THE STRUCTURE OF THE STAR-FORMING CLUSTER RCW 38

The Spitzer Space Telescope has observed the RCW 38 cluster with the IRAC instrument at four different epochs as part of PID 20127. This observing strategy purposed to fully cover the ∼30' × 30' field while recovering the inner ∼15' × 15' field with shorter exposures to overcome the high background of the central core. The first epoch, Epoch1, was taken on 2006 June 5 and utilized cluster mode to map the outer field in a spiral pattern with each pointing offset by ∼270'', with 12 s frames and five dithers per position with the cycling dither turned on. The map ends before the interior region of ∼6' × 6' was observed to avoid the build-up of short-term latents on the image. The second epoch, Epoch 2, was taken directly following Epoch 1 on 2006 June 5 but observed a smaller ∼15' × 15' central field of RCW 38. An integration time of 2 s with one dither per position with the cycling dither turned on was used to avoid saturation. The third epoch, Epoch 3, was observed on 2007 February 21 and followed the same spiral pattern over the outer field of ∼30' × 30' surrounding the RCW 38 core, again ending before the core was observed. The last epoch, Epoch 4, directly followed Epoch 3 and was taken on 2007 February 21 and also observed a ∼15' × 15' IRAC overlap region of the RCW 38 core. Table 1 lists the integration times and coverage of the four epochs.

The analysis used the basic calibrated data (BCD) FITS images from the S14.0.0 pipeline of the Spitzer Science Center for Epochs 1 and 2 and the S15.3.0 pipeline for Epochs 3 and 4. These data were combined into mosaics using WCSmosaic, which also removed cosmic ray and other artifacts. Custom IDL routines, part of the clustergrinder package (Gutermuth et al. 2009), were used to clean the mosaics of bright source artifacts. The pixel scale of the mosaics is 12 pixel⁻¹. The calibration uncertainty across the IRAC bands is estimated at 0.02 mag (Reach et al. 2005). In addition, there are ∼5% position-dependent variations in the calibration of point sources in the flat-fielded BCD data; these have not been corrected for in our data. Furthermore, these systematic uncertainties have not been added into the photometric uncertainties reported in this paper. The noise in a 8 × 8 pixel region around the source was calculated to assess the combined instrumental noise, shot noise, and noise from the spatially varying extended nebulosity. Point sources were found using Photvis, based on a heavily modified version of the DAOfind program in the IDLPHOT package (Landsman 1989). Point sources with peak values more than 5σ above the background were considered candidate detections. After identifying the point sources, aperture photometry was obtained using the aper.pro program in the IDLphot package. An aperture radius of 2 pixels was used, and a sky annulus from 2 to 6 pixels was used to measure the contribution from extended emission in the aperture with mmm.pro, also in the IDLphot package. The zero points for these apertures were 17.0879, 16.5843, 15.9790, and 15.1945 mag for the 3.6, 4.5, 5.8, 8.0 μm bands, respectively, in the native BCD image units of MJy sr⁻¹.

In Epoch 1, there were 4195 detections in the 3.6 μm band with uncertainties less than 0.2 mag, and 3346, 1536, and 653 in the 4.5, 5.8, and 8.0 μm bands, respectively, with 302 detections in all four IRAC bands and 816 detections in the three shortest IRAC bands. In Epoch 3, there were 4137 detections in the 3.6 μm band, and 3941, 1596, and 893 in the 4.5, 5.8, and 8.0 μm bands, respectively, with uncertainties less than 0.2 mag, with 415 detections in all four IRAC bands and 1039 in the three shortest IRAC bands. In the smaller deeper field Epoch 2, there were 2106 detections in the 3.6 μm band with uncertainties less than 0.2 mag, and 1605, 713, and 205 in the 4.5, 5.8, and 8.0 μm bands, respectively, with 108 detections in all four IRAC bands and 329 in the three shortest bands. In Epoch 4, there were 1775 detections in the 3.6 μm band with uncertainties less than 0.2 mag, and 1844, 454, and 299 in the 4.5, 5.8, and 8.0 μm bands, respectively, with 116 detections in all four IRAC bands and 311 in the three shortest bands. The sensitivity of the 8.0 μm band is the limiting factor in four-band detections: it suffers from lower photospheric fluxes, higher background emission, and the presence of bright, structured nebulosity across much of the image. The arrangement of the detectors ensures that only a fraction of the sources can have four-band detections in any single epoch (cf. Section 2.4). The detection threshold in the 3.6 μm band was ∼17 mag—just at the hydrogen-burning limit for a 1 Myr old star at the distance of RCW 38 (∼16.5 mag at 3.6 μm; Baraffe 1998).

We observed the RCW 38 cluster on two occasions with the Multiband Imaging Photometer for Spitzer (MIPS; Rieke et al. 2004). The observations were taken as part of PID 20127. The first epoch was observed on 2006 March 3 and had an aim point of $8^{{\rm h}}59^{{\rm m}}5\mbox{$.\!\!^{\mathrm s}$}78$, $\rm -47^{d}30^{m}43\mbox{$.\!\!^{\mathrm s}$}02$ (J2000). The second epoch was observed on 2006 June 18 and had an aim point of $\rm 8^{h}59^{m}5\mbox{$.\!\!^{\mathrm s}$}49$, $\rm -47^{d}30^{m}41\mbox{$.\!\!^{\mathrm s}$}14$ (J2000). In both epochs, the medium scan rate was used and the total map sizes were approximately 05 × 05 including the overscan region. The typical integrated exposure time per pixel is about 2.62 s at 24 μm. The MIPS images were processed using the MIPS instrument team Data Analysis Tool, which calibrates the data and applies a distortion correction to each individual exposure before combining into a final mosaic (Gordon et al. 2005). The data were further processed using various median filters to remove saturation effects and column-dependent background structure. The resulting mosaics have a pixel size of 125 at 24 μm. The 24 μm images suffered from saturation of the central core region and strong extended emission over the field; it was not possible to extract reliable photometry for any objects, instead the 24 μm mosaic was used to search for deeply embedded protostars surrounding the core (cf. Section 5.4).

This paper utilizes the previous Chandra observations of RCW 38, first presented in Wolk et al. (2006) and now available on the ANCHORS archive (AN archive of Chandra Observations of Regions of Star formation⁶). The observations were reprocessed and an updated source detection procedure was applied for the archive, and this reprocessed data set is used here. Originally, the data were processed in 2001 through the CXCDS ver.6.12 pipeline. A wavelet-based source detection algorithm, originally developed for ROSAT, known as PWDetect (Damiani et al. 1996) was then used to identify sources across the entire ACIS-I array. The array was 2 × 2 binned and the threshold significance was set to detect sources between 4 and 5.31 equivalent Gaussian sigmas; the data are searched on scales of 05–16''. With these settings, a false detection rate of <1% is expected. A total of 460 sources were found. The 2 × 2 binning, employed due to memory constraints means the image is undersampled with respect to the maximum resolution of ACIS. Hence, some close sources, on-axis, can be conflated. For the new source extraction, a three-pass method was used, with the CIAO 3.4 Wavdetect tool (Freeman et al. 2002). The three-pass method searches for sources in images binned by 1, 2, and 4 pixels, respectively, using wavelet scales of 2, 4, 8, 16, and 32 pixels on each image. Each image has a size of 2048 × 2048 pixels. This method was designed to optimally take advantage of additional memory more commonly available in 2007. The source limits were again set to limit false detections to about 1%, however Wavdetect has more robust error analysis and tests indicated that this was slightly more sensitive than PWDetect at the nominal 1% false detection probability. The results of the three passes are combined into a single list of 536 sources. This includes 18 sources detected on the S array which was not considered to by Wolk et al. (2006). By comparison the Chandra Source Catalog (Evans et al. 2010) detects 410 sources in the field with a minimum flux threshold of about 10 net counts on axis, which is significantly higher than our equivalent limit. Table 2 lists the identifiers, coordinates, and properties of the 536 sources identified in this reprocessed data set. The table provides the Chandra identifier, source locations, the raw and net number counts (net counts are background subtracted and aperture corrected), plasma temperature (kT), hydrogen column density (N_H), absorbed and unabsorbed X-ray flux (F_X), and flaring statistics.

For each epoch of Spitzer data, the photometry of the four IRAC bands were merged with the 2MASS J, H, and K bands and Chandra X-ray data to form an initial catalog using custom IDL routines. Sources observed in different wavelength bands, which were located within 1'' of each other were merged in the catalog; if while comparing two bands, multiple sources in one of the bands satisfied this criteria, the closest was chosen. The Chandra X-ray data were taken from Wolk et al. (2006), where a full description of the data reduction process was presented. The observation, Chandra ObsID 2556, had a total exposure time of 97 ks, with 536 sources detected in the Chandra field. These catalogs were used in the identification of the YSOs, cf. Section 3.

The FOVs at each IRAC epoch and of the different instruments did not cover the same extent of the cluster, with the IRAC epochs 2 and 4 limited to a 15' × 15' field in the central cluster. The two larger fields, epochs 1 and 3, cover a 30' × 30' FOV of the cluster, with a ∼6' × 6' gap in coverage to avoid saturation in the cluster core. Furthermore, the IRAC detectors are split into two groups, channels 1 and 3 (3.6 and 5.8 μm) and channels 2 and 4 (4.5 and 8.0 μm), whose FOVs are offset from one another by 65. The Chandra field is the smallest overall, covering a region 17' × 17', centered on IRS 2. The X-ray sample is thus limited to the central region of the cluster, and data are not available for all IR sources in the catalog. The 2MASS data are not spatially constrained. A final photometric catalog was then created by matching the sources detected in the four initial catalogs to within 1''. This catalog was used in the variability study, cf. Section 5.2.

YSOs can be identified by their excess emission at IR wavelengths. This emission arises from reprocessed stellar radiation in the dusty material of their natal envelopes or circumstellar disks. The infrared identification of YSOs is carried out by identifying sources that possess colors indicative of IR excess and distinguishing them from reddened and/or cool stars (Megeath et al. 2004; Allen et al. 2004; Gutermuth et al. 2004). The main limitation of this method is the contamination from extragalactic sources such as PAH-rich star-forming galaxies and active galactic nuclei (AGNs), and galactic sources such as asymptotic giant branch (AGB) stars; both of these groups have colors similar to YSOs. Four color–color diagrams were used to determine the color excess of the sources: two IRAC diagrams: [3.6]–[4.5] versus [5.8]–[8.0] and [3.6]–[4.5] versus [4.5]–[5.8], and two IRAC-2MASS diagrams: J − H versus H − [4.5] and H − K versus K − [4.5]. Figure 3 shows the four diagrams for each IRAC epoch (Ep. 1–4; note that the 2MASS photometry is the same throughout). In the following analysis we required photometry to have uncertainties <0.2 mag in all bands used for a particular color–color diagram. The numbers of sources for each diagram which satisfy this criteria are, respectively, in Epoch 1: 302, 816, 2748, 3007; in Epoch 2: 108, 329, 1084, 1342; in Epoch 3: 415, 1039, 3068, 3478; and in Epoch 4: 116, 311, 1195, 1534.

Zoom In Zoom Out Reset image size

3.1.1. IRAC Color–Color Diagram

3.1.2. IRAC-2MASS Color–Color Diagrams

3.1.3. IRAC Three-band Color–Color Diagram

Due to the limiting factor of requiring an 8 μm detection to place a source on the IRAC CCD, we also utilized the IRAC three-band CCD [3.6]–[4.5] versus [4.5]–[5.8] to identify sources with excess emission but with photometry only in the 3.6–5.8 μm bands. The criteria for young excess sources on this diagram are presented in Gutermuth et al. (2009). Contamination from shocked emission and PAH contamination is dealt with by the selection criteria; however, some AGN interlopers may remain. Sources with 3.6 μm magnitudes fainter than 16 mag were removed as possible AGNs. In Epoch 1, 101 sources were identified as having an excess using this method, of which 24 were not previously known from the other IR diagrams. In Epoch 2, 145 sources were identified, of which 27 were new. In both Epochs 3 and 4, 101 objects were identified, with 26 and 25 new sources, respectively. Combining the four epochs, a total of 33 additional YSOs were identified using this method.

In addition to this method, the color–magnitude diagram of [3.6] versus [3.6–5.8] was employed to identify possible cluster members not identified on the previous diagrams due to weak or no excess emission from the disk or a lack of photometry at 8.0 μm. Figure 4 shows the [3.6] versus [3.6–5.8] color–magnitude diagram, with field stars shown as black dots, the previously identified YSOs as gray circles, and the newly selected candidates as dark gray stars. Those sources with 8.0 ⩽ m_3.6 ⩽ 14.0 and a color more than 1σ greater than [3.6–5.8] > 0.4 were selected. This region delimited the color–magnitude space of the previously identified YSOs on this diagram. We identified a further 177 candidates with the appropriate colors. Given that contamination cannot be identified using the contamination removal methods outlined in Gutermuth et al. (2009), we consider these to be tentative candidate members only and do not include them in the following discussions, see Section 5 and Tables 3 and 4. The coordinates and photometry of the 177 candidates are presented in Table 5.

Zoom In Zoom Out Reset image size

3.1.4. Completeness

The RCW 38 cluster lies at a distance of 1.7 kpc and exhibits a highly variable background across the extent of the cluster. Completeness limits for the Spitzer photometry were assessed using a method of inserting artificial stars into the mosaics and then employing our detection algorithms to identify them. Estimates for the 90% completion limits for all epochs in each of the IRAC bands were 13.5, 13.0, 12.0, 10.5 mag, respectively. These limits correspond to an approximate stellar mass of 0.5 M_☉ at 3.6 μm and an age of 1 Myr (Baraffe 1998). Figure 5 shows the magnitude histograms of the 3.6 μm photometry at each of the four epochs; all detected sources with uncertainties ⩽0.2 are shown by the upper black curve. The red histograms show the sources with detections in bands required to place them on at least one of the IR-excess diagrams. The sample of identified YSOs is shown by the green histogram. The blue histogram shows the subsample of identified YSOs that possess an X-ray detection.

Zoom In Zoom Out Reset image size

The hydrogen-burning limit, at 1.7 kpc, is ∼17 mag at 3.6 μm and an age of 1 Myr (Baraffe 1998). Though the photometry is sensitive to this magnitude, at this distance source crowding and the bright, highly varying background become serious issues and, as can be seen from the histogram of identified YSOs, our identification threshold is closer to 15 mag at 3.6 μm, or ∼0.15 M_☉ at 1 Myr (Baraffe 1998). When considering objects that can be placed on the CCDs we are limited to ∼13.5 mag at 3.6 μm (or 0.5 M_☉). The ratio of the number of sources with m_3.6 ⩽ 13.5 mag that can be placed on the color–color diagrams to the total number of sources with m_3.6 ⩽ 13.5 mag (and correcting this number for completeness by dividing by the fraction of artificial stars recovered in each magnitude bin) is 1778/1883 or 94% for Epoch 1, 993/1094 or 90% for Epoch 2, 1822/1966 or 92% for Epoch 3, and 867/975 or 89% for Epoch 4. These fractions represent upper limits on the completeness of the sample, as the number of contaminating field stars is rising with increasing magnitude, while the number of YSOs does not show an equivalent rise.

The DeRose et al. (2009) study demonstrates the loss of detections toward the RCW 38 nebula in the IRAC data. DeRose et al. (2009) identify 483 sources with uncertainty <0.2 in at least one near-IR band in the central 0.5 pc surrounding IRS 2; in comparison, we detect 31 objects in at least one IRAC band with uncertainty <0.2 in the same region. The measured stellar surface density in DeRose et al. (2009) is 10 times that found with Spitzer due to the higher angular resolution available with the VLT and the reduced confusion with nebular emission achieved toward the RCW 38 nebula in the near-IR.

A more conservative estimate for the 90% completeness of the PMS membership of the cluster (those with IR-excess emission) can be made from the green histograms of Figure 5 as 12 mag at 3.6 μm, or ∼2 M_☉ at 1 Myr (Siess et al. 2000). This estimate does not account for extinction, which will move completeness to even higher masses. Further, completeness varies across the field due to the varying background emission, being particularly low in the central region surrounding IRS 2. Since the fraction of sources with IR excess varies with mass and age (Lada 1987; Hernadez et al. 2006), the completeness with respect to all PMS stars cannot be estimated with the available data.

YSOs possess elevated levels of X-ray emission, with luminosities, L_X, of $\frac{ L_{{\rm Xbol}} }{ L_{X\odot } } \sim 10^{3.5}$ that can be used to distinguish them from foreground or background field stars (Feigelson & Montmerle 1999; Feigelson & Kriss 1981). We utilize this property to identify YSOs that do not have emission from a dusty disk (evolutionary class III) and would otherwise be indistinguishable from field stars. Protostars (class 0/I) and PMS stars with disks (class II) with elevated X-ray emission may also be detected. Wolk et al. (2006) observed the core of RCW 38 in a 96 ks observation. The Chandra image of RCW 38 covered a 17' × 17' FOV centered on the cluster core and IRS 2, thus X-ray data are not available over the entire spatial extent of the sources in our catalog. The 2006 catalog contained 460 detections of which 294 were determined from their X-ray detection and NACO near-IR photometry (DeRose et al. 2009) to be likely young members of the cluster. In this paper, we make use of the reprocessed and updated catalog, available on the ANCHORS online archive, to compare to the IR sources to identify likely cluster members.

There were 536 X-ray detections in the catalog, a mixture of YSOs, AGNs, and dMe stars. For each epoch, the coordinates of the X-ray sources were matched to the nearest IR detection to within a radius of 1''. In Epoch 1, 175 X-ray detections were identified with IR sources, 203 in Epoch 2, 190 in Epoch 3, and 198 in Epoch 4. In total, 226 X-ray sources were identified with IR sources. This number is less than the 294 X-ray detections matched in the Wolk et al. (2006) study since the resolution of the Spitzer data is much lower than Chandra in the central 5' where ∼50% of the X-ray sources are located. The lower limit of the X-ray luminosity detectable in RCW 38 with Chandra can be estimated by comparison with the Chandra Orion Ultradeep Project (COUP) data, see Feigelson et al. (2005), as log₁₀L_X ≈ 30.1 erg s⁻¹, assuming a distance of 1.7 kpc to RCW 38, an exposure time of 97 ks, and a median value of log₁₀(N_H) of 22.3. We match the X-ray sources with IR counterparts to reduce the contamination from AGNs; an X-ray source with no IR counterpart is assumed to be AGN contamination.

Of those sources identified in X-rays, 97 were previously identified from the IR-excess diagrams in Epoch 1, thus 78 YSOs were identified using X-rays. In Epoch 2, 136 were identified in the IR and 67 were X-ray identifications. In Epoch 3, 110 were previously known, with 80 X-ray identifications. In Epoch 4, 131 were identified as excess sources, with 67 detected in X-rays. In total, 74 YSOs were identified through their elevated X-ray emission. Of these, 41 had previously been identified in the Wolk et al. (2006) study, so that 33 are newly identified as members in this work. It should be noted that the ACIS field only partially covers the region, and thus the X-ray detection fractions are only representative of the central region and not the outer distributed YSO population. Wider scale X-ray studies of region would allow us to examine the X-ray properties and dispersion of class III objects in the region. The completeness of the X-ray data can be assessed from Figure 5; it should be noted that only 67% of the YSOs from class 0/I to class II lie in the Chandra FOV, thus the completeness of the X-ray data is limited partially by the smaller field. Considering only those young stars located in the Chandra FOV, and those YSOs with m_3.6 < 12, 55% of this sample are detected in X-rays. This percentage remains approximately constant over the range of 3.6 μm magnitudes from 8 < m_3.6 < 12. Hence, we estimate the X-ray YSO detection completeness to be 55% for objects with m_3.6 < 12.

The four epochs of data were combined, yielding a total of 679 candidate members of the cluster over the IRAC overlap field. Of these, 32 were found to be contaminants: AGN or PAH-rich galaxies, leaving 647 in the IRAC fields. The numbers of YSOs identified in each of epochs 1–4 were 369, 358, 504, and 377. On trimming this catalog to include only those sources in the IRAC overlap field of one or more epochs, the total number of cluster members was found to be 624. Of these, 226 were detected in X-rays by Chandra, including 74 that identified as YSOs by their elevated Chandra X-ray emission. The membership of RCW 38 is estimated to be 90% complete to 12 mag at 3.6 μm or ∼2 M_☉ at 1 Myr (Siess et al. 2000). As will be shown, ∼40% of objects in each evolutionary class are detected in X-rays, thus we are possibly missing ∼200 class III members in the ACIS field. We also identified a further 177 candidate cluster members, which would bring the total membership identified here to 801.

By scaling the Harvey et al. (2006) results to our 0.2 deg² field, we estimate that there is of order 1 AGB contaminant in our catalog. We estimate that one or less of the class III sources may be a dMe contaminant.

In recent years numerous studies have been undertaken to investigate the emission properties of PMS stars and protostars in the higher energy X-ray region of the spectrum (Wolk et al. 2006; Getman et al. 2005; Winston et al. 2010). In developed, hydrogen-burning stars, X-ray activity arises from magnetic fields generated as a result of shear between the core radiative zone and the outer convective zone. In young stars, X-ray emission is thought to arise from mechanisms such as magnetic disk-locking between the star and disk (Hayashi et al. 1996; Isobe et al. 2003; Romanova et al. 2004), accretion onto the star (Kastner et al. 2002; Favata et al. 2003, 2005), or alternative dynamo models for coronal emission (Küker & Rüdiger 1999; Giampapa et al. 1996). The elevated levels of X-ray emission in PMS stars have been shown to decrease with stellar age (Feigelson et al. 2005), as is observed on the main sequence.

Of the 433 YSOs located within the Chandra field, 226 (52%) had confirmed X-ray detections. Nine of these were class 0/I protostars (9/22 or 41% of the class 0/I in the Chandra FOV), 18 were flat spectrum (18/58 or 31% of the flat spectrum sources), and 125 were class II detections (125/279 or 45%). The remaining 74 sources did not exhibit excess emission and were identified as class III members of the cluster solely on their X-ray detection; hence 74/624 or 12% of the members were identified solely by Chandra. To within 1σ there is no apparent trend in rate of detection with evolutionary class, with a ∼40% detection rate for all classes. This detection rate is likely a function of the sensitivity of the Chandra observations. Since we do not possess an independent method of identifying class III young stars we cannot establish the fraction of class III sources detected in X-rays. As we shall subsequently argue the X-ray properties of class II and class III objects appear to be similar, we therefore assume that the fraction of class III objects is also similar to that of the class II. Here we neglect any suppression of X-ray activity in class II stars due to accretion (Telleschi et al. 2007). We thus estimate the total number of class III members to be (74/44.8  ±  4% [percentage of CII detected in X-rays])/60.5% [correction for Chandra coverage] or 272⁺²⁶ _{− 23}, assuming a disk fraction of 67% (see following discussion), for an estimated cluster membership of 823⁺²⁶ _{− 23} objects with mag_3.6μm > 12.

Of the 226 sources identified as young stars, 138 were previously identified as young members in the Wolk et al. (2006) study. The remaining 88 consisted of 43 which were identified as YSOs by comparison to the IR catalogs in this study and 45 which were new X-ray detections included in this catalog. Of the 74 class III sources, 41 were previously identified as members by Wolk et al. (2006), the remaining 33 class IIIs are newly identified in this study by their X-ray emission and IR photometry.

The disk fraction of the YSOs, the ratio of the number of X-ray detected disk-bearing PMS stars (class II) to the total number of X-ray detected PMS stars (class II and III) was found to be 125/(125 + 74) = 63% ± 6%. On including the X-ray detected protostellar sources (class 0/I and flat spectrum) this fraction rises to (9 + 18 + 125)/(9 + 18 + 125 + 74) = 67% ± 5%. The disk fractions are calculated based on the assumption of completeness to 12 mag at 3.6 μm, or 2 M_☉, and similarity in the detection rates of class II and III sources. These fractions are consistent with a cluster age of ∼1 Myr (Hernadez et al. 2007). Both are significantly higher than the 29% ± 3% found by DeRose et al. (2009) using VLT/NACO in a ∼0.5 pc area surrounding IRS 2. By considering only those of our sources falling in the same field (two class II and three class III) we find a disk fraction, 2/2 + 3 = 40%  ±  28%. Expanding to a 1 pc from the IRS 2 binary, the disk fraction becomes 17/17 + 7 = 71%  ±  17%, which is comparable to the region as a whole. These results suggest that proximity to the massive IRS 2 stars may have resulted an increased rate of disk processing in the surrounding lower mass YSOs.

There are no discernible trends in N_H against kT with evolutionary class, the class II and III objects are similar. The class 0/I and flat spectrum sources have marginally higher column densities; this would be indicative of their surrounding infalling natal envelopes. The X-ray luminosity of PMS stars is known to vary with bolometric luminosity (Casanova et al. 1995). As was previously found by the authors for Serpens and NGC 1333 (Winston et al. 2007, 2010) and by Hernadez et al. (2007) in σ Ori, when using the J band as a proxy for the bolometric luminosity in RCW 38 there is no difference in the dependence of L_X on L_bol between class II and III. Figure 8 (right) shows the X-ray luminosity (at an assumed distance of 1.7 kpc) plotted against the dereddened J-band magnitude. A linear fit to the 16 class II sources provides log (L_X) = −0.338  ±  0.027(J_dered) + 35.39  ±  0.31, while the fit to 8 class II shows log (L_X) = −0.331  ±  0.042(J_dered) + 35.32  ±  0.52. This reinforces the suggestion that the mechanisms of X-ray generation are similar for both class II and III and is the basis for our estimate of the class III population. On the left of Figure 8, the hydrogen column density is plotted against the extinction measured at K_s band, A_K. Only those detections with X-ray count in excess of 100 were considered when examining emission properties to ensure reliable estimates, though in Figure 8 (left) the open symbols also show sources with >50 counts.

Zoom In Zoom Out Reset image size

While this work was not explicitly designed to test gas absorption, N_H, versus dust extinction, A_V ≈ 10 × A_K, the X-ray data support the use of a simple model of an absorbed thermal plasma to measure N_H to within 30% (Wabs*APEC; Morrison & McCammon 1983; Smith et al. 2001). We can then examine the data for any trends in the N_H/A_K ratio. Historical measurements of N_H/A_V range from approximately 2.2 × 10²¹ cm⁻² (Ryter 1996) derived from O-star absorption, to roughly 1.6 × 10²¹ cm⁻² (Vuong et al. 2003) derived from a well-behaved sample of PMS stars in the ρ Ophichus cluster. Previously, in the lower mass clusters Serpens and NGC 1333, the authors reported on a decreased N_H to A_K ratio, of ∼0.6 × 10²² (Winston et al. 2007, 2010), than that quoted for the diffuse ISM in Vuong et al. (2003) of 1.6 × 10²². This is thought to be due to the agglomeration of grains in cold dense clouds. However, in RCW 38 the sources are found to fit very well to the ratio of 1.6 × 10²², consistent with the Vuong et al. (2003) value for the local ISM and nearby molecular clouds, and with the previous measurement for the region of 2.0 × 10²² (Wolk et al. 2006). The class II have a median ratio of N_H = 1.66 ± 0.85 × 10²²A_K from 16 sources, while the class III have a median ratio of N_H = 1.52 ± 0.49 × 10²²A_K from 8 objects. The difference may arise from the much higher mass of the RCW 38 cluster. Smith et al. (1999) found that the gas-to-dust mass ratio toward IRS 1 was 10⁴:1, indicating dust depletion with respect to the ISM. The more active stars may efficiently irradiate the dust in the intercluster medium preventing grain growth or aggregation of dust grains, which would be more likely in the colder medium of the less massive clusters (Cardelli et al. 1989; Stepnik et al. 2003). A larger scale survey of the region, including the outer subclusters, would allow us to ascertain if this difference is more dependent on nascent environment or on the presence of OB stars in the vicinity.

Young stars vary photometrically over time. One of the first criteria, in fact, by which young T Tauri stars were identified was their photometric variability. This variability occurs across all wavelengths from the UV through optical to IR and into the high energy X-ray regime. The causes are manifold; XU Ori/FU Ori flaring, stellar spots, stellar pulsation, and in the IR, heating of the circumstellar disk (Carpenter et al. 2001). In RCW 38 we have four epochs of observations to search for known YSOs exhibiting variability and for young diskless class III members not detected in X-rays. Given the different regions and exposure times of the outer field and central pointings and the consecutive observations of epochs 1 and 3, and 2 and 4, only the variance of sources between epochs 1 and 2, and 3 and 4 were examined. Further, only the IRAC four-band overlap subsample was considered. Thus, we are searching for variability on timescales of ∼1 year. A variable source was defined as one having a difference in magnitude in one or more IRAC bands of >3σ above the combined uncertainties.

Of the 13,950 sources in our sample, we identified 72 candidate variable stars. Figure 9 shows the variation of the 72 variable stars at each of the four IRAC bandpasses for the two timeframes considered. Forty-eight of these were previously identified as YSOs due to excess emission; variability is noted in the final column of Table 3. The remaining 24 objects were not otherwise found to be YSOs; from an examination of their SEDs it was found that two exhibit excess emission characteristic of circumstellar disks but were not detected in enough bands to be placed on the color–color diagrams, while the other 22 are candidates to be diskless class III young stars. The coordinates, identifiers, and photometry of these 24 sources are given in Table 6. Of the 24 objects, 9 were found to also be O-star candidates and are marked as such in Table 6. Of those 48 previously identified YSOs found to be variable, 7 were class 0/I (three X-ray sources), 12 flat spectrum (4 X-ray), and 29 were class II (7 X-ray). No known class III were identified as variable. These numbers represent fractions of variables per class of 30% ± 11% for class I, 14% ± 4% for flat spectrum, and 7% ± 1% for class II. The variability fractions for the X-ray detected sample of YSOs were consistent with that of the non-X-ray detected sample. The protostellar objects have a combined variability fraction of 18% ± 4% compared to the 7% ± 1% for the disk-bearing class IIs. Such high variability is consistent with larger and/or more frequent accretion events during the protostellar phase.

Zoom In Zoom Out Reset image size

A study of the properties of these variable sources found that they were evenly spatially distributed across the IRAC overlap field; this is not unexpected as variability is a universal property of young stars. The variables exhibit a similar range of extinctions as the YSOs not found to be varying and there was no trend found between the extinction at K band and the extent of the variance. There was no statistically significant difference in detection of variance in X-ray luminous objects or in the X-ray luminosity of the variable members. There is some indication that the percentage of variables decreases with evolutionary class, with ∼20% of the protostellar YSOs varying compared with ∼7% of the class II. None of the previously detected class III sources were found to vary. The extent of variance also appears to show some decrease with evolutionary class: the mean variance of the class 0/I was 0.43 ± 0.31, that of the flat spectrum 0.26 ± 0.16, the class II 0.27 ± 0.17, and those variables showing no infrared excess emission 0.12 ± 0.12. The trend is not strongly significant but is consistent with a decrease in the scale of accretion events in the later evolutionary stages.

RCW 38 is one of the nearest massive star-forming regions with an estimated membership of more than 1000 members. Previously, Wolk et al. (2006) identified 31 candidate OB stars in a 5 pc² region centered on (and including) IRS 2 using an absolute K-band magnitude ⩾−0.35, for a 2.7 M_☉ at 0.5 Myr (Siess et al. 2000). Twenty were located in the central ∼1 pc. Following Knodlseder (2000) we use the K versus J − K color–magnitude diagram to identify the massive stars in the subsample of objects within the IRAC four-band overlap region. We identify 29 candidate O stars and a total of 604 OB star candidates. The coordinates and photometry of the 29 O-star candidate sources are given in Table 7. Of the 29 objects, 9 were found to overlap with the variable sources and are marked as such in Table 7. An off-field region was subtracted from the cluster to account for field star contamination in the cluster; subtracting the off-field sources yielded an estimate of 26 O stars in the RCW 38 region; thus, we estimate that ∼3 of our 29 objects may be contaminants. Two of the objects, 158 and 244, are identified as YSOs and are assigned the evolutionary class II, they are X-ray detected and were also previously identified as candidates by Wolk et al. (2006) as 112 and 396, respectively. The small overlap (2 of 31) between the two lists arises from two causes. First, the smaller FOV of the Wolk et al. (2006) VLT/NACO observations and their higher sensitivity in the cluster center, where we require a detection in at least one of the IRAC bandpasses. Second, the Wolk et al. (2006) study searches for OB star candidates with M_K ⩽ −0.35, while here M_K ⩽ −3.65. Of the remaining 31 sources, a further 19 are classified as YSOs, and 2 more had IRAC detections in too few bands to classify. Seven were not detected as point sources with IRAC. IRS 2, the known O5.5 binary, is not resolved in the IRAC mosaics and does not appear in the candidate list.

Previously, the total population of RCW 38 was estimated as being of order one thousand members. An estimate of the total membership of the RCW 38 region has been presented in Section 5.1 as 823⁺²⁶ _{− 23} objects with mag_3.6μm > 12. In comparison, Wolk et al. (2011) estimate the total population of Trumpler 16, which contains a similar number of massive stars, as being roughly 14,000 young stars. From this we suggest that the total population of young stars forming in the RCW 38 region is closer to 10⁴, making it one of the more massive known clusters within 2 kpc.

The MIPS 24 μm mosaic was examined in a search for embedded protostars or young objects associated with emission regions. Three such emission regions were identified to the N of the cluster along the ridge of PMS stars forming the NW subcluster. The diffuse emission is associated with three stars not previously identified as YSOs, variable, or O-star candidates. Their IRAC colors and SEDs suggest that they lack circumstellar material and that the emission arises from dust heating of the surrounding nebula. The coordinates and photometry of the three stars are given in Table 8.

A deeply embedded object was detected in the central subcluster, at coordinates $\rm 8^{h}59^{m}05\mbox{$.\!\!^{\mathrm s}$}95$, $\rm -47^{d}30^{m}39\mbox{$.\!\!^{\mathrm s}$}7$. It is offset from IRS 2 by ∼35 or ∼6000 AU. This object was detected only in the 8.0 μm band (no 24 μm photometry is available due to saturation); it does not alter its position during the year between the Ep. 2 and Ep. 4 observations and hence it is assumed to be a deeply embedded protostar. It has a magnitude of 4.679 ± 0.099 at 8.0 μm and is faintly visible at 5.8 μm, though not resolved as a point source. It is not detected at 4.5 μm suggesting an absence of shocked molecular hydrogen emission, which would indicate that it is not a jet feature. Therefore, we assume that this object is an embedded early protostar and may be embedded behind the O5.5 binary. A second object was also identified only at 8.0 μm but was found to move from Ep. 2 to Ep. 4 and hence considered a contaminating foreground source.

The spatial distribution of YSOs in a cluster provides an insight into the fragmentation processes leading to the formation of protostellar cores, evidence for triggered star formation and the subsequent dynamical evolution of the stars as they evolve from the protostellar to the main sequence (Megeath et al. 2004). While DeRose et al. (2009) examined the young stars in the inner core of the cluster at near-IR wavelengths, finding no evidence of subclustering, we will examine the extended molecular cloud. By including the longer mid-IR wavelengths we can obtain a clearer determination of the IR-excess emission, and thereby more accurately classify the YSOs as protostellar, disk-bearing class II or diskless class III. These data provide a deeper understanding of the star formation history of the extended RCW 38 region.

5.5.1. Spatial Distribution of Identified YSOs

The spatial distribution of the young stars in the cluster in each evolutionary class was examined using a nearest neighbor technique. We define the nearest neighbor distance as the projected distance to the nearest YSO of the same class, assuming a distance to RCW 38 of 1.7 kpc. For each evolutionary class, 10K random distributions of stars were generated, with the number of stars equal to the number of objects in the given evolutionary class. For the class III sources, the randomly distributed stars were constrained to fall within the Chandra FOV; for all other classes, the random distribution covered the region of the IR field. The resulting nearest neighbor distributions of the observed YSOs and the random distributions were compared for each class using the Kolmogorov–Smirnov test. The probabilities that the random and observed distributions were drawn from the same parent distribution are listed in Table 9. The distributions of the class 0/I, flat spectrum, and class II are found to be non-random. The class III sources show some evidence of mixing; this is likely a statistical effect of the smaller Chandra FOV.

Table 9. K-S Test Probabilities of YSOs in RCW 38 by Evolutionary Class

Evolutionary Class	Class 0/I	Flat Spectrum	Class II	Class III	Random
Class 0/I	...	0.31	1.2 × 10−5	0.016	6.9 × 10−3
Flat spectrum	0.31	...	1.3 × 10−9	0.02	5.5 × 10−6
Class II	1.2 × 10−5	1.3 × 10−9	...	2.6 × 10−3	2 × 10−7
Class III	0.016	0.02	2.6 × 10−3	...	0.26

Download table as:  ASCIITypeset image

Figure 10 shows the nearest neighbor differential distributions by evolutionary class of the young stars. The open histograms show all sources in that class, while the filled histograms show the subsample of X-ray luminous stars. The upper plot shows the random distribution, averaged over the 10 K runs. The median spacing of the protostellar members was larger than that of the PMS objects. The median separation over the field for the class 0/I sources was 0.644 pc, and of the flat spectrum 0.433 pc. In comparison the PMS class II members had a median spacing of 0.249 pc, and the class III 0.346 pc.

Zoom In Zoom Out Reset image size

Figure 11 shows the spatial distribution of all identified YSOs as a function of their evolutionary class in the upper plots. The lower plots shows the spatial distribution of the young stars, again as a function of evolutionary class, in an expanded region centered on IRS 2. The detection of class III members is limited by the smaller FOV of the Chandra ACIS instrument. Evidence of subclustering and separate protostellar formation sites can be observed in the upper panels of Figure 11—indicating that a unique center of star formation is unlikely. An arc or filament of protostellar objects is also observed surrounding the central IRS 2 cavity. The formation of these young stars may have been influenced by their proximity to the IRS 2 binary. The spatial distributions of the class II and III stars surrounding IRS 2 do not trace this arc/filamentary structure. A SIMBA 1.2 mm map of the IRS 2 region is overplotted in the lower left of Figure 11, and shows a ridge of emission which is spatially coincident with the filament of protostars (Vigil 2004). This implies that the protostars remain embedded in their natal, dense dust and gas while the PMS stars have dispersed through the cluster.

Zoom In Zoom Out Reset image size

5.5.2. Subclustering: Surface Density

The stellar surface density of the YSOs was also calculated to determine the underlying structure of the cluster; the resulting map is shown in Figure 12. The local surface density at each point on a uniform grid was calculated following the method outlined in Casertano & Hut (1985):

$$\begin{eqnarray*} &&\sigma (i,j) = \frac{N-1}{\pi r_{N}^{2}(i,j)}, \end{eqnarray*}$$

where r_N is the projected distance to the Nth nearest cluster member. In this case, the stellar surface density was calculated with a grid size of 100 × 100 using N = 18, the distance to the 18th nearest neighbor, to smooth out smaller scale structure. Further, Casertano & Hut (1985) found that the uncertainty goes as σ/(N − 2)^1/2, so by taking N = 18, the uncertainty is 25%. The surface density plots show three subclusters surrounding the main IRS 2 core: a large subcluster to the SW, one to the NW, and a possible small grouping to the NE. The central cluster itself appears to have three density peaks which are likely real, though small, overdensities in the YSO population. One of these peaks corresponds to the region immediately surrounding IRS 2. The central subcluster coincides with the eastern cloud of Yamaguchi et al. (1999). The SW subcluster is coincident with the western cloud implying continued star formation in that region.

Zoom In Zoom Out Reset image size

5.5.3. Subclustering: Minimum Spanning Tree

Another method of examining the spatial distribution and structure of a cluster is the minimum spanning tree (MST). The MST is defined as the set of branches or lines connecting a set of points such that the sum of the branch lengths is minimized and there are no closed loops. Figure 13 shows the MST for the YSOs in RCW 38. The MST provides a better method of characterizing the separation lengths of sources as all points are connected in a single network, whereas the nearest neighbor technique tends to create small subgroupings due to pairing of adjacent points. The MST also has the advantage of not smoothing out geometry in the substructures. By assuming a critical MST branch length, the length where a break occurs in the cumulative distribution of branch lengths, substructures within the cluster can be defined. Following Gutermuth et al. (2009), in Figure 14 we plot the distribution of branch lengths in the MST to obtain the characteristic branch length, found to be 0.412 pc. This was done by applying two linear fits to the data using the IDL routine linfit and minimizing the χ² uncertainty of the two fits. In Figure 13, the black branches are those with branch lengths less than the characteristic branch lengths while the gray branches are those with longer lengths. Those stars joined by black branches can be said to trace subclusters with the extended region. We note that the central core has two large subclusters to the NW and SW, and a third smaller subcluster to the NE, containing only eight stars, which may be a developing region.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Extending the usage of the MST, Cartwright & Whitworth (2004) formulated the Q parameter, to quantify and distinguish between smooth large-scale radial density gradients and multiscale or fractal subclustering. This method provides a statistical method of characterizing structure in stellar clusters. The Q parameter is defined as $\bar{m}/\bar{s}$, where $\bar{m}$ is the normalized mean MST branch length, $\bar{m}(N_{{\rm total}}A)^{1/2}/(N_{{\rm total}}-1)$, and $\bar{s}$ is the normalized mean separation, $\bar{s}/R_{{\rm cluster}}$. A Q parameter equal to 1 indicates a smooth radial distribution of sources, while values closer to zero indicate subclustering. For the RCW 38 region observed with IRAC, $\bar{m} = 0.2219$ pc and $\bar{s} =$ 0.5407 pc, yielding a value of Q equal to 0.41. This low value indicates that subclustering defines the structure and spatial distribution of the YSOs. DeRose et al. (2009) found a value of Q in the smaller central IRS 2 region of 0.84, indicating that no subclustering was present on this scale. This result is consistent with our own since the DeRose et al. (2009) study covered an area of ∼0.5 × 0.5 pc², corresponding to the density peak containing IRS 2 within the central subcluster. On this smaller scale, any structure may be masked by the dynamical evolution of the young stars. On the larger scale the underlying subclustering of the complex is evident.

5.5.4. The Subclusters

The RCW 38 region has extended star formation over the 30' × 30' IRAC field, a larger area than previously assumed. The central cluster where IRS 1 and 2 are located is surrounded by three subclusters and distributed star formation. Table 10 presents the populations of the subclusters, their median extinction and IRAC slope, and their protostellar and variable fractions.

Table 10. Properties of RCW 38 Subclusters

Cluster	Total	C0/I	FS	CII	CIII	Var	OB	AK	deredαirac	Protostars	Variables
										(%)	(%)
NE	11	1	1	9	...	1	1	0.684 ± 0.382	−0.996 ± 0.802	18.2 ± 12.8%	9.1 ± 9.1%
NW	62	...	9	53	...	8	1	0.608 ± 0.225	−1.078 ± 0.671	14.5 ± 4.8%	12.9 ± 4.5%
SW	41	...	4	37	...	3	1	0.530 ± 0.244	−1.146 ± 0.541	9.7 ± 4.9%	7.3 ± 4.2%
CN	255	15	39	154	47	24	6	0.836 ± 0.747	−0.957 ± 1.295	25.9 ± 3.5%	9.4 ± 1.9%
Dist.	222	7	31	157	27	36	20	0.701 ± 0.448	−1.056 ± 0.808	19.5 ± 3.2%	16.2 ± 2.7%

Note. C0/I: class 0/I; FS: flat spectrum; CII: class II; CIII: class III; Var: variables; OB: OB star candidates; A_K: extinction at K band; ^deredα_irac: slope of IRAC SED.

Download table as:  ASCIITypeset image

Each of the subclusters contains variable stars; the NE subcluster contains just 1, the NW contains 8, the SW contains 3, and the central subcluster contains 24. The remaining 36 form part of the distributed population. While the four subclusters contain very similar fractions of variables, the distributed population contains ∼50% more. The protostellar fractions of the NE and NW subclusters and the distributed population are similar at ∼18%, while the SW subcluster fraction is appreciably lower (∼10%) and the central cluster is significantly higher at ∼26%. The OB star candidates are distributed throughout the extended cluster region, cf. Figure 15, with 20/29 not located in one of the subclusters. Each of the four subclusters contains at least one massive star candidate (the central cluster contains at least 35: four new candidates in this study and 31 from Wolk et al. 2006), indicating that star formation may have been initiated independently in each of the subcluster regions, with a subsequent generation (possibly due to triggering) in the central region around IRS 2.

Zoom In Zoom Out Reset image size

The median K-band extinctions measured toward each subcluster and the distributed population are similar to within 1σ, consistent with all of the regions being at the same distance. Likewise, the median IRAC SED slopes, _deredα_IRAC, are similar. This suggests a similar age and evolutionary history across the whole region. This further supports the idea that star formation is not propagating through the region, but rather that the molecular cloud has undergone collapse and is forming stars in fractal subclusters. Further spectroscopic observations may yet find a trend toward older ages in the distributed population similar to that observed in Serpens and NGC 1333 (Winston et al. 2009). The higher protostar fraction in the central cluster, rather than implying that the outer subclusters and distributed population are older per se, may be due to a second generation of protostars formed via triggering due to the IRS 2 O5.5 binary.

The northwest subcluster of YSOs corresponds to the 2MASS cluster Obj 36 (Dutra & Bica 2001), which is associated with the reflection nebula vdBH-RN43 (van de Bergh & Herbst 1975) located at 08^h58^m04^s, −47^d22^m50^s in the region of RCW 38. The association of this cluster with the main RCW 38 cluster has not previously been established. However, from the Spitzer observations, the subcluster does not appear to be physically isolated from the RCW 38 core; a band of YSOs stretches between the two. Visually, the structure of the molecular cloud appears contiguous between the two regions, cf. Figure 1. The ¹³CO observations of Yamaguchi et al. (1999) also show emission extending between the regions. The K-band extinctions, A_K, and magnitudes at J band and 3.6 μm were compared to those of the core and other subclusters; no difference was found in either the range or median of the extinction or magnitudes between the NW subcluster and the rest of the region. This indicates that Obj 36 lies at a similar distance to the RCW 38 core and that it can be considered a subcluster in the extended RCW 38 star formation region.