A CENSUS OF X-RAY NUCLEAR ACTIVITY IN NEARBY GALAXIES

Supermassive black holes (BHs) are now believed to exist in all massive galaxies with a spheroidal component (Magorrian et al. 1998; Merritt & Ferrarese 2001b; Kormendy 2004). Low-mass galaxies tend to harbor a nuclear star cluster, whose mass is also related to the mass of the spheroidal component (Ferrarese et al. 2006). But in some low-mass galaxies, nuclear BHs have also been identified (Seth et al. 2008; Filippenko & Ho 2003; Jímenez-Bailón et al. 2005; Greene & Ho 2004, 2007; Dong et al. 2007; Greene et al. 2008). Thus, it is still a topic of active investigation whether there is a mass threshold between spiral galaxies containing a nuclear BH or a nuclear star cluster, to what extent they coexist, and what the mass relation is between the two nuclear systems when they are both present. It is also still debated whether there is a Hubble-type threshold for the presence of a nuclear BH, that is, whether late-type spiral galaxies without a spheroidal component can have nuclear BHs, and if so, whether the BH formation mechanisms and growth processes are different from those of nuclear BHs in massive spheroidals (Seth et al. 2008; Wang & Kauffmann 2008).

The main difficulty in resolving these questions is determining which galaxies host nuclear BHs. In particular, kinematic mass determinations require prohibitively high spatial resolution at the low-mass end. When a kinematic mass determination of the central dark mass is not available (i.e., in all but a few cases), the strongest evidence of the presence of a nuclear BH comes from its accretion-powered activity (Salpeter 1964; Lynden-Bell 1969). Active galactic nucleus (AGN) activity can be identified either from an emission-line optical/UV spectrum from the nuclear region, from a pointlike X-ray nuclear source, or from a flat-spectrum radio core; surveys in different bands lead to different selection biases. Statistical studies of the AGN population provide direct information on what fraction and what types of galaxies contain a nuclear BH, and indirect constraints on their masses and accretion rates.

Optical spectroscopic studies such as the Palomar Survey (Ho et al. 1997a, 1997b) and the Sloan Digital Sky Survey (SDSS; Hao et al. 2005a) suggest (Ho 2008) that ≈10% of nearby galaxies are Seyferts, ≈20% are "pure" low-ionization nuclear emission-line regions (LINERs), and another ≈10% are "transition objects" whose spectra are intermediate between those of pure LINERs and H ii regions (Heckman 1980). This means that overall, at least 4/10 of nearby galaxies contain a currently active nuclear BH. This is only a lower limit to the population of currently active nuclear BHs; another 40% of galaxies have an emission-line nucleus (H ii nucleus). H ii nuclei are powered by a compact star-forming region, but in some cases they may also harbor a weakly accreting BH. Optical surveys also show that Seyferts, LINERs, and transition objects are mostly found in early type galaxies: ≈60% of E to Sb galaxies have such nuclear signatures; conversely, almost all late-type disk galaxies are dominated by H ii nuclei, and ≲20% of them have evidence of an accreting nuclear BH (Ho 2008). Thus, it is more difficult to obtain reliable BH demographics in late-type galaxies (containing smaller BHs) using optical spectroscopy, due to confusion from star formation. Spitzer's mid-infrared spectroscopic studies (Satyapal et al. 2008) have recently proved more effective at finding nuclear BHs in late-type galaxies, with an AGN detection rate possibly four times larger than that suggested by optical spectroscopic observations.

Radio-band surveys are not yet as complete or as sensitive as the Palomar or Sloan optical surveys. Nonetheless, arcsec-resolution Very Large Array (VLA) surveys at 5 GHz (Wrobel & Heeschen 1991), 8.4 GHz (Nagar et al. 2005), and 15 GHz (Filho et al. 2006) have confirmed the presence of AGN signatures (nonthermal radio cores) in ≈30%–40% of early type galaxies, mostly Seyferts and LINERs. The radio emission in LINERs is mainly confined to a compact core or the base of a jet (subarcsec size); Seyferts are more often accompanied by extended (arcsec-size) jetlike features (Nagar et al. 2005). Somewhat in contrast with the optical classification, radio detections of compact cores in transition objects are much rarer; this suggests that perhaps ≈50% of transition objects are not AGNs.

X-ray surveys are an underutilized and very promising tool to further our understanding of low-level nuclear activity in the local universe. X-ray emission probes regions much closer to the accreting BH than, for example, optical emission lines. And a direct study of the nuclear X-ray source allows better constraints on its mass accretion rate and output power, stripped off the often messy or ambiguous optical-line phenomenology of the host galaxy. Subarcsec spatial resolution and astrometric accuracy are required for the study of a low-luminosity AGN, for at least two reasons: to confirm that an X-ray source is pointlike and coincident with the optical/radio nucleus and to separate the pointlike, nuclear X-ray source from the surrounding unresolved emission (mostly from diffuse hot gas and faint X-ray binaries), which is often present in galactic cores. Only Chandra can provide such resolution and astrometric accuracy. There are still no complete, unbiased Chandra X-ray surveys of galaxies in the local universe, mostly because time-allocation constraints lead to an implicit bias toward X-ray luminous targets. However, a number of Chandra studies have targeted a sizeable fraction of nearby LINERs (about half of the LINERs in the Palomar survey; Ho 2008). They have confirmed that most LINERs and transition objects contain an X-ray luminous, accreting BH (Satyapal et al. 2004; Dudik et al. 2005; Pellegrini 2005; Satyapal et al. 2005; Flohic et al. 2006; González-Martín et al. 2006). More specifically, ≈75% of LINERs have an X-ray nucleus, and this fraction goes to 100% for the subsample of LINERs that also have a detected radio core (Ho 2008, and references therein). In such Chandra studies, the typical detection limit for pointlike nuclear X-ray sources is ∼10³⁸ erg s⁻¹. This is directly comparable to the detection limit of nuclear Hα emission in the Palomar survey, ∼10³⁷ erg s⁻¹, because, in typical Seyferts, L_X ∼ 10L_Hα (Ho 2008). For such a low-luminosity AGN, radio and X-ray studies are in principle less affected by contamination from possible surrounding star-forming regions than studies based on photoionized Hα emission. It is still not known whether or what fraction of H ii nuclei in late-type spirals also have an X-ray active nuclear BH, and whether there is any correlation with the presence and strength of a large-scale bar. Pioneering studies (Ghosh et al. 2008; Desroches & Ho 2009) suggest that a few nearby late-type spirals with an optically classified H ii nucleus may harbor a low-luminosity AGN with L_X≈ a few 10³⁷ erg s⁻¹. The drawback of X-ray studies is that at such low luminosities, it becomes extremely difficult to determine for any individual galaxy whether the nuclear source is the accreting supermassive BH or an unrelated X-ray binary in the dense nuclear region or in a nuclear star cluster.

X-ray surveys, in combination with other bands, can do more than just refine our statistical census of active nuclei across the Hubble sequence: they can be used to obtain a physical understanding of the process of accretion across the whole range of mass accretion rates. A fundamental issue that deserves investigation is whether there is a continuous distribution of nuclear X-ray activity from the most luminous quasars (L_bol ∼ L_Edd) to run-of-the-mill Seyferts (L_bol ∼ (10⁻⁴–10⁻²)L_Edd), LINERs (L_bol ∼ (10⁻⁶–10⁻⁴)L_Edd), transition objects (L_bol ∼ (10⁻⁷–10⁻⁶)L_Edd) all the way down to "quiescent" galactic nuclei such as the BH in the Milky Way or in nearby elliptical galaxies (L_bol ≲ 10⁻⁹L_Edd), or, instead, whether there is a series of clear thresholds along this sequence, corresponding to fundamental changes in the accretion mode and inflow structure. Such transitions may or may not coincide with the "canonical" accretion states found in stellar-mass BHs. Moreover, thresholds in the X-ray properties may or may not coincide with optically classified classes of AGNs.

One such physical transition may be between accreting BHs with and without an optically thick accretion disk. The presence or absence of disk signatures such as Compton reflection and a broad Fe line in the X-ray spectra can provide strong constraints, in parallel with the presence or absence of a UV bump in the spectral energy distribution (SED). It was suggested (Ho 2008) that the disappearance of the inner disk marks the transition between Seyferts and LINERs. A related issue yet to be properly understood in low-luminosity AGNs is the redistribution of the accretion power among a radiative component (UV/X-rays), a mechanical component (radio jets), and an advected component; AGNs become radio louder at lower accretion rates, in agreement with the trend seen in Galactic BHs (Fender et al. 2004; Merloni et al. 2003). Order-of-magnitude estimates of the accretion rate based on the mass-loss rate from evolved stars and the gravitational capture rate of hot gas from the interstellar medium (ISM) typically lead to large overestimates of the nuclear BH luminosity, especially in LINERs and transition objects (Ho 2008). Therefore, such objects are an ideal test for radiatively inefficient accretion models such as Advection-Dominated Accretion Flows (ADAF) or Radiatively Inefficient Accretion Flows (RIAF; Narayan & Yi 1995; Quataert & Narayan 1999; Narayan 2002; Yuan & Narayan 2004). X-ray studies also allow quantitative comparisons between the amount of gas that is used for star formation in the nuclear region of late-type spirals and that used for fuelling a possible accreting BH (or an upper limit to that accretion rate), and how this ratio depends on the Hubble type and bar structure.

Another physical problem that can be addressed with X-ray surveys is the validity of the unified model (Antonucci 1993) at low accretion rates and low luminosities (L_X ∼ 10³⁸–10⁴¹ erg s⁻¹). The standard unified model is based on the presence of a geometrically and optically thick parsec-scale structure ("dusty torus") that blocks our direct view of the innermost disk in high-inclination sources (Type 2 AGN). The nature and physical structure of the obscuring material are still controversial. In alternative to the parsec-scale torus scenario, it was suggested that the absorption could instead be due to an optically thick disk wind, launched from a few 100 Schwarzschild radii (Elvis 2000; Elvis et al. 2004; Murray et al. 1995; Murray & Chiang 1998). X-ray spectral surveys of low-luminosity AGNs are crucial to determine whether the thick absorber disappears along with the disk signatures (which would point to an intimate connection between the two components) and at what luminosities. We already know that there are no dusty tori around extremely faint BHs such as that in our Galaxy. Chandra and XMM-Newton studies show low absorbing column densities and weak or undetected narrow Fe Kα emission in a large fraction of nearby LINERs, which suggests a direct, unobstructed view of the nucleus (Ho 2008). In contrast, strong Fe Kα emission (Cappi et al. 2006) and a nearly continuous distribution of absorbing columns (Panessa et al. 2006) suggest that Seyferts are more gas-rich than LINERs.

To address these issues, we have collected and analyzed the archival Chandra data available for nearby galaxies (see Section 2 for the selection criteria of our sample). In this paper, we present preliminary results of our X-ray population study, focussing in particular on the luminosity distribution and on the dependence of the absorbing column density on the nuclear luminosity.

In general, we have two choices when selecting a sample of nearby galaxies for X-ray population studies of their nuclei. We can select all targets observed by Chandra within a certain distance or flux detection limit; this sample will probably be biased in favor of X-ray luminous galaxies (or galaxies with starbursts or with some other X-ray peculiarities), which are more likely to be selected as targets. Alternatively, we may select a limited subsample of such targets, which may be regarded as complete (or at least unbiased) according to some optical criteria.

For our work, we have started from an optically/IR-selected sample previously used by Swartz et al. (2008, 2009) for a statistical X-ray study of ultraluminous X-ray sources (ULXs) population. This sample is defined as a volume-limited set of galaxies within 14.5 Mpc that are both contained in the Uppsala Galaxy Catalog (UGC; Nilson 1973) with photographic magnitude m_p < 14.5 mag, and in the Infrared Astronomical Satellite (IRAS) catalogs (Fullmer & Londsdale 1989; Moshir et al. 1993) with a flux f_FIR ⩾ 10^−10.3 erg cm⁻² s⁻¹, where f_FIR/10⁻¹¹ = 3.25S₆₀ + 1.26S₁₀₀ erg cm⁻² s⁻¹ (Rice et al. 1988). Here, S₆₀ and S₁₀₀ are the total flux densities, expressed in Jy, at 60 and 100 μm respectively. The IRAS catalogs are complete to approximately 1.5 Jy for pointlike sources. The UGC contains all galaxies north of B1950 δ = −2°30'. This combined selection criteria favor nearby, predominantly optically bright galaxies with at least a modest amount of recent star formation. The sample is known to exclude smaller dwarf galaxies in the neighborhood (Swartz et al. 2008). Most ellipticals would be included by the selection criteria; however, there are not many northern-hemisphere nearby ellipticals (this bias is further discussed in Swartz et al. 2009). There are 140 galaxies in this complete sample. However, only 116 have been observed with Chandra; the rest only have XMM-Newton and ROSAT data, which are less suitable for our study due to their larger point-spread functions (PSFs). Nonetheless, we estimate that the bias introduced by the lack of Chandra coverage for those 24 galaxies out of 140 is minimal. Henceforth, we will refer to this subsample of 116 galaxies as the optical/IR sample, for simplicity. The morphological classification of these 116 galaxies is 3% ellipticals, 44% early type spirals (S0 through Sb), 42% late-type spirals (Sc through Sm), and 11% dwarfs/irregulars. The optical-line classification is 9% Seyfert, 11% LINERs, 11% transition objects, 29% H ii nuclei, and 40% unknown.

We have then used a larger sample, defined by J. Liu (2009, in preparation): it includes all Chandra/ACIS targets within 15 Mpc, contained in the Third Reference Catalog of Galaxies (RC3; de Vaucouleurs et al. 1991), which is complete for nearby galaxies having apparent diameters ⩾1' at the D25 isophotal level and total B-band magnitudes m_B < 15.5 mag, in both the Northern and Southern sky. We have taken all the Chandra/ACIS targets within this sample (as of the end of 2007); the exposure times range from 500 s to 120 ks. This selection adds another 71 galaxies to those already contained in the optical/IR sample. The Liu sample is slightly biased in favor of the brighter and larger galaxies, as expected; it is essentially an X-ray-selected sample, because only ∼10% of the RC3 galaxies have been Chandra targets. From the morphological point of view, the Liu sample has an overabundance of early type spirals, which somewhat compensates the bias of the optical/IR sample in favor of late-type galaxies; moreover, it includes a few southern-hemisphere ellipticals, lacking from the optical/IR sample.

In summary, there are altogether 187 nearby galaxies with publicly available Chandra observations (as of 2007 December) included either in the optically/IR-selected sample or in the X-ray-selected sample, within 15 Mpc (full list in Table 1).⁵ More than half (104/187) of these galaxies are also included in the optical Palomar sample. The morphological classification of these 187 galaxies is 8% ellipticals, 43% early type spirals (S0 through Sb), 36% late-type spirals (Sc through Sm), and 13% dwarfs/irregulars. The optical-line classification is 8% Seyfert, 11% LINERs, 10% transition objects, 40% H ii nuclei, and 31% unknown. For simplicity, we will call this sample of 187 galaxies the "extended sample."

For each of the 187 Chandra target galaxies in our extended sample, we searched for a pointlike X-ray source at or close to the nuclear position. For the nuclear coordinates, we generally referred to the positions listed in the NASA/IPAC Extragalactic Database (NED⁶). In most cases, they come from the Two Micron All Sky Survey (2MASS) catalog (Skrutskie et al. 2006); the semimajor axes of the 95% confidence ellipse vary between ≈10 and 15. In a few cases, the NED positions come from the SDSS Data Release 6⁷ (semimajor axes ≈ 05) or from VLA radio observations (e.g., in the case of NGC 4203, with semimajor axes ≈ 01).

We used the Chandra Interactive Analysis of Observations software (CIAO) V4.1 for filtering and analyzing the event files. For each Chandra observation, we checked and screened out exposure intervals corresponding to background flares. We used the standard task wavdetect to identify sources in the nuclear region of each galaxy. The morphologies of the nuclear regions in the Chandra images can be loosely grouped into four classes (Figure 1): (I) a dominant nuclear source, (II) a nuclear source embedded in extended, unresolved emission, (III) unresolved emission in the nuclear region but no pointlike source, and (IV) no nuclear source at all.⁸ Eighty-six of the 187 galaxies in our sample belong to class I and II, that is, have a pointlike X-ray source within 1'' of the independently identified nuclear position (details in Section 3.1). For all the nuclear sources, we extracted spectra (in the 0.3–8 keV band) from circular regions of radius 15 (which include ≈90% of the counts), and corresponding background and response files using the CIAO task psextract. We used XSPEC Version 12.1 (Arnaud 1996) for spectral modeling. We used the Cash statistics (Cash 1979) for sources with ≲200 counts and the χ² statistics (with suitably binned data) in the other cases. In addition, some galaxies have been the target of detailed Chandra studies in the literature: in these cases, we have used their spectral results to complement our analysis.

Zoom In Zoom Out Reset image size

3.1. Census of Nuclear X-ray Sources

3.2. Spectral Modeling and Luminosity Distribution

There are about 15 sources with ≈4–10 net counts: these are significant detections because the expected positions are independently known, the background level is very low, and there are no other pointlike sources nearby. We used the Bayesian method of Kraft et al. (1991) to confirm the significance of these detections. For sources with such a low number of counts, we fitted their spectra with a simple power-law model with fixed galactic absorption (no intrinsic absorption). For other sources with more counts, we used an absorbed power-law model with free intrinsic column density N_H. A power law with cold absorption generally provides a good fit for most of the sources. However, about 20 nuclear sources clearly require an additional ionized absorber and/or an optically thin thermal-plasma emission component, which we have included in our spectral fits. Figure 2 shows the distribution of the fitted photon indices for the optical/IR sample and the extended sample. Most of the nuclear sources have a photon index ≈1.5–2.0, which is the typical range for AGNs. Figure 3 shows the distribution of the intrinsic N_H for the two samples (which are once again consistent with each other). It is clear that most nuclear sources are not heavily obscured; very few of them can be classified as Type 2, in the unified scheme (see Section 3.4).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We then used the fitted values of the photon index and absorption to estimate the intrinsic 0.3–8 keV isotropic luminosities of all 86 nuclear sources in the extended sample (Table 4 and Figure 4). The cumulative luminosity distribution is consistent with a power law with a slope of ≈−0.45 above ≈2 × 10³⁸ ergs s⁻¹, which we estimate as the completeness limit, based on the exposure times and detection thresholds for the shortest observations in our sample. Because of the different exposure times and distances of our galaxies, in addition to nonconstant background levels in the nuclear regions, it is not meaningful to give a detection threshold for the whole sample.

Table 4. Galactic Nuclei with Pointlike X-ray Emission

Galaxy	L0.3-8a (1039 ergs s−1)	NHb (1022 cm−2)	Modelc	id (°)	MBH (107 M☉)	Methode	rEddf	log($\dot{m}$)g
IC342	0.37	0.09+0.03−0.05	ARP	20	0.25	MS	1.2E−06	−3.3
IC396	2.5	0.35+0.61−0.35	AP	47	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC45	0.021	<0.6	AP	47	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC253	2.0	20+13−9	[1]	86	0.94	MS	1.6E−06	−3.2
NGC404	0.031	<0.11	AP	0	0.06	MS	4.0E−07	−3.8
NGC598 (M33)	0.8	0.45+0.01−0.01	AP,[2]	56	<1.5 × 10−4	S,[3]	> 0.004	> − 1.4
NGC628 (M74)	0.18	<0.06	AP	0	0.5	BB,[4]	2.7E−07	−3.7
NGC891	0.0045	⋅⋅⋅	P	84	0.23	MS	1.5E−08	−4.2
NGC925	0.59	<0.2	AP	54	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC1012	165	44+82−21	AP	61	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC1023	0.73	<0.11	AP	72	4.4	S,[5]	1.3E−07	−3.9
NGC1058	0.055	⋅⋅⋅	P	16	0.012	MS	3.4E−06	−2.7
NGC1068 (M77)	300	≲0.03	[6]	29	1.50	M,[7]	1.6E−04	−2.0
NGC1291	2.0	2.0+0.7−0.7	AP,[8]	28	7.4	MS	2.1E−07	−3.7
NGC1493	0.54	0.03+0.22−0.03	AP	0	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC1637	0.12	0.56+0.10−0.08	[9]	39	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC1672	1.0	10+30−10	[10]	37	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC1808	12	3.1+0.8−0.7	[11]	50	4.1	MS	1.7E−06	−3.2
NGC2500	0.35	⋅⋅⋅	P	0	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC2681	0.61	0.04+0.04−0.03	ARP	0	1.2	MS	4.0E−07	−3.5
NGC2683	9.0	13+28−13	AHP	79	1.5	MS	4.6E−06	−2.9
NGC2787	3.2	0.13+0.06−0.06	AP	52	4.1	G,[12]	6.0E−07	−3.4
NGC2841	0.89	0.11+0.10−0.09	AP	64	6.3	BB,[13]	1.1E−07	−3.9
NGC3031 (M81)	100	0.09+0.02−0.02	AP,[14]	60	6.8	BB,[15]	1.1E−05	−2.7
NGC3077	0.060	1.6+0.8−0.6	AP	43	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC3115	0.28	0.01+0.09−0.01	AP	66	92	S,[16]	2.4E−09	−4.4
NGC3125	0.04	<0.8	AP	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC3184	0.02	0.21+0.34−0.21	AP	26	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC3344	0.36	⋅⋅⋅	P	23	0.16	BB,[4]	1.8E−06	−3.1
NGC3368 (M96)	0.19	⋅⋅⋅	P	50	3.2	BB,[4]	4.5E−08	−4.1
NGC3377	0.39	0.17+0.15−0.13	AP	54	10	S,[17,18]	3.0E−08	−4.1
NGC3379 (M105)	0.28	0.03+0.1−0.03	AP	25	13.5	S,[19]	1.4E−08	−4.2
NGC3384	0.55	0.12+0.55−0.12	AP	65	1.6	S,[17,20]	2.4E−07	−3.7
NGC3412	0.044	⋅⋅⋅	P	59	0.87	MS	4.0E−08	−4.1
NGC3489	0.66	<0.5	AP	56	1.0	MS	5.0E−07	−3.5
NGC3507	0.22	0.04+0.17−0.04	ARP	37	0.79	MN	2.2E−07	−3.7
NGC3521	0.15	0.13+0.30−0.11	AP	61	0.26	MN	4.4E−07	−3.5
NGC3556 (M108)	0.023	⋅⋅⋅	P	81	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC3593	0.21	⋅⋅⋅	P	69	0.63	BB,[4]	2.6E−07	−3.7
NGC3623 (M65)	0.22	⋅⋅⋅	P	81	1.3	BB,[4]	1.4E−07	−3.8
NGC3627 (M66)	0.912	⋅⋅⋅	P	65	1.3	BB,[4]	5.6E−07	−3.5
NGC3998	280	<0.01	AP	36	60	S,[21]	3.6E−06	−2.7
NGC4026	0.4	0.20+0.41−0.20	AP	83	21	S,[22]	1.5E−08	−4.2
NGC4111	7.6	5.7+1.7−1.6	ARP	87	4.0	MS	1.4E−06	−3.2
NGC4136	0.15	0.31+0.52−0.31	AP	0	0.04	BB,[4]	2.9E−06	−3.0
NGC4138	200	7.7+1.9−1.6	AP	58	3.2	MS	4.9E−05	−2.2
NGC4203	18	1.4+1.2−0.8	AHRP	26	5.8	MS	2.4E−06	−3.1
NGC4258 (M106)	31	4.1+0.6−0.5	AP	71	3.9	M,[23]	6.1E−06	−2.7
NGC4314	0.068	0.19+0.65−0.19	AP	15	1.0	BB,[4]	5.2E−08	−4.1
NGC4321 (M100)	0.15	<1.5	ARP	37	0.45	MS	2.6E−07	−3.7
NGC4342	0.48	0.10+0.11−0.10	AP	58	33	S,[13]	1.1E−08	−4.2
NGC4343	0.25	⋅⋅⋅	P	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC4395	5.2	9.2+0.8−0.8	AHP	38	0.036	R,[24]	1.3E−04	−2.0
NGC4414	0.21	⋅⋅⋅	P	50	1.2	MS	1.4E−07	−3.8
NGC4419	1.8	<0.5	AP	75	0.8	MS	1.7E−06	−3.2
NGC4527	0.78	<0.7	AP	68	16.7	MS	3.6E−08	−4.1
NGC4548 (M91)	8.4	2.3+6.8−2.3	AP	37	3.6	MS	1.8E−06	−3.1
NGC4559	1.2	0.31+0.45−0.31	AP	69	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC4561	4.3	<0.05	AP	25	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC4564	0.48	0.14+0.37−0.14	AP	62	5.6	S,[17,20]	6.4E−08	−4.0
NGC4565	4.5	0.23+0.02−0.03	AP	90	2.9	MS	1.2E−06	−3.2
NGC4594 (M104)	17	0.18+0.03−0.03	AP	79	100	BB,[4]	1.3E−07	−3.8
NGC4636	0.21	<0.04	[25]	44	7.9	MS,[15]	2.0E−08	−4.2
NGC4670	0.79	0.09+0.49−0.09	AP	31	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC4697	0.03	<0.06	[26]	44	17	S,[17,20]	1.0E−09	−4.6
NGC4713	0.18	⋅⋅⋅	P	53	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC4725	0.54	1.6+0.4−0.4	AHBP	43	3.4	MS	1.2E−07	−3.9
NGC4736 (M94)	1.0	0.27+0.39−0.20	AP, [27]	33	2.2	MS	3.5E−07	−3.7
NGC4945	20000	425+25−25	[28]	90	0.14	M,[29]	0.11	∼10
NGC5055 (M63)	0.34	0.08+0.05−0.08	AP	55	0.87	MS	3.0E−07	−3.6
NGC5102	0.006	<1.5	AP	71	3.0	MS	1.6E−09	−4.5
NGC5128 (Cen A)	600	10.0+0.6−0.6	AP,[30]	43	20	G,[31]	1.9E−05	−2.5
NGC5194 (M51a)	200	560+400−160	[32,33]	64	0.71	MS	2.2E−04	−2.1
NGC5195 (M51b)	0.8	0.11+0.08−0.09	AP,[34]	46	3.9	MS	1.6E−07	−3.9
NGC5236 (M83)	0.26	0.10+0.14−0.06	AP,[35]	24	1.3	S,[36]	1.5E−07	−3.8
NGC5253	0.1	0.6+0.1−0.1	[37]	77	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC5457 (M101)	0.1	0.03+0.08−0.03	AP,[38]	0	0.24	MS	3.2E−07	−3.8
NGC5879	24	18+18−13	ARP	73	0.25	MS	7.4E−06	-2.9
NGC6503	0.086	0.5+1.7−0.5	AP	74	0.037	MS	1.8E−06	−3.1
NGC6946	0.1	0.06+0.10−0.06	AP	42	2.7	MN	2.9E−08	−4.1
NGC7013	11.4	6.5+3.1−2.6	AHP	76	0.54	MS	1.6E−05	−2.5
NGC7320	0.17	0.22+0.71−0.22	AP	60	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
NGC7331	0.52	0.09+0.21−0.09	AP	68	3.0	MS	1.4E−07	−3.8
NGC7457	0.18	⋅⋅⋅	P	56	0.35	S,[17,20]	4.0E−07	−3.5
PGC24175	0.87	0.04+0.10−0.04	AP	32	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
PGC50779	20	0.44+0.47−0.21	[39]	65	0.17	M,[40]	9.1E−05	−2.1

Notes. ^aEmitted luminosity in the 0.3–8 keV band, inferred from the best-fitting spectral model. ^bIntrinsic neutral-hydrogen column density, inferred from the best-fitting spectral model. For sources with less than 10 net counts, such estimates are not meaningful and no value is listed; the luminosity of these sources is estimated assuming Galactic LOS absorption only. ^cOur spectral models are coded as follows. P: power law with Galactic absorption; AP: power law with Galactic and intrinsic absorption; ARP: two-component model, with an absorbed optically thin thermal plasma and an absorbed power-law component; AHP: power law absorbed by both neutral hydrogen and an ionized absorber; AHRP: two-component model with an optically thin thermal-plasma component and a power-law component, absorbed by both neutral hydrogen and an ionized absorber. AHBP: two-component model with a disk blackbody and a power-law component, absorbed by both neutral hydrogen and an ionized absorber. When a reference number is given, we used spectral-analysis results from the literature. ^dInclination angle of the host galaxy, from de Vaucouleurs et al. (1991) ^eMethods used for estimating the nuclear BH masses (assuming that late-type galaxies contain BHs), as follows. S: stellar kinematics; G: gas kinematics; M: kinematics of water-maser clumps; R: reverberation mapping; MS: M–σ relation (Terashima et al. 2002); MN: M–Sersić index relation (Graham & Driver 2007); BB: BH mass–bulge mass relation (Dong & De Robertis 2006). ^fX-ray Eddington ratio, defined as L_0.3−8/L_Edd. ^gAccretion parameter required for the inferred X-ray luminosity, assuming the radiatively inefficient ADAF model; we used the grid of ADAF solutions plotted in Merloni et al. (2003). From our definition of $\dot{m} \equiv \dot{M}c^2/L_{\rm Edd}$, the Eddington luminosity corresponds to $\dot{m} \sim 10$. References. [1]: Weaver et al. 2002; [2]: Plucinsky et al. 2008; [3]: Gebhardt et al. 2001; [4]: Dong & De Robertis 2006; [5]: Bower et al. 2001; [6]: Young et al. 2001; [7]: Greenhill & Gwinn 1997; [8]: Irwin et al. 2002; [9]: Immler et al. 2003; [10]: Jenkins et al. 2008; [11]: Jímenez-Bailón et al. 2005; [12]: Sarzi et al. 2001; [13]: Cretton & van den Bosch 1999; [14]: Swartz et al. 2003; [15]: Merritt & Ferrarese 2001a; [16]: Emsellem et al. 1999; [17]: Gebhardt et al. 2003; [18]: Kormendy et al. 1998; [19]: Gebhardt et al. 2000; [20]: Pinkney et al. 2003; [21]: Bower et al. 2000; [22]: Gultekin et al. 2009; [23]: Herrnstein et al. 1999; [24]: Peterson et al. 2005; [25]: Posson-Brown et al. 2006; [26]: Wrobel et al. 2008; [27]: Pellegrini et al. 2002; [28]: Done et al. 2003; [29]: Greenhill et al. 1997; [30]: Evans et al. 2004; [31]: Thatte et al. 2000; [32]: Terashima & Wilson 2001; [33]: Fukazawa et al. 2001; [34]: Terashima & Wilson 2004; [35]: Soria & Wu 2002; [36]: Marconi et al. 2001; [37]: Summers et al. 2004; [38]: Pence et al. 2001; [39]: Smith & Wilson 2001; [40]: Greenhill et al. 2003

Download table as:  ASCIITypeset images: 1 2

Zoom In Zoom Out Reset image size

3.3. Physical Identification of the Nuclear X-ray Source Population

The main problem affecting X-ray surveys of low-luminosity (L_X ≲ 10³⁹ erg s⁻¹) nuclear BH activity is the possibility of confusion with X-ray binaries, in the overlapping range of luminosities. A typical example is the nuclear X-ray source in M 33 (Gebhardt et al. 2001; Dubus et al. 2004). Detailed investigations of individual galactic nuclei (Ghosh et al. 2008) have highlighted the difficulties and ambiguities of any such identifications. On a statistical basis, the slope and normalization of the cumulative luminosity distribution in our extended sample (Figure 4) are also consistent with the luminosity distribution of a population of high-mass X-ray binaries, in a galaxy or ensemble of galaxies with a total star formation rate (SFR) of ≈25M_☉ yr⁻¹ (Grimm et al. 2003, Equation (5)), for which we would also expect an integrated Hα luminosity L_Hα ≈ 3 × 10⁴² erg s⁻¹. This could be a priori consistent with our sample, if the majority of our 187 galaxies were in the high-luminosity tail of the H ii nuclei distribution (Ho et al. 1997c). However, there are a few arguments in support of a nuclear BH identification for the majority of our 86 nuclear X-ray sources.

First, we limited our sample to X-ray sources located within 1'' (typically, ≈30–70 pc for the majority of our galaxies) of the independently known nuclear position. When we considered the annulus between 1'' and 5'' around the nuclear position (a projected area 24 times larger), we found only ≈50 sources, identified with the same criteria used for the 86 nuclear sources. Such a concentration of pointlike X-ray sources exactly at the nucleus but not in its immediate surroundings is much cuspier than typical projected stellar densities (except when the galaxy has a nuclear star cluster) or densities of the SFR. And the argument is even stronger if we consider that high-mass X-ray binaries may easily disperse over ≳100 pc in their lifetime, owing to proper motion; thus, it would be even more unlikely to find most of them exactly at the nuclear position.

Second, our cumulative luminosity distribution is consistent with an unbroken power law up to ∼10⁴² erg s⁻¹, where we run out of galaxies in our sample (Figure 4). Instead, the luminosity distribution of high-mass X-ray binaries (including ULXs) shows a downturn at ≈10⁴⁰ erg s⁻¹ (Grimm et al. 2003; Swartz et al. 2004). A scenario where our nuclear-source population is composed of low-luminosity AGNs for L_X ≳ 10⁴⁰ erg s⁻¹ and high-mass X-ray binaries for L_X ≲ 10⁴⁰ erg s⁻¹, coincidentally with the same normalization, appears highly contrived.

Third, the majority of the nuclear sources (57 out of 86) are detected in the earlier-type galaxies of our sample (E to Sb), that is, in galaxies where the nuclear region is part of a massive, old spheroidal component (Section 3.1). If the contamination from high-mass X-ray binaries were significant, we would expect to find more sources in late-type spirals. Ellipticals and massive spheroidals do have a high stellar density in their cores, with old populations, so they may have low-mass X-ray binaries LMXBs near the nuclear position. But a significant contribution of LMXBs to our source population above 2 × 10³⁸ erg s⁻¹ is also ruled out, because their cumulative luminosity distribution would be much steeper (Gilfanov 2004; Swartz et al. 2004). In addition, we used the following argument to estimate the possible contribution of LMXBs in the nuclear region of ellipticals and spheroidal bulges. From surface brightness profiles (Gebhardt et al. 2003; Lauer et al. 1995), we can estimate the total luminosity and hence the total stellar mass of characteristic galaxy types within, say, 15 (∼50–100 pc for our distance range), that is, the radius of our source extraction region around each nuclear position. This is of course a function of the Hubble type and galaxy history, among other things, but M ∼ 10⁸M_☉ is an order-of-magnitude estimate for massive ellipticals and a conservative upper limit for spheroidal bulges. From the population studies of Gilfanov (2004), we expect ∼0.02 sources with X-ray luminosities ≳2 × 10³⁸ erg s⁻¹, from an old stellar population with a stellar mass ∼10⁸M_☉. A similar estimate can be obtained directly from the growth curves plotted in Figure 3 of Gilfanov (2004) for a sample of nearby elliptical galaxies and spheroidal bulges; these plots also suggest the presence of ≲0.1 sources above 10³⁸ erg s⁻¹ within 15, scaled to the range of distances of our sample. Since our main statistical results are based on sources above a completeness limit ≈ 2 × 10³⁸ erg s⁻¹, the contamination of at most two or three X-ray binaries in our whole sample of early type galaxies is not a significant problem.

Taken together, the previous arguments strongly support our identification of the X-ray source population as active nuclear BHs—without ruling out the possibility of stellar-mass interlopers in some limited cases, such as M 33, where stellar kinematics suggest a BH mass less than 1500M_☉ (Gebhardt et al. 2001). The distribution of X-ray photon indices ≈1.5–2.0 is also in the typical range for AGNs, and rules out, for example, thermal emission from compact starburst nuclei.

Another possibility we may consider is that each of our nuclear sources could instead be the integrated emission from many, fainter sources in a very compact star-forming nucleus. After all, many of our sample galaxies are classified as H ii nuclei. This scenario is already implausible given the predominant detection of X-ray cores in earlier-type galaxies, as noted earlier. But we can test this possibility more quantitatively by comparing the nuclear X-ray and Hα luminosities. Fifty-three of the 86 galaxies with an X-ray core in our sample are also included in the optical spectroscopic Palomar survey of Ho et al. (1997a), which provides nuclear Hα fluxes. Both the Hα and the 0.3–8 keV luminosities of a star-forming region are proportional to the current or recent SFR (assuming that it has stayed approximately constant over the last ∼10 Myr): L_Hα ≈ 1.3 × 10⁴¹ SFR (M_☉ yr⁻¹) erg s⁻¹ (Kennicutt 1998) and L_0.3−8 ≈ 1.0 × 10⁴⁰ SFR (M_☉ yr⁻¹) erg s⁻¹ (Grimm et al. 2003). Thus, we expect L_0.3−8 ∼ 0.1L_Hα for a starburst-dominated compact nucleus. Instead, we expect L_0.3−8 ∼ 1–25L_Hα for typical low-luminosity AGNs (Ho 2008; Flohic et al. 2006). The results for our sample (Figure 5) are more consistent with the AGN scenario.

Zoom In Zoom Out Reset image size

3.4. Intrinsic Absorption and Luminosity

From our spectral fitting of individual sources, we found that the intrinsic column density N_H appears to be correlated with the unabsorbed X-ray luminosity (Figure 6, left panel); more luminous sources tend to be more obscured, while most of the fainter sources are consistent with column densities ≲ 10²¹ cm⁻² or with only line-of-sight (LOS) Galactic absorption. There is also a dependence on the viewing angle of the host galaxy (Figure 6, right panel), as expected, but it is less strong than the luminosity dependence. In terms of the unified AGN classification scheme, we found few or no "Type 2" nuclear BHs below an X-ray luminosity ≈ 10⁴⁰ erg s⁻¹. Only ≲10% of the nuclear sources with L_X≲ a few 10³⁹ erg s⁻¹ are obscured by a neutral column density N_H > 10²² cm⁻². But obscured sources represent two-thirds (10 out of 15) of those with L_X ≳ 10⁴⁰ erg s⁻¹. This apparent trend of higher absorption at higher luminosities is opposite to what seems to happens in more luminous AGNs, with unabsorbed X-ray luminosities ∼10⁴²–10⁴⁶ erg s⁻¹ (Hasinger 2008; Gilli et al. 2007; Ueda et al. 2003); see also Hopkins et al. (2009) for an alternative explanation of the luminosity dependence.

Zoom In Zoom Out Reset image size

Before attempting a physical explanation for the observed trend, we need to check that it is not the result of observational bias. We have already excluded from the absorption–luminosity plot (Figure 6) all sources detected with ⩽10 counts. One possibility might be that we lose all sources with N_H ≳ 10²² cm⁻² and L_X ≲ 10³⁹ erg s⁻¹ because they would not have enough counts to be detected. However, using simulated X-ray data and spectral models, we verified that this is not the case; even for N_H ≈ 10²³ cm⁻², most sources with a typical photon index Γ ≈ 1.7 and L_X ≈ 10³⁹ erg s⁻¹ would have enough counts in the 2–8 keV band to be detected. An alternative possibility is that most of the fitted values of N_H in faint sources have been underestimated (correspondingly, their photon indices and intrinsic luminosities would also have been underestimated). To test this scenario, we stacked the spectra of all sources with ⩽50 counts; this is roughly equivalent to stacking all spectra of nuclear sources with L_X ≲ 10³⁹ erg s⁻¹ (most of the galaxies in our sample are at distances ∼10–15 Mpc). We then fitted the coadded spectrum with an absorbed power-law model; we obtain (Figure 7) that the best-fitting Γ ≈ 1.7, and N_H,tot ≈ 5 × 10²⁰ cm⁻², which is only slightly higher than the average Galactic absorption for the sources in the stacked sample. We conclude that the coadded spectrum confirms a very low intrinsic absorption for the least luminous nuclear BHs.

Zoom In Zoom Out Reset image size

Furthermore, we conducted a Monte Carlo simulation to test whether the apparent absorption–luminosity correlation could have been spuriously introduced during spectral fitting (e.g., due to an overestimate of the intrinsic luminosity for sources with higher N_H). We assumed that the unabsorbed flux for a simulated sample of 100 sources has a uniform distribution (in log scale) from 10⁻¹³ to 10⁻¹⁰ erg cm⁻² s⁻¹ in the 0.3–8 keV band (similar to the range of fluxes typical of our real sample of nuclear sources), and their intrinsic N_H has a uniform distribution (also in log scale) from 10²⁰ to 10²⁴ cm⁻² (Figure 8). In other words, we assumed no intrinsic correlation between luminosity and N_H. We also assumed an absorbed power-law spectrum with Γ = 1.7 for every source. We then fitted the simulated spectra (using the Cash statistics) with absorbed power-law models, leaving Γ and N_H as free parameters, and we estimated the unabsorbed fluxes implied by the best-fitting model for each source. The fitted parameters do not show any correlation between unabsorbed fluxes and N_H or any significant bias in the inferred fluxes (Figure 8).

Zoom In Zoom Out Reset image size

We conclude that the absence of Type 2 low-luminosity nuclei and the positive correlation between intrinsic absorption and luminosity up to L_X ≈ 10⁴² erg s⁻¹ are probably real. This is opposite to the negative correlation between absorption and luminosity known to exist at higher luminosities (Hasinger 2008; Gilli et al. 2007; Ueda et al. 2003). We also searched for a possible galaxy-morphology dependence of the intrinsic column density, for example, whether late-type spirals have systematically higher absorption than ellipticals. We find that all Hubble types appear dominated by unobscured sources, although the small number of sources in each class does not permit us to draw stronger conclusions (Figure 9). The nuclei of early type spirals in our X-ray detected sample seem to be the most obscured: 15 of the 42 nuclear sources detected in early type spirals have N_H ⩾ 10²² cm⁻²; only 3 of the 20 nuclear sources in late-type spirals, and none of the 6 elliptical nuclei have N_H ⩾ 10²² cm⁻². However, this may still be due to small-number statistics in each class. A larger X-ray sample of galaxies will be needed to study the Hubble-type dependence for a given range of luminosities and BH masses.

Zoom In Zoom Out Reset image size

3.5. Eddington Ratios

Among the 86 galaxies with an X-ray core in our extended sample, 31 already have a measurement of their nuclear BH mass in the literature, based on stellar kinematics, gas kinematics, water masers, or reverberation mapping (see detailed references in Table 4). For another 31 galaxies, stellar velocity dispersions of their cores are available from Hyperleda,⁹ with typical uncertainties ∼10%–20%; in these cases, we used the M_BH–σ relation (Tremaine et al. 2002) to estimate their BH masses. For three other cases, only photometric observations are available: we constrained their BH masses via the M_BH–S$\acute{\rm e}$rsic index relation (Graham & Driver 2007). Thus, we have 65 nuclear BHs in our sample with a mass estimate out of 86 candidate X-ray cores (Table 4).

The elliptical galaxies in our sample have larger inferred BH masses (M_BH ∼ 10⁸M_☉) and low X-ray luminosities (L_X ≲ 10³⁹ erg s⁻¹), as expected. Late-type spirals also have low X-ray luminosities (L_X ≲ 10³⁹ erg s⁻¹) but the lowest BH masses (M_BH ∼ 10⁵–10⁷M_☉). The most luminous nuclear BHs in the local universe (L_X ∼ 10⁴¹–10⁴² erg s⁻¹) are found in early type spirals (Figure 10, top panel). X-ray nuclei of barred galaxies are less luminous than those of nonbarred galaxies, for a given range of BH masses (Figure 10, middle panel). This is partly due to the higher fraction of strongly barred galaxies in late-type spirals, which tend to have weaker nuclear emission. Optically classified Seyferts have the most luminous X-ray nuclei, as expected (Figure 10, bottom panel). There is only a very weak positive correlation (slope ≈ 0.26 and correlation coefficient ≈ 0.19) between the BH mass and X-ray luminosity. If the X-ray luminosity scaled with the Bondi accretion rate (Hoyle & Lyttleton 1939) onto the nuclear BH, we would expect L_X ∼ M²_BHρ(∞)/c³_s(∞), where ρ(∞) and c³_s(∞) are the gas density and sound speed outside the accretion radius. The higher gas density in the nuclear region of spiral galaxies compensates for their lower BH masses. In contrast, the lower abundance of gas available for accretion in elliptical galaxies is offset by higher BH masses. Because the luminosity is almost independent of the BH mass, there is a negative correlation (slope ≈ −0.41 and correlation coefficient ≈ −0.49) between BH masses and X-ray Eddington ratios r_Edd ≡ L_0.3−8/L_Edd (Figure 11). In local-universe ellipticals, L_0.3−8 ∼ 10⁻⁹–10⁻⁸L_Edd; in late-type spirals, L_0.3−8 ∼ 10⁻⁵ to 10⁻⁴L_Edd. The only outlier is the late-type, barred (SBcd) Seyfert-2 galaxy NGC 4945, perhaps the only "true" AGN in our sample (L_0.3-8 ≈ 2 × 10⁴³ erg s⁻¹ ∼0.1L_Edd). This is also the galaxy with the highest intrinsic absorption (N_H ∼ 10²⁴ cm⁻²); see Done et al. (2003) and Itoh et al. (2008) for detailed X-ray studies of this AGN.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Estimating the mass accretion rate from the observed luminosities requires the assumption of a model for the radiative and total efficiency. In our local-universe sample, even the most luminous nuclei (except for NGC 4945) are likely to be in the low-radiative-efficiency regime, which is thought to set in for L_bol ∼ 10L_X ≲ 0.01L_Edd, by analogy with stellar-mass BHs (Esin et al. 1997; Jester 2005). We adopted the ADAF scenario, which includes contributions to the emitted flux from disk blackbody (truncated outer disk), synchrotron, bremsstrahlung, and inverse Compton (Narayan et al. 1997, 1998). ADAF spectral models scale with the BH mass and the dimensionless accretion parameter $\dot{m} \equiv L_{\rm bol}/\left(\eta L_{\rm Edd}\right) = \dot{M} c^2/L_{\rm Edd} \propto \dot{M}/\dot{M}_{\rm Edd}$, where η is the radiative efficiency (η ∼ 0.1 for efficient accretion).¹⁰ Other physical information is included in three (dimensionless) parameters: the ratio of gas to magnetic pressure, the viscosity parameter, and the fraction of viscous heating that goes into the electrons. A useful table of band-limited X-ray luminosities for a grid of ADAF spectral models, as a function of the BH mass and accretion parameter (with standard assumptions for the other three parameters), was calculated by Merloni et al. (2003). We used their grid values and interpolations to estimate $\dot{m}$ of our sample nuclei, from their observed X-ray luminosities and indirectly inferred BH masses. We obtain a range of accretion parameters $\dot{m} \sim 10^{-5}$–10⁻², and we have plotted them as a function of the BH mass, Hubble type, and optical spectroscopic classification (Figure 12).

Zoom In Zoom Out Reset image size

We studied the X-ray nuclear activity of nearby galaxies (distance <15 Mpc), in a range of luminosities intermediate between low-luminosity AGNs and "normal" nonactive galaxies. More specifically, we chose a complete sample of optically/IR-selected Northern galaxies, as defined in Swartz et al. (2008); most of the galaxies in this sample were observed by Chandra/ACIS for a ULX survey. We then extended that sample with another ∼70 galaxies (also at distances less than 15 Mpc) with publicly available Chandra/ACIS data; the extended sample contains 187 galaxies. The main results presented in this paper are as follows.

1. We made a census of weakly active nuclei in the local neighborhood, down to a completeness limit L_X ≈ 2 × 10³⁸ erg s⁻¹ in the 0.3–8 keV band. Eighty-six out of 187 galaxies have a pointlike nuclear X-ray source, coincident (within 1'') with the radio or infrared nuclear position. The presence of an X-ray core depends strongly on the Hubble type: ≈60% of early type galaxies (E to Sb) contain an X-ray core, but only ≈30% of later-type galaxies. About 90% of optically classified Seyfert and LINERs have a nuclear X-ray source with L_X ≳ 2 × 10³⁸ erg s⁻¹; this fraction drops to ≈60% for transition objects and ≈30 % for H ii nuclei. The AGN demographics from our Chandra survey is consistent with the AGN demographics inferred from optical spectroscopic studies, for example the Palomar survey (Ho 2008, and references therein). Our Chandra survey suggests that pointlike nuclear X-ray emission is a reliable indicator of BH activity in normal galaxies, especially in cases where optical signatures of BH accretion may be swamped by the surrounding stellar emission.

2. Spiral and elliptical galaxies within 15 Mpc are detected with a continuous range of nuclear luminosities from ∼10³⁸ erg s⁻¹ to ∼10⁴² erg s⁻¹. There is no gap between low-luminosity AGNs and BH activity in normal galaxies. The cumulative luminosity distribution can be fitted by a power law with a slope of ≈−0.5. Because of the relatively small number of galaxies in our sample, we cannot yet accurately determine whether the luminosity distribution of these faint nuclei matches the slope and normalization of the luminosity distribution of fully fledged AGN (L_X > ∼ 10⁴² erg s⁻¹). However, we are currently studying a larger Chandra sample of galaxies (within 40 Mpc) and we will address this issue in a follow-up paper.

3. For each individual X-ray core with luminosities ∼10³⁸–10³⁹ erg s⁻¹, it is always very difficult to distinguish between a nuclear BH and an unrelated, luminous X-ray binary in the nuclear region. However, on a statistical basis, we have discussed various reasons why we think that the majority of detected sources are nuclear BHs rather than X-ray binaries (Section 3.1). Most of the X-ray sources are exactly coincident with the nuclear position, and there are much fewer sources in the annulus between 1'' and 5'' from the radio/optical nucleus. Besides, the luminosity distribution has no hint of a break at L_X ≈ 10⁴⁰ erg s⁻¹, as we would expect from a population of high-mass X-ray binaries. The ratio between nuclear Hα luminosity (when available, from the Palomar survey) and X-ray luminosity is more typical of low-luminosity AGNs than of star-forming regions with young X-ray binaries.

4. We fitted the spectra of each source, assuming absorbed power-law models with two free parameters: the photon index and the intrinsic absorption. The photon index is consistent with the expected value for AGNs (Γ ≈ 1.5–2). The intrinsic column density N_H is positively correlated with the emitted X-ray luminosity; among the population of fainter nuclear BHs, very few or none can be classified as Type 2 (highly obscured). This is the opposite of the trend known for luminous AGNs, with L_X > 10⁴² erg s⁻¹ (Hasinger 2008; Gilli et al. 2007; Ueda et al. 2003). It is still not clear what produces the absorption (a geometrically thick parsec-scale torus? An optically thick disk wind?), and hence we cannot determine from the data available what causes the fraction of obscured sources to be highest for sources with L_X ∼ 10⁴²–10⁴³ erg s⁻¹, and to decrease at both lower and higher nuclear BH luminosities. At high luminosities, it was suggested that the thick torus gets progressively evaporated or ablated by the radiation flux from the central object (Hasinger 2008; Menci et al. 2008). On the other hand, X-ray faint nuclei have less gas available for accretion. They may not be surrounded by a torus at all; or the torus may collapse and become geometrically thin, so that only galaxies seen perfectly edge-on would be classified as Type 2; or, instead, the dramatic decrease in absorption may be caused by the suppression of the optically thick disk wind. Alternatively, it was suggested (Hopkins et al. 2009) that a large fraction of sources with L_X ∼ 10⁴²–10⁴⁴ erg s⁻¹ have been erroneously classified as obscured, because of optical dilution effects and because of the transition from a standard-disk accretion geometry to a radiatively inefficient flow. Both effects would make these sources less prominent or invisible in the UV/optical band, and slightly harder in the X-ray band, thus making them appear "obscured" even though they are not. Based on these arguments, it was proposed that the fraction of truly obscured sources can be as low as 20%, independent of luminosity (Hopkins et al. 2009). However, an opposite result was obtained by Reyes et al. (2008), based on an optically selected sample of quasars from the SDSS; they found that at least half of the quasars in the nearby universe (z ≲ 0.8) are truly obscured, particularly in the high-luminosity population. In our sample of sources, the absorbing column densities are estimated directly from the fitted X-ray spectra, and do not rely on hardness ratios or optical/X-ray flux ratios, thus reducing the bias discussed by Hopkins et al. (2009). Thus, we suggest that the high fraction (10 out of 15) of obscured sources at X-ray luminosities >10⁴⁰ erg s⁻¹ (0.3–8 keV band) may be really due to a higher density of absorbing gas or dust around the nuclear BH (whatever its geometry) in that moderate luminosity range. The fraction of obscured nuclei seems to depend more directly on luminosity rather than the Hubble type. In our sample, slightly more nuclei detected in early type spirals seem to be obscured (N_H ⩾ 10²² cm⁻²), compared with the obscured fractions in ellipticals and late-type spirals; however, we do not have enough galaxies to draw strong conclusions. We are planning further work with a larger sample of galaxies to address this issue.

5. Having argued that the apparent luminosity dependence of the obscured fraction of nuclei is truly due to changes in the column density of the absorbing medium, we can still look for a causal link between such changes and the disappearance of the standard disk at low accretion rates (or other transitions in the geometry of the accretion flow). For example, the disk may disappear when it is no longer fed by a large torus or, vice versa, optically thick winds may be suppressed when the standard disk turns into an optically thin ADAF. It was already noted (Ho 2008) that the decrease or suppression of the intrinsic absorption at low luminosities often coincides with the disappearance of optical signatures of a standard accretion disk. If the apparent or real changes in the obscuration fraction are due to a sharp standard-disk/ADAF transition at $\dot{m} \sim 0.1$, we should not expect a trend in our sample, because all but one of our sources are almost certainly below this threshold, in the radiatively inefficient regime. Instead, we see changes in the fraction of obscured sources over the X-ray luminosity range ∼10³⁹–10⁴² erg s⁻¹ and $\dot{m} \sim 10^{-4}$ to 10⁻². This suggests a more gradual evolution in the amount of absorbing material below the radiatively inefficient threshold or perhaps a more gradual disappearance of the outer accretion disk at low accretion rates. It was also recently suggested (Zhu et al. 2008; Wang & Zhang 2007) that a dusty torus (produced during major galaxy mergers or via secular evolution processes) can provide the main source of fuel in a self-regulated way: when it is completely evaporated by the radiation from the central source, the AGN phase turns off and the supermassive BH becomes inactive; this scenario is among those consistent with our observational findings.

6. We confirm that there is no positive correlation between the presence of a bar and the X-ray luminosity of the nucleus. Large-scale bars are highly effective in delivering gas to the central few hundred parsecs of a spiral galaxy and therefore enhance the probability and rate of star formation in the nuclear region (Heller & Shlosman 1994). However, it was observed from optical/IR surveys (Ho et al. 1997d; Hunt & Malkan 1999; Laurikainen et al. 2004) that the presence of a bar seems to have no impact on either the frequency or strength of AGN activity with the possible exception of narrow-line Seyfert 1 galaxies, which may be preferentially associated with barred spirals (Ohta et al. 2007). In fact, our results suggest a negative correlation, with strongly barred galaxies having a less active nuclear BH than nonbarred/weakly barred galaxies, for a given range of BH masses and Hubble types (we do not have narrow-line Seyfert 1s in our sample). This preliminary but intriguing result will have to be tested more strongly on our larger Chandra sample we are currently studying. If confirmed, it may suggest that there is less gas reaching the supermassive BH when there is circumnuclear star formation (i.e., in most strongly barred galaxies). In this scenario, galaxies may cycle through phases of dominant nuclear star formation (when a bar is present) and phases dominated by nuclear BH accretion. One possible explanation could be that nuclear star formation in spiral galaxies actively disfavors gas inflows toward the nuclear BH—most of the gas not used for star formation may be blown away by supernova-powered outflows. In addition, there may be a mismatch between the characteristic timescales of bar formation and disruption (≈a few 10⁸ yr; Combes & Elmegreen 1993) and the longer timescales of nuclear BH accretion. In this scenario (Hunt & Malkan 1999), bars and associated nuclear star formation would appear first and nuclear activity later after the bar-driven star formation has subsided.

7. There is an anticorrelation between the BH mass and the Eddington ratio; elliptical galaxies in the local universe have higher BH masses but much lower accretion parameters, corresponding to X-ray luminosities ∼10⁻⁸L_Edd. When late-type spirals have an X-ray nuclear source, their luminosities are ∼10⁻⁴L_Edd, because they have a larger supply of gas. In any case, all classes of nuclear sources in our sample are expected to be in the radiatively inefficient regime, dominated either by energy advection (e.g., ADAF) or by nonradiative output channels (jets). We are planning further work to determine the radio core luminosity of the nuclear X-ray sources detected in our sample.

We thank Luis Ho for stimulating discussions, particularly during his visit to Beijing, and the anonymous referee for his/her insightful suggestions. S.N.Z. acknowledges partial funding support by the Yangtze Endowment from the Ministry of Education at Tsinghua University, Directional Research Project of the Chinese Academy of Sciences under project no. KJCX2-YW-T03, the National Natural Science Foundation of China under grant Nos. 10521001, 10733010, 10725313, and 973 Program of China under grant 2009CB824800. R.S. acknowledges a UK–China Fellowship for excellence and a Leverhulme Trust Fellowship (through University College London); he also thanks the School of Physics at the University of Sydney for their hospitality and support during the completion of this work.