Imaging the Circumnuclear Region of NGC 1365 with Chandra Junfeng Wang, G. Fabbiano, M. Elvis, and G. Risaliti1 Harvard-Smithsonian Center for Astrophysics, 60 Garden St, Cambridge, MA 02138 9 [email protected]; [email protected]; [email protected]; 0 0 [email protected] 2 n J. M. Mazzarella, J. H. Howell, and S. Lord a J 3 Infrared Processing and Analysis Center, California Institute of Technology, MS 100-22, Pasadena, CA 91125 ] A G [email protected]; [email protected]; [email protected] . h p ABSTRACT - o r t s We present the first Chandra/ACIS imaging study of the circumnuclear re- a [ gion of the nearby Seyfert galaxy NGC 1365. The X-ray emission is resolved into 1 point-like sources and complex, extended emission. The X-ray morphology of v the extended emission shows a biconical soft X-ray emission region extending ∼5 7 9 kpc in projection from the nucleus, coincident with the high excitation outflow 2 0 cones seen in optical emission lines particularly to the northwest. Harder X-ray . 1 emission is detected from a kpc-diameter circumnuclear ring, coincident with the 0 star-forming ring prominent in the Spitzer mid-infrared images; this X-ray emis- 9 0 sion is partially obscured by the central dust lane of NGC 1365. Spectral fitting : v of spatially separated components indicates a thermal plasma origin for the soft i X extended X-ray emission (kT = 0.57 keV). Only a small amount of this emission r can be due to photoionization by the nuclear source. Detailed comparison with a [OIII]λ5007 observations shows the hot interstellar medium (ISM) is spatially anticorrelated with the [OIII] emitting clouds and has thermal pressures com- parable to those of the [OIII] medium, suggesting that the hot ISM acts as a confining medium for the cooler photoionized clouds. The abundance ratios of the hot ISM are fully consistent with the theoretical values for enrichment from Type II supernovae, suggesting that the hot ISM is a wind from the starburst circumnuclear ring. X-ray emission from a ∼ 450 pc long nuclear radio jet is also detected to the southeast. 1Current Address: INAF-Arcetri Observatory, Largo E, Fermi 5, I-50125 Firenze, Italy – 2 – Subject headings: galaxies: active — galaxies: starburst — galaxies: individ- ual(NGC 1365) — X-rays: galaxies — X-rays: ISM 1. Introduction Understandingactivegalacticnucleus(AGN)–galaxyinteractionandfeedbackisofgreat consequence for the study of both galaxy and AGN evolution (e.g., Silk & Rees 1998; Kauff- mann et al. 2003; Hopkins et al. 2006). The intense ionizing radiation, relativistic jets, and winds from the AGNs interact with the interstellar medium (ISM) of their host galaxies, and may either stimulate or suppress star formation (e.g., Sazonov, Ostriker & Sunyaev 2004). Visible signatures of direct AGN-host interaction include kpc-scale regions with [OIII] and Hα line emission, known as the extended narrow line region (ENLR), which have been found in many nearby Seyfert galaxies (Schmitt et al. 2003; Veilleux et al. 2003). Soft X-ray emis- sion is seen associated with the ENLR, providing an opportunity to obtain X-ray diagnostic of the physical properties of the interacting ISM (e.g., Young et al. 2001; Yang et al. 2001; Ogle et al. 2000, 2003; Bianchi et al. 2006; Evans et al. 2006; Kraemer, Schmitt & Crenshaw 2008). Another important link between AGNs and starbursts is that molecular gas accreted to the nuclear region may induce a starburst (Elmegreen 1994), feed a central AGN, or both. AGN-related shocks may further stimulate star formation (e.g., Scoville 1992; Gonz´alez Del- gado et al. 1998; Veilleux et al. 1995; Veilleux 2001). Star formation in the disks and galactic winds in hot halos of spiral galaxies can be traced by the diffuse X-ray emission as well (e.g., Tyler et al. 2004, Strickland et al. 2004; Swartz et al. 2006; Warwick et al. 2007). NGC 1365, the object studied in this paper, is ideal for these investigations, since it hosts an AGN and also has IR emission coincident with active, intensive nuclear star formation. NGC 1365 is a nearby (D = 18.6±1.9 Mpc, 1′′ ∼ 90 pc; Madore et al. 1998; Silbermann et al. 1999) archetype barred spiral galaxy (SBb(s)I; Sandage & Tammann 1981), and likely a member of the Fornax cluster (Jones & Jones 1980). A thorough review of the NGC 1365 galaxy is given in Lindblad (1999). It hosts vigorous star formation in the circumnuclear region and a variably obscured Seyfert 1.5 nucleus1 (Veron et al. 1980; Hjelm & Lindblad 1996; Risaliti et al. 2007). The spectacular symmetric main spiral arms extending from the ends of a strong bar are apparent in the XMM-Newton Optical Monitor (OM) images (Figure 1). Dark dust lanes run across the nuclear region and partially obscure the nucleus. NGC 1365 has long been known to exhibit a biconical outflow (Burbidge & Burbidge 1Otherclassificationsexistintheliterature. Forexample,Turner,Urry&Mushotzky(1993)andMaiolino & Rieke (1995) classified the NGC 1365 nucleus as a Seyfert 2 and 1.8, respectively. – 3 – 1960; Phillips et al. 1983; Storchi-Bergmann & Bonatto 1991). A ∼ 5′′ long (∼450 pc) 100◦ wide conical [OIII]λ5007 emission line region (ELR) is present to the southeast (SE) and to the northwest (NW) of the nucleus (Hjelm & Lindblad 1996; Kristen et al. 1997; Veilleux et al. 2003), which suggests an AGN ionization cone, although a starburst driven outflow explanation has also been advanced (Komossa & Schulz 1998). In the Hubble Space Telescope (HST) images, the inner 3′′ of the conical outflow seen in [OIII] is resolved into a number of small clouds and larger agglomerations (Kristen et al. 1997). A high excitation [NII]λ6583/Hα region is aligned with the [OIII] structure, but appears to be rectangular (Veilleux et al. 2003). Sandqvist et al. (1995) show that the inner parts of the [OIII] region contain a nuclear radio jet, extending 5′′ SE along the galaxy minor axis. Note that the spatial properties of the NGC 1365 galaxy and the outflow cones are well determined from these multiwavelength studies (see Lindblad 1999 review and references therein): the position angle (PA) of the line of nodes is ≈ 220◦, and the inclination angle of the galaxy is 40◦; the [OIII] cone extends from the nucleus out of the galactic plane with its symmetry axis closely aligned with the rotation axis of the galaxy, with a full opening angle of about 100◦; the counter cone to the NW is partially obscured by the absorbing dust in the galaxy plane. Another prominent characteristic of NGC 1365 is the circumnuclear star-forming ring with a diameter of ∼ 14′′ (∼ 1.3 kpc; Kristen et al. 1997). The kpc-size nuclear vicinity contains bright optical “hot spots” (Sersic & Pastoriza 1965; clearly visible in the UV im- ages shown in Figure 1). The circumnuclear regions are resolved into many compact super star clusters (Kristen et al. 1997), forming an elongated ring surrounding the AGN. Its minor axis is parallel to both the bi-cone axis and the galaxy minor axis, suggesting the possibility of a collimating torus (Sandqvist 1999). High resolution mid-infrared imaging of the ring (Galliano et al. 2005) also unveils a circumnuclear population of point-like sources coincident with bright CO spots (Sakamoto et al. 2007) and non-thermal radio continuum sources (Sandqvist et al. 1995; Forbes & Norris 1998), which are interpreted as embedded young massive star clusters. The biconical ELR/outflow and the circumnuclear ring should be readily detectable in the Chandra images as hinted from the spatially extended X-ray emission in the ROSAT HRI image (Stevens et al. 1999). In this paper we present a high spatial resolution X-ray imaging study of the complex circumnuclear region of the active galaxy NGC 1365 with the Chandra X-ray Observatory. NGC 1365 has received a great deal of attention as a result of X-ray observing campaigns to monitor the obscuration of the AGN (e.g., Risaliti et al. 2005a,b; Risaliti et al. 2007). Comparing to observations of NGC 1365 from previous generation X-ray satellites and the XMM-Newtonobservatory, theChandraimagesofferanunprecedented viewoftheinnermost regions of NGC 1365, which is critical to disentangle the different emission components. – 4 – 2. Observations and Data Reduction NGC1365wasfirstimagedin2002withtheback-illuminatedchipoftheAdvancedCCD Imaging Spectrometer spectroscopy array (ACIS-S; Garmire et al. 2003) positioned at the focalpoint ofChandra X-ray ObservatoryHighResolutionMirror Assembly (vanSpeybroeck et al. 1997; Weisskopf et al. 2002). It was further monitored with six Chandra/ACIS-S observations in “1/4 window” mode during April 2006, resulting in a total exposure time of ∼100 ks (see Table 1 for details; the overlap region common to these observations is outlined in Figure 1). The data products were analyzed with the Chandra X-Ray Center (CXC) CIAO v4.0 software and HEASOFT v6.4 package2. The new release of calibration files CALDB3 v3.4.3 were used. The data were processed by the CXC (ASCDS version 7.6.7-7.6.9); verification and validation of the data products showed no anomalies. Following the standard Chandra ACIS data preparation thread, we reprocessed all level 1 data to create new level 2 events for consistency. A preliminary run of CIAO tool wavdetect (Freeman et al. 2002) is done for point source detection in the central 8′ region of each observation. A 10′ radius circle centered at the S3 aimpoint was used to query the Naval Observatory Merged Astrometric Dataset (NOMAD) catalogue (Zacharias et al. 2004) and the extracted optical sources are searched for counterparts to Chandra detections. Using optical point sources in the field detected in X-rays, the absolute astrometry for Observation ID (ObsID) 3554 is measured to be accurate to within 0′.′5. By comparing the positions of the bright point like X-ray sources in common between observations, we then corrected the absolute astrometry for any relative offsets between the 2006 data sets and ObsID 3554. The relative shifts were ∼ 1 pixel. Toremoveperiodsofhighbackground, foreachObsIDweextractedalightcurveforchip S3, excluding any bright sources in the field. The light curves for 2006 observations showed no strong flares. However, a period of enhanced background was present in the last quarter of ObsID 3554, and 1.6 ks exposure of affected data was screened out. Finally all seven observations were merged into a single event file after correcting the relative astrometric offsets. Detailed study of the point source population is deferred to a future paper (J. Wang et al. in preparation), which will also use the XMM-Newton observations of NGC 1365. In addition to Chandra data, ancillary UV and mid-IR images were also used in this paper. NGC 1365 was observed with the Spitzer Space Telescope (Werner et al. 2004) as 2Seehttp://cxc.harvard.edu/ciao/andhttp://heasarc.gsfc.nasa.gov/lheasoft/formoreinfor- mation. 3http://cxc.harvard.edu/caldb/ – 5 – part of a survey of luminous infrared galaxies (LIRGs; L ≥ 1011L )4. Observations with IR ⊙ the Infrared Array Camera (IRAC; Fazio et al. 2004) were dithered to correct for cosmic ray events and bad pixels, and raster mapping was used to construct mosaics of the galaxy at 3.6, 4.5, 5.8 and 8.0 µm (Spitzer AOR #12346880). Details of the IRAC observations and dataprocessing aredescribed byMazzarella etal. (2008,inpreparation). TheXMM-Newton OMdatawere aquired fromXMM-Newton ObsID #0205590301(PI:G.Fabbiano), using the Processing Pipeline System (PPS) products processed with Science Analysis System (SAS; version 6.5.0). The UV images shown in Figure 1 used the following OM filters (Mason et al. 2001): U (λ ∼ 3472˚A), UVW1 (λ ∼ 2905˚A), and UVM2 (λ ∼ 2298˚A). 0 0 0 3. The X-ray Color Images and Morphology Smoothed images of the NGC 1365 region were first created without removing the point sources. This enabled us to examine the spatial distribution of the point sources with respect to the galaxy. We extracted images from the merged data in the commonly used soft band (0.3–0.65 keV), the medium band (0.65–1.5 keV), and the hard-band (1.5–7 keV). Note that the medium band encompasses the emission from the Fe-L blend. The 0.3 keV low-energy boundary is the lowest energy where ACIS remains well calibrated, and the high energy 7 keV cut-off is chosen to limit the contribution from the background. To derive exposure-corrected images, exposure maps were created for individual obser- vations and bands, then reprojected to create the combined exposure maps matching the merged data. The band-limited images were then divided by the appropriate exposure maps and adaptively smoothed. This process enhances faint extended features while preserving the high resolution of brighter features. In order to prevent large variations in the adaptive smoothing scales between the bands with different counts, we applied to the soft and hard bands the same smoothing scales calculated for the medium band image, which has the best signal to noise ratio (S/N). The significance was chosen between 2.4 and 5σ above the local background (corresponding to the smoothing Gaussian kernel of varying scales between 1 and 40 pixels). Figure 2 presents the “false color” composite image of the central 3′ × 3′ region of NGC 1365, where the soft, medium, and hard band smoothed images are shown in red, green, and blue, respectively. Besides the bright nucleus (marked as “N” in Figure 2), some 40 point sources are present. The bright off-nuclear X-ray source towards the lower right corner is a well-studied, variable ultraluminous X-ray source (ULX) NGC 1365 X-1 (Komossa & Schulz 1998), most likely a 50–150M accreting black hole (Soria et al. 2007). ⊙ 4http://goals.ipac.caltech.edu/spitzer/Spitzer.html – 6 – Figure 2 shows the circumnuclear ring (mainly pink; ∼ 14′′ in diameter), intersected by the obscuring dust lane (blue). This ring is emphasized with an alternative set of energy bands shifted to higher energies images in Figure 3, in which the soft band (red) is 0.3–1.5 keV, the medium band (green) is 1.5–3.0 keV, and the hard band (blue) is 3.0–7.0 keV; the same smoothing was applied as for Figure 2. The hard emission (presumably from young supernovae remnants and X-ray binaries) is less affected by the obscuration from the dust lane, tracing the extent of the star forming ring. Note that with this energy band selection the (broader) soft band now has the best S/N of the three bands and so the smoothing scales from this band were used to smooth the images in the other two bands. This alternate band selection is only used here in this paper. Comparison with Spitzer data clearly relates the hard X-ray ring to the circumnuclear star-forming ring prominent in the infrared. Figure 4 shows the Chandra X-ray image of the diffuse emission (from Figure 6) in NGC 1365 (green/blue), together with the mid-IR (8 µm) view of the galaxy (red). Figure 5 zooms in to the central 1 arcmin, which shows a composite mid-IR image of the circumnuclear ring and a composite mid-IR/UV image of the same region with an overlay of the medium-band X-ray emission. The circumnuclear ring is compact and maintains regularity close to the center. Such circumnuclear star formation rings are commonly observed in strongly barred spirals and assumed to be dynamically associated with the Inner Lindblad Resonance (ILR)—inflowing gas accumulates to form a massive star forming ring between the two inner resonances (see Lindblad 1999). The Inner ILR and Outer ILR in NGC 1365 is ∼ 3′′ and ∼ 30′′ from the nucleus, respectively (Lindblad et al. 1996), hence the ring is located ∼ 4′′ outside of the Inner ILR. We note a hot-spot located ∼ 22′′ (∼ 2 kpc) NE of the nucleus, which shows co-spatial X-ray, UV, and mid-IR emission. IR filaments bridge the ring and this object, suggesting that it is related to the galaxy, instead of being a background object, and may be associated with the bar. The spatial correlation between the X-ray and UV emission is apparent. Because of the obscuring dust lane, which is traced by its bright IR emission, a bright ridge of X-ray/UV emission with hot-spots appears coincide with the edge of the IR ring ∼ 10′′ (1 kpc) N and NW of the nucleus, probably related to the discrete regions of active star formation associated with the ring. To study the large scale diffuse emission, we followed the procedure of Baldi et al. (2006a) to create mapped-color images with point sources removed. Given the short indi- vidual exposure times, we performed point source detection in the combined exposures in the full band (0.3–7.0 keV). The point-like ACIS detections including the nucleus were then removed from the data, and the resulting holes were filled by interpolating the surrounding background emission. The same procedure to define source regions and background regions – 7 – employed by Baldi et al.(2006a) was followed. The results were visually compared with the band-limited images and the smoothed images to ensure that no apparent individual point sources were missed. In the circumnuclear ring region, a few “point” sources detected by wavdetect appear extended. These could be either association of unresolved point sources or bright patches and knots of diffuse emission. To check whether they affect the spectral analysis of the circumnuclear ring, we created separate event files without removing these sources with uncertain classification. These event files were used in spectral extraction and fitting described later in this paper. The images with point sources removed and holes filled in the three bands (0.3–0.65 keV; 0.65–1.5 keV; and 1.5–7 keV) were adaptively smoothed, then combined to create a “mapped-color” image of the diffuse emission shown in Figure 6. The mapped color image shows complex morphology and striking substructures that have not been seen in lower resolution X-ray images previously. TheoverallmorphologyoftheX-rayemissionisratherconsistentwiththeopticalpicture described in § 1: 1. Extended soft X-ray emission appears to resemble a bicone (NW and SE) structure with the apex at the nucleus. The SE structure is more extended (∼ 0.8′ measured from the nucleus) and visually closer to a cone-shape. Its inner region is brighter and appears to contain substructure. The NW cone is only bright at the base close to the nuclear region (∼ 0.4′), with fainter emission extends towards far NW. Both cones are incomplete, possibly due to obscuring material in the galaxy plane and bulge. The measured projected half opening angle of the cone in the X-ray image is approximately ∼ 55◦. The X-ray cones appear to be edge-brightened, which is consistent with the hollow cones seen in the optical (Lindblad 1999). 2. Harder X-ray emission is associated with the circumnuclear star-forming ring, which appears pink and clumpy in the image. 3. The hardest X-ray emission, which appears as a ∼ 4′′-wide blue band intersecting the top portion of the circumnuclear ring, is coincident with the dust extinction lane of the galaxy. 4. The Line-Strength Map Aquickexaminationofthespectraextractedfromthediffuseemissionshowthepresence oftheOxygen+Iron+Neon(O+Fe+Ne)emissionblend(0.6–1.16keV),theMagnesium(Mg)- – 8 – XI (1.27–1.38 keV) and the Silicon (Si)-XIII (1.75–1.95 keV) lines (Figure 9). To identify regions of strong apparent line emission and to select possible anomalous abundance regions for spectral analysis, we generated an emission line map, following Fabbiano et al. (2004) and Baldi et al. (2006a), to highlight regions of prominent emission lines in the hot ISM. Three band-limited images were created to encompass the lines noted above. A “con- tinuum template” was extracted using the 1.4–1.65 keV and 2.05–3.05 keV bands which are not visibly contaminated by emission lines, and was used to determine the underlying continuum in the emission line bands. All point-like sources were removed and the holes were filled with background as described in § 3. The O+Fe+Ne 0.6–1.16 keV band image contains the largest number of counts and so was used to set the smoothing scales for the other line and continuum images. To determine the continuum contribution in each line image, we extracted the spectrum of the entire diffuse emission of NGC 1365 and fitted the continuum with a simple two- componentthermalbremsstrahlung model,excluding thethreeenergyrangeswithstrongline emission. The best fit model provides the appropriate scaling factors between the continuum level in our continuum template and the continuum in a given line band. The scaling factors are 2.5, 0.2, and 0.19 for the 0.6–1.16 keV, 1.27–1.38 keV, and 1.75–1.95 keV band, respectively. Eachlineimagewascontinuum subtractedafterweighting thecontinuum image by these scaling factors, and the resulting images were combined to create a mapped-color line-strength image (Figure 7). Figure 7 shows that the emission-line structure of the hot ISM in NGC 1365 is also quite complex. Silicon emission (blue) appears most pervasive in the dust lane; this could be due to obscuration of the softer energy photons in this region, and this possibility is explored further below. Magnesium lines (green) and O+Fe+Ne emission (red) are prevalent in most regions. Part of the NW diffuse region shows enhanced O+Fe+Ne emission. Noticeably there is an area (orange in color) south of the star-forming ring that is lacking in silicon line emission. 5. Spectral Analysis 5.1. Data Extraction The mapped-color X-ray image (Figure 6) and the line-strength map (Figure 7) demon- strate that the diffuse emission of the hot ISM in NGC 1365 is rich in spatial and spectral features. These figures suggest structure in the hot ISM, possibly the effect of varying absorption columns, plasma temperatures, and metal enrichment. – 9 – As outlined in Figures 8 and 7, we identified five separate regions based on morphology and three regions based on apparent abundance pattern, to perform spectral analysis of the X-ray emission from the hot ISM. The region “ring” refers to the circumnuclear ring, excluding the nucleus (see § 3 and Figure 3). The region “diffuse” refers to all the diffuse emission, excluding theareainside theouterlimit ofthering. Itisfurther divided into“inner diffuse” (the bright portion of the cones) and “outer diffuse” (the faint diffuse emission). In addition, for the ring area we also extracted photons from the “ring” region including the wavdetect selected point sources that appear extended (see § 3). We dubbed the region and the corresponding extraction “ring+”. We did not further divide the diffuse emission into sub-regions(e.g.,theringisseparatedbythedustlaneintonorth-ringandsouth-ring)tohave good statistics for the spectral fitting. Regions related to the emission line features noted in Figures 7 were also created, with descriptive names (“high-Si”, “low-Si”, and “high-Fe”). We emphasize here that, even though the extended emission is referred to as “diffuse”, the X-ray emission is not bonafide diffuse plasma. There is evidence for extensive clumpiness and X-ray hot-spots. The “diffuse” X-rays include a mixture of hot ISM and unresolved point sources (XRB/SNR) associated with regions of very active star formation. The hard power law component seen in the spectral fit (see § 5.2) is likely related to the XRB/SNR components. We extracted spectra separately from the individual ObsIDs, for each diffuse emission regions described above. We excluded all the point source areas above 3σ described in § 3. We also extracted background counts representative of the field from three circular source-free regions (each with a 0.4′ radius, resulting in a total area of 1.5 arcmin2) well outside the galaxy, but within the ACIS-S3 chip in all seven observations (Figure 1). A new CIAO script specextract5 was used to extract the region spectra for multiple observations. This script extracts the spectrum and creates area-weighted Response Matrix Files (RMF) and Ancillary Response Files (ARF) for each region. For each region of interest, all source spectra and background spectra were co-added and the response files were combined with their appropriate weights using the FTOOLS software6. The resulting spectra were then grouped so as to have signal to noise ratio S/N ≥ 10 in each energy bin. 5See CIAO thread http://cxc.harvard.edu/ciao/threads/specextract/. 6http://heasarc.gsfc.nasa.gov/docs/software/ftools/ – 10 – 5.2. Thermal and Power Law Models The extracted spectra were analyzed with XSPEC v12.0 (Arnaud 1996), using combi- nations of absorbed optically-thin thermal emission and power law components to fit the data. We used the Astrophysical Plasma Emission Code (APEC) thermal-emission model (Smith et al. 2001) that utilizes improved atomic data from the Astrophysical Plasma Emis- sion Database7. The absorption column was modeled with tbabs (Wilms et al. 2000), which includes updated photoionization cross section and abundances of the ISM, as well as a treatment of interstellar grains and the H molecule. 2 Wefirstconsideredabsorbedsingle-temperatureAPECmodelsaswellastwo-temperature APEC models. A single line-of-sight absorption column was used in all cases, equivalent to Galactic line-of-sight column (1.4 × 1020 cm−2; Kalberla et al. 2005) combined with the intrinsic absorption within NGC 1365, which was free to vary. When the fit required un- measurably small absorption (N ≪ 1020 cm−2), we froze N at the Galactic column. As H H the spectrum of the integrated diffuse emission clearly shows prominent MgXI (∼1.3 keV), OVII-NeIX/X-Fe-L complex (0.6-1.2 keV), and SiXIII (∼1.8 keV) line features, the abun- dances of these elements were left to vary freely. When fitting with the two component models, the Oxygen abundance was fixed at the solar value as it is poorly constrained in our data and consistent with solar abundance. A power law component is introduced to the single-temperature model to allow for the possible integrated emission of unresolved X-ray binaries that are too faint to be detected individually. A simple absorbed power law model give poor fits (χ2 > 4) to all the extracted spectra. To fit the spectra of the three extreme ν line ratio regions (Figures 7), we only allow the abundances of Mg, Si, and Fe to vary, given the limited number of extracted counts. For each of the regions used in the spectral analysis, the net counts (background sub- tracted) in the full band, the reduced χ2 (χ2 over the numbers of degrees of freedom dof, χ2), the resulting best-fit temperature(s) kT (keV), absorbing column density N (cm−2), ν H and power law photon index Γ (if a power law component is invoked) are summarized in Ta- ble 2, together with the abundances of Oxygen, Neon, Magnesium, Silicon, and Iron relative to the solar values. The last column of Table 2 gives the adopted “best-fit model”, which is chosen based on having the smallest χ2. If more than one model can fit the spectrum ν well and the χ2 values are close (difference less than 0.1), a simpler model with less free ν parameters is preferred. The errors are quoted at 1.65σ (90% confidence interval) for one interesting parameter. The spectra of the regions, together with the best-fit models and the data−(best-fit-model) residuals, are shown in Figure 9. 7http://cxc.harvard.edu/atomdb/