Chandra Limits on X-Ray Emission Associated with the Supermassive Black Holes in Three Giant Elliptical Galaxies

Elliptical galaxy nuclei are the sites of the largest black holes known but typically show little or no nuclear activity. We investigate this extreme quiescence using Chandra X-Ray Observatory observations of the giant elliptical galaxies NGC 1399, NGC 4472, and NGC 4636. The unique Chandra imaging power enables us to place upper limits of 7.3, 15, and 28 × 10-9LEdd for the ~108-109 M☉ black holes in NGC 1399, NGC 4472, and NGC 4636, respectively. The corresponding radiative efficiencies in this band are 4.1, 24, and 620 × 10-6 using Bondi accretion rates derived from the Chandra hot interstellar gas surface brightness profiles. These limits are inconsistent with basic advection-dominated accretion flow models for NGC 1399 and NGC 4472, indicating accretion onto the black hole at ≲10% of the Bondi rate.


INTRODUCTION
Knowledge about the frequency and masses of supermassive black holes (SMBHs) in galactic nuclei and the characteristics of their associated radiation deepen our understanding of galaxy formation and clarify its connection with the active galactic nucleus (AGN) phenomenon (e.g., Blandford 1999). While the intense energetics and rapid variability in the most luminous active galaxies form the empirical foundation of the SMBH/ AGN paradigm, such extreme objects are rare-especially in the local universe.
Nevertheless, there is now a consensus that most, if not all, galaxies host SMBHs. This is based, in part, on the high frequency of low-level nuclear activity in nearby galaxies (Ho, Filippenko, & Sargent 1997) and on statistical arguments connected to the cosmic X-ray background (e.g., Fabian & Iwasawa 1999). However, the strongest evidence for the ubiquity of SMBHs comes from dynamical studies. Magorrian et al. (1998) found that, of 32 galaxies that could be fitted with an axisymmetric dynamical model, 26 (29) require the presence of a central massive (∼10 8 -10 10 M ) dark object at 95% (68%) , confidence. More unambiguous and accurate measurements based on Hubble Space Telescope (HST) stellar or ionized gas spectroscopy confirms the presence of SMBHs while deriving systematically lower SMBH masses (Gebhardt et al. 2000a;Merritt & Ferrarese 2001a).
Although kinematically studied luminous AGNs are found to host SMBHs (Gebhardt et al. 2000b;McLure & Dunlop 2000), the converse is not true: activity in most SMBH host galaxies is modest or absent. This can be quantified by considering the ratio of the nuclear luminosity to the Eddington luminosity ( , where the black hole 38 L p 1.25 # 10 M Edd SMBH mass is measured in solar units). The most powerful M SMBH AGNs have ratios on the order of unity, as expected from the theory of thin disks accreting at their maximal rate. However, this ratio is less than 10 Ϫ7 and 10 Ϫ4 , respectively, in the two 1 Also with the University of Maryland Department of Astronomy. 2 Also with the Universities Space Research Association. most securely measured SMBHs: Sagittarius A* at the Galactic center Reid 1999) andNGC 4258 (Gammie, Narayan, &Blandford 1999). If these SMBHs accrete at the Bondi (1952) rate, they must do so in a fundamentally different-and much less radiatively efficient-way than in luminous AGNs. The theory of advection-dominated accretion flows (ADAFs; e.g., Narayan & Yi 1994) provides a predictive formalism for explaining the observed quiescence in the nuclei of Sgr A*, NGC 4258, and other systems (Fabian & Rees 1995;Mahadevan 1997).
Elliptical galaxies are the sites of the most luminous quasars and also host the most extreme cases of quiescent SMBHs. Although they contain nuclear black holes with the highest estimated masses and are observed to contain an ample supply of fuel in the form of X-ray-emitting hot interstellar gas, they often emit less than (Loewenstein et al. 1998;Di Ϫ6 10 L Edd . Radio/X-ray observations of some systems are inconsistent with predictions of the simplest models of an ADAF accreting at the Bondi rate but can be accommodated by variations that include outflow (Blandford & Begelman 1999;Di Matteo et al. 1999) or convection (Narayan, Igumenshchev, & Abramowicz 2000;Quataert & Gruzinov 2000).
Prior to the launch of the Chandra X-Ray Observatory, Xray searches for nuclear point sources in elliptical galaxies were complicated, in the imaging domain, by centrally concentrated hot gas that dominates the soft X-ray flux and, in the spectral domain, by an extended hard component consisting of X-ray binaries. Thus, the previous sensitivity to SMBH-associated Xray emission, for even the nearest giant elliptical galaxies, was ∼10 40 ergs s (Roberts & Warwick 2000;Sulkanen & Bregman Ϫ1 2001) since neither the ASCA SIS/GIS nor the ROSAT PSPC/ HRI had the necessary combination of sensitivity and spatial and spectral resolution to probe more deeply.  claim to detect a very hard nuclear X-ray component in six elliptical galaxies with characteristics consistent with ADAF/outflow models, given the observed radio  on the X-ray centroids for NGC 1399 and NGC 4472 and on the near nuclear source for NGC 4636. emission (Di Matteo et al. 1999). However, these results are based on a nonunique decomposition of ASCA spectra.
Because of the presence of bright stellar cores and dust, the best approach for constraining the level of SMBH activity is via high angular resolution X-ray observations (e.g., George et al. 2001): for a typical AGN X-ray/optical ratio, the Chandra detection threshold corresponds to a B magnitude of ∼24, while elliptical galaxy central surface brightnesses range from ∼11 to 16 mag arcsec . This is further demonstrated by deep Chan-Ϫ2 dra surveys in combination with optical follow-up. Lowluminosity AGNs, most of which are not optically detected, dominate the X-ray background, and nonthermal emission is found to be common-at any given time, ∼10% of all bright galaxies host ergs s AGNs (Mushotzky et al. 2000;41 Ϫ1 L տ 10 X Barger et al. 2001). There is also a high frequency of nuclear X-ray point sources associated with local AGNs, although often at levels that overlap in luminosity with X-ray binaries . Chandra has detected unresolved nuclear emission associated with quiescent SMBHs in M31 at ∼ Ϫ9 5 # 10 L Edd (Garcia et al. 2000

Observations, Data Reduction, and Analysis
NGC 1399, NGC 4472, and NGC 4636 were observed with the Chandra Advanced CCD Imaging Spectrometer chip S (ACIS-S) detector for 40-60 ks during Cycle 1. Results on the diffuse hot gas emission and the X-ray binary population, and a more detailed data reduction description, are included elsewhere (e.g., Angelini, Loewenstein, & Mushotzky 2001). The latest revised standard pipeline processing was used for NGC 4472 and NGC 4636, while an earlier version in combination with additional bad pixel removal, gain map correction, and quantum efficiency map correction was used for NGC 1399. The final useful exposure times are 56, 40, and 52 ks for NGC 1399, NGC 4472, and NGC 4636, respectively. NGC 1399 was approximately centered on the S3 chip; NGC 4636 and NGC 4472 were positioned at the standard ACIS-S aimpoint. The 1.25-7 keV images of the central are shown for all 50 # 50 three galaxies in Figure 1. The positions of the optical nuclei from archival HST Near-Infrared and Multi-Object Spectrometer (NICMOS; NGC 4472 and NGC 4636) or Wide Field Planetary Camera 2 (WFPC2; NGC 1399) images are marked. The brightest point sources in the images are easily visible and are indicated in the images; their count rates are in the 0.005-0.008 counts s range. The bright X-ray core in NGC Ϫ1 4472 is resolved and has a thermal spectrum. Only in the case of NGC 4636 is there a possible coincidence of the nucleus with a Chandra X-ray source (just to the southwest of the nuclear position indicated in the right panel of Fig. 1), with a 0.3-7 keV count rate of ∼0.004 counts s (see below). The  Figure 2. For illustration purposes, the NGC 4636 profile is centered on the near-nuclear source. No excess emission from unresolved nuclear emission is apparent for the other two galaxies.
The position of the nucleus in NGC 1399 is identified based on the registration of globular cluster X-ray sources with their WFPC2 counterparts (Angelini et al. 2001); the resulting HST and Chandra centroids are consistent. Similarly, the nucleus of NGC 4636 is located based on registration with a bright NICMOS source and with the USNO catalog (Monet et al. 1998) and is found not to be consistent with the position of the near-nuclear source (Fig. 1, right panel). For NGC 4472, we adopt the unregistered HST centroid that is consistent with the peak in the soft thermal emission seen in the broadband Chandra image (Fig. 1, middle panel). The 3 j upper limits to any nuclear X-ray point-source emission are derived following Gehrels (1986). Counts are accumulated in inner ("source") and outer ("background") concentric annuli with radii that are varied to obtain the tightest upper limit. The Chandra calibration database enclosed energy function at the appropriate offaxis angle is used, and an exposure map correction is applied for NGC 1399, which was centered off-axis. The assumed position of the SMBH is allowed to vary in a box (cor-3 # 3 responding to a very conservative adopted Chandra/HST relative systematic pointing uncertainty) centered on the assumed nuclear positions, and the highest upper limit in this box is adopted. These boxes encompass the peaks of the diffuse emission for NGC 4472 and NGC 1399. The diffuse hot gas emission is implicitly assumed to be locally flat, on average, when treated as background, consistent with the flat central surface brightness profiles (Fig. 2). Luminosity upper limits are derived in both the 0.3-7 and 2-10 keV energy bands assuming a slope 1.5 power-law spectrum-the limits increase by ∼2 for slope 0.5. Due to reduced dilution from the diffuse hot gas emission, the 2-10 keV count rate limits are 3-5 times lower than in the softer band; however, because of the declining effective area, the energy flux upper limits are comparable. The 0.3-7 keV detection limit varies from a few times 10 37 ergs s in regions Ϫ1 of low surface brightness hot interstellar medium (ISM) emission to a few times 10 38 ergs s in the core of NGC 1399.

Ϫ1
The 2-10 keV limits-all derived within 1Љ of the optical nucleus-are displayed in Table 1; these correspond to 28, 23, and 16 counts for NGC 1399, NGC 4472, and NGC 4636, respectively-comparable to the brightest sources in the subimages shown in Figure 1. Table 1 also includes less conservative, but perhaps more realistic, limits calculated by fixing the positions at the assumed nuclear positions described above.
The inner gas density profile is estimated by fitting the pointsource-subtracted surface brightness profiles with b-models, out to some maximum radius, followed by an analytical deprojection with conversions obtained from spectral fits to the inner regions (best-fit temperatures , 0.8, and 0.6 keV for kT p 0. 8 NGC 1399, NGC 4472, and NGC 4636, respectively). The quality of the fit declines as the maximum radius is increased because of deviations from the b-model, but the profiles are sufficiently concentrated that the central density estimates-4.6, 3.2, and g cm for NGC 1399, NGC 4472, and NGC 4636,

Accretion Rates and Eddington Ratio Limits
Individual black hole masses for the three galaxies studied here have been estimated by Magorrian et al. (1998) from twointegral dynamical modeling of ground-based kinematical data. A more accurate value can be derived using the correlation with bulge velocity dispersion j derived for galaxies with limits based on HST spectroscopy. Therefore, we adopt SMBH masses from Merritt & Ferrarese (2001b) ( Table 1); these are 3-5 times lower than in Magorrian et al. (1998). The resulting 2-10 keV luminosities in units of the Eddington luminosity and the Bondi (1952)

THE NATURE OF THE ACCRETION FLOW
The conservative upper limits that we derive are 1-2 orders of magnitude lower than the luminosities estimated by Allen et al. (2000) using ASCA spectral decomposition. Direct Chandra images (Angelini et al. 2001) indicate that most of the hard X-ray emission spatially unresolved by ASCA originates in Xray binaries. Since upper limits on the nuclear X-ray emission and estimates of the Bondi accretion rate have each decreased by ∼10-30 from Di  (the latter because of lower black hole mass estimates), 3 a reexamination of the nature of the accretion flow is in order.
For our estimated black hole masses and Bondi accretion rates, we have computed spectra using the techniques described in ; we aim to test the basic ADAF model and so do not include outflows or convection. The resulting predicted 2-10 keV luminosities, , are With the revised black hole masses and accretion rates, we find that the radio flux predicted by the ADAF model is less than the observed flux for NGC 1399 and NGC 4636 but is in excess for NGC 4472; thus, the radio flux problem highlighted by Di Matteo et al. (1999) is somewhat ameliorated. Finally, a preliminary HST measurement of the nuclear far-UV emission in NGC 1399 (R. W. O'Connell et al. 2001, in preparation)ergs s -also conflicts with 38 Ϫ1 nL ∼ 7 # 10 n pure-inflow ADAF models, although dust may be an issue at this wavelength. Our observations require a modification of the standard ADAF/Bondi model such that the gas density near the SMBH is reduced by at least a factor of ∼10. This could be due to the role of strong outflows or convection in an ADAF. Alternatively, the central regions of these galaxies may not be in steady state due to heating by supernovae or by episodic accretion onto the central SMBH (e.g., Ciotti & Ostriker 1997). In addition, the cooling time at the accretion radius (Table 1) is less than R Bondi 10 7 yr for all three galaxies; therefore, inflow with mass deposition could be important (Brighenti & Mathews 1999 If the decline in the accretion rate is due to outflow, this raises the question of the ultimate fate of this gas as it encounters the dense hot ISM in the core. Many questions about stability, intermittency, and the regulation of gas flows in the inner regions of elliptical galaxies remain to be addressed. The Chandra isophotes in the cores of NGC 1399 and NGC 4472 appear round and undisturbed, although there is an apparent spiral-like structure in the NGC 4636 diffuse emission.
The more stringent limits required to constrain the nature of the central accretion flow further will be difficult. Our observations indicate greater than 0.01 super-Eddington X-ray binaries per square arcsecond near the nucleus, so that detections below ∼10 39 ergs s cannot be unambiguously identified as Ϫ1 AGNs. Although the hot gas emissivity precipitously declines at energies above 2 keV, so does the Chandra sensitivity, restricting the detection limit in gas-rich systems. Tighter limits can be obtained on gas-poor systems that, although of interest, may be fuel-starved and would thus shed little light on the accretion process. It would be especially useful to spectrally decompose the inner ∼1Љ to place a limit on the fraction of non-ISM emission. This would appear as a hard excess and would require a very deep exposure because of the declining Chandra effective area mentioned above.
We are grateful to Dave Davis and Una Hwang for assistance with data reduction and to Wayne Landsman for sharing unpublished HST results. We made use of the HST archive, the NASA/IPAC Extragalactic Database (NED), and interactive data language function-fitting routines developed by Craig Markwardt.