 ± 0.5, roughly one magnitude brighter than the characteristic galaxy magnitude M*_R=−22.75 (Metcalfe et al. 1998). According to the shape of their spectral energy distribution (SED), BLLs are broadly distinguished into two types (see Padovani & Giommi 1995): those for which the SED peaks in the near-infrared/optical and the γ-ray MeV regions [called low-frequency peaked BL Lacs (LBL Lacs)], and those for which the SED peaks in the UV/X-ray and the γ-ray TeV energies [called high-frequency peaked BL Lacs (HBL Lacs)]. As shown by Urry et al. (2000), the host galaxy properties of HBLs and LBL Lacs are indistinguishable, and therefore the two subclasses will not be separated for this analysis.

Table 1.

Host galaxy properties of BLLs: (a) name of the source; (b) redshift; (c) absolute R host galaxy magnitude; and (d) black hole mass in units of M_? in logarithmic scale. (For reference, see Urry et al. 2000; Scarpa et al. 2000.)

The RLQ sample

Because there is not a homogeneous and large set of HST observations for RLQs, we have constructed a sample of 18 RLQs from the merging of three different subsets [Bahcall et al. 1997; Kirhakos et al. 1999 (subset BK); Boyce et al. 1998 (BO); Dunlop et al. 2003 (D)]. Bahcall et al. (1997) and Kirhakos et al. (1999) studied eight RLQs in the F606W and F555W filters and in the redshift range 0.158 < z < 0.367; Boyce et al. (1998) reported the analysis for five sources with 0.223 < z < 0.389 in the F702W filter. The largest subsample was investigated by Dunlop et al. (2001), who reported host galaxy properties for 10 RLQs, with 0.1 < z < 0.25 observed in the F675W filter. As in the case of BLLs, an elliptical model is always a good representation for the host galaxies. The average properties of the three subsamples are reported in Table 2. Our evaluations of M_R are consistent with absolute values reported by the quoted authors when galactic extinction, filter correction and evolution correction are taken into account. Because these subsets have statistically indistinguishable host luminosity distributions, we have merged these subsamples to construct a representative sample of RLQs (see also Treves, Carangelo & Falomo 2002), taking average values for objects observed twice. The combined data set therefore consists of 18 objects with redshift in the range 0.158 < z < 0.389, 〈z〉= 0.26 ± 0.07 and 〈M_R〉=−24.04 ± 0.4. In Table 3, we give the redshift (z), the assumed galactic extinction in the R-band (A_R), the host galaxy apparent magnitude (R) and the absolute magnitude (M_R) for each source. In Fig. 1, we compare the redshift distributions for the RLQ and BL Lac samples.

Table 2.

Average properties of three subsamples of RLQs: (a) the sample (BK =Bahcall et al. 1997 and Kirhakos et al. 1999; Bo =Boyce et al. 1998; D =Dunlop et al. 2003); (b) number of objects; (c) average redshift; and (d) average absolute host galaxy magnitude.

Table 3.

Host galaxies properties of RLQs: (a) name of the source; (b) redshift; (c) adopted extinction in the R band; (d) apparent host galaxy R magnitude; (e) absolute R host galaxy magnitude; (f) black hole mass in units of M_⊙ in logarithmic scale; and (g) reference (BK =Bahcall et al. 1997 and Kirhakos et al. 1999; Bo =Boyce et al. 1998; D =Dunlop et al. 2003).

Figure 1.

Redshift distribution for (a) RLQs; (b) BLLs; and (c) BLL matched samples (see text).

Comparison of the host luminosities of BLLs and RLQs

In our sample, both BLLs and RLQs have been mostly discovered as counterparts of radio and/or X-ray sources. Therefore, the objects considered here were selected on the basis of the nuclear properties, and — because there is not a significant correlation between the nuclear and host galaxy luminosity (Urry et al. 2000; Percival et al. 2001; Dunlop et al. 2003) — we can consider the distribution of the host galaxy luminosity as being unbiased by selection effects. Moreover, the homogeneous treatment of the data attests to a reliable comparison of host luminosity between the two classes (BLLs and RLQs). The exiguity of the RLQ sample, however, remains the main limitation of this comparison.

We find that the average absolute magnitude of RLQs is about 0.5 mag brighter than that of BLLs. The difference is illustrated in Fig. 2, where we compare the cumulative absolute magnitude distributions of the hosts for the two samples (a Kolmogorov–Smirnov test indicates that they are statistically different at the >99 per cent level). Because the two samples span slightly different redshift ranges, we checked that this did not affect our result. If we consider a subsample of BLLs with redshift distribution matched with that of RLQs (see Fig. 1c), we find 〈M_R〉_BLL (matched) =−23.54 ± 0.47, thus confirming our finding. Given the homogeneity of the data analysis and the procedure for the selection of the objects, we believe that this difference is not biased. We note, however, that a larger number of objects (in particular, of RLQs) is required to confirm this result on a firm statistical basis.

Figure 2.

Comparison of the cumulative host galaxy absolute magnitude (R-band) distributions of RLQs (dotted line) and BLLs (solid line).

To further compare the luminosity distributions of the host galaxies, we constructed the host galaxy luminosity function (HGLF) for the two subsets of objects. To set the normalization of the HGLFs, we simply assume the space density of both class of objects as derived from studies of complete samples. For BLLs, we use the value Φ₀= 10⁻⁵ Mpc⁻³ mag⁻¹ of the Fanaroff–Riley I (FR I) radio galaxies luminosity function at M_R=−22.8 given by Padovani & Urry (1991), under the assumption that FR I radio galaxies are the parent population of BLLs (e.g. Urry & Padovani 1995). For RLQs, we took the value of the luminosity function (LF) of close-by radio-quiet QSOs (Köhler et al. 1997; Grazian et al. 2000) to be M_B=−25.1, which corresponds to the average value of the nuclear magnitude for our sources, and scaled it by a factor of 10 to account for the ratio between radio-quiet quasars (RQQs) and RLQs (e.g. Moderski, Sikora & Lasota 1998). This yields Φ₀= 2.3 × 10⁻⁹ Mpc⁻³ mag⁻¹.

In Fig. 3, we show the HGLFs of BLLs and RLQs compared with that of inactive ellipticals (Metcalfe et al. 1998). To quantify the differences in shape of the HGLFs, we fitted the luminosity distributions of the host galaxies with a modified Schechter function Φ=KΦ_S (L/L*)^β, where K is a constant and Φ_S is the Schechter function for elliptical galaxies (Metcalfe et al. 1998): Φ_S=Φ*(L/L*)^α exp(−L/L*), assuming Φ*= 8.5 × 10⁻² Mpc⁻³, α=−1.2 and L*= 2.25 × 10⁴⁴ erg s⁻¹ (Metcalfe et al. 1998). The best fit to the HGLF was estimated, minimizing χ² for the function Φ. We find β= 2.7 ± 0.2 for BLLs, β= 3.6 ± 0.3 for RLQs. The shapes of the two HGLFs are somewhat different, but only at the 2σ level.

Figure 3.

The HGLF of RLQs and BLLs (filled points) compared with the fit (solid lines) with a modified Schechter function (see text). A slightly different value of β is found for BLLs (β= 2.7) and RLQs (β= 3.6). The dotted line is the fit to RLQ data normalized to the BLL data. The dashed curve represents the luminosity function of elliptical galaxies of Metcalfe et al. (1998).

This suggests that a given elliptical has a probability of having a radio-loud active nucleus that depends on the galaxy luminosity. Moreover, one can argue that the steepness of this behaviour depends on the intrinsic luminosity of the nucleus, as hinted by the different value of β for BLLs and RLQs.

It therefore turns out that both types of radio-loud active galaxies exhibit remarkably different distributions with respect to normal ellipticals, with clear preference for more luminous (and massive) galaxies to show nuclear activity. This behaviour disagrees with that found by Wisotzki, Kuhlbrodt & Jahnke (2001) for the host galaxies of radio-quiet QSOs. In fact, the shape of the HGLF of radio-quiet QSOs appears to be consistent with that of ordinary inactive early-type galaxies.

Mass of the Central Black Hole

Based on dynamical studies of nearby early-type galaxies, it was shown that there is a linear relation between the luminosity of the spheroidal component of a galaxy (L_bulge) and the mass of the central black hole (M_BH) (e.g. Kormendy & Gebhardt 2001, and references therein). This correlation has a scatter of ∼ 0.4 in log M_BH that can be ascribed mainly to the errors of measurements of M_BH and to the uncertainties in disentangling the bulge from the disc component of the galaxies. Nevertheless, it can be used to estimate M_BH for our objects, provided that consistent and reliable host galaxy luminosities are used.

To avoid systematic effects, it is important that the adopted absolute magnitude of the galaxy is homogeneous (in terms of spectral band, adopted cosmology, extinction correction, filter, etc.) with that used to derive the M_BH–L_bulge relation. To satisfy this requirement, we used the relationship between M_BH and M_R derived by Bettoni et al. (2003) for 20 inactive ellipticals, assuming the same calibrations of M_R we have adopted here.

From the analysis of the inactive sample of ellipticals, Bettoni et al. find that  

(1)

which is used to derive M_BH from the absolute (total) magnitude M_R of ellipticals (H₀= 50 km s⁻¹ Mpc⁻¹). This relation has an rms scatter of 0.38 in log M_BH, and it is similar to that derived by McLure & Dunlop (2002).

Because the host galaxies in the samples considered here are all bona fide ellipticals, we used relation (1) to derive M_BH for the two samples of radio-loud AGN (BLLs and RLQs), and report the value for each object in Tables 1 and 2, respectively. The uncertainty in the estimated black hole mass is dominated by the scatter of relation (1), while the uncertainties in the host galaxy magnitude are usually smaller.

The distributions of M_BH for the two samples are shown in Fig. 4. The two classes exhibit an average difference of a factor of ∼2 in M_BH, as a consequence of the different average host luminosity. We find that the average values of M_BH are 〈log(M_BH/M_⊙)〉= 8.75 ± 0.25 and 9.02 ± 0.20 respectively for BLLs and RLQs.

Figure 4.

Comparison of the black hole mass (derived from the host galaxy luminosity, using the relation between M_BH and bulge luminosity of nearby ellipticals) distributions of RLQs and BLLs.

A complementary method for the derivation of BH masses is based on the relation between the BH masses and σ of the host galaxy (Gebhardt et al. 2000; Ferrarese & Merritt 2000). This has been applied to a small number of nearby BLLs (Falomo et al. 2002, 2003; Barth et al. 2002, 2003). As shown by Falomo et al. (2003), there is a good agreement in the results obtained with the two techniques. No estimates of the BH mass of RLQs are available from σ because of the lack of measurements.

Wu, Liu & Zhang (2002) and Woo & Urry (2002) have derived estimates of the BH mass of BLLs using the M_BH–σ relation, where the latter quantity was inferred from measurements of the effective surface brightness and the effective radius of the host galaxy, and by assuming these are linked to σ through the Fundamental Plane (FP) relation. Although in principle this method could work, we believe it is less accurate than the direct use of the M_BH–M_bulge relation. In fact, in addition to the uncertainty in the measured quantities (surface brightness μ_e and effective radius R_e), which are much larger than the total magnitude of the galaxy, one has to take into account the uncertainty due to the intrinsic scatter of the FP relation and that of the M_BH–σ relation. Comparison of our BH masses (see Table 1) with those derived by Woo & Urry (2002) or Wu et al. (2002) indicates remarkable differences in M_BH for many objects, mainly because of the incorrect evaluation of the velocity dispersion. In some cases (e.g. 1807+698, 1104+384), the poor estimate of the velocity dispersion via the FP method has been clearly confirmed by direct measurements of σ (Barth et al. 2003; Falomo et al. 2003).

We also note that in the Woo & Urry (2002) estimates, additional errors are derived from a mistreatment of the FP parameters, because — instead of the average surface brightness (〈μ〉_e)– the isophotal surface brightness was used.

Discussion

The analysis of HST images for low-redshift BLLs and RLQs has shown that, for both types of active nuclei, the host galaxies are very luminous ellipticals. On average they are ∼1–2 mag more luminous than the typical galaxy luminosity (M*_R∼−22.75: Metcalfe et al. 1998). After homogeneous treatment of the data, we also found that host galaxies of RLQs are systematically more luminous by ∼0.5 mag than BL Lac hosts. Although this result does not seem to depend on the selection of the objects, a larger sample of RLQs is needed to reach a firm conclusion. We have shown that the distribution of the host galaxy luminosity exhibits a marked drop towards less luminous galaxies, and that it is somewhat different for the two classes (BLL and RLQ). This indicates that such a kind of nuclear activity occurs preferentially (or lasts longer) in massive galaxies. The different distributions of host luminosity for BLLs and RLQs may simply reflect the very large range of intrinsic nuclear luminosity (about two orders of magnitude; see below). High-power nuclear activity like that observed in RLQs can occur only in the most luminous and massive galaxies, and is therefore a rare event. On the other hand, low-power nuclear activity like that observed in BLLs (or in radio galaxies believed to be identical objects unaffected by beaming effects) can also be present in galaxies with intermediate luminosities.

With the assumption that the galaxy luminosity is correlated with the central BH mass, the host galaxy luminosity can be translated into central BH masses. It therefore turns out that, within a factor of 2, BLLs and RLQs have similar BH masses, but their total intrinsic nuclear luminosities are remarkably different. In addition to the higher observed nuclear/host ratio of RLQs with respect to BLLs, we have to take into account the fact that — while we consider RLQs as being basically unbeamed — for BLLs, a substantial beaming factor is present (δ∼ 15: see Ghisellini et al. 1998; Capetti & Celotti 1999). The intrinsic nuclear luminosities therefore differ by about a factor of 100. This implies a dramatic difference in the Eddington ratio ξ_E=L/L_E, where L_E= 1.25 × 10³⁸(M_BH/M_⊙) erg s⁻¹ (see also O'Dowd et al. 2001; O'Dowd, Urry & Scarpa 2002; Treves et al. 2002). Based on the estimated total QSO luminosity of L ∼ 3 × 10¹² L_⊙ (e.g. Elvis et al. 1994) and assuming BH masses of 1– 5 × 10⁹ M_⊙, we find that RLQs may be emitting at rates of 10 per cent or higher than their Eddington power, while BLLs are always emitting at regimes that are much lower than L_E.

According to the unification schemes of radio-loud AGNs (e.g. Urry & Padovani 1995), BLLs are radio galaxies, the jet of which is closely oriented toward the observer. Based on arguments of number density, luminosity functions and unbeamed properties (as the extended radio luminosity and the global properties of the host galaxies), the parent population of BLLs is likely to be formed by FR I radio galaxies with some contamination by FR II sources (Padovani & Urry 1990; Wurtz, Stocke & Yee 1996; Falomo & Kotilainen 1999; Cassaro et al. 1999; Urry et al. 2000). Under this hypothesis, the BH mass of BLLs (〈log(M_BH/M_⊙)〉= 8.75 ± 0.25) and of the parent (unbeamed) objects must be identical. In Fig. 5, we compare the distribution of the BH mass for our sample of BLLs with that of low-redshift radio galaxies from the sample of Govoni et al. (2000) (see also Bettoni et al. 2003). The two distributions are rather similar, although the most massive BHs in luminous FR I radio galaxies do not appear to have counterparts in the known BLLs. The average values of M_BH are 〈log(M_BH/M_⊙)〉= 9.04 ± 0.30 and 8.78 ± 0.35 for FR I and FR II radio galaxies, respectively.

Figure 5.

Black hole mass distribution of our sample of BLLs (top panel) compared with that of FR I (middle panel) and FR II (bottom panel) radio galaxies studied by Bettoni et al. (2003). The open histograms in the middle and bottom panels refer to the distribution of the whole sample of radio galaxies (see text).

Finally, we wish to note that, in addition to the mass, the other parameter which characterizes a black hole — and that may play a relevant role in the observed phenomenology — is the BH spin. It has been suggested that the spin energy is responsible for the jet emission L_j, and therefore for the development of the radio emission (e.g. Blandford 2002; Dunlop et al. 2003). The spin is clearly not directly measurable, but it could be deduced from an estimate of L_j and the BH mass using the Blandford & Znajek (1977) formula. Whereas L_j could be obtained from the spectral energy distribution (e.g. Tavecchio et al. 2000, 2002), the BH mass may come through the procedures described in this work.

Acknowledgments

This work has received partial support under contracts COFIN 2001/028773, ASI-IR-115 and ASI-IR-35.