X-RAY STAR CLUSTERS IN THE CARINA COMPLEX

As outlined in the previous section, we cannot consider the full CCCP sample of 14,368 X-ray sources in the Carina region for two reasons: a fraction of the sources are foreground or background contaminants, and the survey has inhomogeneous sensitivity due to characteristics of the Chandra telescope and the survey design. Both of these effects are greatly reduced in the spatially complete sample of 3220 probable Carina members above a photon flux limit described above.

Figure 1 shows the distribution of this spatially complete Carina member sample. Panel (a) shows the individual star positions, while panel (b) shows the smoothed source surface density using a normal kernel with standard deviation σ = 30''. Assuming a distance of 2.3 kpc to the Carina Nebula, this corresponds to a Gaussian kernel with σ ≃ 0.3 pc or full width at half-maximum (FWHM) = 0.8 pc. We select the density threshold for cluster identification to be the third contour (0.003 in the normalized convolution of the kernel with the data) in this surface density map. The 20 surface density peaks above this threshold are labeled in the figure and listed in Table 1 as the principal X-ray-selected clusters in the region. We adopt the nomenclature "CCCP-Cl" for "Chandra CCCP Cluster" to identify these clusters (Column 1). The cluster positions (Columns 2and3) represent the surface density peaks; due to non-spherical distributions, these may not correspond to mean or median positions of the member stars. When a larger σ = 40'' kernel is used, the Tr 16 substructure becomes less distinct, and when a smaller σ = 20'' is used, some of the small groups (Section 3.2) begin to emerge.

Zoom In Zoom Out Reset image size

While the spatially complete sample (3220 bright probable Carina members) allows a spatially unbiased identification of rich clusters, we use the full sample (10,728 probable Carina members) to estimate properties of each cluster because the CCCP survey sensitivity variations are generally small on intra-cluster scales. Figure 2 shows the distribution of the full sample superposed on the contours of clusters found from Figure 1. The figure also outlines the individual ACIS fields; they give a clear warning that the spatial variations in the full sample can be due to instrumental effects (off-axis sensitivity degradation and overlapping exposures) as well as to intrinsic star clustering in the Carina complex. Our use of a σ = 30'' (0.3 pc) Gaussian kernel and omission of fainter X-ray sources also blurs substructure within the rich Tr 14, 15, and 16 clusters. More complete analyses of the structure of these clusters are given by Ascenso et al. (2007), Wang et al. (2011), and Wolk et al. (2011), respectively.

Zoom In Zoom Out Reset image size

To estimate the star population N of each cluster, we give the number of sources in the spatially complete sample within some boundary around the cluster peak locations to be cluster members. The contours, counted from the lowest contour in Figure 1(b), are listed in Column 4 and N is given in Column 5 of Table 1. For example, contour "2m" represents the second lowest contour where "m" is appended when the contour was modified to produce a closed boundary. Observed cluster populations range from N ≃ 10 to over 1000 X-ray-selected stars.

In PMS populations, useful estimates of line-of-sight interstellar gas column densities, N_H, can be obtained from the observed median X-ray event energy statistic (Feigelson et al. 2005; Getman et al. 2010). Table 1 reports the average and standard deviation of this statistic, 〈MedE〉 (Column 7), for the N_ME (Column 6) cluster members for which median event energy is available (sources with >4 net counts extracted). Visual absorption, A_V, can be estimated from the gas column densities, N_H, inferred from 〈MedE〉 assuming standard gas-to-dust ratios (Vuong et al. 2003). These A_V estimates, ranging from A_V ≃ 2 to 10 mag, are reported with low precision in Column 8.

Correspondences to previously known optical and near-infrared star clusters (Column 9) are established visually from maps given by Feinstein (1995) and Smith & Brooks (2008). Note that Trumpler 16 is divided into seven clusters in our analysis. Associations with mid-infrared clusters in the South Pillars region are based on the Spitzer IRAC study of Smith et al. (2010). A ⋆ symbol indicates whether an X-ray cluster lies in the region occupied by the South Pillars and the large Collinder 228 star cluster discussed in Section 5.2. The final column of Table 1 lists the dominant massive stars within or near each cluster with spectral types obtained from Gagné et al. (2011) or SIMBAD. These are labeled a, b, ..., y to identify them in Figure 2 where they are plotted as triangles. These are far from a complete list of OB stars, but give a general indication of the cluster richness in the region. The X-ray OB population in the CCCP, including their multiplicity and X-ray properties, is discussed in detail by Gagné et al. and Povich et al. (2011a).

While the smoothed map of Figure 1 is effective in finding clusters with scales around 05–5' (0.3–3 pc) in a spatially unbiased fashion, it misses some small groups that are evident in visual examination of the full sample. We therefore constructed a smoothed density map of the full sample with a 10'' Gaussian kernel and made a list of structures above a subjectively chosen density threshold. This new list of small groups is restricted to regions lying outside the boundaries of the principal clusters defined in Column 4 of Table 1. Note that, due to instrumental spatial sensitivity variations across the field using the full sample (Section 2), the search for sparse groups cannot be considered complete to some fixed flux limit. The group list also depends on the choice of smoothing kernel; a larger 15'' kernel merges some of the closer groups into larger structures.

The resulting list of 31 small X-ray-selected groups of probable Carina members is presented in Table 2; Columns 4–7 have the definitions as in Table 1. The group member tallies, N, are measured in 40'' diameter circular regions centered on the local peaks of the smoothed map. These groups are designated "CCCP-Gp" to distinguish them from the larger and richer CCCP-Cl clusters. The groups are both compact and sparse, exhibiting only 5 < N < 12 X-ray members. Typical member surface densities are 50–100 stars pc⁻². Most lie in the southern portion of the mosaic, and many have counterparts among the mid-infrared clusters identified by Smith et al. (2010) from Spitzer IRAC images of the South Pillars region. The early-type stars in this region are often considered to be members of an extended open cluster Collinder 228 (Feinstein et al. 1976), so one might choose to associate these small X-ray groups with Collinder 228 (Section 5.2).

The clusters and groups defined in Tables 1 and 2 contain ∼3000 and ∼300 stars in the full sample, respectively. But more than twice as many stars lie outside these cluster and group boundaries in large regions with enhanced stellar surface densities. We have divided these unclustered stars into three spatial regions shown in Figure 1(b): Region A around and between Trumpler 14 and 15; Region B within and south of the Trumpler 16 complex; and Region C over the Collinder 228 (South Pillars) area. Region D represents all other widely distributed stars in the full sample; this group has >5000 probable Carina members. All 14,368 CCCP sources lie in one of these four Regions; for example, Region A includes members of Trumpler 14 and 15 as well as members in the surrounding area. The observed group member tallies and average absorptions of these regions are listed at the end of Table 1.

The three principal clusters—Trumpler 14, 15, and 16—are distinct from each other, with strong peaks in stellar surface density separated by substantial dips. The Chandra data do not support the suggestion that Trumpler 14 and 16 are contiguous, separated only by a dust lane (Feinstein et al. 1973). A broad bridge of stars is seen between Trumpler 14 (CCCP-Cl 1) and Trumpler 15 (CCCP-Cl 8), which is part of the Region A large-scale density enhancement. This bridge strongly indicates that Trumpler 15 is part of the Carina complex rather than the chance superposition of a more distant cluster as suggested by several authors (e.g., Walborn 1995). CCCP results on Trumpler 15 and its relation to the Carina complex are discussed in detail by Wang et al. (2011).

Each of these three rich clusters has internal structure which cannot be readily summarized by simple terms such as a circular or elliptical shape. In both Trumpler 14 and 15, the peak in stellar surface density is offset from the center of outer density contours by 0.5–1 pc, and the distribution of lower mass members is not symmetrically distributed around the dominant O stars. The internal structure of Trumpler 15 is discussed by Wang et al. (2011). The stellar structures may be related to nearby molecular cloud cores 10 and 11 studied by Yonekura et al. (2005). Spherically symmetric models (for example, as discussed by Ascenso et al. (2007) for Trumpler 14) will not capture the full complexity of the star distribution.

The internal structure of Trumpler 16 is qualitatively different from Trumpler 14 and 15, as it does not have a single surface density peak. Although our analysis divides it into seven clusters (CCCP-Cl 3, 4, 6, 9, 10, 11, and 12), this number of subclusters identified depends strongly on the smoothing kernel. The structure seen in Figure 1 can be described as hierarchically clustered. For example, one could choose to define a single cluster centered at (10^h44^m50^s, − 59°43') with diameter 11' (7.3 pc) containing several subclusters, none of which lies at the cluster "center." The curved southern boundary of the cluster may, in part, be caused by absorption in the V-shaped dust lane that crosses the middle of the Carina field. See Figure 1 of Albacete-Colombo et al. (2008) for the spatial relationship between the dust lane and the X-ray population. Evans et al. (2003) and Albacete-Colombo et al. (2008) give detailed Chandra studies of the X-ray properties of Trumpler 16 stars. CCCP results on Trumpler 16 are discussed in detail by Wolk et al. (2011).

The Figure 1 map clarifies the long-standing confusion regarding the status of less rich clusters in the Carina complex (Section 1). We find that Collinder 232 east of Trumpler 14 is a distinct, compact cluster (CCCP-Cl 5), confirming the conclusion of Carraro et al. (2004). It is clear that Collinder 232 is not just part of the halo of Trumpler 14. Note that, as in the richer Trumpler clusters, its lower-mass stars are not centered on its dominant O stars. Infrared studies show that Collinder 232 stars have a wide range of absorptions and a considerable population of very young protostars (Preibisch et al. 2011; Povich et al. 2011b).

Several researchers have commented that Collinder 228 in the South Pillars region is not a distinct cluster, but may be an extension of Trumpler 16 (e.g., Walborn 1995; Smith & Brooks 2008). The X-ray source distribution in Figure 1 does not support this interpretation, as the stellar surface density of Trumpler 16 shows a distinct near-circular boundary with diameter around 11'.

Instead, the X-ray source distribution in the southern part of the Carina complex supports the idea that Collinder 228 is a loose association of young stars and sparse groups distributed over a very large (>20' diameter) region centered around QZ Car at (10^h444 − 60°00'), as described by Feinstein et al. (1976). Collinder 228 might be considered to be the stars associated with the dusty South Pillars cloud structures (Smith et al. 2000, 2010). Our Region C of enhanced X-ray star density would then trace the lower-mass stars in the inner 10' of the cluster.¹⁶ Examination of Figure 2 shows that the low-mass star distribution attributable to Collinder 228 is not uniform but exhibits distinct clumping. The most prominent grouping is around (10^h456, − 59°58') which includes CCCP-Cl 13, 17 (the Treasure Chest), 18, and CCCP-Gp 21, 22, 23, 24, 25, 26, and 27. Another concentration is a ∼8' diameter region around (10^h440, − 60°00') where we find CCCP-Gp 8, 9, 10, 11, 12, 13, 14, 16, and CCCP-Cl 2. Group 14 is centered on the well-studied multiple supergiant system QZ Car (O9.7 I + O8 III, V = 6.3) studied by Parkin et al. (2011). Other groups are distributed up to ∼20' south of the Treasure Chest including CCCP-Cl 16, 19, 15, and CCCP-Gp 18. These stellar groupings are also independently found in the Spitzer mid-infrared study of Smith et al. (2010). We caution that the Collinder 228 region considered here covers several ACIS fields, so the X-ray sensitivity is not uniform. In particular, sparse groups of lower-mass stars that happen to fall far off-axis in the ACIS observations may be missed.

Collinder 234 is a poorly studied cluster lying ∼3' southeast of CCCP-Cl 12, the highest surface density spot of X-ray stars in the Trumpler 16 complex. It is reportedly as a concentration of ≳10 stars with diameter 4' in the 2MASS survey centered at (10^h452, − 59°45') (Dutra et al. 2003). A cluster at the Collinder 234 location is not confirmed in the X-ray sample of PMS stars. The CCCP sample does include several optically bright late-O stars (Gagné et al. 2011, and SIMBAD): LS 1870 (V = 10.0, O9 III); MJ 484 (V = 10.0, O9 III); MJ 496 (V = 11.0, O8.5 V); HD 93343 (V = 9.6, O8 V + O 7-8.5 V); MJ 516 (V = 9.3, O8 V + O9.5 V); MJ 517 (V = 9.3, O7 V + O8 V + O9 V); and MJ 535 (V = 9.3, O5.5-O6 V(n)((fc)) + B2 V-III). Except for a sparse subcluster of a few lower-mass stars within a few arcseconds of MJ 496, there are no X-ray emitting PMS stars in the vicinity of these massive stars. This is quite extraordinarily different from other young clusters. The Orion Nebula Cluster (ONC) has a similar distribution of late-O stars and would have ∼360 PMS stars detected in the CCCP were it to lie in the Carina Nebula (Townsley et al. 2011a).

Bochum 11 to the southeast is confirmed in the X-ray cluster list (CCCP-Cl 20) as a sparse cluster with low stellar surface density. Examination of Figure 2 suggests that it consists of two star groupings in a ∼6' (4 pc) diameter region: one containing CCCP-Cl 20 and CCCP-Gp 31 around (10^h14^m16^s, − 60°07'); and the other containing CCCP-Gp 28, 29, and 30 around (10^h46^m52^s, − 60°05'). Group 28 lies ∼30'' northwest of the spectroscopic binary HD 93576 (V = 9.7, O9 IV) and appears associated with the mid-infrared source MSX6C G287.9816-00.8724. Group 30 is centered on HD 93576 and appears associated with MSX6C G287.9475-00.8846.

Highly luminous and extended mid-infrared sources, such as IRAS sources with bolometric luminosities in the L ∼ 10⁴–10⁶ L_☉ range, and compact radio H ii regions are relatively rare in the Carina complex compared to other active high-mass star-forming regions. This demonstrates that few embedded clusters with luminous OB stars are present. However, two infrared sources associated with current star formation appear as X-ray-selected clusters.

First, CCCP-Cl 17 is the Treasure Chest, a compact cluster with bright mid-infrared emission dominated by an O9.5 star producing the compact H ii region MSX G287.84-0.82 (Hägele et al. 2004; Smith et al. 2005). About 100 members are in the full X-ray sample. Their average X-ray median energy is 1.8 keV, only slightly higher than the rich lightly obscured clusters; this is surprising considering the A_V ∼ 10–50 found from infrared studies of some Treasure Chest members. This suggests that Treasure Chest stars have a wide range of absorptions.

Second, a sparse grouping of CCCP sources can be seen on the eastern side of the Treasure Chest cluster near the compact mid-infrared source IRAS 10439-5941 = MSX6C G287.84-0.82, a very bright illuminated pillar in mid-infrared images of the Carina Nebula (Smith et al. 2000; Rathborne et al. 2004). It is a faint radio continuum compact H ii region and is associated with a molecular globule with column density $N_{\rm H_2} \simeq 9 \times 10^{21}$ cm⁻² and mass 30–200 M_☉. The infrared source has a very high bolometric luminosity 3.7 × 10⁶ L_☉ and likely represents the collective emission of an embedded OB cluster. The CCCP survey has detected a few of these stars.

Third, Group 20 just south of Trumpler 16 is centered 30'' west of IRAS 10430-5931, an embedded protostar with bolometric luminosity ∼1 × 10⁴ L_☉ accompanied by a compact group of lower-mass stars near the edge of the bright-rimmed molecular globule MSX6C G287.63-0.72. This small star-forming region is discussed in detail by Megeath et al. (1996). The strongest X-ray source here is a high-L_x PMS star presented by Evans et al. (2003).

Other sparse groups of X-ray sources that did not satisfy our criteria for producing the lists in Tables 1 and 2 may be present. For example, a few sources appear to be associated with the mid-infrared source IRAS 10441-5949 = MSX6C G287.8893-00.9316, which is modeled as a dusty proto B0 star with bolometric luminosity ≃3 × 10⁴ L_☉ by Rathborne et al. (2004).

The inferred fraction of embedded X-ray stars is relatively low compared to other large-scale star formation regions studied with Chandra. For example, the M 17, Rosette, and NGC 6634 complexes each have hundreds of embedded X-ray stars, a significant fraction of the population in the associated revealed central clusters (Broos et al. 2007; Wang et al. 2009; Feigelson et al. 2009). The comparison with the Rosette Molecular Cloud is particularly valuable, as the CCCP ACIS exposures have sensitivities similar to the RMC maps presented by Wang et al. where a number of obscured clusters, each with dozens of X-ray stars, were readily detected. Were a similar ratio of obscured-to-revealed stars present in the Carina complex, thousands of heavily absorbed stars would have been detected with cluster/group 〈MedE〉 values in the range 2–4 keV. Nonetheless, a moderate population of nascent stars is present: ∼1400 infrared-excess young stellar objects are found from Spitzer IRAC maps, most of which are not detected in X-rays (Povich et al. 2011b).

• 1.  
  CCCP-Cl 14 lies southeast of Trumpler 16, centered on the eclipsing binary O5.5V star V662 Car (Figure 2(c)). Sanchawala et al. (2007) first reported an X-ray-selected cluster here, identifying 10 of the 41 CCCP sources we assign to the cluster. Lying 2' northeast of the C¹⁸O cloud core 12, its CCCP members have a somewhat higher absorption (A_V ≃ 10 mag; Table 1) than that reported by Sanchawala et al. (2007) for the other principal clusters in the Carina complex.
• 2.  
  CCCP-Gp 18, lying at the southern tip of the mosaic, has only five faint X-ray source members, but is extraordinary in two ways (Townsley et al. 2011a). Three of the sources lie within ∼1'' of each other, and the group's median energy, 〈MedE〉>5 keV, is equivalent to A_V ∼ 125 mag. Chandra observations of star-forming regions have, on rare occasions, detected sources with median energies around 5–6 keV. Examples include several stars near the Becklin–Neugebauer hot core in Orion (Grosso et al. 2005), several stars in NGC 6334 I(N) and nearby embedded clusters (Feigelson et al. 2009), and two Class 0/I protostars in nearby star-forming cores (Hamaguchi et al. 2005; Getman et al. 2007). However, there have been no reports of a molecular cloud, diffuse or compact infrared emission, or other indicator of current star formation at the location of Group 18 (10^h44^m519, − 60°25'09'').
• 3.  
  CCCP-Gp 4 and 7 can be seen as low contours ∼10' northwest of Trumpler 14 in Figure 1. Group 4 is associated with the faint mid-infrared source MSX6C G287.2238-00.5339 and the C¹⁸O cloud core 9 with estimated molecular gas mass 1400–2200 M_☉ and column density $\log N_{\rm H_2} \simeq \break 21.9$ cm⁻² (Yonekura et al. 2005). Three prominent 2MASS stars are also present, two of which are young stars with infrared-bright disks (Mottram et al. 2007). Group 4 members have a high 〈MedE〉 equivalent to A_V ∼ 11, consistent with being embedded in this cloud. These groups are discussed in Povich et al. (2011b).

The origin of the dispersed distribution of young stars discussed in Section 3.3 is not obvious, and several causes may be present. The stars in Regions A and B may be associated with the Trumpler 14, 15, and 16 clusters, either as primordial low density halos or as members ejected by dynamical interactions. Evidence for a stellar halo around Trumpler 14 also emerged from the near-infrared study of the inner 3' by Ascenso et al. (2007). The star distribution in Region C was discussed in Section 5.2. A few groups are associated with current star formation in the Southern Pillars, and we suspect that other members were formed in similar molecular structures that have now dissipated. Further discussion of this possibility appears in a recent Spitzer study of stars in the Southern Pillars (Smith et al. 2010) and in Chandra studies of star formation around nearby H ii regions (e.g., Getman et al. 2007).

The widely distributed stars (Region D) do not exhibit any concentrations or association with prominent young clusters. Their distribution is likely to extend beyond the ∼50 pc CCCP mosaic. These most likely represent an older population of PMS stars, much as the Orion Nebula region (Ori OB1d) is surrounded by older, more dispersed clusters (Orion OB 1a–c in Bally 2008). A widely distributed population of massive stars was also found in a Chandra study of the NGC 6334 complex (Feigelson et al. 2009). The idea of a previous generation of star formation whose most massive members have produced supernovae in Carina is strongly indicated by the strong diffuse X-ray emission (Townsley et al. 2011b), the presence of a young neutron star in Region D east of Trumpler 16 (Hamaguchi et al. 2009), and the absence of early O-stars in Tr15.

We can quantify the size of the populations in the four large-scale Regions by constructing X-ray luminosity functions (XLFs) of CCCP sources classified as Carina members. These are shown in Figure 3 for each Region, where a small number of highly luminous sources associated with known OB stars (Skiff 2010) are removed to facilitate comparisons among lower-mass populations. Approximate observed X-ray luminosities for individual sources are obtained by multiplying the photon flux F_t,photon (Section 2) times the median energy and the factor 4πd² where d = 2.3 kpc. We then apply an absorption correction based on the median energy value, as described by Getman et al. (2010). This correction is not large in most cases, but is not available for the faintest sources. The Carina region XLFs are compared to the XLF of 839 lightly obscured low-mass members of the ONC obtained in the Chandra Orion Ultradeep Project (COUP; Getman et al. 2005; Feigelson et al. 2005). The membership and initial mass function of the ONC is measured (e.g., Muench et al. 2002), so it can serve as a calibrator for more distant clusters. We assume that the ONC population is 2800 stars, obtained by Hillenbrand & Hartmann (1998) within a radius of 2.06 pc. The total population, including brown dwarfs and outer regions of the cluster, may be substantially larger.

Zoom In Zoom Out Reset image size

The CCCP XLFs have a similar shape to the COUP XLF above log L_x ≃ 30.5 erg s⁻¹; the XLFs in this luminosity regime are roughly a power law dN/d(log L_x) ∝ (−1.0 to − 0.9) × log L_x. We then examine the vertical offsets between the fitted power laws in Figure 3 to estimate the total populations of young stars in each region. The results are Region A has 10 ONCs (28,000 stars); Region B has six ONCs (17,000 stars); Region C has five ONCs (14,000 stars); and the outlying areas in Region D have 16 ONCs (45,000 stars). The total population in the CCCP field is then about 104,000 stars down to the brown dwarf limit.

Caution is needed in interpreting these estimates of total Carina member populations. The XLFs will be significantly affected by the strong spatial variations of the survey sensitivity on these scales which combine several ACIS pointings. We have further assumed that the classification of individual CCCP sources into Carina members and contaminants has no error and that none of the "unclassified" sources are members. Both of these assumptions are not accurate (Broos et al. 2011b). CCCP sources with very few photons do not have estimated X-ray luminosity values and are missing from the lower end of the XLFs. The 104,000 stars estimate undercounts heavily obscured populations that are not well represented in the CCCP survey (Povich et al. 2011b). Nonetheless, because these estimates of total young stellar population are scaled to the ONC where the membership is well established, they represent the estimates of the full populations down to the brown dwarf limit, not just the X-ray detected sample.