Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/~kim/pap/2007_N821_I.pdf Äàòà èçìåíåíèÿ: Mon Dec 6 20:20:06 2010 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 00:19:58 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: ï ï ï ï ï ï ï ï ï ï ï ï ï

The Astrophysical Journal, 667:731 Y 748, 2007 October 1
# 2007. The American Astronomical Society. All rights reserved. Printed in U.S.A.

A

A DEEP CHANDRA LOOK AT THE LOW-LB ELLIPTICAL NGC 821: X-RAY BINARIES, A GALACTIC WIND, AND EMISSION AT THE NUCLEUS
S. Pellegrini,1 A. Baldi,2 D. W. Kim,2 G. Fabbiano,2 R. Soria,
2, 3

A. Siemiginowska,2 and M. Elvis

2

Received 2006 December 29; accepted 2007 June 7

ABSTRACT The relatively nearby (distance = 24.1 Mpc) elliptical galaxy NGC 821, hosting a central massive black hole but inactive at all wavelengths, was observed with Chandra for a total exposure time of 230 ks in order to search for nuclear emission and gas available for accretion. Within its optical image, 41 sources were detected, with spectral properties typical of low-mass X-ray binaries ( LMXBs). The fractions of LMXBs in the field and in globular clusters were determined, together with their X-ray luminosity function ( XLF ), down to L(0:3 Y 8 keV ) ¼ 2 ; 10 37 ergs sþ1. At the galactic center a source of L(0:3 Y 8 keV ) ¼ 6 ; 10 38 ergs sþ1 was detected for the first time, slightly extended. Its spectral shape is quite hard ( þ ¼ 1:49×0::14 ), without intrinsic absorption. It is surrounded by three sources with specþ0 13 tral shapes typical of LMXBs and luminosities on the brightest end of the XLF. One is consistent with being pointlike; the others could be the superposition of a few point sources and /or truly diffuse emission, with one resembling a jetlike feature. Diffuse emission was detected out to R $ 30 00 and comes mostly from unresolved LMXBs, with a minor contribution from other types of stellar sources. Different lines of investigation consistently provide no evidence for hot gas. Hydrodynamical simulations show that stellar mass losses are driven out of NGC 821 in a wind sustained by Type Ia supernovae, but also hot accreting gas within a very small inner region. A companion paper presents further observational results from Spitzer and the Very Large Array ( VLA), and possible accretion modalities for this central massive black hole. Subject headings: galaxies: elliptical and lenticular, cD -- galaxies: individual ( NGC 821) -- galaxies: nuclei -- g X-rays: galaxies -- X-rays: ISM Online material: color figures

1. INTRODUCTION NGC 821 is an isolated elliptical galaxy at a distance of 24.1 Mpc (Smith et al. 2004; see Table 1), with very regular optical isophotes of disky shape ( Bender et al. 1994; Lauer et al. 2005 ) and an old and metal-rich stellar population typical of elliptical galaxies ( Proctor et al. 2005 ). No cold ( H i) or dusty interstellar medium ( ISM ) has been revealed in it ( Lauer et al. 2005; Sarzi et al. 2006 ). NGC 821 is also one of those $30 nearby galactic spheroids for which a central massive black hole ( MBH ) has been claimed based on resolved dynamical studies ( Ferrarese & Ford 2005 ). The mass of this MBH is 8:5 ; 107 M ( Table 1), and its Eddington luminosity is LEdd $ 1:1 ; 10 46 ergs sþ1 . However, this MBH is extremely quiescent at all wavelengths (see also Pellegrini et al. 2007 ) and represents an excellent example of the class of ``inactive'' nuclei that is very common in the local universe (e.g., Pellegrini 2005a; Ho 2005 ) and key to our understanding of the mechanisms of active galactic nucleus (AGN ) evolution and nuclear feedback, which have been fundamental for shaping our universe (e.g., Springel et al. 2005 ). Since it is nearby, NGC 821 is a prime target for observing a quiescent nucleus and its surroundings. X-ray emission is a key symptom of nuclear activity resulting from accretion onto a MBH ( Rees 1984), and the hot ISM that could provide a source of fuel is readily visible in the X-rays (e.g., Fabian & Canizares 1988; Loewenstein et al. 2001; Pellegrini 2005a). Previous observations of NGC 821 with the Rontgensatellit (ROSAT ) PSPC had placed an upper limit of ¨
Astronomy Department, Bologna University, Italy ; silvia.pellegrini@ unibo.it. Harvard-Smithsonian Center for Astrophysics, 60 Garden Street , Cambridge , MA 02138. 3 Mullard Space Science Laboratory, University College London, Holmbury St. Mary, UK.
2 1

2:8 ; 1040 ergs sþ1 (O'Sullivan et al. 2001) on the total X-ray luminosity, suggesting a low content of hot gas (e.g., Kim et al. 1992). With the subarcsecond resolution of Chandra,we could attempt to resolve the different components of the X-ray emission, i.e., the hot gas, the nucleus, and the population of LMXBs ( Fabbiano et al. 2004, hereafter F04). This previous shallow (39 ks) Chandra observation, performed in 2002, revealed diffuse emission in the central galactic region, possibly from a hot ISM. We did not detect any source that could be unequivocally identified as the galactic nucleus (for which we derived LX /LEdd < 10þ7 ), but an intriguing S-shaped feature crossing the center of the galaxy was observed. Explanations for this feature included either a weak two-sided X-ray nuclear jet or a hot gas filament, at a temperature higher than that of the surrounding gas. This opened the possibilities that the accretion power could end mostly in mechanical rather than radiative power ( Di Matteo et al. 2003; Pellegrini et al. 2003b; Fabbiano et al. 2003), or that the accretion flow may be disrupted by nuclear feedback (e.g., Ciotti & Ostriker 2001; Omma et al. 2004; see also Soria et al. 2006a, 2006b). In this paper we report the results of the analysis of deep Chandra observations (for a total exposure of nearly 230 ks) aimed at defining the properties of the various sources in the central galactic region. With the deep Chandra exposure, we can constrain all the components (nucleus, hot gas, unresolved sources) contributing to the emission, a necessary and critical step when the goal is to understand the accretion process. For example, we can detect a significant portion of the LMXB population, which was barely visible in the 2002 data, and thanks to the calculation of its XLF we can now discriminate between the contributions to the diffuse emission of the unresolved LMXBs and of the hot ISM (e.g., Kim & Fabbiano 2004). Moreover, determining the properties of the hot ISM also gives clues on the way the ISM 731


732

PELLEGRINI ET AL.
TABLE 1 NGC 82 1: M a in Pr ope r tie s
0 BT a (mag)

Vol. 667

Type

a

Db ( Mpc) 24.1

log LB (LB,) 10.27

Sizea (arcmin) 2.57 ; 1.62

R ec (arcsec, kpc) 43.9, 5.1

ed (km sþ1) 209

NHe (cmþ2) 6.2 ; 10
20

MBHf (107 M) 8.5 ô 3.5

100 ( pc) 117

E6 ...........................

11.72

a 0 Type, BT , and size from de Vaucouleurs et al. (1991, hereafter RC3). The size gives the major and minor axis of the D25 ellipse, that is, the 25.0 B magnitude per square arcsecond isophote. The position angle is 25 ( RC3). b Distance D from Tonry et al. (2001). c Effective radius R e in the R band (from Soria et al. 2006b; see also x 5.3). d Effective stellar velocity dispersion (averaged over Re ) from Pinkney et al. (2003). e Galactic hydrogen column density ( Dickey & Lockman 1990). f Gebhardt et al. (2003) reported a value of 3:7×2::4 ; 107 M , later revised to the value given here ( Richstone et al. 2004) that is considered more reliable ( K. Gebhardt þ0 8 2006, private communication).

evolves in this galaxy, and more generally in galaxies of similar optical luminosity (e.g., Pellegrini & Ciotti 1998; Sansom et al. 2006; David et al. 2006 ). A companion paper ( Pellegrini et al. 2007 ) presents an observational campaign aimed at determining the properties of the nuclear emission of NGC 821 at various wavelengths, including more sensitive proprietary VLA and Spitzer Infrared Array Camera ( IRAC ) observations, together with archival HST observations. The paper is organized as follows. In x 2 we describe the Chandra observations and data preparation, and in x 3we present the X-ray properties of the whole LMXB population of the galaxy, also making use of Hubble Space Telescope (HST ) Wide Field Planetary Camera 2 ( WFPC2) images to identify those in field and in globular clusters. In xx 4 and 5 we report the results of the X-ray analysis for the nuclear region and the diffuse emission, respectively; in x 6 we discuss the implications of our observations; and in x 7 we summarize our results. 2. CHANDRA OBSERVATIONS AND DATA PREPARATION NGC 821 was observed with Chandra ACIS-S ( Weisskopf et al. 2000) seven times between 2002 November and 2005 June, for a total exposure of 230 ks ( Table 2). Here we consider both the archival observations (ObsIDs 4408 and 4006 in Table 2), already presented in F04, and the new unpublished data obtained in 2004 and 2005. The satellite telemetry is processed at the Chandra X-Ray Center (CXC ) with the standard data processing (SDP) pipelines in order to correct for the motion of the satellite and to apply instrument calibration. The data used in this work were reprocessed in custom mode with version 7.6.0 of the SDP, to take advantage of improvements in processing software and calibration not available at the time of the original processing. Verification of the data products showed no anomalies. The relative astrometry of the different
TABLE 2 Chan dra ACIS-S Observing Log for NGC 821 Exposure ( ks) 25.3 13.7 28.0 40.1 32.4 50.1 40.1

observations was corrected by using detected sources (above a 3 threshold) as reference points.4 The data products were then analyzed with the CXC CIAO version 3.0.1 software and the HEASARC XSPEC package. CIAO data model tools were used for data processing, such as screening out bad pixels and producing images in given energy bands. Calibration files in CALDB version 3.1.0 of 2005 June were used. No significant background flares were observed in these data, so no further screening was necessary. A time-dependent gain correction5 was applied to the SDP level 2 event files of the first two observations before further analysis. This correction was already included in the SDP level 2 event files of the other observations supplied by the CXC. The seven observations were then co-added to produce a deep image, using the CIAO task mergeall, which takes as input the event files and the relative aspect solution (ASOL) files and reprojects the coordinates of all the observations onto the first one. Below we report the results of the analysis of these data. 3. THE LMXB POPULATION Figure 1 shows the merged X-ray image, with the D25 ellipse of NGC 821 overplotted (this ellipse is the isophote of the 25.0 B magnitude per square arcsecond brightness level; Table 1); also marked are the X-ray sources detected by the CIAO task wavdetect. Of the 104 sources detected within the ACIS-S3 CCD, 41 lie inside the D25 ellipse. Except for minor contamination from interlopers (see xx 3.2 and 5.6), and four particular sources in the central region (discussed in x 4), these sources are pointlike and represent the LMXB population of NGC 821 (see Fabbiano [2006] for a review on X-ray binary populations of galaxies). For the pointlike sources within D25, counts were extracted within a circle centered on the wavdetect-determined source position, and background counts were estimated locally in an annulus surrounding the source. We chose the source-extraction radius to be the 95% encircled energy radius at 1.5 keV (that is, varying as a function of the off-axis angle), with a minimum of 3 00 near the aim point. Similarly, the background was estimated for each source from an annulus surrounding it, with inner and outer radii of 2 and 5 times the source radius, respectively. When nearby sources are found within the background region, they are excluded before measuring the background counts. Net count rates were then calculated with the effective exposure (including vignetting) for both the source and background regions. Errors on counts were derived
4 5

ObsID 4408............................ 4006............................ 5692............................ 6314............................ 6310............................ 6313............................ 5691............................

Date 2002 2002 2004 2005 2005 2005 2005 Nov 26 Dec 1 Dec 4 Jun 20 Jun 21 Jun 22 Jun 23

See http://cxc.harvard.edu /cal /ASPECT/align _evt. See http://cxc.harvard.edu /contrib/alexey/tgain /tgain.html.


No. 2, 2007

DEEP CHANDRA OBSERVATION OF NGC 821

733

the galactic center. Column (1) gives the IAU name (following the convention CXOU Jhhmmss.s +/þddmmss); columns (2) and (3) give the J2000.0 source position, and column (4) its associated uncertainty at the 95% confidence level, determined based on the source net counts and off-axis angle, from the empirical formula given by Kim et al. (2007 ); column (5 ) gives the projected distance R from the galactic center ; column (6 ) gives the net counts and the 1 error; columns (7 ) and (8) give the flux and luminosity in the 0.3 Y 8 keV band, derived using the energy conversion factor described in x 3.2; and columns (9) Y (14) give the hardness ratio ( HR), the C21 and C 32 colors defined above, and the respective error determined at a 1 significance level. 3.1. Spectral Properties and Optical Identification Figure 2 shows the C32YC21 distribution of the point-source population; the grid gives the location of spectra described by power laws with given values of the photon index þ and intrinsic column density NH. In agreement with previous spectral studies of LMXB populations (e.g., Kim et al. 1992; Irwin et al. 2003; Fabbiano 2006 ), most well- defined colors fall near a typical þ ¼ 1:5 Y 2:0 spectrum, with no intrinsic absorption. Three of the sources in Table 3 have been detected with a number of net counts sufficiently high for a specific spectral analysis (they are marked in Table 3). We performed such analysis, extracting from the merged event file the spectra for these sources, using an extraction radius of 200 ; this radius was chosen to avoid contamination from nearby fainter sources. Instrumental response files, weighted by the exposure time of each individual observation, were also generated; the background spectrum was derived from a source-free region outside the optical body of the galaxy. The results of the spectral analysis ( Table 4; Fig. 3) confirm that these sources are well described by power laws with low-to-moderate intrinsic absorption, as is typical for the LMXB population. Their intrinsic 0.3 Y 8 keV luminosities are 4:1 ; 10 38 , 6:9 ; 10 38 , and 8:9 ; 10 38 ergs sþ1. With the aid of HST WFPC2 images, we then investigated possible associations of our detected point sources with LMXBs belonging to the galactic field or to globular clusters (GCs). Kundu & Whitmore (2001) studied the globular cluster population of NGC 821 using HST WFPC2 images in the V ( F555W ) and I ( F814W ) filters; these optical images cover a large fraction ($60%) of the D25 ellipse of NGC 821. Kundu & Whitmore (2001) identified 105 GC candidates whose positions, magnitudes, and colors were kindly provided to us by A. Kundu (2007, private communication). This list of GCs was cross-correlated with the X-ray source list of Table 3, in order to look for GC/ X-ray source coincidences. After correction for a small shift of the optical image (see x 4.2 for more details), six matches were found well within the 95% uncertainty radius of the X-ray position; another potential match has quite a distorted shape and uncertain magnitude in the optical (CXOU J020817.7+105907 ), since it is located at a chip edge, and another one has an offset of 1.24 times the X-ray positional uncertainty. These eight sources are marked in Table 3 (see also Fig. 8). From a V þ (V þ I ) diagram of the w ho l e sa m pl e o f 10 5 G C c a nd i da t es , o n e c an s e e t h a t L MX B s are preferentially found in the brightest clusters (4 matches reside in the 5 brightest GCs, and 6 in the 12 brightest ones), following a general trend already reported (e.g., Sarazin et al. 2003, Kim et al. 2006b) that is explained as the probability of hosting a LMXB increasing with the GC luminosity. The LMXBs of NGC 821 also tend to be associated with the redder GCs, in agreement with previous works (e.g., Kundu et al. 2007 ), although the

Fig. 1.-- Merged ACIS-S image of NGC 821, with the D25 ellipse of Table 1 overplotted in light gray, and the X-ray sources detected by wavdetect marked in dark gray with circles of 300 radii (x 3). [See the electronic edition of the Journal for a color version of this figure.]

following Gehrels (1986 ). When the source-extraction regions of nearby sources overlap, to avoid an overestimate of their source count rates, we calculated the source counts from a pie sector, excluding the nearby source region, and then rescaled them based on the area ratio of the chosen pie to the full disk. Once the correction factor is determined, the same factor can be applied to correct counts in all energy bands. For a small number of sources that overlap with nearby sources in a more complex way (e.g., overlap with more than two sources), instead of correcting the aperture photometry, we used the source cell determined by wavdetect to extract the source counts. The range of net detected counts for the pointlike sources is $8 Y 300, corresponding to 0.3Y8 keV luminosities of 2 ; 10 37 to 8 ; 10 38 ergs sþ1 when using the energy conversion factor described in x 3.2. Most of the sources are too faint for a detailed spectral analysis; therefore, their hardness ratio and X-ray colors were calculated in order to characterize their spectral properties. The X-ray hardness ratio is defined as HR ¼ (H þ S )/(H × S ), where S and H are the net counts in the 0.5Y2.0 and 2.0 Y 8.0 keV bands, respectively. Following the prescription of Kim et al. (2004), the X-ray colors are defined as C21 ¼ log (C1/C2) and C32 ¼ log (C2/C3), where C1, C2, and C3 are the net counts in the energy bands of 0.3 Y 0.9, 0.9 Y 2.5, and 2.5 Y 8 keV, respectively. These counts were corrected for the temporal quantum efficiency variation, referring them all to the first observing epoch (2002 November; Table 2), and for the effect of the Galactic absorption (using the value in Table 1). By definition, as the X-ray spectra become harder, the HR increases and the X-ray colors decrease. For faint sources with a small number of counts, the formal calculation of the HR and colors often results in unreliable errors, because of negative net counts in one band and an asymmetric Poisson distribution. Therefore, we applied a Bayesian approach developed by Park et al. (2006 ), which models the detected counts as a nonhomogeneous Poisson process in order to derive the uncertainties associated with the HR and colors. The main properties of the point sources detected in the merged observation within the D25 ellipse are summarized in Table 3, where the sources are listed in order of increasing distance from


TABLE 3 Properties of the X-Ray Point-source Popu lati on within D25

CXOU Name (1) +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +10 +11 +11 +10 +10 +10 +10 +10 +11 +10 +11 +11 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 59 00 00 59 59 59 58 59 00 59 00 00 46.5 37.7 48.4 41.5 45.7 48.7 35.3 39.1 40.7 50.2 26.6 42.9 52.8 30.1 56.2 26.5 51.3 24.4 22.4 48.9 57.9 12.9 52.6 09.9 59.1 50.6 03.6 08.9 04.6 19.3 00.0 55.4 00.2 07.8 07.7 41.1 44.1 0.40 0.40 0.39 0.21 0.40 0.15 0.42 0.43 0.49 0.36 0.45 0.67 0.41 0.35 0.44 0.57 0.48 0.55 0.15 0.43 0.35 0.58 0.14 0.51 0.27 0.30 0.47 0.35 0.44 0.43 0.54 0.30 0.53 0.31 0.28 0.35 0.48 6.24 6.75 6.95 7.16 8.16 9.44 9.53 12.74 14.21 14.51 16.39 16.53 17.38 17.60 18.42 18.43 20.05 21.45 24.76 25.46 27.17 28.80 31.95 32.01 32.30 33.26 37.02 39.19 39.42 41.70 45.41 46.64 49.23 51.53 60.28 61.17 65.73 20.6 29.2 21.0 77.2 20.3 164.9 20.4 20.6 15.5 25.3 23.8 8.2 19.9 34.1 42.0 10.4 11.7 13.0 214.2 25.1 26.6 13.7 301.0 19.2 44.6 43.3 11.4 30.8 29.0 26.4 17.1 52.8 13.6 43.6 60.9 24.4 15.0 ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô ô 6.1 8.5 6.1 10.5 6.1 14.3 7.4 6.0 5.3 6.6 7.3 4.3 5.9 8.6 10.5 6.2 6.1 6.7 16.1 7.4 7.4 8.1 18.6 8.6 8.7 8.0 5.9 7.7 7.5 7.4 6.6 9.2 6.1 8.5 9.6 7.2 6.4 8.0 11.3 8.1 29.9 7.9 63.9 7.9 8.0 6.0 9.8 9.2 3.2 7.7 13.2 16.3 4.0 4.5 5.0 83.0 9.7 10.3 5.3 116.7 7.4 17.3 16.8 4.4 11.9 11.2 10.2 6.6 20.4 5.3 16.9 23.6 9.5 5.8 0.56 0.79 0.56 2.08 0.55 4.44 0.55 0.55 0.42 0.68 0.64 0.22 0.54 0.92 1.13 0.28 0.31 0.35 5.76 0.68 0.72 0.37 8.10 0.52 1.20 1.17 0.31 0.83 0.78 0.71 0.46 1.42 0.37 1.17 1.64 0.66 0.41 ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 0.19 0.24 þ0.43 þ0.19 þ0.51 þ0.27 þ0.53 þ0.02 þ0.62 0.02 þ0.46 þ0.16 þ0.22 þ0.05 þ0.43 þ0.26 0.12 þ0.13 þ0.45 þ0.31 þ0.85 0.54 þ0.27 0.48 þ0.21 0.03 þ0.01 þ0.95 þ0.31 þ0.02 þ0.57 þ0.40 0.26 þ0.56 þ0.36 þ0.30 þ0.73 0.52) 0.46 ) þ0.04) þ0.11) þ0.05 ) þ0.20) þ0.09) 0.30) þ0.17 ) 0.23) þ0.12) 0.25 ) 0.12) 0.23) þ0.14) 0.51) 0.56 ) 0.33) þ0.35 ) 0.10) þ0.31) 1.39) þ0.22) 0.94) 0.03) 0.20) 0.87 ) þ0.39) þ0.02) 0.27 ) þ0.06 ) þ0.16 ) 1.34) þ0.22) þ0.20) 0.36 ) þ0.09) 0.29 0.52 0.39 0.58 0.24 0.58 0.84 0.34 0.74 0.79 1.23 1.18 0.41 0.30 0.54 þ0.06 0.70 0.50 0.42 0.15 0.12 þ0.37 0.43 0.09 0.55 0.18 þ0.35 0.24 0.77 0.52 0.41 0.57 þ0.69 0.26 0.68 þ0.49 0.11 þ0.52 þ1.00 þ0.52 þ0.53 þ0.36 þ0.59 þ0.56 þ0.31 þ0.64 þ0.79 þ1.00 þ1.00 þ0.58 þ0.35 þ0.53 þ0.14 þ1.00 þ1.00 þ0.45 þ0.28 0.01 þ0.36 þ0.44 þ0.70 þ0.62 þ0.54 0.10 þ0.31 þ1.00 þ0.56 þ0.20 þ0.46 0.66 þ0.19 þ0.72 1.00 0.24 þ0.74, þ1.00, þ0.74, þ0.63, þ0.60, þ0.64, þ0.90, þ0.54, þ0.87, þ0.96, þ1.00, þ1.00, þ0.78, þ0.51, þ0.70, þ0.76, þ1.00, þ1.00, þ0.51, þ0.49, þ0.25, þ0.77, þ0.49, þ0.86, þ0.74, þ0.63, þ0.45, þ0.42, þ1.00, þ0.81, þ0.61, þ0.59, (0.20, (þ0.35, (þ0.82, (0.66, (þ0.29, þ0.27 ) þ0.68) þ0.27 ) þ0.43) þ0.10) þ0.51) þ0.24) þ0.05 ) þ0.36 ) þ0.57 ) þ0.94) þ0.90) þ0.34) þ0.08) þ0.38) 0.45 ) þ0.64) þ0.58) þ0.38) þ0.02) 0.21) 0.02) þ0.39) þ0.37 ) þ0.45 ) þ0.35 ) 0.68) þ0.02) þ0.73) þ0.29) 0.21) þ0.32) 0.84) þ0.03) þ0.60) 1.00) 0.72) (þ0.13, (0.00, (þ0.99, (þ0.36, (þ1.24, (þ0.38, (þ1.44, (þ0.37, (þ1.30, (þ0.19, (þ0.99, (þ0.65, (þ0.61, (þ0.36, (þ0.83, (þ1.43, (þ0.32, (þ0.72, (þ0.53, (þ0.96, (þ2.11, (0.01, (þ0.35, (0.14, (þ0.47, (þ0.16, (þ1.03, (þ2.23, (þ0.67, (þ0.34, (þ1.67, (þ0.68, (þ0.65, (þ1.06, (þ0.54, (þ1.39, (þ2.06, (þ0.09, (0.07, (0.07, (0.46, (þ0.07, (0.48, (0.38, (þ0.01, (0.32, (0.36, (0.76, (0.60, (0.07, (0.14, (0.31, (þ0.67, (0.06, (þ0.01, (0.35, (þ0.07, (0.05, (þ1.25, (0.37, (þ0.44, (0.37, (0.04, (þ1.14, (0.09, (0.37, (0.13, (0.02, (0.37, (þ1.81, (0.13, (0.51, (þ0.95, (þ0.27,

R.A. (J2000.0) (2)

Decl. (J2000.0) (3)

à (arcsec) (4) Count (6 ) FX (7 ) LX (8) HR (9) Error (10) C21 (11) Error (12) C32 (13) Error (14)

R (arcsec) (5 )

J020821.4 +105946 a................... J020820.7+105937 b ................... J020821.2 +105948 b .................. J020821.6 +105941b ................... J020821.6 +105945 b .................. J020821.5 +105948b,c ................. J020821.6 +105935 a................... J020822.0 +105939 a................... J020822.1+105940 b ................... J020821.9 +105950 b .................. J020820.7+105926 a ................... J020822.2+105942 a ................... J020822.0 +105952b ................... J020820.2+105930b ................... J020821.9 +105956 .................... J020821.8 +105926 .................... J020819.9 +105951b ................... J020820.2 +105924b ................... J020822.2 +105922c ................... J020822.8 +105948 .................... J020819.6 +105957 b .................. J020820.9 +105912b ................... J020823.1+105952c ................... J020820.8+105909 b ................... J020822.9 +105959 .................... J020823.3 +105950 .................... J020819.1+110003a.................... J020823.0 +110008 .................... J020822.0 +105904b ................... J020818.7+105919b ................... J020819.9 +105900b ................... J020820.7+105855 a ................... J020819.3 +105900d ................... J020824.1+110007 ..................... J020817.7+105907 a ................... J020820.2+110041d ................... J020822.5 +110044 ....................

2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8

21.4 20.7 21.2 21.6 21.6 21.5 21.6 22.0 22.1 21.9 20.7 22.2 22.0 20.2 21.9 21.8 19.9 20.2 22.2 22.8 19.6 20.9 23.1 20.8 22.9 23.3 19.1 23.0 22.0 18.7 19.9 20.7 19.3 24.1 17.7 20.2 22.5

0.75 ) 1.34) 0.77 ) 0.78) 0.59) 0.67 ) 1.85 ) 0.76 ) 1.39) 1.58) 2.10) 2.12) 0.83) 0.72) 0.79) 0.50) 1.93) 1.47 ) 0.50) 0.42) 0.49) 0.26 ) 0.50) 0.77 ) 0.92) 0.39) 0.17 ) 0.49) 1.55 ) 1.21) 1.02) 0.78) þ0.14) 0.47 ) 0.93) þ0.15 ) 0.57 )

Notes.-- Units of right ascension are hours, minutes and seconds, and units of declination are degrees, arcminutes and arcseconds. Units of à (the positional uncertainty) and R (the distance from the galactic center) are arcseconds. The positional uncertainty is based on the statistical error of the X-ray data alone (x 3), while the absolute positional error is discussed in x 4.2. Units of FX are 10þ16 ergs sþ1 cmþ2, and those of LX are 1038 ergs sþ1; both quantities have been absorption-corrected for the Galactic column density and refer to the 0.3Y8 keV band. See x 3 for more details (e.g., on the definitions of the HR and the colors C21 and C23, and the calculation of their errors). a The position of this source is coincident with that of a candidate globular cluster found in the HST WFPC2 images; see x 3.1. The offset between the optical and X-ray position is always well within à, except for the first of these eight sources (J020821.4+105946 ), which has an offset of 1.24à. b This is a field LMXB candidate (see x 3.1). c A specific spectral analysis has been done for this source (x 3.1 and Table 4). d This X-ray source may correspond to a faint background galaxy, from an inspection of the HST images discussed in x 3.1 (A. Kundu 2007, private communication).


DEEP CHANDRA OBSERVATION OF NGC 821

735

Fig. 2.--X-ray colors of point sources, with error bars (left ) and without error bars (right) (x 3.1). The grid indicates photon index values of 0, 1, 2, 3, and 4 ( from bottom to top), and NH values ( from right to left )of10 20,10 21,5 ; 10 21 (dashed line), and 1022 cmþ2. Gray circles correspond to sources at R < 10 00 ; black circles correspond to sources between R ¼ 10 00 and the ellipse D25. Filled circles are sources with net counts >30; open circles are sources with net counts <30. [See the electronic edition of the Journal for a color version of this figure.]

small number of sources here does not make this result statistically significant. From the HST WFPC2 images above, in the overlap region with the D25 ellipse, 17 X-ray sources have no optical counterpart at a 50% completeness limit of 24.1 V magnitudes (A. Kundu 2007, private communication);6 these can be considered field LMXB candidates and are marked in Table 3. In the following we further limit the analysis to sources located at radii larger than 1000 , to avoid possible completeness problems near the galactic center (see x 3.2). Then, it turns out that $30% of the X-ray sources lie in GCs. From Chandra studies this fraction is known to increase with the GC specific frequency of the galaxy, or the morphological type (e.g., Sarazin et al. 2003), and goes from $20% in S0 galaxies to 30% Y50% in E galaxies (e.g., Fabbiano 2006 and references therein). The fraction estimated here for NGC 821 is therefore close to what is expected, given also its morphological type ( E6 ). The fraction of GCs that host an X-ray source is $6.6%, a value broadly consistent with what typically has been
6 This V- magnitude value together with those quoted in x 4.2 have been computed by Kundu & Whitmore (2001) assuming the Burstein & Heiles (1982) reddening.

observed previously from Chandra studies (3% Y5%; Sarazin et al. 2003; Kim et al. 2006b; see also Kundu et al. 2007 ), but somewhat higher, likely because of the larger depth of our X-ray observation compared to the previous analyses. Finally, there is no statistically significant difference between GC and field LMXBs. The present sample of X-ray sources is too small to investigate other statistical properties, such as the X-ray spectral difference between LMXBs in red and blue GCs or the role of galactocentric distance. 3.2. The LMXB X-Ray Luminosity Function For the point sources falling within the D25 ellipse, we estimated an XLF that in differential form can be expressed as þ dN LX ¼k : Ï 1÷ dLX 10 38 ergs sþ1 To construct the XLF, we calculated fluxes and luminosities of the point sources in the 0.3Y8 keV band with an energy conversion factor ( ECF ) corresponding to an assumed power-law spectral shape with þ ¼ 1:7 and Galactic NH (see Fig. 2 for a justification of this assumption). The ECF was calculated with the arf

TABLE 4 Spectr al Analysis Result s for th e Brightest S ources in Table 3 wabs( pow) NH (1021 cmþ2) .............................................. þ ..................................................................... 2/dof ............................................................. Flux (10þ15 ergs cmþ2 sþ1) ........................... Luminosity (10 38 ergs sþ1)............................ J020821.5+105948
×0 8 2:2þ1::0 1:98×0::42 þ0 28 8.1/4 4.5 4.1

J020822.2+105922 0:5×1::3 þ0 5 ×0 17 1:21þ0::24 5.2/7 9.7 6.9

J020823.1+105952 1:5×0::5 þ0 5 1:58×0::13 þ0 15 7.6/11 10.9 8.9

Notes.--The NH values are in addition to the Galactic one. Fluxes are observed values, while luminosities are corrected for absorption; both are calculated for the 0.3 Y 8 keV band. Errors give the 68% confidence interval for one interesting parameter. See also x 3.1.


736

PELLEGRINI ET AL.

Vol. 667

Fig. 4.-- Cumulative XLF of point sources detected in the merged ACIS-S image within the D25 ellipse, with a central circle of 10 00 radius excluded (x 3.2). Black squares with error bars show the bias-corrected function ; gray circles are the uncorrected function; the histogram gives the best-fit single power law; the dotted line shows the expected number of cosmic X-ray background sources from the log NYlog S relation of ChaMP plus CDF data ( Kim et al. 2007 ). [See the electronic edition of the Journal for a color version of this figure.]

Fig. 3.-- Chandra ACIS-S spectra of the brightest point sources in Table 3, together with their best-fit power law models ( Table 4). From top to bottom: J020821.5 +105948, J020822.2 +105922, and J020823.1+105952 (see also x 3.1).

(auxiliary response file) and the rmf (redistribution matrix file) generated for each source in each observation. As was done for the colors, we took into account the temporal QE variation by calculating the ECF in each observation and then taking an exposure-weighted mean ECF. The ECF over the 0.3 Y 8 keV band varied by $2.5% between 2002 and 2005. We then corrected the XLF to eliminate biases and incompleteness effects, which could affect the lowest luminosity range, causing an artificial break ( Kim & Fabbiano 2003, 2004; Kim et al. 2006a). Our procedure for correcting XLFs derived from Chandra data was developed

and first applied to NGC 1316 by Kim & Fabbiano (2003; see the appendix therein, where the method is discussed in detail) and was then applied to a sample of early-type galaxies by Kim & Fabbiano (2004). A similar procedure had also been used to correct the XLF of the Antennae galaxies ( Zezas & Fabbiano 2002). In brief, we simulated $20,000 point sources, added them one by one to the observed image, and then ran wavdetect to determine whether the added source was detected. With this procedure, we correct simultaneously for incompleteness near the detection threshold, Eddington bias ( Eddington 1913), and source confusion. In the simulations, we assumed a typical XLF of differential slope of  ¼ 2; however, the adopted slope does not affect the results significantly, and neither does the assumed radial distribution of point sources (which we assume follow the optical light, as observed in many elliptical galaxies; see Fabbiano 2006 ). We excluded central sources with R < 10 00 in Table 3 because of the large photometric error caused by confusion with other overlapping sources, and by the possible presence of some diffuse emission; these conditions make the incompleteness corrections uncertain. With the above procedure, we established the 90% completeness limit (i.e., the luminosity at which 10% of sources would not be detected inside the D25 ellipse, excluding the central 10 00 )to be LX ¼ 3 ; 10 37 ergs sþ1 (or FX ¼ 4:3 ; 10þ16 ergs cmþ2 sþ1)in the 0.3 Y 8 keV band. We can reliably correct the XLF to a luminosity of $2/3 of the 90% limit. The resulting cumulative XLF [i.e., the number of sources N (> LX )] is shown in Figure 4. From the log NYlog S relation derived for the ChaMP plus CDF data ( Kim et al. 2007 ), we then estimated that the number of expected cosmic background sources (mostly AGNs) falling within the D25 ellipse and brighter than the 90% completeness limit is three, corresponding to a source contamination of 7%. To establish the functional form of the XLF, we fitted the biascorrected differential XLF (eq. [1]) with single power laws, using both the Cash and 2 statistics in the CIAO Sherpa modeling and fitting application. We found that a single power law represents


No. 2, 2007

DEEP CHANDRA OBSERVATION OF NGC 821

737

well the XLF, with a best-fit slope  ¼ 2:0 ô 0:3, compatible with other power-law fits of LMXB XLFs ( Kim & Fabbiano 2004; Kim et al. 2006a) and a best-fit amplitude k ¼ 8:4×3::1 . þ2 5 Even without the correction for incompleteness, which is significant only at the point corresponding to the faintest fluxes (LX ¼ 2 ; 10 37 ergs sþ1), a single power law is a good fit all the way down to LX ¼ 3 ; 10 37 ergs sþ1.Figure4maysuggest alow luminosity break around LX $ 5 ; 10 37 ergs sþ1. This break was suggested to be a universal feature of LMXB populations (Gilfanov 2004) but is not observed in all cases ( Kim et al. 2006a). For NGC 821 there is no compelling evidence for such a break, given the size of the error bars and the good fit with a single power-law model. The combined LMXB XLF ( Kim & Fabbiano 2004; Gilfanov 2004) shows a high luminosity break at LX > (5:0 ô 1:6) ; 10 38 ergs sþ1, which may originate from the presence of neutron star and black hole binary populations. We cannot constrain here the high-luminosity shape of the XLF because of the small number of luminous sources (only two sources have LX > 5 ; 10 38 ergs sþ1). In contrast to the uniform slope, the amplitude of the XLF varies widely from galaxy to galaxy, reflecting the varying content of LMXBs per unit galactic luminosity. For their sample of early-type galaxies, from integration of the XLF above LX; min ¼ 10 37 ergs sþ1, Kim & Fabbiano (2004) estimated the total LX due to LMXBs and the average ratio LX /LB ¼ (0:9 ô 0:5) ; 10 30 ergs sþ1 Lþ;1 . For NGC 821 this ratio is 0:4 ; 10 30 ergs sþ1 B Lþ;1 (see x 5.5), lower than the mean value by 1 . This ratio is B known to correlate with the total specific frequency of globular clusters, but this quantity unfortunately is not available for NGC 821. Based on the LX /LB ratio, we would expect NGC 821 to have a relatively poor GC population. 4. THE CENTRAL REGION OF NGC 821YRESOLVED SOURCES Figure 5 shows the central region of the image obtained from the merged data, in three spectral bands (0.3 Y1, 1Y2, and 2Y 4keV ) that cover the energy range where most of the counts are detected. From these three images in different spectral bands we also obtained an adaptively smoothed (using the CIAO task csmooth) color image ( Fig. 6 ), where the data were smoothed using scales ranging from 1 to 20 pixels (0.5 00 Y1000 ). Given the small extent of the region considered, no exposure correction was needed. Figure 6 shows diffuse emission at the center, which follows the optical shape of the galaxy: it is elongated in the northwest-southeast direction, roughly aligned with the optical major axis (see also Table 1 and Figs. 1 and 8). Most of the diffuse emission is concentrated well within one R e (that is, 43.9 00 , while the size of Fig. 6 is 55 00 ). Its color indicates emission mostly in the soft and medium energy bands. Harder emission from a few central pointlike sources is also visible. Below we concentrate on the X-ray analysis of the sources detected in the central region; in x 5 we report the analysis of the diffuse emission for the central region and the whole galaxy. 4.1. Detection and Spatial Properties We can now establish with higher significance than in F04 the morphology of the central X-ray emission. In particular, F04 had identified three possibly connected extended central emission regions (that they called S1 through S3) and a pointlike source to the northeast of this grouping ( labeled the NE source). The wavdetect task detects all these emission regions in the merged data, plus another source that we call S4 ( Fig. 7 ). F04 had found the spatial distribution of photons of the S1YS3 sources to be more extended

than expected for pointlike sources seen through the Chandra mirrors, and suggested that they may be part of an S-shaped jet or filament, possibly connected with low-level or past nuclear activity. With the longer exposure, it is apparent ( Fig. 7 ) that the emission of S3 may be pointlike, suggesting that this source is a luminous LMXB, but the other sources are more extended. In order to establish the spatial properties of S1Y S4, following F04 we compared the spatial distribution of their counts with that of the on-axis image of the quasar GB 1508 +5714, a hard pointlike source (Siemiginowska et al. 2003). GB 1508 +5714 gives a good representation of the Chandra ACIS-S point-spread function ( PSF ) for our analysis, since it lies at the same distance from the aim point as S1Y S4 and is similarly hard (see x 4.3). The image of GB 1508+5714 contains 5300 counts within 200 of the centroid of the count distribution; the ratio of counts within the 100 Y 200 annulus to those in the central circle of 100 radius is ratio( PSF ) ¼ 0:043 ô 0:001 (1 ). The isolated NE source in this central field ( Fig. 7 ) has a total of 173 source counts and ratio( NE ) ¼ 0:123 ô 0:083, a value consistent with that of our reference quasar, within the 1 error. The total number of source counts for S1Y S4 are 184, 250, 326, and 84, respectively; the analogous ratios for the backgroundsubtracted counts, around the centroids determined by wavdetect, are ratio(S1) ¼ 0:59 ô 0:11, ratio(S2) ¼ 0:77 ô 0:10, ratio(S3) ¼ 0:20 ô 0:06, and ratio(S4) ¼ 0:80 ô 0:15. Therefore, the spatial distribution of the counts from S1Y S4 is extended when we compare it with that of the quasar. If, more conservatively, we use as a benchmark the NE source, we find that S3 is pointlike, while S1, S2, and S4 are all significantly more extended than the NE source (at 3.5, 4.6, and 4 respectively). Although the significance of the NE source is such that its ratio is less well determined than that of the quasar, this source would be affected by similarly small amounts of smearing in the merging process as in the S1Y S4 sources. [ Note that in F04 the northeast source had a ratio( NE ) ¼ 0:057 ô 0:054, consistent within the errors with the value we find here, although the nominal value is smaller; this justifies using NE as a benchmark.] In conclusion, excluding source S3, the counts of S1, S2, and S4 cannot come entirely from a point source. While the previous Chandra observation was suggestive of a central emission feature intrinsically elongated, possibly mostly due to diffuse emission ( F04), the present imaging analysis suggests a cluster of extended sources. With these data it remains possible that each source is truly extended or produced by several pointlike components, or alternatively, due to a pointlike component embedded in truly diffuse emission. 4.2. Position of the Nucleus (S2) From the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006 ), the center of NGC 821 is at R:A: ¼ 02h 08m 21:14s , decl: ¼ ×10 59 0 41:7 00 (J2000.0), with an uncertainty of 1.2500 (at 95% confidence), as reported in the NASA / IPAC Extragalactic Database ( NED). Considering also the Chandra aspect uncertainty7 (the 90% uncertainty circle has a radius of P0.600 , and the 68% circle has P0.2500 ), this position agrees with that of the extended source S2 determined by wavdetect, which is at R:A: ¼ 02h 08m 21:10s , decl: ¼ ×10 59 0 41:6 00 (J2000.0), with an offset of just 0.600 . In addition, we compared the Chandra and HST positions by using the archival WFPC2 F555W and F814W filter images (also used by Kundu & Whitmore 2001; x 3.1). First we improved the astrometry of the WFPC2 images, refitting the coordinate grid to
The Chandra absolute astrometric accuracy is discussed at http://cxc.harvard .edu /cal /ASPECT/celmon.
7


738

PELLEGRINI ET AL.

Vol. 667

Fig. 5.-- Co-added Chandra ACIS-S image of a central field of 25 00 ; 25 00 in three bands: 0.3 Y1(top left), 1Y2(top right), and 2Y 4 keV (bottom left ). North is up and east is to the left. The bar below the figure gives the correspondence between the color scale and the number of counts per pixel (the size of 1 pixel is 0.500 ). See x 4 for more details. [See the electronic edition of the Journal for a color version of this figure.]

all sources with positional errors P0.300 in the 2MASS All-Sky Catalog of Point Sources8 and the USNO-B1.0 catalog ( Monet et al. 2003). We found five coincidences between the WFPC2 and the USNO-B1.0 sources, two of which are also 2MASS sources; with these, we improved the HST absolute astrometry to P0.200 in the nuclear region. We then compared the recalibrated WFPC2 images with the Chandra image and identified eight sources with optical / X-ray coincidences, not including the nuclear source S2 ( Fig. 8). Of these, seven lie inside the D25 ellipse and have been already picked up in x 3.1 as associated with globular clusters, with an offset smaller than the uncertainty in the X-ray position ( Table 3); however, for the present astrometric analysis, we excluded one of them (CXOU J020817.7+105907 ) located almost
8

See http://irsa.ipac.caltech.edu /applications /Gator.

at a chip edge. The remaining six coincidences are shown with white circles in Figure 8. They have a nearly pointlike appearance, as expected for GCs at the distance of NGC 821, and their V magnitudes range from 20.9 to 22.4; i.e., they are among the brightest GCs of NGC 821 ( Kundu & Whitmore 2001; also discussed in x 3.1). The eighth source with optical / X-ray coincidence is much brighter ( by almost one magnitude in the V band ) than the brightest GC, and lies outside the D25 ellipse (it is located at R:A: ¼ 02h 08m 17:72s , decl: ¼ ×11 00 0 31:22 00 [J2000.0]); it is unresolved and could be a background AGN, as there is a higher chance of a background source outside the optical image. There is no systematic shift or rotation between the revised HST and Chandra positions. For each of the seven coincidences, the displacement between the optical and X-ray positions is 0.2200 , and the rms displacement for the whole sample is ¼ 0:14 00 .The


No. 2, 2007

DEEP CHANDRA OBSERVATION OF NGC 821

739

relative uncertainty between HST and Chandra is therefore good to 0.14 00 , which suggests that the Chandra pointing was in fact very accurate. In the recalibrated WFPC2 image the optical nucleus is located at R:A: ¼ 02h 08m 21:13s ,decl: ¼ ×10 59 0 41:8 00 (J2000.0), that is, within 0.200 of the 2MASS position. As shown by Figure 9, the WFPC2 nucleus is 0.500 off the wavdetect center of S2, a displacement larger than the average 0.1400 offset, which can be attributed to the fact that S2 is an extended source. In conclusion, within the accuracy of the best relative astrometry available, S2 is coincident with the optical center. 4.3. Spectral Analysis of Sources S1Y S4 We extracted the spectra of sources S1Y S4 from the merged event file, using the extraction regions shown in Figure 7; the background spectrum was derived from a source-free region outside the optical body of the galaxy but still inside the ACIS- S3 chip. We also generated instrumental response files, weighted by the exposure time of each individual observation. Prior to fitting, we binned the spectra in order to have at least 20 total counts per energy bin, with the exception of the weakest source (S4) that was rebinned at 10 counts per energy bin.9 We then fitted the spectra with an absorbed power-law model [ XSPEC model: wabs( powerlaw)] and an optically thin thermal model corrected for absorption [ XSPEC model: wabs(apec),withabundance Z ¼ 0:5 in solar units, and the solar abundance values of
9 When grouping with different criteria, or using the Cash statistics instead of 2 statistics, the spectral results for this source do not change significantly.

Fig. 6.-- Co-added smoothed ``true-colo r '' image of a central (see details in x 4). The red, green, and blue colors correspond to and 2Y 4 keV bands, respectively. North is up and east is to the l emission follows the optical shape of the galaxy (see Table 1;

55 00 ; 55 00 field the 0.3 Y1, 1Y2, eft. The diffuse Figs. 1 and 8).

Fig. 7.--ACIS-S image in the 0.3 Y 6 keV band of the central 40 00 ; 25 00 region of NGC 821, with the sources and their extent detected by the CIAO task wavdetect (x 4.1). North is up and east is to the left. Here 1 pixel = 0.2500 . The semimajor and semiminor axes for S1Y S4 are 2.5 00 ; 1.5 00 , 1.7 00 ; 1.5 00 , 1.2 00 ; 1.2 00 , and 1.2 00 ; 1.000 , respectively. [See the electronic edition of the Journal for a color version of this figure.]


740

PELLEGRINI ET AL.

Vol. 667

Fig. 8.-- HST WFPC2 image of NGC 821, with the positions of the Chandra ACIS- S sources detected by wavdetect marked with gray circles. The six white circles show the best optical / X-ray coincidences falling within the D25 ellipse (not including the nuclear source) that were found in x 3.1 and were also used for the astrometric analysis of x 4.2. HST isophotes for the central galactic region are also overlaid. [See the electronic edition of the Journal for a color version of this figure.]

Anders & Grevesse 1989]; the absorption was allowed to be a free parameter. The low number of counts did not allow for fits with composite spectral models. The results are summarized in Table 5. The spectral analysis indicates a hard spectral shape (thermal emission of kT k 3 keV or power-law emission with þ ¼ 1:4 Y 2:3), with little or no intrinsic absorption. The resulting column densities NH are consistent with the Galactic value (6:2 ; 10 20 cmþ2) for most of the fits in Table 5; if NH is fixed to be Galactic, the best-fit spectral parameters are substantially unchanged within the uncertainties. The spectra of S1Y S4 along with their best-fit power-law models are shown in Figure 10. The spectral properties of S1YS4 are overall consistent with those typical of LMXBs (see, e.g., x 3.1), even though these sources look extended (x 4.1). The source at the nucleus (S2) seems to be harder than the other sources. Table 5 lists the source luminosities calculated from the bestfit parameters. S2 has a luminosity of about 6 ; 10 38 ergs sþ1; since it is extended, a nuclear point source, if it contributes to the emission of S2, will have a smaller luminosity. Since we have now established that S2 is coincident with the nucleus (x 4.2), it is reasonable that a fraction of the S2 emission originates from the MBH. However, an estimate of this fraction would not be unique, because it will require some model assumptions (e.g., one could model S2 as a point source plus some extended emission in various ways). F04 derived a 3 upper limit for the nuclear lumi-

nosity of LX < 4:6 ; 10 38 ergs sþ1 in the 0.3 Y10 keV band; they assumed that all of the detected emission in the central S-shaped feature (x 4.1) was truly extended, and calculated the limit from the largest emission found in a 1 00 ; 1 00 sliding cell moved across this feature. In order to avoid model-dependent results, we adopt as an upper limit to the MBH emission the value derived with the same procedure used in F04, with the 1 00 ; 1 00 cell slid over and around S2. This gives a 3 upper limit of LX < 2:8 ; 10 38 ergs sþ1 in the 0.3 Y 8 keV band, and LX < 1:8 ; 10 38 ergs sþ1 in the 2 Y10 keV band. 5. THE DIFFUSE EMISSION AND THE HOT ISM CONTRIBUTION Estimating the amount of gaseous emission in NGC 821 is important, because this hot gas could be a source of fuel for the nuclear MBH (see x 1). The ACIS- S adaptively smoothed mapped color image of the NGC 821 field ( Fig. 6 ) shows diffuse emission at the center, roughly aligned with the major axis of the galaxy. However, this diffuse emission is not by itself proof of the presence of a hot ISM, since an important contaminant in a galaxy like NGC 821, which has a low hot gas content (x 1), is the undetected portion of the LMXB population (see, for example, the case of NGC 1316; Kim & Fabbiano 2003); another contaminant could be the emission of normal stars (first discussed for this type of galaxy in Pellegrini & Fabbiano 1994).


No. 2, 2007

DEEP CHANDRA OBSERVATION OF NGC 821

741

Fig. 9.--ACIS-S image of NGC 821, with gray circles of 0.700 radii marking the positions of the CIAO wavdetect sources (1 pixel = 0.500 ), and the white circles corresponding to GCs (as in Fig. 8). The galactic isophotes are also overlaid. In the recalibrated WFPC2 image, the galactic center is located at R:A: ¼ 02h 08m 21:13s , decl: ¼ ×10 59 0 41:8 00 (J2000.0) (see x 4.2) and falls within the circle of the Chandra source S2. [See the electronic edition of the Journal for a color version of this figure.]

TABLE 5 Sp ectral Analysis Results for t he Central Sources S 1YS4 Parameter S1 S2 S3 S4 82 ô 9

Net counts .................................. 178 ô 14 246 ô 16 324 ô 18 wabs( pow) NH (10 21 cmþ2) .......................... þ ................................................. 2/dof ......................................... Flux (10þ15 ergs cmþ2 sþ1) ....... Luminosity (10 38 ergs sþ1)........ 1:7×0::8 þ1 0 ×0 35 1:80þ0::23 7.3/6 5.8 4.7 wabs(apec) NH (1021 cmþ2) .......................... kT ( keV ) .................................... 2/dof ......................................... Flux (10þ15 ergs cmþ2 sþ1) ....... Luminosity (10 38 ergs sþ1)........ <1.1 5:3×4:: þ1 7.0/6 5.7 4.3
7 8

<0.5 1:49×0::14 þ0 13 11.5/9 8.6 6.0

2:6×0::5 3:0×1::3 þ0 6 þ1 5 ×0 52 2:23×0::23 2:28þ0::34 þ0 17 15.8/12 2.1/5 9.1 2.3 9.0 1.9

In this section we report the results of the analysis of this diffuse emission. First, we analyze the merged data to estimate its spatial and spectral properties (xx 5.1 and 5.2). Since our deep data allow for an accurate characterization of the LMXB population (x 3), we then use these results to estimate how much of the diffuse emission could come from unresolved LMXBs below the detection threshold, comparing (1) the shape of the radial profiles observed for the diffuse emission, the resolved point sources, and the stellar optical light (x 5.3); (2) the radial profile of the soft emission contributed by LMXBs and the total observed soft profile (x 5.4); and (3) the total luminosity of the diffuse emission with that expected from unresolved LMXBs, using the XLF derived in x 3.2 (x 5.5).Finally,in x 5.6, we derive the expected contribution from other types of stellar sources. 5.1. Radial Profile After having excluded all the detected sources, we derived a radial profile of all the remaining counts in the 0.3 Y 6 keV energy band; at energies > 6 keV the background contribution becomes dominant, and the addition of these photons would only increase the error bars. The profile was centered on the nucleus and included 30 circular annuli (each $500 wide). The resulting radial profile is shown in Figure 11, where the presence of emission is evident as far as $2000 Y 3000 from the nucleus ($2.3 Y 3.5 kpc at the galaxy

<0.5 8:5×114 þ3: 11.9/9 8.1 5.6

0:9×0::6 þ0 5 4:3×1::0 þ0 8 18.4/12 9.7 7.6

1:4×1:: þ1 3:4×2:: þ0 1.9/5 2.4 1.7

5 3 2 9

Notes.--The NH values are in addition to the Galactic one. Fluxes are observed values, while luminosities are corrected for absorption; both are calculated for the 0.3 Y 8 keV band. Errors give the 68% confidence interval for one interesting parameter.


742

PELLEGRINI ET AL.

Vol. 667

Fig. 10.-- Chandra ACIS-S spectra of the extended sources S1Y S4, together with their best-fit power-law models ( Table 5). From left to right, the panels refer to S1 and S2 (top), and S3 and S4 (bottom). See x 4.3 for more details.

distance). The flattening at larger radii is consistent with the expected background level. Therefore, diffuse emission is detected out to a radius that lies well within the optical extent of the galaxy, even within one effective radius ( Table 1). 5.2. Spectral Analysis of the Diffuse Emission We analyzed the spectra of three circles centered on the galactic center, of progressively larger radius: 500 and 1000 , to investigate the gas presence in the circumnuclear region, and 30 00 , which is the outermost radius at which diffuse emission is detected. Prior to their extraction, we removed regions including detected sources. The background was estimated from a source-free circular area of 50 00 radius located outside the galaxy. We fitted the spectra with single-component models, such as a power law ( XSPEC model wabs*pow), as expected from the integrated LMXB emission, and a thermal APEC model ( XSPEC model wabs*apec), to account for the emission of a hot gas component; we then used a composite power-law and APEC model [wabs(apec+pow)]. The results are summarized in Table 6; errors quoted below give the 68% confidence interval for one interesting parameter. The spectrum of a central circle of 500 radius is well fitted by a single absorbed power-law model, with a photon index þ ¼ 1:77 ô 0:17 and NH consistent with the Galactic value, which is in good agreement with the spectral parameters of LMXBs (x 3.1). The thermal model also gives an acceptable fit, provided that the temperature is high (kT $ 5 keV at the best fit), which again is what is expected for LMXBs ( Fabbiano 2006 ). Next we attempted the wabs(apec+pow) spectral model, to probe for the presence of

Fig. 11.-- Radial profile of the 0.3 Y 6 keV emission, after removal of sources detected by wavdetect (x 5.1). The flattening at radii k2500 is due to field background. Vertical bars give the ô1 uncertainty.


No. 2, 2007

DEEP CHANDRA OBSERVATION OF NGC 821
TABLE 6 Sp ectral Analysis of the Dif fuse Emission ( P rojected Re gions)

743

Radius (arcsec) 5.................................. 10................................ 30................................

Net Counts 166 ô 13 175 ô 16 581 ô 48 1: 1: 1: 1: 1: 1:

þ
× 77þ × 61þ × 81þ × 57þ × 74þ × 59þ 0:17 0:17 0:23 0:21 0:13 0:13 0:12 0:15 0:14 0:14 0:09 0:17

(10

þ15

F (0.3 Y 8 keV ) ergs cmþ2 sþ1) 8.4 ô 0.8 8.6 ô 1.2 9.8 ô 0.7 10.0 ô 1.4 17.9 ô 1.3 ×3 1 16:9þ2::3

L (0.3 Y 8 keV ) (10 38 ergs sþ1) 6.7 ô 0.7 6.7 ô 1.0 7.9 ô 0.6 7.8 ô 1.1 14.3 ô 1.10 13:0×2::40 þ1 70

kT ( keV ) .. . ×0 16 0:21þ0::20 .. . ×0 08 0:25þ0::06 .. . ×0 18 0:59þ0::18

(10

þ15

F (0.3 Y 8 keV ) ergs cmþ2 sþ1) . .. :6 0:37×1318 þ0: . .. 0:69×0::35 þ0 32 . .. 1:06×0::42 þ0 71

L (0.3 Y 8 keV ) (10 38 ergs sþ1) . .. :2 0:44×1622 þ0: . .. 0:75×0::38 þ0 35 . .. 0:96×0::38 þ0 64

2/dof 8.6/9 7.4/7 25.0/19 21.9/17 78.2/64 71.9/62

Notes.--NH is fixed at the Galactic value of 6:2 ; 10 20 cmþ2 ( Table 1); in the apec spectral model, the abundance is fixed at the solar value. Fluxes are observed; luminosities are intrinsic. Errors give the 68% confidence range for one interesting parameter.

a hot gas component, in addition to the LMXB emission that is clearly dominant. The NH was fixed at the Galactic value and the abundance to the solar value, which is an average for the stellar population in the central galactic region ( Proctor et al. 2005 ). At the best fit, þ ¼ 1:61×0::23 and kT ¼ 0:21×0::16 keV; however, this þ0 20 þ0 21 fit does not represent a statistically significant improvement with respect to the simple power-law model, as was established via a ``calibration'' of the F-test through simulations, following the prescriptions10 discussedinProtassov et al.(2002;see their x 5.2). The analysis of the spectrum of a central circle of 10 00 radius gave similar results ( Table 6). This spectrum is well described by a power law consistent with the average spectrum of LMXBs ( þ ¼ 1:81 ô 0:13 and Galactic NH ), and again a thermal component is not statistically required. We then analyzed the spectrum of the diffuse emission from the whole galaxy (i.e., from within a radius of 30 00 ). Again, a simple power-law model gives an acceptable fit, with þ ¼ 1:74 ô 0:14. If a thermal component is added, with solar abundance and Galactic NH, then þ ¼ 1:59×0::09 and kT ¼ 0:59 ô 0:18 keV; howþ0 17 ever, this thermal component is not required statistically. The 68% confidence upper limit on its luminosity is L(0:3 Y 8 keV ) < 1:34 ; 10 38 ergs sþ1. Based on the shallower pointings of 2002, David et al. (2006 ) estimated a 90% confidence upper limit on L (0:5 Y 2 keV ) of 7:0 ; 10 38 ergs sþ1 for a thermal component. In the analysis above, the presence of a small amount of thermal emission confined to the central region might be overwhelmed by the emission of unresolved LMXBs along the line of sight. Therefore, we also attempted the deprojection of the spectral data by using the technique implemented within XSPEC, for which the spectra extracted from annuli centered on the nucleus are compared with the spectra expected from the superposition along the line of sight of the emission coming from the corresponding spherical shells (as successfully done for the Sombrero bulge; Pellegrini et al. 2003a). The central region was therefore divided in an inner circle of 500 radius, a surrounding annulus with outer radius of 10 00 , and another surrounding annulus of outer radius of 20 00 . The spectra of each of these regions were jointly fitted with the spectral models used above. Unfortunately, the outermost annulus did not have enough counts to add meaningful spectral information; therefore, only the two inner regions were deprojected. The simple power-law model ( projct*wabs*powerlaw), with the photon index constrained to be the same in the two regions, gives an
10 The basic steps of the procedure consist of fitting the observed data with the simple ( pow) and composite (apec+pow) models, getting the F-statistic for these fits, simulating a large number of spectra for the simple best-fit pow model with XSPEC, fitting each of the simulated data sets with both the simple and the composite model and recording the F-statistic, and getting the distribution of the simulated F-statistic to compute a p-value ( Protassov et al. 2002) that establishes whether the addition of the thermal spectral component is significant or not.

acceptable fit for þ ¼ 1:68þ0:15 and NH ¼ 2:5×8::0 ; 10 20 cmþ2 þ2 5 (inner region) and <4:3 ; 10 20 cmþ2 (outer region). The addition of a thermal component [ XSPEC model projct*wabs(apec+ pow), with solar abundance and Galactic NH]gives a best-fit þ ¼ 1:73×0::17 and temperatures of kT ¼ 0:17×0::06 keV for the outer þ0 15 þ0 07 annulus, and kT ¼ 0:03×0::09 keV for the inner circle. The latter þ0 03 value is too small to be meaningful; therefore, the fit was repeated with a fixed þ ¼ 1:7 (typical of LMXBs; x 3.1). The outer temperature remained unchanged, and a more meaningful temperature was obtained for the inner circle (kT ¼ 0:18 keV ), but it was basically unconstrained. Again, the addition of the thermal component was not statistically significant. At the best fit, the luminosity of the thermal component is $5% and $13% that of the power law, respectively, for the inner and outer regions. In summary, the analysis of the projected regions and of the central deprojected regions never requires the presence of a soft thermal component at a statistically significant level. The luminosity of a soft thermal component inserted in the fit is much lower than that of the power law (<1/10 at the best fit) for the spectra of the 500 , 1000 , and 3000 circles. 5.3. Radial Profiles of Detected Sources, Diffuse Emission, and Galactic Optical Emission Figure 12 shows the comparison of the radial profiles of three quantities: (1) the background-subtracted diffuse emission in the 0.3 Y 6 keV band, (2) the number density of resolved point sources, and (3) the R-band emission of NGC 821. The latter, which gives the distribution of the stellar surface brightness, is well described by a de Vaucouleurs (1948) law, I (R) / exp fþ7:669½(R/43:9 00 )1/4 þ 1g,where R is in arcseconds (Soria et al. 2006b). All three profiles follow the same radial trend. The agreement between the radial profile of resolved point sources and that of the stellar light is not surprising (see, e.g., Kim & Fabbiano 2003; Fabbiano 2006 ). The similarity of these profiles with that of the diffuse emission provides instead support to (or at least is consistent with) the idea that also the diffuse emission mostly comes from undetected LMXBs. 5.4. Simulated ( LMXB) and Observed Soft Radial Profiles If the diffuse emission is mostly due to unresolved LMXBs, as suggested by the analysis of xx 5.2 and 5.3, we would expect its spectrum to have a power-law shape typical of the LMXB population at all radii. Any deviation is suggestive of a localized additional hot ISM component. In order to further check the lack of a soft thermal component indicated by the previous analysis at all radii, we made the following test. We derived from observations the radial profile of the diffuse hard emission in the 1.5 Y 6 keV band; in the hypothesis that it

×0:26


744

PELLEGRINI ET AL.

Vol. 667

Fig. 12.-- Radial profiles of detected point sources (black), in units of number per square arcminute; background-subtracted diffuse emission over 0.3 Y 6 keV ( gray), in units of counts per pixel; and galactic R-band emission (dashed lines; arbitrarily normalized ). See x 5.3 for details. [See the electronic edition of the Journal for a color version of this figure.]

is produced by a distributed source whose spectrum is a power law of þ ¼ 1:7 and Galactic NH , we then used it to derive a ``simulated'' radial profile in the 0.3Y1.5 keV band. The chosen spectrum represents the emission of unresolved LMXBs, in accordance with the results of x 3.1. In Figure 13 we compare the simulated soft profile with the observed profile of the diffuse emission in the same energy band. The agreement between the two profiles is good, consistent with the idea that the observed soft emission is due substantially to unresolved binaries. Only within the central circle of 1000 radius could an additional contribution from a soft source be present, but it is not revealed by the spectral analysis (x 5.2). 5.5. The Unresolved Emission Implied by the XLF The emission from undetected LMXBs can be recovered from the XLF derived in x 3.2 specifically for NGC 821, as the difference between the total luminosity expected from LMXBs based on the XLF, LX (total), and the luminosity due to all detected sources, LX (detected). In the 0.3 Y 8 keV band, LX (detected ) ¼ 6:5 ; 10 39 ergs sþ1 for the sources falling within the D25 ellipse. For the same region, LX (total) is not just given by an integration of the XLF, since the XLF was derived for the D25 ellipse with a central circle of 10 00 radius excluded (x 3.2). The LX (total) for the whole D25 ellipse is then obtained by rescaling the luminosity resulting from the XLF by the ratio between the luminosity of detected sources in the two regions (i.e., the D25 ellipse and this ellipse without the central circle).11 This procedure relies on the assumptions that (1) the XLF for R < 10 00 and R > 10 00 is the same, and (2) LX (total )/LX (detected) is also the same in these two regions. While (1) is reasonable, (2) is less valid, since more sources would be hidden in the inner region. Therefore, both LX (total )
This integration assumes that the differential XLF has the same slope ($2) down to LX ¼ 10 37 ergs sþ1 and is done over the range LX ¼ (10 37 Y10 39 )ergs sþ1. Given the slope value, sources of luminosities below 10 37 ergs sþ1 do not contribute significantly to the result; for example, decreasing the lower boundary by a factor of 2 would increase the resulting total luminosity by $15%. In addition, a lowluminosity break in the XLF is also expected (e.g., Kim et al. 2006a).
11

Fig. 13.-- Radial distribution of the diffuse 0.3 Y1.5 keV emission (in black) that is implied by the observed 1.5 Y 6 keV emission, when assuming for it the spectral model described in x 5.4. In gray, the observed profile in the same 0.3 Y1.5 keV band is also shown. The flattening at large radii is due to field background. Vertical bars give the ô1 uncertainty. [See the electronic edition of the Journal for a color version of this figure.]

and LX (undetected ) ¼ LX (total ) þ LX (detected) are actually lower limits. For the D25 ellipse, LX (total) is 7:6 ; 10 39 ergs sþ1, and LX (undetected ) ¼ 1:1 ; 10 39 ergs sþ1. It is interesting now to compare the unresolved LMXB emission with the luminosity of the diffuse emission derived from the spectrum of a central circle of 30 00 radius (i.e., 1:4 ; 10 39 ergs sþ1; x 5.2). Therefore, we repeated the procedure above to calculate LX (undetected) for this circle, and it turned out that LX (undetected ) ¼ 7:4 ; 10 38 ergs sþ1. This value is slightly lower than the spectral luminosity of the diffuse emission; however, as noted above, it is likely an underestimate of the true value. Also, considering the uncertainty in the normalization of the XLF (k ¼ 8:4×3::1 ; x 3.2), LX (undetected) can be as large as þ2 5 2:7 ; 10 39 ergs sþ1. In summary, for the same circle of R ¼ 30 00 , the spectral luminosity of the diffuse emission and the XLF-based unresolved luminosity are close, and they agree within the uncertainties. 5.6. Other Stellar Sources Since we are constraining the origin of the diffuse emission in NGC 821 better than it has been possible so far in previous analyses of low LX /LB ellipticals, we check also for the amount of diffuse emission that can be accounted for by stellar sources other than LMXBs. In the old stellar population of an elliptical galaxy, these include coronae of late-type main-sequence stars, RS Canum Venaticorum ( RS CVn) systems, and supersoft X-ray sources ( Pellegrini & Fabbiano 1994). The latter are characterized by blackbody emission with effective temperatures of 15 Y 80 eV and X-ray luminosities up to a few ; 10 38 ergs sþ1, and are more frequent in late-type and irregular galaxies ( Kahabka & van den Heuvel 1997; Di Stefano & Kong 2004). NGC 4697 is the only elliptical where they have been identified so far (Sarazin et al. 2001). In the X-ray color-color plot of NGC 821 ( Fig. 2), supersoft sources would fall in the upper right corner, and there is no such source with


No. 2, 2007

DEEP CHANDRA OBSERVATION OF NGC 821

745

>30 counts; one or two sources with less than 30 counts have a high C32 color, but with a large error. The contribution of supersoft sources will not be considered further here. Stellar coronae of main-sequence stars are all sources of thermal X-ray emission (LX % 10 26 Y10 30 ergs sþ1), indicative of coronal plasmas at temperatures of $106Y107 K (e.g., Schmitt et al. 1990). These luminosities are faint in comparison to accretionpowered stellar sources, but main-sequence stars of spectral types G, K, and M are present in a very large number ( hereafter, respectively, NG , NK and NM ). These Ni can be recovered from the initial stellar mass function ( IMF ) (M ) ¼ AM þ2:35 (Salpeter 1955 ), where the scale factor A ¼ 1:67LB,with LB in LB; ,for a 12 Gyr old stellar population with a 0.5 solar metallicity ( Maraston 2005 ), as is suitable for NGC 821 ( Proctor et al. 2005 ). Then R Msup Ni ¼ A Minf M þ2:35 dM , where Minf and Msup are, respectively, 0.7 M and the main-sequence turnoff mass of 0.9 M (as is appropriate for NGC 821; Maraston 2005 ) for G stars; 0.5 and 0.7 M for K stars; and 0.1 and 0.5 M for M stars. Integration gives NG ¼ 0:6LB , NK ¼ 1:1LB , and NM ¼ 24:5LB ; therefore, the collective M dwarf X-ray emission is by far the most important, because the average X-ray luminosity of a G, K, and M star ( hereafter LX;M ) is comparable ( Kuntz & Snowden 2001). Taking LX; M P 1:4 ; 10 27 ergs sþ1 in the 0.3 Y 8 keV band,12 the collective M dwarf emission in NGC 821 is then NM LX; M P 6:4 ; 10 38 ergs sþ1. From a central circle of 3000 radius, for the de Vaucouleurs optical profile of x 5.3, the M dwarf emission is P2:5 ; 10 38 ergs sþ1,which is P18% of the luminosity derived from the spectrum of the diffuse emission from the same region (x 5.2). However, recent claims favor an IMF flatter than the Salpeter one at the low-mass end, as for the Kroupa (2001) IMF, which is / Mþ1.3 for 0:08 M /M 0:5. In this case the scale factor A ¼ 1:69LB ( Maraston 2005 ), the integration gives NM ¼ 4:3LB , and the collective M dwarf emission becomes P1:1 ; 10 38 ergs sþ1 for the whole galaxy and P0:5 ; 10 38 ergs sþ1 for a circle of 3000 radius. Therefore, it can account for P3% of the luminosity of the diffuse emission. RS CVn systems are chromospherically active objects, consisting of a G or K giant or subgiant, with a late-type main-sequence or subgiant companion ( Linsky 1984). They are the most X-rayluminous late-type stars (LX $ 10 29 Y10 31:5 ergs sþ1), and their spectra can be well modeled by a thermal plasma with two temperatures (of average kT ¼ 1:3 and 0.18 keV; Dempsey et al. 1997 ). From Dempsey et al.'s Figure 3, the average 0.3 Y 8 keV luminosity of a system is P3 ; 10 30 ergs sþ1.13 The number of RS CVn systems is a fraction of the number of giants and subgiants (Ngiant and Nsubg , respectively) expected to be present in NGC 821, which is Nsubg × Ngiant % 0:04LB .14 By taking half this number for those giants or subgiants that are also in binary systems, and another factor ($0.2) for those that become RS CVn, the collective contribution is P0:1(Nsubg × Ngiant ) ; (3 ; 10 30 ) ergs sþ1 ¼ (1:2 ; 10 28 )LB ergs sþ1. This is P2:2 ; 10 38 ergs sþ1 for the whole galaxy, and P8:9 ; 10 37 ergs sþ1 from acircle of 3000 radius, which is P6% of the total diffuse emission
12 The ``P'' sign comes from assuming thermal emission of kT ¼ 1keV when converting LX; M in the ROSAT band ( Kuntz & Snowden 2001) to the 0.3 Y 8keV band. The proper spectral description consists of two thermal components: one with T $ (2Y 4) ; 106 K and the other with T $ 107 K; the latter usually has a larger (and often much larger) emission measure (Giampapa et al. 1996 ). 13 Again, the ``P'' sign comes from assuming that most of the emission is due to the higher temperature component. 14 Ngiant and Nsubg are the products of the time spent by stars in such evolutionary phases by the specific evolutionary flux for a 12 Gyr old stellar population (which is fairly independent of the IMF; Maraston 2005 ) and by LB. Low-mass metal-rich stars spend 7 ; 108 yr in the G/ K subgiant phase, and 5 ; 108 Gyr on the red giant branch (e.g., Renzini 1989).

within the same region. A somewhat higher percentage ($15%) is derived in the 2Y10 keV band when adopting the emissivity per unit stellar mass of these sources in the Galactic plane (Sazonov et al. 2006 ) and assuming an average stellar mass-to-light ratio in the B band of $6.5 (x 6.1). This higher percentage could be in part the result of a higher emissivity in the Milky Way, due to younger stellar ages than that estimated for NGC 821 (see above). In conclusion, the collective emission of stellar sources other than LMXBs from the whole galaxy is P3:3 ; 10 38 ergs sþ1;as expected, this is much smaller than the collective LMXB's emission (7:6 ; 10 39 ergs sþ1; x 5.5), but it is even smaller than its unresolved fraction (i.e., LX ½undetected¼ 1:1 ; 10 39 ergs sþ1; x 5.5). Also, these sources can contribute up to $10% of the diffuse emission detected within a radius of 30 00 (x 5.2). 5.7. Conclusions on Hot Gas Presence All the different lines of investigation undertaken in this section consistently indicate a largely dominating LMXB contribution to the diffuse X-ray emission. Its spectrum does not require soft thermal emission, which is limited to <1:34 ; 10 38 ergs sþ1 (<10% of the total diffuse luminosity) at 68% confidence. The observed radial profiles of the stellar light, the resolved point sources, and the diffuse emission agree within the errors, suggesting that the diffuse emission has a stellar origin. The observed radial profile in the soft band is consistent with that expected from the sources (the LMXB population) producing the hard emission, except perhaps for the innermost (R P 10 00 ) region. Finally, the luminosity of undetected sources derived from the bias-corrected XLF agrees with the spectral luminosity of the diffuse emission, within the uncertainties. A fraction of the diffuse emission (P10%) can also come from stellar sources other than LMXBs. 6. DISCUSSION In the previous sections, we have reported the results of the analysis of a deep Chandra pointing (for a total exposure of nearly 230 ks) aimed at detecting the nuclear emission and setting stringent constraints on the presence of hot gas on both the circumnuclear and the galactic scale. The deep Chandra image revealed a significant portion of the LMXB population, for which the X-ray luminosity function was derived. This, together with a spectral and imaging analysis of the diffuse emission, severely constrained the presence of a hot ISM (x 5). At the galactic center, in addition to diffuse emission (which we can explain as predominantly due to unresolved LMXBs), we find four sources of LX ¼ (2 Y 9) ; 10 38 ergs sþ1. We have been able to establish that one of these sources (S2) is coincident with the galactic center. Three of these sources, including S2, are extended, with one of them as long as 500 (S1; Fig. 7). These sources could be truly extended, or made of one or more pointlike components, possibly also embedded in truly diffuse emission. Their spectral shape is consistent with that of LMXBs, but in NGC 821 there are only two other sources as bright as these, outside the central circle of 10 00 radius ( Fig. 4). This supports the idea that more than one LMXB is contributing to the emission of each source. An alternative possibility is that they belong to a continuous complex feature, made of LMXBs (that may also lie there due to a projection effect), a nucleus ( possibly with associated extended emission), and an outflow/jet. S1, in particular, has a linear structure suggestive of a jet ( Fig. 7). Our VLA observations (x 1) help better address these possibilities and are presented in the companion paper ( Pellegrini et al. 2007 ), together with a discussion on whether / how accretion proceeds in this nucleus, based on all the available observational evidence. We estimated here a 3 upper limit to the 0.3 Y 8 keV emission of a


746

PELLEGRINI ET AL.

Vol. 667

pointlike source associated with the MBH of <2:8 ; 10 38 ergs sþ1, which makes this one of the quietest MBHs studied with Chandra (e.g., Pellegrini 2005b), with LX /LEdd < 2:5 ; 10þ8 . In the following we discuss the possibility that there may be undetected hot gas in NGC 821 (x 6.1). Hidden hot gas on the nuclear scale could be a source of fuel for the MBH; furthermore, we check whether the lack of detection on the galactic scale agrees with expectations for the evolution of the ISM in this elliptical. 6.1. Is There Hot Undetected Gas? The present analysis does not detect hot gas available for accretion at the nucleus; coupled with the absence of ISM observed at other wavelengths (x 1), the lack of fuel for the MBH could be the simplest explanation for the very low levels to which the nuclear emission has been constrained. However, an aging stellar population continuously returns gas to the ISM via stellar mass losses (e.g., Ciotti et al. 1991; David et al. 1991); as a minimum, the circumnuclear region should be replenished with this fuel. Below we discuss the possible presence of this ISM. The lack of detection of hot gas on the galactic scale can be expected from simple energetic considerations, comparing the energy made available per unit time by Type Ia supernova explosions (LSN) with the energy required to steadily extract from the galactic potential well the stellar mass losses produced per unit time (Lgrav; Ciotti et al. 1991), for the whole galaxy lifetime. The LSN comes directly from the observed SN Ia's explosion rate in ellipticals (Cappellaro et al. 1999), rescaled for LB in Table 1; this rate was higher in the past (e.g., Greggio 2005 ). The Lgrav derives from the stellar mass loss rate and the galactic mass profile, which can be well modeled by the superposition of two Hernquist (1990) density profiles:15 one for the stars with total mass Mö and one for the dark matter with mass Mh . This model is then tailored onto NGC 821 taking the observed LB , Re , central stellar velocity dispersion ( Table 1), and Mh /Mö ratio. Anisotropic Jeans models reproducing the observed velocity dispersion profiles of stars and planetary nebulae out to 5Re constrain the range of (Mh × Mö )/LB to be 13 Y17 ( Romanowsky et al. 2003). For a plausible stellar mass-to-light ratio Mö /LB (Gerhard et al. 2001; Napolitano et al. 2005 ), one derives Mh /Mö $ 1:2. Finally, the stellar mass loss rate is distributed as the stars, and derives by summing the mass ejected by each star as a function of its mass ( Renzini & Ciotti 1993) for a Kroupa IMF, scaled for NGC 821 as described in x 5.6; this produces a total mass loss rate of 0.25 M yrþ1 after 12 Gyr, in good agreement with the values estimated for ellipticals from Infrared Space Observatory (ISO) observations (Athey et al. 2002). The resulting LSN $ (3 Y 4)Lgrav during the galaxy lifetime, with LSN $ 4Lgrav at present; therefore Type Ia SNe have always provided the heating to drive the stellar mass losses in a galactic wind (neglecting the effect of radiative cooling that is not expected to be important for very low gas densities typical of a wind ). However, this is a global energetic calculation, and the galaxy may host regions with LSN > Lgrav and regions with LSN < Lgrav . For centrally peaked mass distributions like those described above, numerical simulations of hot gas evolution showed that this is indeed the most frequent case ( Pellegrini & Ciotti 1998): the ISM is driven out of the outer galactic regions in a wind, while it is inflowing within a stagnation radius that can be largely different from
This model gives a very good approximation of the de Vaucouleurs (1948) law, which fits the light profile of NGC 821 (x 5.3). The radial distribution of the dark halos of ellipticals is not well constrained by the observations. Theoretical arguments favor a peaked profile (Ciotti & Pellegrini 1992; Evans & Collett 1997 ). High-resolution numerical simulations ( Navarro et al. 1996 ) produce a central density distribution equal to that of the Hernquist model.
15

galaxy to galaxy. To explore whether this situation applies to NGC 821, we ran hydrodynamical simulations specific for the galaxy model described above, plus the additional gravitational attraction contributed by its central MBH (see Pellegrini & Ciotti [2006] for more details on the numerical code and the time-evolving input quantities). The central grid spacing was set to 5 pc to allow for a better sampling of the inner regions, and the gas flow evolution was followed for 12 Gyr, an age comparable with the stellar age of NGC 821 ( Proctor et al. 2005 ). The flow kept in a partial wind all the time, and at the end its total LX $ few ; 10 36 ergs sþ1 in the 0.3 Y 8 keV band, well below our observational limit on the hot gas. The bulk of the hot gas was outflowing, and directed toward the center from within a radius of $25 pc. At the smallest radius at which the flow is well resolved (10 pc), the mass inflow rate is Min % few ; 10þ5 M yrþ1; the gas density is ne % 10þ3 cmþ3, one-third of the density value calculated from the emission measure of the ``best fit'' thermal component for the central circle, in the deprojected analysis of the diffuse emission (this component is not statistically required; x 5.2). Note however that the inflowing region is very small and fully included within the extent of the S2 source (x 4.1). By varying the Mh /Mö ratio, the SN Ia's rate, the age of the galaxy, and the stellar mass loss rate within the limits im posed by observational uncertainties, the value of Min keeps within the range (2Y7) ; 10þ5 M yrþ1. The implications for the nuclear emission coming from this estimate of Min are dis cussed in Pellegrini et al. (2007 ). Here we just note that Min ranges from 10þ3 to a few ; 10þ5 M yrþ1 in the past $10 Gyr, and its integration over this time gives a total mass accreted at the center of $2 ; 106 M , that is, 2:3 ; 10þ2 of the observed MBH mass ( Table 1). Therefore, accretion of stellar mass losses was not effective in building this MBH mass. Finally, we comment on the fact that, by using Chandra observations, the mass accretion rate of galactic nuclei is customarily estimated from the analytic formula of Bondi (1952), valid for spherically symmetric accretion from a nonrotating polytropic gas with given density and temperature at infinity (e.g., Loewenstein et al. 2001; Soria et al. 2006a). Infinity is replaced with an accretion radius racc ¼ 2GMBH /cs2 ( Frank et al. 2002), where the sound speed cs / (kT )1/2 is calculated as closely as possible to the MBH. Even in those cases where kT can be observed close to the galactic center, so that one can be confident to derive an approximate estimate of racc , there are additional ingredients that must be considered in the estimate of the mass accretion rate that were not included in the Bondi (1952) treatment: (1) the presence of mass and energy sources (i.e., stellar mass losses and SN Ia heating), (2) the presence of cooling, and (3) the fact that racc is not a true ``infinity'' point, since the gas here experiences a pressure gradient. The simulations described here take into account all these aspects and should give a more reliable mass accretion rate than the Bondi theory. However, they have some limits too. At these very low Min values and very small inflowing regions, the detailed shape of the galactic mass profile becomes important. In the case of NGC 821, the observed profile is steeper than modeled here at the very center (Gebhardt et al. 2003), which should produce a larger Min. The true ac cretion rate could be somewhat higher than Min also if even the stellar mass losses within the innermost radius resolved by the simulations are to be accreted. But further pursuing the flow behavior closer to the MBH with the simulations used here presents a problem: the discrete nature of the stellar distribution becomes important, since the accretion time ($3 ; 105 yr from 10 pc, in the simulations) becomes comparable to (or lower than) the time required for the stellar mass losses to mix with the bulk flow ( Mathews 1990), and the time elapsing between one SN Ia event and the next.


No. 2, 2007 7. SUMMARY

DEEP CHANDRA OBSERVATION OF NGC 821

747

We have observed the nearby inactive elliptical galaxy NGC 821, known to host a central MBH, with deep Chandra pointings, in order to put strong constraints on its nuclear emission and the presence of gas available for accretion. Our results can be summarized as follows. 1. We detect 41 sources within the optical size ( D25 ) of the galaxy; excluding four particular sources at the center, these represent the LMXB population of NGC 821. Their X-ray colors, and the spectral analysis of the three brightest ones, are consistent with spectral shapes described by power laws of photon index in the range þ ¼ 1:5Y2:0, without significant intrinsic absorption. There is no clear evidence for supersoft sources. 2. In the overlap region between D25 and the field of view of previously taken HST WFPC2 images, there are six clear associations of LMXBs with globular clusters, plus two additional marginal matches. These LMXBs reside in the brightest GCs of the galaxy. Excluding the two marginal matches, $30% of the X-ray sources lie in GCs, and $6.6% of the GCs have a LMXB, broadly in agreement with the percentages found for ellipticals in previous works. 3. The XLF of these sources is well fitted by a simple power law down to LX ¼ 3 ; 10 37 ergs sþ1 (or to LX ¼ 2 ; 10 37 ergs sþ1 after completeness correction), with a slope (  ¼ 2:0 ô 0:3) similar to that found previously for other ellipticals. The ratio between the total LX due to LMXBs and the galactic LB is lower than the average found for early-type galaxies, but still within the observed scatter. 4. At the position of the galactic center a source (S2) is detected for the first time. Its spectral shape is quite hard ( þ ¼ 1:49×0::14 ), þ0 13 without intrinsic absorption. S2 is, however, slightly extended, and a 0.3 Y 8 keV upper limit of 2:8 ; 10 38 ergs sþ1 is derived for a pointlike nuclear source, one of the smallest established with Chandra. 5. Three other sources, with a spectral shape typical of LMXBs, are detected in the central galactic region; they are as bright as the brightest sources in the XLF. Only one (S3) is consistent with being pointlike; the other two (S1 and S4) are extended, and could be due to the superposition of a few LMXBs and /or truly diffuse emission. The morphology of S1 resembles a jetlike feature.

6. Diffuse emission is detected out to a radius of $3000 . A few independent lines of investigation, exploiting the spectral and imaging capabilities of Chandra, consistently indicate that this diffuse emission is due to unresolved LMXBs and provide no evidence for hot gas, either close to the center or on a larger scale. The spectral analysis gives a (1 ) limit of L(0:3 Y 8 keV ) < 1:34 ; 10 38 ergs sþ1 on any soft thermal component within a radius of 30 00 . 7. Other unresolved stellar sources (mostly M dwarfs and RS CVn systems) could contribute to the diffuse emission by P10%. The general scaling of their collective LX with the galactic LB was also derived. 8. Numerical simulations of the hot gas evolution for a galaxy model tailored on NGC 821 show that the bulk of the gas is driven out in a wind for the whole galaxy lifetime, due to the heating provided by Type Ia supernovae. While this may have been expected based on simple energetic calculations, the simulations also show that the gas is accreting toward the center from within a very small inner region. This gas flow pattern is expected to be common in low-LB ellipticals, due to their cuspy central mass distribution. Gaseous accretion alone at the rates given by the simulations cannot have formed the nuclear MBH of NGC 821. In a companion paper ( Pellegrini et al. 2007 ), we present further observational results for the central region of NGC 821, obtained with Spitzer and the VLA, and discuss various possibilities for the nature of the accretion process in this very low luminosity nucleus.

We thank F. Ferraro for useful information related to x 5.6, the referee for useful comments, and A. Kundu for kindly providing the list of globular clusters used in x 3.1 and information on possible optical counterparts of the X-ray sources in Table 3. S. P. acknowledges partial financial support from the Italian Space Agency ASI (Agenzia Spaziale Italiana) through grant ASI-INAF I /023/05/0. Partial support for this work was provided by the NASA Chandra Guest Observer grant GO5-6110X and by the Chandra X-Ray Center NASA contract NAS8-39073. The data analysis was supported by the CXC CIAO software and CALDB. We have used NASA NED and ADS facilities, and have extracted archival data from the Hubble Space Telescope archive.

REFERENCES Anders, E., & Grevesse, N. 1989, Geochim. Cosmochim. Acta, 53, 197 Fabbiano, G., Baldi, A., Pellegrini, S., Siemiginowska, A., Elvis, M., Zezas, A., Athey, A., Bregman, J., Bregman, J., Temi, P., & Sauvage, M. 2002, ApJ, 571, & McDowell, J. 2004, ApJ, 616, 730 ( F04) 272 Fabbiano, G., et al. 2003, ApJ, 588, 175 Bender, R., Saglia, R. P., & Gerhard, O. E. 1994, MNRAS, 269, 785 Fabian, A. C., & Canizares, C. R. 1988, Nature, 333, 829 Bondi, H. 1952, MNRAS, 112, 195 Ferrarese, L., & Ford, H. 2005, Space Sci. Rev., 116, 523 Burstein, D., & Heiles, C. 1982, AJ, 87, 1165 Frank, J., King, A., & Raine, D. 2002, Accretion Power in Astrophysics Cappellaro, E., Evans, R., & Turatto, M. 1999, A&A, 351, 459 (Cambridge: Cambridge Univ. Press) Ciotti, L., D'Ercole, A., Pellegrini, S., & Renzini, A. 1991, ApJ, 376, 380 Gebhardt, K., et al. 2003, ApJ, 583, 92 Ciotti, L., & Ostriker, J. P. 2001, ApJ, 551, 131 Gehrels, N. 1986, ApJ, 303, 336 Ciotti, L., & Pellegrini, S. 1992, MNRAS, 255, 561 Gerhard, O., Kronawitter, A., Saglia, R. P., & Bender, R. 2001, AJ, 121, 1936 David, L. P., Forman, W., & Jones, C. 1991, ApJ, 369, 121 Giampapa, M. S., Rosner, R., Kashyap, V., Fleming, T. A., Schmitt, J. H. M. M., David, L. P., Jones, C., Forman, W., Vargas, I. M., & Nulsen, P. 2006, ApJ, 653, 207 & Bookbinder, J. A. 1996, ApJ, 463, 707 Dempsey, R. C., Linsky, J. L., Fleming, T. A., & Schmitt, J. H. M. M. 1997, Gilfanov, M. 2004, MNRAS, 349, 146 ApJ, 478, 358 Greggio, L. 2005, A&A, 441, 1055 de Vaucouleurs, G. 1948, Ann. d'Astrophys., 11, 247 Hernquist, L. E. 1990, ApJ, 356, 359 de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Jr., Buta, R. J., Paturel, Ho, L. C. 2005, Ap&SS, 300, 219 G., & Fouque, P. 1991, Third Reference Catalogue of Bright Galaxies ( New Irwin, J., Athey, A., & Bregman, J. 2003, ApJ, 587, 356 York: Springer) ( RC3) Kahabka, P., & van den Heuvel, E. 1997, ARA&A, 35, 69 Dickey, J. M., & Lockman, F. J. 1990, ARA&A, 28, 215 Kim, D. W., & Fabbiano, G. 2003, ApJ, 586, 826 Di Matteo, T., Allen, S. W., Fabian, A. C., Wilson, A. S., & Young, A. J. 2003, ------. 2004, ApJ, 611, 846 ApJ, 582, 133 Kim, D. W., Fabbiano, G., & Trinchieri, G. 1992, ApJ, 393, 134 Di Stefano, R., & Kong, A. 2004, ApJ, 609, 710 Kim, D. W., et al. 2004, ApJ, 600, 59 Eddington, A. S. 1913, MNRAS, 73, 359 ------. 2006a, ApJ, 652, 1090 Evans, N. W., & Collett, J. L. 1997, ApJ, 480, L103 Kim, E., Kim, D. W., Fabbiano, G., Lee, M. G., Park, H. S., Geisler, D., & Fabbiano, G. 2006, ARA&A, 44, 323 Dirsch, B. 2006b, ApJ, 647, 276


748

PELLEGRINI ET AL.
Protassov, R., van Dyk, D. A., Connors, A., Kashyap, V. L., & Siemiginowska, A. 2002, ApJ, 571, 545 Rees, M. 1984, ARA&A, 22, 471 Renzini, A. 1989, in Evolutionary Phenomena in Galaxies, ed. J. Beckman & B. Pagel (Cambridge: Cambridge Univ. Press), 422 Renzini, A., & Ciotti, L. 1993, ApJ, 416, L49 Richstone, D., et al. 2004, preprint (astro-ph /0403257 ) Romanowsky, A. J., et al. 2003, Science, 301, 1696 Salpeter, E. E. 1955, ApJ, 121, 161 Sansom, A. E., O'Sullivan, E., Forbes, D., Proctor, R. N., & Davis, D. S. 2006, MNRAS, 370, 1541 Sarazin, C., Irwin, J., & Bregman, J. 2001, ApJ, 556, 533 Sarazin, C., Kundu, A., Irwin, J., Sivakoff, G. R., Blanton, E., & Randall, S. W. 2003, ApJ, 595, 743 Sarzi, M., et al. 2006, MNRAS, 366, 1151 Sazonov, S., Revnivtsev, M., Gilfanov, M., Churazov, E., & Sunyaev, R. 2006, A&A, 450, 117 Schmitt, J., Collura, A., Sciortino, S., Vaiana, G., Harnden, F. R., & Rosner, R. 1990, ApJ, 365, 704 Siemiginowska, A., Smith, R. K., Aldcroft, T. L., Schwartz, D. A., Paerels, F., & Petric, A. O. 2003, ApJ, 598, L15 Skrutskie, M. F., et al. 2006, AJ, 131, 1163 Smith, R. M., Martinez, V. J., & Graham, M. J. 2004, ApJ, 617, 1017 Soria, R., Fabbiano, G., Graham, A., Baldi, A., Elvis, M., Jerjen, H., Pellegrini, S., & Siemiginowska, A. 2006a, ApJ, 640, 126 Soria, R., Graham, A., Fabbiano, G., Baldi, A., Elvis, M., Jerjen, H., Pellegrini, S., & Siemiginowska, A. 2006b, ApJ, 640, 143 Springel, V., Di Matteo, T., & Hernquist, L. 2005, ApJ, 620, L79 Tonry, J. L., et al. 2001, ApJ, 546, 681 Weisskopf, M., Tananbaum, H., Van Speybroeck, L., & O'Dell, S. 2000, Proc. SPIE, 4012, 2 Zezas, A., & Fabbiano, G. 2002, ApJ, 577, 726

Kim, M., Wilkes, B. J., Kim, D. W., Green, P. J., Barkhouse, W. A., Lee, M. G., Silverman, J. D., & Tananbaum, H. D. 2007, ApJ, 659, 29 Kroupa, P. 2001, MNRAS, 322, 231 Kundu, A., Maccarone, T. J., & Zepf, S. E. 2007, ApJ, 662, 525 Kundu, A., & Whitmore, B. C. 2001, AJ, 121, 2950 Kuntz, K. D., & Snowden, S. L. 2001, ApJ, 554, 684 Lauer, T. R., et al. 2005, AJ, 129, 2138 Linsky, J. L. 1984, in Cool Stars, Stellar Systems and the Sun, ed. S. L. Baliunas & L. Hartmann ( Berlin: Springer), 244 Loewenstein, M., Mushotzky, R. F., Angelini, L., Arnaud, K. A., & Quataert, E. 2001, ApJ, 555, L21 Maraston, C. 2005, MNRAS, 362, 799 Mathews, W. 1990, ApJ, 354, 468 Monet, D. G., et al. 2003, AJ, 125, 984 Napolitano, N. R., et al. 2005, MNRAS, 357, 691 Navarro, J. F., Frenk, C. S., & White, S. D. M. 1996, ApJ, 462, 563 Omma, H., Binney, J., Bryan, G., & Slyz, A. 2004, MNRAS, 348, 1105 O'Sullivan, E., Forbes, D. A., & Ponman, T. J. 2001, MNRAS, 328, 461 Park, T., Kashyap, V. L., Siemiginowska, A., van Dyk, D. A., Zezas, A., Heinke, C., & Wargelin, B. J. 2006, ApJ, 652, 610 Pellegrini, S. 2005a, ApJ, 624, 155 ------. 2005b, MNRAS, 364, 169 Pellegrini, S., Baldi, A., Fabbiano, G., & Kim, D. W. 2003a, ApJ, 597, 175 Pellegrini, S., & Ciotti, L. 1998, A&A, 333, 433 ------. 2006, MNRAS, 370, 1797 Pellegrini, S., & Fabbiano, G. 1994, ApJ, 429, 105 Pellegrini, S., Siemiginowska, A., Fabbiano, G., Elvis, M., Greenhill, L., Soria, R., Baldi, A., & Kim, D. W. 2007, ApJ, 667, 749 Pellegrini, S., Venturi, T., Comastri, A., Fabbiano, G., Fiore, F., Vignali, C., Morganti, R., & Trinchieri, G. 2003b, ApJ, 585, 677 Pinkney, J., et al. 2003, ApJ, 596, 903 Proctor, R. N., Forbes, D. A., Forestell, A., & Gebhardt, K. 2005, MNRAS, 362, 857