CHANDRA/ACIS-I STUDY OF THE X-RAY PROPERTIES OF THE NGC 6611 AND M16 STELLAR POPULATIONS

Figure 1 shows the image of the Eagle Nebula in the I band obtained with the Wield Field Imager mounted on the 2.2 m telescope at ESO (Guarcello et al. 2007). The contours mark the dust emission detected at 8.0 μm with Spitzer/IRAC (Indebetouw et al. 2007; Guarcello et al. 2009), marking the Pillars of Creation, the Column V, and the SFO30 cloud (the outer contours delimit the size of the used IRAC image). The rotated squares represent the fields observed with Chandra/ACIS-I. The archival observation of 78 ks (Linsky et al. 2007) is centered on NGC 6611 (the c-field), and it includes the Pillars of Creation and the SFO30 cloud. The two new observations of 80 ks (P.I: Guarcello) are centered on Column V (the e-field), and on a group of Class I and embedded Class II stars (the ne-field) that is younger than 1 Myr.

Zoom In Zoom Out Reset image size

X-ray image reduction and source detection in all the fields are described in Guarcello et al. (2010b), where detected X-ray sources (1158, 363, and 315 in the c-, e-, and ne-fields, respectively) with their optical–NIR counterparts have been used to classify young stellar objects (YSOs) in this region. Briefly, for each observation a Level 2 event file has been obtained using the CIAO 4.0³ tool acis-process-events, retaining only the events incompatible with cosmic rays and removing the 05 event position randomization added by the standard Chandra data processing pipeline. Source detection has been performed with PWDetect⁴ (Damiani et al. 1997) adopting a threshold corresponding to 10 spurious detections. The total number of detected X-ray sources is 1755 (81 sources are in the overlapping region between the c-field and the e-field). As described in Guarcello et al. (2010b), the mass limit of the X-ray sources with optical counterparts (for which it was possible to estimate the mass) is equal to 0.2 M_☉. Figure 1 shows the photon extraction regions of each detected source. The clustering of sources in NGC 6611 is evident, but this image clearly shows that the outer parts of the nebula are also well populated by young stars.

Photon extraction has been made using the IDL software ACIS Extract⁵ (AE; Broos et al. 2010), which uses TARA,⁶ CIAO, FTOOLS,⁷ and MARX⁸ packages. AE calculates the point-spread function (PSF) for each source, using it to define a photon extraction region as the region encompassing 90% of the PSF evaluated at 1.49 keV. Background events have been extracted for each source in circular annuli centered on the sources, applying a mask for source counts defined in two iterations (first a circular area covering 99% of the local PSF and then with a more accurate mask region). The extraction regions of crowded sources are reduced to avoid overlapping each other, down to 40% of the local PSF in the most crowded regions. In this first phase, AE uses the source positions provided by the source detection algorithm (PWDetect in our case). Once both the PSF and the background for each source are evaluated, AE computes new source positions by correlating the source images with the local PSF model. The user has the possibility to use these new positions or retain the set of positions provided by the source detection algorithm. As suggested by the AE manual, we updated the positions of the sources observed with an off-axis larger than 5' (further coordinate corrections of Δ_α = −018 and Δ_σ = −005 have been applied to cross-correlate the X-ray catalog with the optical–infrared catalog; see Section 2.2). AE then repeats the computation of the PSF and the background and the photon extraction using the new positions. For each source, AE provides information such as the net observed counts, the light curves, and a probability that the observed light curve is constant based on a Kormogorov–Smirnov test; the most relevant is summarized in the electronic catalog described in the Appendix.

Detected X-ray sources are classified in Guarcello et al. (2010b) according to the properties of their optical/infrared counterparts. Among the 834 sources with disks, selected by their NIR excesses with respect to the expected photospheric emission, 219 have an X-ray counterpart and are classified as "disk-bearing members." We classified as "disk-less members" 964 X-ray sources that are younger than 10 Myr and have an extinction higher than 2.6^m, according to their position in the optical color–magnitude diagram. A total of 76 sources are older than 10 Myr and/or have extinction smaller than 2.6^m. These sources are most likely foreground main-sequence stars detected in X-rays, which are then classified as "foreground sources." A total of 504 X-ray sources are not identified with a known stellar counterpart. In Guarcello et al. (2010b), we compared our count-rate limit (6.25 × 10⁻⁵ counts s⁻¹), properly converted into a limit flux in the 0.5–10 keV energy band, with the log N vs. log S distribution of extragalactic sources in the ELAIS field shown in Puccetti et al. (2006), estimating an upper limit of the observed extragalactic X-ray sources equal to 193.5. We then expect that about 300 X-ray sources without optical–infrared counterparts can be very embedded sources associated with M16 (a significant fraction of them lie in the regions with high extinction, such as the ne-field) or background sources lying in our Galaxy. Since we cannot discern among these hypotheses, we simply classified them as "background sources" and have not considered them in our study of the global X-ray properties of the M16 population.

Figure 2 shows the spatial distributions of these four classes of X-ray sources. The disk-bearing stars are clearly clumped in NGC 6611 toward Column V and in the embedded northeast cluster. The disk-less members are more numerous and are distributed across the whole cloud. Also, the background sources are sparse in the field, but the high concentration of them in the regions with the highest extinction and embedded ongoing star formation activity (the ne-field, eastward the Column V, and in the tip of the Pillars of Creations and Column V) is consistent with the idea that most of these sources are very embedded young stars associated with M16.

Zoom In Zoom Out Reset image size

The determination of the absorption corrected X-ray luminosity (L_X), as well as the plasma temperature (kT) and hydrogen column density (N_H), requires the analysis of the X-ray spectra. AE provides both the source and background spectra, the redistribution matrix files, and the ancillary response files. We fit the observed spectra with thermal plasma (with both one and two temperatures) and power-law models. We use the APEC ionization-equilibrium thermal plasma code (Smith et al. 2001), assuming the sub-solar elemental abundances of Maggio et al. (2007). The absorption was treated using the WABS model (Morrison & McCammon 1983). The one-temperature (1T) thermal model was applied to all the sources with more than 25 counts, while the two-temperature (2T) thermal model was applied to each source with more than 80 counts. The power-law model has been applied to those sources with hard spectra for which the best-fit thermal model predicts a plasma temperature kT > 5 keV. In order to avoid false convergences due to local minima in the χ² space, we applied different sets of initial values to each model. When more than one model has been used for a given source, we chose the best model by the χ² probability and visual inspection of the spectrum.

In the c-field we were able to obtain a good spectral fit for 441 sources, 114 in the e-field and 60 in the ne-field (out of a total of 1158, 363 and 315 sources, respectively), for a total of 575 sources with a good spectral fit (40 sources in the overlapping regions). For 33 sources the best-fit model is a two-temperature thermal plasma model. For a total of 20 sources the best-fit model is a power-law spectrum: 15 of them are classified as "background sources," and the remaining 5 are classified as "disk-less members."

Spectral analysis cannot be performed for faint sources. For them, an estimate of the X-ray luminosity can be provided by the analysis of the photon energy quantiles (Hong et al. 2004). This method is similar to the analysis of the X-ray hardness ratio adopted in several works, but it does not suffer the drawback of the choice of the energy bands used to define the "X-ray colors." Adopting the definition by Hong et al. (2004), if x% is the percentage of the total counts with energy below the value E_x%, the quantile Q_x is defined as

$$\begin{equation} Q_x=\frac{E_{x\%}-E_{{\rm lo}}}{E_{{\rm up}}-E_{{\rm lo}}}, \end{equation} \tag{ 1 }$$

where E_lo and E_up are the lower and upper limits of the energy range, respectively, 0.5 keV and 8.0 keV in our case. We used the median energy (i.e., the 50% quartile) and the 25% and 75% quartiles, which are combined in the following independent variables: log (Q₅₀/(1 − Q₅₀)) and 3 × (Q₂₅/Q₇₅). In order to estimate the values of N_H and kT for each source, the observed quantile variables have been compared (by interpolation) with a theoretical grid of thermal models with sub-solar abundances and a grid of power-law models with spectral indices typical of active galactic nuclei (AGNs; Brandt et al. 2001).

Figure 3 shows the thermal and power-law grids and the quantiles of the X-ray sources (only the sources with a significance larger than 2 are shown; see the Appendix) for the four groups of sources defined in Section 2.2. The thermal models have a plasma temperature ranging from 0.3 keV to 10 keV and hydrogen column density from 0.01 × 10²² cm⁻² to 6 × 10²² cm⁻²; the power laws have the same hydrogen column density range and a photon index ranging from 0.1 to 2.0. The actual grids used for the interpolations are finer than those shown in Figure 3.

Zoom In Zoom Out Reset image size

The disk-bearing and disk-less sources (upper panels) are mostly concentrated in the region with 1.2 keV ⩽ kT ⩽ 4 keV and from 0.1 × 10²² cm⁻² ⩽ N_H ⩽ 1.1 × 10²² cm⁻², corresponding to a wide range of visual extinction. Both the background and foreground sources are mostly outside the thermal grid. The distribution of the latter suggests values of N_H smaller than those used for the grid. The former mostly show harder spectra and are, in most cases, compatible with power-law models. However, from the grid it is not possible to discriminate the real extragalactic sources from the embedded cluster members, given the degeneracy between very hard thermal spectra and power laws.

A close inspection of Figure 3 shows the presence of eight background sources in the area of the soft thermal spectra corresponding with kT  <  0.5 keV. Even if they are not very bright (the median net counts of this sample is 19), the probability associated with the hypothesis that they are spurious detections is negligible (see the PBS parameter, explained in the Appendix). Five of these sources lie in the c-field and have a median photon energy between 1.3 keV and 1.8 keV. It is not obvious what these sources are. One intriguing possibility is that some of them are compact objects associated with M16 and are remnants of the explosion of a massive star. The possibility that M16 hosted a supernova explosion in the past has been suggested by Flagey et al. (2011) to explain the hot dust (∼70 K) they observed in the cloud. However, NGC 6611 is not old enough to firmly claim that the most massive members already exploded as supernovae, and the median photons energy of these eight sources is higher than that of the candidate neutron stars recently identified in the Carina complex by Townsley et al. (2011b).

In total, 846 sources lie inside the thermal grid. For these sources the individual values of N_H and kT have been calculated by interpolating their quantile variables in the thermal grid and then their X-ray luminosity, L_X, using the distance adopted for the cluster (1750 pc; Guarcello et al. 2007) and correcting for the individual absorption. To estimate the uncertainties in the X-ray stellar parameters from the grid, we calculated the errors in their photons energy quantiles, following the procedure of Hong et al. (2004), and propagated to the independent quantile quantities. Then, we repeated the interpolation with the grid four times, each time with a different combination of sum and subtraction of the errors to the quantile quantities. In this way, for each star and for each parameter we had four slightly different values. For each nominal value (i.e., those derived from the grid without adding or subtracting the errors) of the three quantities, the upper error has been defined as the difference between the maximum of these four measures and the nominal value, the lower error as the difference with the minimum value. The median errors for the three parameters are $\sigma _{N_{{\rm H}}}=0.7\,{\rm cm}^{-2}$, $\sigma _{k_T}=0.4\,{\rm keV}$, and $\sigma _{\log (L_X)}=0.3\,{\rm erg\,s^{-1}}$.

A total of 312 sources lie inside the power-law grid, and, as shown in Figure 3, 82% of them are background sources (with a wide range of hydrogen column density) and disk-less stars. There are 346 sources lying outside both grids. Most of them have 3Q₂₅/Q₇₅ smaller than those of the grid models. Hong et al. (2004) suggested that low significance sources can fall in the area of the diagram with 3Q₂₅/Q₇₅ < 0.8. In fact, the sources in this region of the diagram have a median of 12 counts, indicating they mostly fall in this region because of low counts.

The main aim of this paper includes the comparison between the X-ray properties of disk-less and disk-bearing stars, the study of the global X-ray properties of the M16 population, and their comparison with those of the Orion members. A reliable comparison requires knowledge of the upper limit of X-ray luminosity for the undetected members. The problem of the incompleteness of the sample of Class III objects of M16 will be discussed later, since we do not know the X-ray-undetected Class III stars. This is not the case for the disk-bearing members, since we have detected in X-rays 219 sources with infrared excesses out of a total of 834 previously identified sources. The upper limit of the photon count rate at the positions of the undetected disk-bearing stars is calculated with PWDetect, adopting the same threshold significance used for source detection. We have converted these upper limit count rates to X-ray luminosity using as a conversion factor the median of the ratio between the X-ray luminosities and photon count rates for the detected members.

X-ray luminosities and their uncertainties have been obtained from the spectral or quantile analysis with the same criteria adopted for plasma temperature and hydrogen column density (Section 3). In addition, for the sources without a good spectral fit and that are outside the thermal grid in the diagrams in Figure 3, they have been inferred directly from the count rates, using as a conversion factor the median L_X/(count_rate) ratio found for the other sources (which ranges from 7 × 10³³ to 1 × 10³⁴ in the three fields).

The total luminosity of M16 is equal to log (L_X) = 34.90 erg s⁻¹, with an individual source median value of log (L_X) = 30.4^+0.1_−0.6 erg s⁻¹. In this estimate, we excluded a candidate disk-less member with very high luminosity log (L_X) = 34.1^+0.1_−0.5 erg s⁻¹. This luminosity, in fact, is apparently too high to be correct, since it is two orders of magnitude higher than the L_X of the second most luminous star in our sample. Its spectrum has been fitted with a two-temperature thermal model, with a soft primary component (kT₁ = 0.11 ± 0.01 keV) and a harder secondary component with kT₂ = 1.7 ± 0.6 keV, but the emission measure of the hard component is 10⁵ times fainter than that of the soft component. One possibility is that this is a foreground source with an overestimated distance or a background high-mass X-ray binary, but both the N_H value (equal to 4.45 × 10²² cm⁻²) and its position in the optical color–magnitude diagrams suggest a young and embedded star lying in the BRC-SFO30.

Figure 6 compares the X-ray luminosity calculated from the quantile analysis with that from the spectral fits, for the sources in our sample with a good significance and a spectral fit with good χ². The estimate of L_X from the quantile analysis agrees with that from the spectral analysis for most of the stars in Figure 6, with the exception of few outliers. These outliers mostly lie in the low kT–high N_H region of the grid, where the low sensitivity of the ACIS detector to the soft part of the spectrum results in large uncertainty in estimates of N_H and kT.

Zoom In Zoom Out Reset image size

One of the aims of this paper is to verify whether the XLF of M16 confirms the universality of the XLFs in clusters younger than a few Myrs, as a consequence of the very slow decline of the X-ray luminosity for PMS stars younger than 10 Myr (Preibisch & Feigelson 2005). Several authors compared the XLFs of young clusters with that observed in the Orion Nebula Cluster (ONC) in the COUP survey, which is the most complete X-ray observation of a young cluster, complete down to 0.1 M_☉ (Getman et al. 2005). A good agreement was found in IC348 and NGC 1333 (2–3 Myr; Feigelson et al. 2005; Winston et al. 2010), M17 (∼1 Myr; Broos et al. 2007), NGC 6357 (∼2 Myr; Wang et al. 2007), NGC 2244 (2–3 Myr; Wang et al. 2008), and Cyg OB2 (3–5 Myr; Wright et al. 2010). In this paper we adopt the same approach, comparing the XLF of M16, where the X-ray-emitting population has an age ranging from <1–3 Myr (Guarcello et al. 2007), with that of the ONC.

Figure 7 shows the XLF we obtained for M16, together with that of the COUP sources in Orion. The latter is almost constant for log (L_X) < 30 erg s⁻¹, while the M16 XLF drops below this limit, which turns out to be the completeness limit of our observations. For luminosities higher than the completeness limit, the XLF of M16 can be fitted with a power law with index Γ = −0.85 ± 0.09, which is consistent with that of the Orion XLF (Γ = −0.93 ± 0.08; Feigelson et al. 2005) in the same luminosity range. In this calculation, we have not considered the OB stars detected in X-rays, since the Orion OB population is significantly different from that of M16.

Zoom In Zoom Out Reset image size

The comparison between the XLFs of Orion and M16 can help us to estimate the level of incompleteness of our sample of members down to the subsolar limit. To match the Orion XLF with the M16 XLF, we have to scale the former by 1.7. The total Orion population with M > 0.1 M_☉ counts 1600 members (Getman et al. 2005). The match between the two XLFs suggests a total M16 population of 2700 members, larger than the number of members we identified in our previous work (1907 in total, incomplete for a low-mass regime). This is a rough estimate of the still unidentified M16 population, since it depends on several properties of M16 and ONC, such as the physical extent and absorption characteristics in the respective sky areas that are not compared properly here.

In young PMS stars, L_X is generally observed to increase with the stellar mass, due to the fact that the dynamo mechanism in PMS stars is in the saturated regime, which means that it is independent of the rotational velocity and it depends only on the stellar mass and bolometric luminosity (Preibisch & Feigelson 2005). To study the L_X versus mass relation in M16, we estimated the masses of M16 members with optical and X-ray detections, first obtaining their intrinsic optical photometry using the extinction map of M16 obtained in Guarcello et al. (2010b), then finding both stellar mass and age by interpolating the dereddened V and V−I with the colors expected from the grid of isochrones of Siess et al. (2000). In this study, we focused only on the Class III stars, since the mass estimate from the optical photometry in Class II objects can be affected by the presence of the disk (Guarcello et al. 2010a).

Figure 8 shows the L_X versus mass scatter plot for Class III stars in M16 and Orion (Getman et al. 2005). In each mass bin, there is a large spread of L_X of about 1.5–2 orders of magnitude. This spread is mostly statistical, but binarity can also play a role (Feigelson et al. 2002). The distribution of the low-mass stars in M16 (i.e., with M ⩽ 0.8 M_☉) follows the L_X versus mass relation valid in Orion. In fact, a linear fit in this range of mass gives

$$\begin{equation} \log (L_X) = (30.9 \pm 0.1) + (1.1 \pm 0.3)\cdot \log (M/M_{\odot }) \end{equation} \tag{ 2 }$$

with a slope consistent with that found in Orion (Preibisch & Feigelson 2005). Given the consistency of the two XLFs and assuming the universality of the initial mass function (IMF), the fact that the low-mass population of M16 and ONC follows the same L_X versus mass relation is not surprising.

Zoom In Zoom Out Reset image size

In Figure 8, the distribution of stars more massive than 0.8 M_☉ is totally different, being flatter (with a slope of 0.4 ± 0.2) than that in the low-mass regime. Similar behavior has been observed in other young clusters. Our data do not allow us to understand the nature of the flattening of the L_X versus mass distribution observed at about 1 M_☉. The most likely explanation is that it is due to changes in stellar interior. Feigelson et al. (2003), following the models of Palla & Stahler (1993), have shown as in the mass range between 2.4 < M/M_☉ < 3.9, very young PMS stars have already developed a radiative core, surrounded by a convective mantle, where deuterium burning takes place. The efficiency of the magnetic dynamo and the X-ray activity level of the star strongly depend on the precise boundary between these two regions, and this boundary changes with the age and accretion history of the PMS phase. For even more massive stars, the mechanism for the production of X-rays changes drastically.

It is still debated whether the X-ray activity in PMS stars can be affected by the presence of a circumstellar disk. Several authors have compared the X-ray luminosity of Class III and Class II members of young clusters, and the general picture remains unclear. In some cases, such as Chamaleon I (Feigelson et al. 1993), ρ Ophiuchi (Casanova et al. 1995), and IC348 (Preibisch & Zinnecker 2002), the same level of X-ray activity in disk-less and disk-bearing members has been found; while in other clusters, such as the Taurus–Auriga region (Stelzer & Neuhäuser 2001), NGC 1893 (Caramazza et al. 2012), and NGC 2264 (Flaccomio et al. 2006), the X-ray luminosity of Class II objects is significantly lower than that in Class III stars. In the ONC the first attempt to find any difference in X-ray activity between these two classes of PMS stars was made by Gagne et al. (1995), who found no significant difference. Later, Flaccomio et al. (2003) discovered a different emission level in accreting and non-accreting stars in Orion. This result was reinforced by Preibisch & Feigelson (2005), taking advantage of the completeness of the member selection with the deep COUP data. These authors found no difference in X-ray emission between stars with and without infrared excesses, detected with a diagnostic based on the L photometric band, but they found a significant difference between non-accreting and accreting stars selected using the Ca ii line emission. Their finding suggests that the effects on the X-ray activity in disk-bearing stars are not due to the presence of the disk itself, but due to active accretion. In fact, in the ONC study of Prisinzano et al. (2008), where the disk-bearing stars have been selected by means of their infrared excesses, the level of X-ray emission in Class II and Class III stars is only marginally different.

The present study of the stellar population of M16 can help shed some light on the effects of the circumstellar disk in the X-ray emission in young stars, taking advantage of the accurate selection of both Class II and Class III members. However, a reliable comparison between the X-ray luminosity distributions of Class II and Class III objects can be attempted only if the two samples are complete. The Class II sample (219 stars) can be easily completed using the upper limits of L_X of the undetected Class II objects (see Section 2.5). These upper limits, together with the X-ray luminosity of detected Class II and Class III objects, are shown in the left panel of Figure 9 as a function of stellar mass.

Zoom In Zoom Out Reset image size

We cannot complete the Class III sample with the same approach adopted for Class II stars, since we do not have a representative sample defined independently of X-ray information. However, we can estimate the fraction of Class III members with X-ray emission below our completeness limit using the COUP data of Orion. Our sample of Class III objects, as well as the Class II, is more massive than 0.2 M_☉ (Guarcello et al. 2010b), and in Orion 15% of the Class III objects with M > 0.2 M_☉ have log (L_X) < 30 erg s⁻¹. Given the similarities between the X-ray properties of the two clusters, we can suppose that this is also valid for M16, i.e., that the undetected Class III stars more massive than 0.2 M_☉ are 15% of the detections (124 sources). Regarding their upper L_X limit, since our aim is to verify whether the X-ray activity in Class III sources is higher than that in Class II objects, we can consider the most conservative case, resulting in the lowest L_X distribution for Class III stars: that the upper limit of the X-ray luminosity of the missing Class III stars is equal to the lowest L_X measured for Class III objects in the COUP survey (Getman et al. 2005), equal to log (L_X) = 27.89 erg s⁻¹. We then calculated the XLFs for the "completed" samples of Class III (detected Class III sources plus 124 sources with log (L_X) = 27.89 erg s⁻¹) and Class II objects (detected Class II objects plus the upper limits of the undetected Class II objects), which are shown in the right panel of Figure 9. The emission level of the completed sample of Class II objects is evidently lower than that of the Class III, and we have verified with the ASURV⁹ (Feigelson & Nelson 1985) statistical package that the probability that the two XLFs are drawn from the same parent distribution is null. In the left panel, a few sources with log (L_X) = 29.5 erg s⁻¹ are present (fainter than our detection limit). They are low significance sources, whose luminosity has been inferred directly from their count rates. In conclusion, in M16 the level of X-ray emission is higher in Class III objects than in disk-bearing stars, supporting the idea that the presence of a disk results in lower X-ray activity. We cannot discern between accreting and non-accreting disks, but it can be expected that, since the average age of the M16 population is ∼1 Myr (Guarcello et al. 2007), a large fraction of stars with disks are still actively accreting.

There is more than one hypothesis to explain the observed X-ray emission level lower in Class II than in Class III stars. The accreting material, for instance, could be responsible for a higher absorption in disk-bearing objects. In fact, the correction we made for the extinction cannot be completely appropriate for disk-bearing stars. For these objects, it would be more appropriate to use an extinction model that takes into account the presence of both interstellar and circumstellar obscuring material (E. Flaccomio et al., in preparation). It is also possible that the accreting material deforms the large-scale stellar magnetic field (Romanova et al. 2004) or that it increases the gas density in the magnetic loops around the stars, resulting in less efficient heating of the material by the energy released during flares (Preibisch & Feigelson 2005), even if this would require a fraction of the stellar surface covered by hot spots larger than the typical values observed in CTTS (Muzerolle et al. 2001). The last possibility is that the accretion phenomenon changes the internal structure of the star, as pointed out by Wuchterl & Tscharnuter (2003), who found that even with small accretion rates the central star is no longer fully convective. A different hypothesis is that the presence of the accreting inner disk is not the cause of the lower X-ray activity but a consequence, since, as discovered by Drake et al. (2009), a high X-ray flux emitted by the central star can enhance the photoevaporation of the disk.

Only one Class I object has been detected in X-rays. Its position is α = 20:33:51.237 and δ = +41:25:10.68, and it has only IRAC and Chandra detections with a high X-ray luminosity log (L_X) = 31.49 erg s⁻¹ derived from the spectral fit with a one-temperature thermal model. This source lies in the embedded cluster in the ne-field, and it is not surprising that it has a very high extinction (N_H = 48.0 × 10²² cm⁻², corresponding to A_V = 60.8^m). The X-ray luminosity of this source is about half an order of magnitude higher than that of the brightest Class I objects in Orion (Prisinzano et al. 2008), but a similar embedded object has been identified in M17 (Broos et al. 2007). However, neither the IRAC nor the X-ray data allow us to discard the possibility that this source is a background AGN, whose locus in the IRAC color–color diagram (used to classify the selected disk-bearing stars) roughly coincides with that of Class I objects. The fit of the source spectrum with a power-law model is statistically acceptable and the photon index we obtained is equal to 2.03, compatible with typical values of AGN spectra. The only valid argument against the extragalactic nature of this source is its spatial correspondence with the embedded NE cluster.

Generally, O and late-B stars are sources of soft (kT ∼ 0.5 keV) X-rays, thought to be produced in a myriad of small shocks in their radiatively accelerated winds (Owocki & Cohen 1999). This emission is usually slowly variable, and, since the wind intensity scales with L_bol, X-ray emission also scales with L_bol, usually approximately as L_X = 10⁻⁷ × L_bol = 10³¹–10³³ erg s⁻¹ (Harnden et al. 1979; Pallavicini et al. 1981). However, a large scatter is usually observed, mostly due to a wide range of shock filling factors and binarity. Usually, a thermal emission spectrum observed in an OB star with kT = 0.5–0.7 keV and with a constant or slightly variable light curve is a signature that the X-ray emission is produced by the self-shocked wind. In some cases, however, an intermediate/hard component in the X-ray emission from O stars has also been observed: a moderate hard component (kT = 2–3 keV) rotationally modulated or with rapid variability (Stelzer et al. 2005) or a very hard thermal component due to shocks in the magnetically confined wind in the stellar equatorial plane in stars with strong magnetic dipole (Babel & Montmerle 1997); wind–wind collision in close massive binaries (Pollock et al. 2005), which can also be a source of soft X-ray emission (Gagné et al. 2011); or a hard non-thermal component caused by inverse Compton scattering of UV stellar photons by relativistic charged particles in the stellar wind (Chen & White 1991).

As listed in Table 1, all the O stars in NGC 6611 have soft spectra (with one exception that will be discussed later), suggesting that in M16 the shocked wind emission is the only mechanism producing X-rays in the O stars. Figure 10 shows the log (L_X) versus log (L_bol) relation for the OB stars in NGC 6611. The best linear fit for stars with L_bol > 10³⁷ erg s⁻¹ (the sample of B stars with log (L_bol) < 37 erg s⁻¹ is strongly incomplete) gives a slope of 0.7 × 10⁻⁷ (dashed line), similar to the L_X ∼ 10⁻⁷ · L_bol relation (represented by the solid line) typical of the shocked wind emission, as found first by Harnden et al. (1979) and Pallavicini et al. (1981) and then observed in other massive star-forming regions (i.e., in the Carina region; Nazé et al. 2011).

Zoom In Zoom Out Reset image size

A significant level of hard emission should be expected in principle from the O4V and O7.5V binary system W205 from the collision of the intense stellar wind emitted from the two components (expected to be ∼2.0 × 10⁻⁶ M_☉ yr⁻¹ and ∼0.35 × 10⁻⁶ M_☉ yr⁻¹; Smith 2006), but the separation between the two stars is too large (Sana et al. 2009 reported a rotation period of hundreds of years) to expect a significant wind+wind X-ray emission.

Only for W175 does the spectral fit suggest the presence of a possible hard component. This system is composed of an O8.5V and an O5V stars, with a third intermediate mass (spectral types A–F) component reported by Duchêne et al. (2001). The separation between the two more massive components is about 1200 AU (Sana et al. 2009). As expected, W175 is very bright in our Chandra observations, with 1258 net counts and log (L_X) = 32.65 erg s⁻¹. The spectral fit with abundances fixed at standard values is very poor, with χ²_ν = 2.06 indicating a statistically unacceptable fit. An acceptable spectral fit (χ²_ν = 0.86) was instead obtained adopting a two-temperature thermal model with abundances for iron and oxygen allowed to vary. The values of N_H, kT, and L_X of W175 obtained with this fit are reported in Table 1, and the abundances found are O = 2.3 ± 0.7 O_☉ and Fe = 0.7 ± 0.2 Fe_☉. In O stars a subsolar abundance of iron can be expected, as predicted, for example, by the model developed by Zhekov & Palla (2007), but in the case of W175 the iron abundance is only marginally subsolar. This spectral fit suggests the presence of a hard component with a temperature kT = 4 ± 2 keV, even if the emission measure of this hot plasma is more than 20 times smaller than that of the cold component, which can be produced in the wind-colliding zone such as those observed in other similar systems (i.e., HD 93129A; Gagné et al. 2011). Figure 11 shows the two-temperature thermal plasma model fitted to the W175 spectrum.

Zoom In Zoom Out Reset image size

In the sample of early-B stars of NGC 6611, the ratio of bolometric flux emitted in X-ray ranges from −5.4 ⩽ log (L_X/L_bol) ⩽ −8.5. This large scatter is related to the variety of mechanisms for X-ray production operating in this family of stars. Among these B stars, 10 have kT < 1 keV and a constant light curve, according to the K-S test performed by AE. Their X-ray emission is then typical of shocked wind emission. On the other hand, 21 early B-stars have a plasma temperature kT  >  1 keV, with 6 of them showing very hard spectra, with kT  >  3 keV. For the B stars with plasma temperature kT  >  1 keV, the K-S test cannot help to discern between the possibility that the hard emission is due to an unresolved low-mass companion or to the action of one of the mechanisms described above. However, among the six stars with kT > 3 keV, five are not spectroscopically classified as close binary systems; for these, the hard X-ray emission cannot be produced by colliding winds. Their light curves produced by AE suggest some flaring activity that will be studied in detail in a forthcoming paper. This emission can come from a low-mass unresolved companion in a wide orbit observed during a flaring activity or from the B stars themselves, since, as proposed by Feigelson et al. (2002), magnetic reconnection events can take place on the surface of young B stars.

One peculiar case is W601. This B1.5V star has been suggested to be very young (0.016 Myr) and in a transition phase between the Class II and the MS phases, with a P Cygni like H_α line typical of the presence of outflow/infall activity (Martayan et al. 2008). This star has been studied in detail by Alecian et al. (2008), who found an accretion rate equal to 10⁻⁴ M_☉ · yr⁻¹, a rotational velocity of 190 ± 10 Km s⁻¹, and an intense mean magnetic field equal to 3 kG, classifying it as a Herbig Be star. With these characteristics, it is not surprising that we detected an intermediately hard component in the W601 spectrum (kT = 1.72 keV). The K-S probability that the light curve is constant is high (0.96).

As expected, the fraction of detected sources in the A to late-B range of spectral classes is very low: 19% of late-B stars and 10% of the A–F stars, with log (L_X) < 30.4 erg s⁻¹, consistent with emission from a low-mass companion following the L_X versus mass relation shown in Figure 8. Evidence that the X-ray emission detected from A-late B stars is due to a low-mass companion has been obtained by De Rosa et al. (2011), who found that multiplicity in a sample of X-ray-detected intermediate-mass stars is about four times higher than in a control sample of stars.

The central cavity contains the core of NGC 6611, with almost all the OB stars discussed in Section 5 and a rich YSO population (385 members detected in X-rays, among which 21% have a disk) with a median age of 1 Myr. This region has the highest fraction of members with disks among all the regions analyzed in this section. The median N_H of the X-ray members is 0.52 × 10²² cm⁻², corresponding to an extinction of A_V = 2.91^m, slightly larger than that estimated by Guarcello et al. (2007) from the optical color–magnitude diagrams (A_V = 2.7^m); the median plasma temperature is kT = 1.95 keV, lower than the median value of the whole population. In the central cavity there is a significant population of candidate background sources (45 sources, 10% of the total number of X-ray sources in this field). It is likely that most of these are really background contaminants, since we do not expect a large embedded population in this field where the molecular cloud has been cleared by the energetic radiation from the OB stars lying in this region.

There are 33 members of the Pillars of Creation (8 with disks), with an age of about 2–3 Myr old in the bottom of the pillar, in the southeast, getting younger in the direction of the tip of the pillars. The median absorption of the X-ray sources falling in this region is typical of the cluster (N_H = 0.45 × 10²² cm⁻²), suggesting that almost all these sources belong to the cluster but they are not physically associated with the pillars. The median plasma temperature is kT = 2.03 keV, lower than the median of the members of the whole cluster. Three X-ray sources are associated with the tip of the largest pillar, called Pillar 1, where a protostar with a mass of 4–5 M_☉ and an intense infrared excess has been found (McCaughrean & Andersen 2002). One (at α = 18:18:50.88 and δ = −13:48:43.48) has been classified as a foreground source, and its N_H, derived from the quantile analysis, is compatible with this classification (N_H = 2 × 10¹⁹ cm⁻², corresponding to an A_V = 0.01^m). The second source is a disk-less member located at α = 18:18:49.78 and δ = −13:48:56.2. It has a spectrum fitted with a one-temperature thermal plasma model, with kT = 1.71 keV and N_H = 0.43 × 10²² cm⁻², which corresponds to the average extinction of the cluster, suggesting that this source is not embedded in the pillar. The only likely X-ray counterpart of the protostar in the tip of the pillar is at α = 18:18:50.31 and δ = −13:48:54.31, with 103 net counts, an absorption N_H = 3.89 × 10²² cm⁻², corresponding to an A_V = 21.7^m, and a plasma temperature of kT = 10.89 keV and log (L_X) = 31.69 erg s⁻¹, suggesting an embedded very active young star with an X-ray thermal spectrum. The X-ray counterpart of this protostar has already been identified by Linsky et al. (2007).

In the other pillar of M16, Column V, fall 32 members (6 with a disk) with an age of about 1 Myr. Their median absorption is N_H = 0.95 × 10²² cm⁻², twice that of the central cavity, with the most obscured sources at the bottom of the pillar. The median value for the plasma temperature is kT = 2.18 keV. Meaburn & Walsh (1986) discovered a bright Herbig–Haro object in the tip of Column V. We detected one X-ray source at less than 1'' away from it, at α = 18:19:04.65 and δ = −13:45:32.31. The quantile analysis revealed a plasma temperature of kT = 3.4 keV, an intermediate absorption (N_H = 0.3 × 10²² cm⁻²), and an X-ray luminosity L_X = 10^30.1 erg s⁻¹. Another candidate embedded protostar of Column V detected in X-rays at α = 18:19:07.08 and δ = −13:45:22.54 can be associated with water masers identified by Healy et al. (2004). This source is without a stellar counterpart in our catalog. It has been observed both in the c-field (32.7 net counts, log (L_X) = 30.6 erg s⁻¹, and a spectrum fitted by a power law) and the e-field (11.8 net counts).

Another region rich in cluster members is BRC SFO30, with 85 members (6 with a disk) detected in X-rays. In the optical and infrared diagrams, these sources are largely extinct (A_V > 5^m). The median absorption (N_H = 1.16 × 10²² cm⁻²) is similar to that of the sources falling in Column V, with a median plasma temperature of kT = 2.67 keV. The distribution of plasma temperature of these stars, however, has a very hot tail, with the 66% terzile at kT = 5.33 keV. Considering all the X-ray sources falling in this region, there is the largest fraction of candidate members without a disk and the lowest of candidate background sources. Since, from infrared studies (Indebetouw et al. 2007; Guarcello et al. 2009) and detection of water masers (Healy et al. 2004), a population of embedded protostars is expected to fall in this region, it is possible that the extinction here is high enough to efficiently absorb all the background sources or that these protostars are not yet X-ray active (the population of this region is on average younger than 1 Myr).

The most extinguished region is the NE molecular cloud, where a young embedded cluster lies (Indebetouw et al. 2007; Guarcello et al. 2009). This region has the highest fraction of X-ray sources without stellar counterparts (82 sources, 52% of all X-ray sources falling in this region). Among the 70 candidate YSOs in this region, 12 have a disk and are classified as embedded YSOs from their IRAC colors (among them are also the Class I sources detected in X-ray discussed in Section 4). It is not surprising that this is the region with the highest median absorption, with N_H = 1.92 × 10²² cm⁻², which is four times larger than the value in the central cavity and corresponds to an A_V = 10.7^m.

The remaining regions are the poorest in candidate cluster members. In the east region the 60 members detected in X-rays (12 with a disk) are older than 1 Myr, with a median absorption similar to the central cavity (N_H = 0.65 × 10²² cm⁻²) and a median plasma temperature of kT = 2.5 keV. The members detected in X-ray falling in the west region (25 sources, none with a disk) have similar ages, an absorption lower than the members in the central cavity (median value N_H = 0.33 × 10²² cm⁻²), and a median plasma temperature of kT = 2.45 keV. The southeast region contains the oldest X-ray-detected cluster members (12 sources, one with a disk, with a median age of about 3 Myr), with the lowest median plasma temperature (kT = 1.72 keV) and a median absorption of the candidate members detected in X-rays smaller than that of the central cavity (N_H = 0.36 × 10²² cm⁻²).