Candidate X-Ray-emitting OB Stars in MYStIX Massive Star-forming Regions

X-ray point source catalogs for the 18 MYStIX regions studied here were produced by Kuhn et al. (2010, 2013b) and Townsley et al. (2014) from archival Chandra Advanced CCD Imaging Camera (ACIS, Garmire et al. 2003) observations using the ACIS extract software package (Broos et al. 2010). The point-source resolution varies from $\lt 1^{\prime\prime}$ within a few arcminutes of the Chandra image axis to several arcseconds at larger off-axis angles toward the edges of the ACIS-I field. A summary of the various X-ray data sets, including the number of $17^{\prime} \times 17^{\prime}$ ACIS-I images and the integration times for each, is given by Feigelson et al. (2013).

MYStIX JHK_s NIR photometry was obtained from images taken with the United Kingdom Infrared Telescope (UKIRT) Wide-field Camera or from the Two-Micron All-Sky Survey (2MASS) point-source catalog (Skrutskie et al. 2006). Some of the data were obtained as part of the UKIRT Infrared Deep Sky Survey (UKIDSS; Lawrence et al. 2007; Lucas et al. 2008). The UKIRT data analysis, modified from the UKIDSS pipeline to treat crowding and nebulosity, is described by King et al. (2013). Compared to 2MASS, the UKIRT photometry provides higher spatial resolution and sensitivity but saturates on bright sources.

Spitzer MIR photometry at 3.6, 4.5, 5.8, and 8.0 μm was provided either by the Galactic Legacy Mid-Plane Survey Extraordinaire (GLIMPSE; Benjamin et al. 2003) or by Kuhn et al. (2013a). The $\sim 2^{\prime\prime}$ spatial resolution of the Spitzer Infrared Array Camera (IRAC; Fazio et al. 2004) is well matched to the 2MASS resolution, while the $\lt 1^{\prime\prime}$ UKIRT resolution is comparable to the on-axis ACIS spatial resolution.

Averaged across all MYStIX star-forming regions, 62% of X-ray sources were matched to predominantly faint NIR or MIR point sources with $\gt 0.80$ counterpart probability (Broos et al. 2013; Naylor et al. 2013). Of these X-ray/IR matches, the vast majority (95% of X-ray-to-NIR matches and 91% of X-ray-to-MIR matches) have apparent separations of $\lt 1^{\prime\prime}$. This gives projected physical separations of $\leqslant 2000\,\mathrm{au}$ for X-ray/IR matches in the majority of MYStIX regions that have heliocentric distances $\leqslant 2\,\mathrm{kpc}$ (Table 1). Chance alignments between physically unrelated X-ray sources and IR stars bright enough to be selected as OB candidates are thus expected to be exceedingly rare, and likely occur only in the cores of the densest MYStIX clusters or toward the edges of ACIS-I fields where the Chandra resolution is degraded. It is possible that in unequal-mass binary/multiple systems—for example, a B-type star with low-mass, T Tauri companions—the X-ray emission could be produced by a lower-mass component, while the IR source is dominated by the primary (e.g., Evans et al. 2011). Such cases may enable the identification of a new candidate OB star based on X-ray emission from a lower-mass companion. While the large majority of NIR counterpart photometry to MYStIX X-ray sources comes from UKIRT, the majority of published OB stars and OB candidates presented here have 2MASS photometry replacing saturated UKIRT sources.

Applying a Bayesian statistical framework to the combined X-ray and IR data sets, Broos et al. (2013) calculated the posterior probability that each MYStIX X-ray source was a young star associated with its parent star-forming region versus a foreground or background contaminant. The final product was the MYStIX Probable Complex Member (MPCM) catalog, which combined three overlapping samples: (1) X-ray sources with spatial distribution, variability, and/or IR counterpart properties consistent with pre-main-sequence (pre-MS) stars (Broos et al. 2011); (2) young stellar objects (YSOs) with MIR excess emission consistent with circumstellar disks and/or envelopes (regardless of an associated X-ray source; MIRES catalog from Povich et al. 2013); and (3) stars with published OB spectral types.⁵

Because the MPCM criteria were based on assumptions of X-ray source properties and NIR brightnesses for lower-mass, pre-MS stars (Broos et al. 2011, 2013), it is possible that some unpublished massive stars were not included as MPCMs or were misclassified as foreground stars because they are systematically brighter in the NIR and often have softer X-ray spectra. We therefore searched the SED Classification of IR Counterparts to MYStIX X-Ray Sources (SCIM-X) catalog (Appendix A of Povich et al. 2013) for candidate OB stars rather than the MPCM catalog. A few of the published OB stars are massive YSOs deeply embedded in their natal clouds and hence present themselves as very bright MIR excess sources because the OB stars themselves heat the surrounding dust. Only SCIM-X SEDs that did not show evidence of excess IR emission above the stellar photosphere at $\lambda \leqslant 4.5\,$μm (identified by SED_flg = −2 or −1) were included in the search. The resulting OB candidate lists hence will not include new candidate embedded massive stars but will include obscured OB stars observed through high foreground dust extinction.

The target massive star-forming regions for this study are listed in Table 1. These are 18 of the 20 MYStIX regions with Galactic coordinates, distance from the Sun, and earliest published spectral type obtained from Feigelson et al. (2013). The Orion Nebula Cluster is omitted because it has been well characterized, and the Carina Nebula OB populations are discussed by P11 and Gagné et al. (2011).

Our primary OB candidate selection methodology follows the procedures described in P11. Povich et al. (2013) fit a family of Castelli & Kurucz (2004) ATLAS9 stellar atmosphere models to the IR SEDs using the weighted least-squares (minimum ${\chi }^{2}$) fitting tool of Robitaille et al. (2007). The extinction law of Indebetouw et al. (2005), parameterized by the visual extinction A_V, was used to apply interstellar reddening to the model spectra as part of the SED fitting process. To be considered for OB candidacy, each star must have measurement in ${N}_{\mathrm{data}}\geqslant 4$ of the seven available photometric bands (J, H, K_s, [3.6], [4.5], [5.8], and [8.0]), including ${K}_{s}\leqslant 14.6$ mag. The fractional uncertainties on [5.8] and [8.0] flux densities were set to a minimum 10% and those on the other bands to a minimum 5% to avoid underestimating systematic errors in the fitting process (Povich et al. 2013).

The statistical criterion for acceptable model fits was ${\chi }_{i}^{2}-{\chi }_{0}^{2}\leqslant {N}_{\mathrm{data}}$, with ${\chi }_{0}^{2}\leqslant {N}_{\mathrm{data}}$ (Povich et al. 2013), where ${\chi }_{i}^{2}$ and ${\chi }_{0}^{2}$ are the goodness-of-fit parameters for the ith and best-fit models, respectively. This is a more stringent criterion than the ${\chi }_{i}^{2}-{\chi }_{0}^{2}\leqslant 2{N}_{\mathrm{data}}$ used by P11 and would have removed six stars from their list of 94 OB candidates. All of the models that fit the data with ${\chi }_{i}^{2}$ values within this limit were kept, so a given SED will have a set of many well-fit models. The range of A_V parameters accepted for our analysis was further restricted below a maximum value for each MYStIX region (Table 1), which we estimated from JHK_s color–color diagrams (CCDs). The bolometric luminosities are based on SED fits to the IR photometry and the assumed distance to each MYStIX star-forming complex (Table 1). The results of this fitting process are a set of self-consistent and well-fit models parameterized by the effective temperature (${T}_{\mathrm{eff}}$), the bolometric luminosity (${L}_{\mathrm{bol}}$), and their corresponding A_V values.

The set of well-fit ${T}_{\mathrm{eff}}$ and ${L}_{\mathrm{bol}}$ model parameters for each SED in the SCIM-X catalog was plotted on a theoretical H–R diagram, as illustrated in Figure 1. The intersection point of the locus of model parameters with the theoretical OB zero-age main sequence (ZAMS) (de Jager & Nieuwenhuijzen 1987; Martins et al. 2005) is marked with crosshairs. A star is selected as an OB candidate if the luminosity at the intersection point is greater than our "cutoff luminosity" of $1\times {10}^{4}\,{L}_{\odot }$, corresponding approximately to a main-sequence spectral type of B1 V or earlier (see also Figure 2 of P11).

Zoom In Zoom Out Reset image size

Two caveats must be borne in mind when interpreting these SED fitting results. The inferred ${L}_{\mathrm{bol}}$ depends on the assumption that the star lies at the (possibly incorrect) distance of the parent MYStIX complex listed in Table 1. A minor, systematic bias is introduced by the adoption of the ATLAS9 spectral models, which assume plane-parallel, stationary atmospheres in local thermodynamic equilibrium and do not include the effects of massive stellar winds. This can lead the model fits to return values that overestimate the true bolometric luminosities, and the effects should be most pronounced for evolved, OB supergiants with significant mass loss (see Section 3.2 of P11). The magnitude of this systematic bias reported by P11 is, however, small enough to be completely irrelevant to OB candidate selection and, as we will show, unimportant compared to other sources of uncertainty in measuring the extinction and luminosity of spectroscopically classified OB stars.

The locus of fits may intersect both the ZAMS and pre-MS evolutionary tracks (Figure 1). Models significantly to the left of the ZAMS can be ruled out on physical grounds. At higher luminosities, the model locus may also intersect the OB giant/supergiant tracks, and this is a plausible alternative. When the locus of models intersects the ZAMS above the cutoff luminosity, it only crosses very early stages of the pre-MS tracks. The exclusion of IR excess sources therefore reduces the likelihood of misidentification, because very young pre-MS stars are expected to be surrounded by dusty circumstellar disks or protostellar envelopes (P11). In most cases, the models with the lowest ${\chi }^{2}$ values (colored blue in the example diagrams) fell near the ZAMS, bolstering our confidence that they are actually hot, massive stars.

Inspection of the NIR color–magnitude diagrams (CMDs) for each MYStIX region, plotted in the lower-right panels of Figure 2, reveals that most regions contain X-ray-detected MPCMs that have bright J-band counterparts but are not previously classified OB stars (blue dots), IR excess sources (red symbols), or OB candidates selected by our SED fitting criteria (magenta boxes). SED fitting reveals that the majority of these NIR-bright stars are inconsistent with hot photospheres, ruling them out as OB candidates. The remaining NIR-bright, X-ray-detected stars constitute a less reliable, secondary sample of possible candidate OB stars that lacked MIR photometry of sufficient quality for SED fitting. There are two main reasons why a legitimate X-ray OB candidate with a good NIR counterpart could lack high-quality MIR photometry: (1) MIR nebulosity can be extremely bright in the MYStIX star-forming complexes, obscuring point sources at longer wavelengths, and (2) the Spitzer/IRAC point source resolution is poorer than that of Chandra or UKIRT, so in crowded regions the X-ray/NIR combination may resolve stars that are confused in the MIR (provided they do not saturate the UKIRT images).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (18 panels)

Tarball of figure set images

We used the NIR CCDs and CMDs (Figure 2) to identify additional OB candidates without SED fitting results in the SCIM-X catalog, based on the NIR colors and J magnitudes of the X-ray-detected stars. Our simple procedure applied the following steps to each of the 18 MYStIX star-forming regions listed in Table 1:

• 1.  
  Identify all MYStIX MPCMs in the region that were not listed as previously published OB stars by Broos et al. (2013) and had no SED fitting result reported in the SCIM-X catalog (SED_FLG = −99; Povich et al. 2013). All sources thus selected automatically have Chandra X-ray detections but cannot have been identified as OB candidates by our primary SED fitting methodology.
• 2.  
  Select the subset of sources from the previous step that fell into the locus of normally reddened main-sequence stars on the NIR CCD (between the blue and orange reddening vectors in Figure 2, lower-left panels). This step ensures that all secondary OB candidates are detected in all three NIR bands (JHK_s). To be conservative, sources that fell within this locus but had photometric uncertainties overlapping the reddened red-giant locus (orange) were excluded from consideration as OB candidates.
• 3.  
  Identify sources from the previous two steps that when dereddened to the OB main sequence (as defined by Pecaut & Mamajek 2013) on the J versus J − H CMD (Figure 2, lower-right panels) have absolute magnitudes ${M}_{J}\leqslant -2.4$ (equivalent to that of a B1 V star or brighter).

The 27 sources in 10 different MYStIX regions satisfying the above NIR criteria are designated as secondary OB candidates and marked by magenta diamonds in all panels of Figure 2. We emphasize that these secondary candidates are very likely to be less reliable than the primary OB candidates identified via SED fitting. As is evident in the CCDs, the locus of reddened main-sequence stars is strongly contaminated by IR excess sources, some of which are bright enough to fall within the region occupied by reddened OB stars on the CMDs. Hence, with only NIR photometry available, we cannot rule out the presence of circumstellar disks,⁶ which could make a lower-mass, pre-MS star appear brighter in the NIR (although we have attempted to minimize this effect by using J to estimate luminosities, as it is less likely to be contaminated by disk emission than H or K_s). Strong photometric variability is frequently observed among YSOs (Morales-Calderón et al. 2011), and a flaring, low-mass YSO could conceivably be classified as an OB candidate in the absence of MIR photometry, which could establish the presence of a disk or reveal a flare by providing a second epoch of photometry. The inferred luminosity of secondary OB candidates (here represented by M_J rather than ${L}_{\mathrm{bol}}$) is more dependent on the dereddening calculation than that of the SED-fitting candidates, because in highly obscured regions MIR photometry is less affected by dust extinction than NIR photometry.

The 328 previously published OB stars included in the MPCM catalog (Broos et al. 2013) within the 18 MYStIX regions analyzed for this study are listed in Table 2. The first three columns displayed in this table⁷ give the name, spectral type, and citation source of the spectral type for each star; in the vast majority of cases the spectral types come from optical or NIR spectroscopy (the only exceptions being a few extremely embedded stars surrounded by MIR-bright nebulosity, for which the spectral types were estimated from the radio continuum or IR luminosity of the associated UC H ii regions).⁸ While this list is not complete, we adopt it as the best available compilation of "known" OB stars across all of these well-studied regions against which to compare and evaluate our X-ray-selected OB candidates. We later note (in Tables 3 and 4) a handful of cases where we have discovered that one of our OB candidates had a published spectral type in the literature that was not included in this compilation.

Table 2.  Previously Cataloged OB Stars among MYStIX MPCMs

Star		References	${T}_{\mathrm{eff}}(\mathrm{ST})$ a	${A}_{V}^{\mathrm{SED}}$	$\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}$	Spitzer Origin and Name		J	H	KS	Originc	X-Ray/	SED
Name	ST	ST	(kK)	(mag)	(${L}_{\odot }$)	(GLIMPSE/K13)b		(mag)	(mag)	(mag)	JHKS	IRE?d	Cand?e
(1)	(2)	(3)	(4)	(5)	(6)	(7a)	(7b)	(8)	(9)	(10)	(15)	(16)	(17)
NGC 6357
Cl Pismis 24 15	O7.5Vz	12	35.0	4.8	4.93	A	G353.0997+00.9090	9.03	8.58	8.35	TTT	X	✓
Cl Pismis 24 10	O9V	1	33.0	5.1	4.72	A	G353.1250+00.8966	9.51	9.02	8.75	TTT	X	✓
Cl Pismis 24 12	B1V	1	26.0	4.8	4.04	C	G353.1691+00.9006	10.50	9.99	9.78	TTT
Cl Pismis 24 3	O8V	1	35.0	6.0	4.96	A	G353.1462+00.8849	9.29	8.77	8.47	TTT	X	✓
Cl Pismis 24 2	O5.5V((f))	1	40.0	5.4	5.37	A	G353.1566+00.8879	8.47	8.00	7.69	TTT	X	✓
Cl Pismis 24 18	B0.5V	1	28.0	6.4	4.52	A	G353.1709+00.8976	10.00	9.31	8.97	TTT	X
Cl Pismis 24 1 SW+NEab	O4III(f+)+O3.5Ifj	5	42.0	6.1	6.22	A	G353.1678+00.8946	6.73	6.18	5.89	TTT	X
Cl Pismis 24 19	B1V	1	26.0	...	...			...	...	...	...
Cl Pismis 24 16	O7.5V	1	36.0	...	...			...	...	...	...	X
Cl Pismis 24 17	O3.5III(fj)	1	44.5	12.5	5.92	A	G353.1689+00.8902	7.82	7.28	6.97	TTT	X
Cl Pismis 24 13	O6.5V((f))	1	38.0	6.0	5.25	A	G353.2038+00.9095	8.81	8.23	7.95	TTT	X	✓
HD 157504	WC6+O7/9	1	37.0	...	...	A	G353.2271+00.8293	7.04	6.53	5.86	TTT	X
[N78] 51	O6Vn((f))	12	41.0	5.1	5.35	A	G353.0724+00.6400	8.52	8.13	7.82	TTT	X	✓
[N78] 49	O5.5IV(f)	12	39.0	6.7	5.87g	C	G353.1107+00.6450	7.52	6.93	6.56	TTT	X
LS 4151	O6/7III	1	38.0	4.4	5.41	A	G353.3189+00.7676	7.93	7.55	7.34	TTT	X	✓
HD 319788	B2III:	1	21.0	1.7	4.22	C	G353.2966+00.5536	8.47	8.34	8.30	TTT

Notes. Spectral type (ST) references: (1) Skiff (2009), (2) Shuping et al. (2012), (3) Ellerbroek et al. (2012), (4) SIMBAD (Wenger et al. 2000), (5) Maíz Apellániz et al. (2007), (6) Schulz et al. (1981), (7) Hoffmeister et al. (2008), (8) Chini et al. (1980), (9) Bik et al. (2012), (10) Hirsch et al. (2012), (11) Sota et al. (2011), (12) Maíz Apellániz et al. (2016).

^aFrom the Martins et al. (2005) calibrations of ${T}_{\mathrm{eff}}$ versus spectral type for O stars, extended to early B stars by Crowther (2005). ^bIn column (8a), K = MYStIX MIR point source from K13; C/A = GLIMPSE pipeline point source from a highly reliable catalog or more complete archive (see Povich et al. 2013 for details). ^cPhotometry from (T)woMASS or (U)KIRT in each NIR band. ^dX = MYStIX X-ray source; Y = definite IR excess emission evident at $\lambda \leqslant 4.5\,\mu {\rm{m}}$; M = "marginal" IR excess emission (or systematics in photometry) affecting the [5.8] and/or [8.0] IRAC bands. SED fitting of "M" sources always excluded the [5.8] and [8.0] bands. SED fitting results for "Y" sources are reported only when the four JHK_S + [3.6] bands could be well fit by photospheric models, except where otherwise noted. ^eIf checked, the star was among the 67 selected independently by our procedure for identifying OB candidates among IR SEDs that were counterparts to MYStIX X-ray sources. The remaining 128 MYStIX OB stars were not selected because they lacked X-ray detections, had insufficient/poor MIR counterpart photometry for SED fitting, or were less luminous than the B1 V equivalent cutoff. ^fThe star is significantly underluminous when compared to theoretical calibrations of luminosity versus spectral type, $\mathrm{log}{L}_{\mathrm{bol}}({ST})-\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}\gtrsim 0.3$ dex. This could indicate the incorrect identification of the MIR source, confusion/saturation affecting IR photometry, an inaccurate published spectral type, or an underestimation of distance to the star. ^gThe star is significantly overluminous compared to theoretical calibrations of luminosity versus spectral type, $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}-\mathrm{log}{L}_{\mathrm{bol}}({ST})\gtrsim 0.5$ dex. This could indicate the incorrect identification of the MIR source, an unresolved multiple system, other confusion/saturation affecting IR photometry, an inaccurate published spectral type, or an overestimation of distance to the star. ^hNo luminosity class specified with ST, but SED fitting indicates a supergiant star. ⁱThe W3(OH) ionizing source SED was fit using the Robitaille et al. (2006) YSO models. ^jPhotometry data point excluded from SED fitting.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Table 3.  Candidate X-Ray-emitting OB Stars from SED Fitting

MOBc		A${}_{V}^{\mathrm{MS}}$ b	${T}_{\mathrm{eff}}^{\mathrm{MS}}$ b	$\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{MS}}$ b	Spitzer	J	H	KS	Origin	Other ID/ST (References);e
Name	X-Ray IDa	(mag)	(kK)	${L}_{\odot }$	Namec	(mag)	(mag)	(mag)	NIRd	IRE?
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(14)	(15)
Lagoon 1	180307.97-242355.1	5.8	28.7	4.4	G005.8929-01.0766	10.09	9.49	9.20	T
Lagoon 2	180322.59-242204.9	3.7	33.2	4.8	G005.9468-01.1098	8.72	8.31	8.14	T
Lagoon 3	180400.80-242334.6	7.3	28.1	4.3	G005.9963-01.2484	10.63	9.82	9.41 T
Lagoon 4	180423.55-242116.8	4.8	25.8	4.0	G006.0721-01.3048	10.29	9.74	9.54	T	OB (Bik15)
Lagoon 5	180438.91-242414.1	4.3	39.2	5.4	G006.0578-01.3799	7.99	7.46	7.29	T	NGC 6530 95/K0III (2)
RCW 38 1	085859.19-472658.0	11.0	26.8	4.2	G351.1645+00.6753	12.22	10.99	10.32	T	OBc (1)
RCW 38 2	085901.68-473037.1	7.9	27.2	4.2	G351.2109+00.6680	11.14	10.33	9.89	T	OBc (1)
NGC 6334 1	172003.34-355824.8	4.0	30.2	4.5	G351.1645+00.6753	9.44	9.04	8.81	T
NGC 6334 2	172012.99-355622.3	8.0	34.9	5.0	G351.2109+00.6680	9.92	9.13	8.68	T	M
NGC 6334 3	172014.17-355456.0	37.9	25.9	4.1	G351.2329+00.6785	21.46	16.54	13.74	U
NGC 6334 4	172015.03-355427.4	14.3	26.0	4.1	G351.2411+00.6807	13.34	11.73	10.87	T
NGC 6334 5	172021.06-354727.7	8.3	37.9	5.2	G351.3487+00.7304	9.57	8.72	8.27	T
NGC 6334 6	172024.87-354415.1	10.1	44.0	5.8	G351.3997+00.7506	9.02	7.94	7.36	T
NGC 6334 7	172026.46-355235.8	36.7	27.8	4.3	G351.2886+00.6668	19.58	15.98	13.44	U
NGC 6334 8	172027.12-355412.3	8.8	30.1	4.5	G351.2679+00.6498	10.87	9.98	9.48	T
NGC 6334 9	172032.79-355145.9	33.3	27.8	4.3	G351.3123+00.6572	18.66	15.38	12.92	U
NGC 6334 10	172040.61-354835.5	36.8	26.2	4.1	G351.3706+00.6657	21.11	16.21	13.53	U
NGC 6334 11	172043.46-354917.2	33.7	30.5	5.6	G351.3667+00.6511	...	12.35	10.17	T
NGC 6334 12	172044.66-354916.7	14.1	37.9	5.2	G351.3691+00.6479	11.34	9.79	8.93	T	M
NGC 6334 13	172048.00-354919.0	10.1	28.0	4.3	G351.3750+00.6382	11.69	10.63	10.05	T
NGC 6334 14	172103.69-354403.5	5.8	29.0	4.4	G351.4771+00.6448	10.18	9.60	9.30	T
NGC 6357 1	172405.60-340709.6	4.3	33.3	4.8	G353.1603+01.0475	9.03	8.63	8.46	T	IRAS 17207-3404/OBc (3, 4)
NGC 6357 2	172434.78-341318.0	5.9	26.1	4.1	G353.1323+00.9068	10.75	10.09	9.79	T	OBc (4)
NGC 6357 3	172440.49-341206.3	5.6	29.7	4.5	G353.1598+00.9017	10.05	9.45	9.14	T	Pis 24-8/OBc (4)
NGC 6357 4	172442.87-340911.6	8.8	31.9	4.7	G353.2046+00.9222	10.66	9.73	9.23	T	OBc (4)
NGC 6357 5	172443.40-340845.7	17.4	34.3	4.9	G353.2116+00.9248	12.86	10.95	9.90	T
NGC 6357 6	172447.56-341048.6	13.3	29.8	4.5	G353.1914+00.8937	12.34	10.87	10.17	T	OBc (4)
NGC 6357 7	172447.86-341517.0	6.0	34.8	5.0	G353.1304+00.8509	9.36	8.75	8.42	T	Pis 24-7/OBc (4)
NGC 6357 8	172448.46-341743.8	5.0	36.7	5.1	G353.0977+00.8263	8.76	8.30	8.03	T	OBc (4)
NGC 6357 9	172455.08-341111.5	7.5	33.0	4.8	G353.2007+00.8687	10.09	9.29	8.87	T	OBc (4)
NGC 6357 10	172457.94-342332.1	5.8	35.1	5.0	G353.0360+00.7450	9.22	8.66	8.38	T
NGC 6357 11	172511.53-342241.6	18.0	30.3	4.5	G353.0739+00.7141	...	11.79	...	U
NGC 6357 12	172528.51-342652.5	5.3	27.0	4.2	G353.0489+00.6268	10.40	9.88	9.64	T	OB (Bik15)
NGC 6357 13	172532.57-342400.0	5.0	36.7	5.1	G353.0964+00.6421	8.72	8.27	8.05	T	OBc (9)
NGC 6357 14	172547.33-342154.1	5.8	25.9	4.1	G353.1538+00.6196	10.74	10.15	9.87	T	OB (Bik15)
NGC 6357 15	172554.16-342156.4	6.2	25.9	4.1	G353.1664+00.5998	10.87	...	9.91	T	OB (Bik15); M
NGC 6357 16	172557.29-341850.8	7.8	32.0	4.7	G353.2152+00.6197	10.33	9.52	9.10	T	OB (Bik15)
NGC 6357 17	172558.08-341606.1	6.8	26.1	4.1	G353.2546+00.6431	11.00	10.26	9.90	T	OB (Bik15)
NGC 6357 18	172558.15-342119.2	8.1	32.1	4.7	G353.1826+00.5944	10.42	9.53	9.09	T	OB (Bik15)
NGC 6357 19	172559.40-341734.3	7.6	38.9	5.3	G353.2368+00.6255	9.18	8.43	8.00	T
NGC 6357 20	172602.07-341757.6	8.9	33.0	4.8	G353.2366+00.6143	10.52	9.52	9.06	T	OB (Bik15)
NGC 6357 21	172603.20-341629.8	6.0	44.0	5.8	G353.2590+00.6248	7.92	7.33	6.98	T
NGC 6357 22	172604.04-341520.0	8.1	33.9	4.9	G353.2768+00.6332	10.17	9.27	8.79	T	OB (Bik15)
NGC 6357 23	172604.25-341557.0	24.3	28.9	4.4	G353.2686+00.6269	15.75	13.16	11.52	U	OB (Bik15)
NGC 6357 24	172604.71-341526.3	13.7	32.7	4.8	G353.2765+00.6303	12.06	10.52	9.68	T	OB (Bik15)
NGC 6357 25	172604.72-341611.2	8.5	26.1	4.1	G353.2663+00.6234	11.47	10.64	10.14	T	OB (Bik15)
Eagle 1	181836.89-132822.9	7.5	37.0	5.2	G017.2267+00.9907	9.70	8.87	8.51	T
Eagle 2	181849.41-134501.9	17.3	32.0	4.7	G017.0059+00.8148	...	11.67	10.52	U
Eagle 3	181900.16-132520.1	5.7	35.1	5.0	G017.3161+00.9316	9.41	8.85	8.49	T
Eagle 4	181925.08-134511.6	35.5	26.1	4.1	G017.0719+00.6863	19.78	16.17	13.75	U
Eagle 5	181927.15-134943.4	9.2	33.7	4.9	G017.0093+00.6433	10.71	9.72	9.23	T
Eagle 6	181928.90-134658.3	18.9	42.1	5.6	G017.0531+00.6588	12.46	10.29	9.13	T
Eagle 7	182005.47-134323.1	16.9	26.9	4.2	G017.1757+00.5564	14.19	...	...	T
M17 1	182008.27-160552.1	10.8	29.8	4.5	G015.0871-00.5738	11.95	10.71	10.11	T
M17 2	182015.63-160613.0	5.6	33.1	4.8	G015.0960-00.6026	9.82	9.18	8.98	T	Late-type (WIRO)
M17 3	182024.10-160840.4	14.8	26.1	4.1	G015.0759-00.6518	13.68	12.26	11.32	U
M17 4	182027.40-161255.0	40.0	30.8	4.6	G015.0198-00.6968	...	16.39	13.62	U
M17 5	182029.12-161247.6	21.6	26.8	4.2	G015.0249-00.7018	15.72	13.45	12.10	U
M17 6	182034.39-160938.7	19.3	43.9	5.8	G015.0809-00.6957	...	10.33	9.09	T	Cl* NGC 6618 B 113
M17 7	182047.67-160136.3	3.0	36.7	5.1	G015.2244-00.6796	8.43	8.23	8.18	T	BD-16 4822/OBc (5), B2 (WIRO)
M17 8	182048.01-161721.0	17.1	26.9	4.2	G014.9935-00.8044	14.27	12.42	11.43	U
M17 9	182051.05-155658.1	3.6	36.2	5.1	G015.2989-00.6550	8.70	8.36	8.23	T	BD-16 4823/G0 (6), Late-type (WIRO)
M17 10	182053.85-160306.5	3.4	35.9	5.1	G015.2139-00.7131	8.70	8.41	8.28	T	Late-type (WIRO)
M17 11	182101.16-160547.4	2.8	31.1	4.6	G015.1882-00.7598	9.24	9.02	8.93	T	LS 4943/OBc (7), O9 (WIRO)
M17 12	182102.66-155858.3	5.9	34.9	5.0	G015.2913-00.7118	9.61	9.03	8.72	T	B3 (WIRO)
M17 13	182112.27-155421.3	5.1	26.9	4.2	G015.3773-00.7095	...	10.20	...	T
M17 14	182115.23-161334.7	4.8	31.7	4.7	G015.1001-00.8707	9.96	9.38	9.19	T	Late-type (WIRO)
M17 15	182115.62-160737.0	13.5	43.1	5.7	G015.1886-00.8254	10.77	9.28	8.45	T
M17 16	182119.78-160619.2	20.9	26.1	4.1	G015.2155-00.8300	15.54	13.37	12.07	U
M17 17	182125.96-160117.8	2.0	30.9	4.6	G015.3009-00.8122	9.00	8.95	8.85	T	BD-16 4832/OBc (7), B2–B4 (WIRO)
M17 18	182127.79-155732.6	2.9	38.2	5.3	G015.3597-00.7895	8.17	...	...	T	BD-16 4834/OBc (7), B3 I (WIRO)
W3 1	022518.56+620116.9	9.3	40.2	5.4	G133.7017+01.1284	9.88	8.94	8.40	T	O5–O7V + O7–O9V (8)
W3 2	022527.44+620506.0	16.9	26.8	4.2	G133.6953+01.1940	14.37	12.61	11.59	T
W3 3	022616.94+620735.1	10.4	31.0	4.6	G133.7707+01.2671	11.57	10.57	10.02	T	B: (8)
W3 4	022638.18+620940.0	3.1	25.9	4.1	G133.7969+01.3144	10.32	9.99	9.90	T	IC 1795 71/G5: (6)
W3 5	022655.52+621050.7	7.1	27.9	4.3	G133.8212+01.3451	11.16	10.49	10.09	T	B1V (8)
W4 1	023113.84+613046.3	2.4	32.1	4.6	G134.5397+00.9109	14.89	14.20	13.88	T	M
Trifid 1	180241.19-224905.1	5.5	34.4	4.9	G007.2180-00.2089	10.23	9.72	9.39	T	O8.5–Early B (Kuhn16)
Trifid 2	180247.58-224820.6	6.5	44.0	5.8	G007.2408-00.2242	8.87	8.31	...	T	O8.5–O9.5 (Kuhn16)
Trifid 3	180250.44-224850.7	5.1	44.1	5.8	G007.2390-00.2377	8.39	7.96	7.72	T	O6III–O8II (Kuhn16)
Trifid 4	180303.82-230105.9	3.8	25.9	4.1	G007.0865-00.3832	11.19	10.69	10.56	T
NGC 3576 1	111101.21-612035.1	7.7	29.2	4.4	G291.1914-00.7816	...	11.00	...	T
NGC 3576 2	111124.01-611722.8	5.5	29.8	4.5	G291.2132-00.7149	11.10	10.41	10.22	T
NGC 3576 3	111131.66-612147.1	6.8	36.1	5.1	G291.2551-00.7771	10.42	9.76	9.41	T	M
NGC 3576 4	111137.40-611914.2	15.9	28.0	4.3	G291.2498-00.7334	14.56	12.77	11.82	T
NGC 3576 5	111155.36-611756.1	8.8	25.8	4.0	G291.2748-00.7000	12.79	11.82	11.38	T
NGC 3576 6	111204.89-612657.3	4.0	25.7	4.0	G291.3486-00.8322	11.31	10.96	10.81	T
NGC 3576 7	111242.75-610422.3	3.4	33.3	4.8	G291.2789-00.4546	9.85	9.50	9.43	T	TYC 8959-1106-1
NGC 3576 8	111249.13-611811.8	4.1	28.2	4.3	G291.3763-00.6642	10.88	10.35	10.20	T
NGC 1893 1	052214.81+332907.5	4.0	32.9	4.8	J52214.82332907.4	10.66	10.24	10.09	T	Cl* NGC 1893 CF 52
NGC 1893 2	052221.53+332851.3	6.5	27.1	4.2	J52221.52332851.2	12.32	11.53	11.24	T	YSOc (10)
NGC 1893 3	052225.03+333106.8	3.4	26.2	4.1	J52225.05333106.4	11.60	11.19	11.09	T
NGC 1893 4	052238.90+332555.5	2.7	25.8	4.0	J52238.87332555.3	11.41	11.23	11.14	T	Cl* NGC 1893 CF 42/A4V (6)
NGC 1893 5	052247.79+332206.6	3.3	29.8	4.5	J52247.76332206.8	10.92	10.58	10.49	T	Cl* NGC 1893 CF 8/A4 (11)
NGC 1893 6	052257.35+332822.4	3.9	26.2	4.1	J52257.33332822.3	11.76	11.26	11.16	T	Cl* NGC 1893 CF 16
NGC 1893 7	052302.94+333215.1	5.3	26.1	4.1	J52302.91333214.6	12.20	11.57	11.33	T	YSOc (10)
NGC 1893 8	052307.15+333251.6	3.3	27.0	4.2	J52307.13333251.4	11.44	11.06	10.96	T	YSOc (10)
NGC 1893 9	052321.21+332944.6	2.7	36.2	5.1	J52321.19332944.4	9.69	9.53	9.43	T	Cl* NGC 1893 CF 32

Notes.

^aChandra X-ray source designation from Broos et al. (2013). ^bThe values in Columns 3–5 were derived from SED fitting under the assumption of single massive stars on the theoretical zero-age main sequence (see text). Formally, the ${T}_{\mathrm{eff}}^{\mathrm{MS}}$, ${A}_{V}^{\mathrm{MS}}$, and $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{MS}}$ reported here are upper limits and should be used only for comparison purposes among this sample (and with the Carina OBc sample from P11). ^cMIR source designation from Kuhn et al. (2013a) or the GLIMPSE point source archive. ^dNIR source from (T)woMASS or (U)KIRT. ^eNotes for Column 15: YSOc = YSO/T Tauri candidate; OBc = photometric massive star candidate; OB (Bik15) = spectroscopically confirmed OB star (A. Bik 2015, private communication); (WIRO) = new spectral type reported from spectroscopic follow-up with WIRO; and (Kuhn16) = from spectroscopic follow-up by Kuhn et al. (2016, in preparation). References in Column 15: (1) Wolk et al. (2006), (2) Rauw et al. (2002), (3) Gvaramadze et al. (2011), (4) Wang et al. (2007), (5) Povich et al. (2009), (6) SIMBAD (Wenger et al. 2000), (7) Reed (2003), (8) Kiminki et al. (2015), (9) Neckel (1984), (10) Prisinzano et al. (2011), (11) Renson & Manfroid (2009). As in Table 2, "M" denotes marginal IR excess sources for which available [5.8] and [8.0] photometry was excluded from SED fitting.

A machine-readable version of the table is available.

Download table as:  DataTypeset images: 1 2 3

The provenance (GLIMPSE Catalog/Archive or Kuhn et al. 2013a) and name (Galactic coordinates) for each Spitzer/IRAC source associated with an OB star are provided in Column 7 of Table 2. The NIR and MIR point source magnitudes used for our SED fitting analysis are provided in Columns 8–14, and the [T]woMASS/[U]KIRT provenance of each magnitude is given in Column 15. In the typeset table the MIR photometry is hidden for compactness, but the JHK_s magnitudes are displayed to facilitate comparisons with the OB candidates in Tables 3 and 4 as well as the CCDs and CMDs in Figure 2. In Column 16 we indicate the presence of a MYStIX X-ray detection (X) and/or marginal/significant IR excess emission (M/Y).

Table 4.  Secondary Candidate X-Ray-emitting OB Stars from NIR Photometry

MOBc Name	X-Ray IDa	A${}_{V}^{\mathrm{MS}}$	${M}_{J}^{\mathrm{MS}}$	J	H	KS	${\sigma }_{J}$	${\sigma }_{H}$	${\sigma }_{{K}_{S}}$	Originb	Other ID/ST (Ref)c
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
Rosette 1	063302.88+043617.0	6.4	−4.14	8.29	7.77	7.68	0.02	0.05	0.02	T	HD 259513/K2?
DR 21 1	203828.79+421038.5	37.3	−4.03	17.36	13.55	11.37	0.01	0.01	0.01	U
DR 21 2	203900.04+421936.5	33.4	−2.53	17.77	14.32	12.38	0.03	0.01	0.01	U	[DS96] I1 (1)
RCW 38 3	085859.79-473222.3	...d	...d	14.44	11.18	9.41	0.03	0.03	0.03	T	YSOc,OBc (2)
RCW 38 4	085902.16-473016.7	14.3	−2.90	12.30	10.88	10.24	0.06	0.04	0.05	T	OBc (3)
RCW 38 5	085904.44-473056.2	19.6	−2.92	13.78	11.79	10.73	0.04	0.02	0.06	T	OBc (3)
NGC 6334 15	172001.37-355639.8	...d	...d	10.94	9.15	8.13	0.02	0.03	0.02	T
NGC 6334 16	172011.86-355157.3	...d	...d	12.74	9.79	8.14	0.02	0.02	0.04	T
NGC 6334 18	172018.85-355424.2	21.5	−3.28	13.93	11.80	10.56	0.05	0.04	0.03	T
NGC 6357 26	172601.88-341631.1	8.0	−3.34	10.07	9.41	9.04	0.03	0.04	0.03	T	OB (Bik15)
Eagle 8	181911.10-132644.0	...d	...d	9.96	7.46	6.20	0.02	0.06	0.02	T
M17 19	182019.25-161326.7	28.2	−2.58	16.87	14.00	12.34	0.01	0.01	0.01	U
M17 20	182021.74-161112.6	24.0	−2.49	15.77	13.34	12.13	0.02	0.01	0.00	U	Cle NGC 6618 B 330
M17 21	182024.32-161352.9	18.8	−3.84	12.98	11.14	10.20	0.02	0.04	0.04	T
M17 22	182024.55-161127.7	22.4	−2.93	14.88	12.62	11.34	0.03	0.01	0.02	U	Cle NGC 6618 B 285
M17 23	182030.57-161102.0	37.6	−4.45	17.65	13.79	11.60	0.02	0.01	0.00	U	[CW98] IRS 22
M17 24	182031.22-160220.8	37.4	−3.35	18.70	14.90	12.75	0.06	0.01	0.01	U
M17 25	182032.15-160157.0	22.8	−3.02	14.90	12.61	11.28	0.01	0.01	0.00	U
W3 6	022529.88+620522.3	27.0	−3.28	15.88	13.21	11.84	0.06	0.03	0.02	T	YSOc (4)
W3 7	022557.38+620547.6	16.0	−3.54	12.53	10.96	10.17	0.04	0.03	0.03	T
W3 8	022712.97+615139.1	6.0	−2.43	10.82	10.32	10.03	0.02	0.02	0.02	T	B1–B2V + B3–B4V (5)
Trifid 5	180229.99-225601.9	29.2	−3.27	17.11	14.17	12.60	0.03	0.01	0.01	U
Trifid 6	180230.38-230134.4	9.2	−2.43	12.32	11.45	11.06	0.03	0.03	0.03	T
Trifid 7	180238.14-230228.6	23.5	−3.26	15.51	13.16	11.99	0.01	0.01	0.01	U
Trifid 8	180244.15-230245.0	13.2	−2.93	12.96	11.66	10.86	0.02	0.02	0.03	T
NGC 3576 9	111157.04-611813.2	13.8	−2.89	13.25	11.90	11.22	0.02	0.04	0.05	T
NGC 3576 10	111213.72-611740.0	14.3	−2.42	13.85	12.44	11.76	0.03	0.02	0.02	T

Notes. Columns 3–10 list data values in magnitude units. The magnitudes in Columns 3 and 4 were derived by dereddening the observed J − H color and J magnitude to the main sequence, assuming the distance to each region listed in Table 1 (see text).

^aChandra X-ray source designation from Broos et al. (2013). ^bNIR source from (T)woMASS or (U)KIRT. ^cNotes for Column 12: YSOc = YSO/T Tauri candidate; OBc = photometric massive star candidate. References in Column 12: (Bik15) A. Bik (2015, private communication), (1) Davis & Smith (1996), (2) Winston et al. (2011), (3) Wolk et al. (2006), (4) Rivera-Ingraham et al. (2011), (5) Kiminki et al. (2015). ^dDereddening to the main sequence would produce M_J more luminous than an O3 V star. This could occur for evolved, luminous stars or unresolved binary/multiple systems; hence no intrinsic source colors can be assumed for the NIR source, and no ${M}_{J}^{\mathrm{MS}}$ or ${A}_{V}^{\mathrm{MS}}$ is reported. ^eThe NIR spectrum shows CO absorption bands (Bik15).

A machine-readable version of the table is available.

Download table as:  DataTypeset image

We have performed two analyses of the SED fitting results on all previously published OB stars in Table 2 with sufficient NIR and MIR photometric detections. First we applied the OB candidate search criteria described in Section 2.2 to identify those known OB stars that would have been selected as OB candidates by SED fitting if they had not already been spectroscopically classified; these 67 O-type and early B-type stars are indicated by check marks in Column 17.

The second SED modeling analysis fixed the ${T}_{\mathrm{eff}}$ parameter to a value ${T}_{\mathrm{eff}}(\mathrm{ST})$ based on the published spectral type, using the calibrations of Martins et al. (2005) for O stars and the Crowther (2005) extension to early B stars (column 4 of Table 2). For the O-type stars in Table 2 we used the different Martins et al. (2005) temperature scales for supergiants (class I or II), giants (class III or IV), and dwarfs (class V, or the default if no published luminosity class was available); for B-type stars of any luminosity class we used a single temperature scale. For unresolved binary/multiple systems, ${T}_{\mathrm{eff}}(\mathrm{ST})$ was chosen based on the spectral type of the most luminous component. With temperature fixed, the extinction ${A}_{V}^{\mathrm{SED}}$ and bolometric luminosity ${L}_{\mathrm{bol}}^{\mathrm{SED}}$ become tightly constrained by the SED modeling; these parameter values are presented in columns 5 and 6, respectively, of Table 2. As P11 showed, the random uncertainties on the model parameters are typically $\gtrsim 0.3$ mag for ${A}_{V}^{\mathrm{SED}}$ and $\lt 5 \%$ for ${L}_{\mathrm{bol}}^{\mathrm{SED}}$ (or $\lt 0.02$ dex in $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}$), although the uncertainties will be greater (particularly on ${A}_{V}^{\mathrm{SED}}$) for stars missing NIR photometry. Cells with no data values in columns 5 and 6 indicate either that there were too few SED points for SED fitting or that no acceptable model fits to the SED were achieved, while values of −99 indicate that there were well-fit models but none consistent with the ${T}_{\mathrm{eff}}(\mathrm{ST})$ for that star in column 4.

Fifteen of the 18 MYStIX regions contained one or more OB stars with SEDs that we could model, and we plot $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}$ versus $\mathrm{log}{T}_{\mathrm{eff}}(\mathrm{ST})$ for all of these stars on a single H–R diagram in Figure 3. This figure can be compared with similar H–R diagrams published by Gagné et al. (2011, their Figure 2) for Carina and Kiminki et al. (2015, their Figure 11) for W3. In spite of differences in methodology for measuring ${L}_{\mathrm{bol}}$ and A_V, the dispersion of the stars with respect to the ZAMS shows both of these single-region H–R diagrams appear qualitatively similar to our H–R diagram of OB stars across different MYStIX regions.

Zoom In Zoom Out Reset image size

Among spectroscopically classified O-type dwarfs, there is generally good correlation with the theoretical ZAMS (Martins et al. 2005), with the majority of stars falling within 0.2 dex in $\mathrm{log}{L}_{\mathrm{bol}}$ of the ZAMS for their spectral type. This supports the conclusion of P11 that even fitting only the 1–8 μm IR SEDs can effectively constrain the luminosities of OB stars in the absence of UBV photometry.⁹

We mark known spectroscopic binaries (or higher-multiple systems) with boxed symbols in Figure 3. Since multiplicity is common among massive stars (Sana et al. 2013) but the fraction of MYStIX OB stars classified as binaries is low, it is probable that a significant fraction of the stars listed in Table 2 have undiscovered multiplicity. An unresolved, equal-mass binary system misclassified as a single star will be overluminous by 0.3 dex, which places it significantly above our observed symmetrical scatter of 0.2 dex around the O-type ZAMS on the H–R diagram. This extra luminosity is sufficient to move a star from the O-type ZAMS to the O giant sequence (Martins et al. 2005). We hence identify six candidate equal-mass binary systems, or alternatively single O-type giants, that may have been misclassified as single O dwarfs (plus signs without boxes located along the O giant sequence in Figure 3). These possible multiple systems are mentioned in the discussion of their respective parent regions in the Appendix.

The O-type giants and supergiants lie significantly above the ZAMS, as expected, with the exception of two O giants in NGC 6357 (see Appendix A.7). We identify five candidate O (super)giants with no published luminosity class (open circles located on or above the O giant sequence in Figure 3), four in M17 (Appendix A.9) and the most luminous known O star in NGC 6334 (Appendix A.6). Our SED fitting measured $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}/{L}_{\odot }\gt 6$ for three O supergiant systems that constitute the dominant ionizing stars in NGC 2362, W4, and the Pismis 24 cluster of NGC 6357. We discuss each of these systems in the appendix, and we reiterate here the caveat that our ${L}_{\mathrm{bol}}^{\mathrm{SED}}$ and ${A}_{V}^{\mathrm{SED}}$ values must be regarded as upper limits, especially for stars with very significant mass loss (P11).

Some individual OB stars are strong outliers in luminosity with respect to their published spectral types, and these are noted in Table 2. Two stars in the Eagle Nebula and four in M17 are among the most extreme examples of SED fitting results that are underluminous compared to their published spectral types, as discussed in the appendix (A.8 and A.9).

Compared to the O stars, the MYStIX B-type stars show far more scatter about the theoretical ZAMS, by nearly one order of magnitude in ${L}_{\mathrm{bol}}$. Similar large scatter on the H–R diagram is evident for early B stars both in Carina (Gagné et al. 2011) and in W3; in the latter case Kiminki et al. (2015) point out that this scatter is dominated by uncertainties in spectral types. While generally we have no information about uncertainties on the spectral types from the literature, we suspect that the difficulty in assigning accurate spectral types to (the often obscured) early B stars dominates the scatter in our sample as well. We therefore cannot accurately identify new candidate binaries of evolved (super)giants among the MYStIX B stars, and we caution the reader that published early B spectral types in obscured star-forming regions should be viewed with some skepticism.

Figure 3 reveals no MYStIX regions that show significant, systematic shifts to higher or lower luminosities for their ensemble of OB stars compared to the expectations from their spectral types. This suggests that the assumed distance to each region (Table 1) is accurate to within $\sim 25 \%$ to 50%, the relatively loose constraints provided by the observed luminosity dispersion about the OB ZAMS.

The 98 candidate MYStIX OB stars satisfying our SED-based selection criteria (Section 2.2) are listed in Table 3, while Table 4 lists the 27 candidate OB stars identified via our secondary NIR selection criteria (Section 2.3). The first two columns in each table give the MYStIX X-ray source designation, which encodes the source position in equatorial (J2000; sexagesimal) coordinates. Within each MYStIX region, candidates are sorted and numbered first by method of identification and then by increasing right ascension; hence, for example, in NGC 6357 the first of the 25 primary X-ray OB candidates from SED fitting is MYStIX OB candidate (MOBc) NGC 6357 1, while the first of the secondary X-ray OB candidates from NIR photometry is MOBc NGC 6357 26.

Columns 3–5 in Table 3 give the results of the SED fitting: the interstellar extinction (magnitudes ${A}_{V}^{\mathrm{MS}}$), stellar effective temperature (${T}_{\mathrm{eff}}^{\mathrm{MS}}$ in kilo-kelvin), and bolometric luminosity ($\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{MS}}/{L}_{\odot }$), assuming the stars lie on the ZAMS (Figure 1). Columns 3 and 4 in Table 4 give the analogous information for our secondary X-ray NIR candidates: ${A}_{V}^{\mathrm{MS}}$ and ${M}_{J}^{\mathrm{MS}}$ from dereddening observed J − H colors and J magnitudes to the ZAMS (Figure 2, CMDs in bottom-right panels). Formally, all of these "MS" extinction, luminosity, and temperature values are upper limits only, pending follow-up spectral typing.

We also provide the IR point source information and photometry used in our analysis and plotted in the CCDs and CMDs of Figure 2. For the SED-fitting candidates (Table 3) the Spitzer MIR source name, the JHK_s and IRAC point source photometry values (the IRAC photometry values are included in the full online version of the table), and the provenance of the NIR photometry (T = 2MASS or U = UKIRT) are listed in columns 9–14.¹⁰ For the NIR photometry candidates we simply give the JHK_S magnitudes and provenance (Table 4).

The final column in each table lists, to the best of our knowledge, the most relevant independent identification, spectroscopic classification, or photometric selection as an OB candidate for each star as well as the citation for the source of this information. We also indicate (with an "M" in column 15 of Table 3) the five OB candidates showing evidence for marginal MIR excess emission in the SCIM-X catalog (Povich et al. 2013).

Four of our candidates in W3 (MOBc W3 1, 3, 5, and 8) were independently identified and confirmed—one as an O-type binary, two as B stars, and one as a B-type binary—by Kiminki et al. (2015). Targeted, NIR spectroscopic follow-up of our candidates using the K-band Multi-object Spectrograph (KMOS) on the Very Large Telescope (A. Bik 2015, private communication) confirmed 1 new OB star in Lagoon (MOBc Lagoon 4) and 12 new OB stars in NGC 6357 (MOBc NGC 6357 12, 14, 15, 16, 17, 18, 20, 22, 23, 24, 25, and 30). Spectroscopic follow-up using the ARCoIRIS instrument of the 4-m Blanco Telescope (M. A. Kuhn et al. 2016, in preparation) has confirmed three new OB stars near the Trifid Nebula (MOBc Trifid 1, 2, and 3). Visible-light spectroscopic follow-up of SED-selected OB candidates in M17 in the 2.3-m Wyoming Infrared Observatory (WIRO) identified five new early-type stars (MOBc M17 7, 11, 12, 17, and 18) but classified four others (MOBc M17 2, 9, 10, and 14) as late-type stars. Six other OB candidates in various regions (MOBc Rosette 1, Lagoon 5, W3 4, and NGC 1893 4 and 5) had previous spectroscopic classifications indicating that they are cool stars. We have chosen to list spectroscopically classified cool stars that are not actual OB candidates, in Tables 3 and 4, as examples of "false positives" selected by our SED fitting criteria, and some of them could still prove to be evolved, massive yellow or red supergiants. Overall, 20 of the 125 candidates have already been spectroscopically confirmed as OB stars, while nine are late-type interlopers.

In Figure 4 we plot the X-ray median energy (${E}_{\mathrm{med}}$, measured across the total 0.5–8 keV Chandra/ACIS-I passband; Kuhn et al. 2013b; Townsley et al. 2014) versus the interstellar extinction (${A}_{V}^{\mathrm{SED}}$ from Table 2 or ${A}_{V}^{\mathrm{MS}}$ from Tables 3 and 4) as a simple basis for comparing the known and candidate OB samples. We further mark the several candidate OB stars that have already been confirmed (blue plus signs) or rejected (red × symbols). ${E}_{\mathrm{med}}$ is affected by both the extinction (since the soft X-rays are preferentially obscured, hardening the radiation) and the intrinsic X-ray spectrum, which can vary greatly across individual massive stars, due to differences in wind shock properties (e.g., Gagné et al. 2011). Notwithstanding the expected variation in intrinsic X-ray hardness, the median energy is clearly correlated with extinction, with both known and candidate OB stars following the same general trend. The relatively large scatter about the trend is commonly seen in plots of X-ray versus IR measures of stellar extinction in star-forming regions (e.g., Vuong et al. 2003; Feigelson et al. 2005; P11).

Zoom In Zoom Out Reset image size

Published O-type through early B-type stars and spectroscopically confirmed OB candidates show characteristically lower ${E}_{\mathrm{med}}$ over the extinction range of $0\lt {A}_{V}^{\mathrm{SED}}\lesssim 6$ mag than the calibration curves of ${E}_{\mathrm{med}}$ versus the X-ray-absorbing column density N_H (converted to A_V assuming ${N}_{H}/{A}_{V}\,=2\times {10}^{21}\,{\mathrm{cm}}^{-2}\,{\mathrm{mag}}^{-1}$; Vuong et al. 2003) for low-mass, pre-MS stars (Getman et al. 2010). For relatively lightly absorbed sources, ${E}_{\mathrm{med}}$ hence reflects the canonical, softer intrinsic X-ray spectra for massive stars compared to those for the young, low-mass stars that dominate the MYStIX source populations. Notwithstanding the scatter, stars classified as B1.5 or later appear systematically harder in ${E}_{\mathrm{med}}$ than earlier OB stars and OB candidates in the same extinction range (recall that our selection criteria only identify the equivalent of B1 V or earlier/more luminous stars), indicating intrinsically harder X-ray spectra than those for earlier-type massive stars. Many of the X-ray-detected MYStIX B stars could therefore be intrinsically fainter, weak-wind sources with unresolved, low-mass pre-MS companions producing the observed X-ray emission (see, e.g., Evans et al. 2011). A detailed analysis of the X-ray spectral properties of the known and candidate MYStIX OB stars is beyond the scope of this study, and our past experience has shown that X-ray spectral modeling does not provide sufficient evidence to confirm or reject candidate OB stars (see Section 4.3).

OB candidates with higher extinction but softer X-ray median energy than the trend for published and confirmed OB stars may be late-type contaminants or incorrect matches between the X-ray source and the IR counterpart. All eight of our primary OB candidates from SED fitting that have been spectroscopically classified as late-type stars (Table 3) cluster near ${E}_{\mathrm{med}}\approx 1\,\mathrm{keV}$ for $2\gtrsim {A}_{V}^{\mathrm{MS}}\gtrsim 6$ mag. Since several previously published MYStIX OB stars also lie in or near this region (Figure 4), we cannot recommend any simple cut on extinction and median energy to flag probable contaminants.

Several candidates appear to be extreme outliers. Three NIR-selected candidates have ${A}_{V}^{\mathrm{MS}}\gt 20$ mag and ${E}_{\mathrm{med}}\lt 2.6\,\mathrm{keV}$ (DR 21 1 and Trifid 5 and 6). All are located well outside of dense or embedded clusters, and they could be chance alignments between an X-ray source and an unassociated, IR field star. For these three sources the MPCM classification by Broos et al. (2013) may be incorrect. MOBc NGC 6334 11 is the sole SED-fitting candidate with similar soft median energy but high extinction (Figure 4), but it is centrally located among the embedded, massive clusters and H ii regions, which increases the likelihood that it is actually a member of the NGC 6334 complex. It is possible that our SED fitting overestimated the reddening and hence the luminosity for this star (see Figure 2(j) and Appendix A.6) or that there was a mismatch between the X-ray and MIR source in this dense cluster.

Six OB candidates identified via SED fitting (MOBc Eagle 4, M17 4, and NGC 6334 3, 7, 9, and 10) and three identified via NIR photometry (MOBc DR 21 2 and M17 23 and 24) have very high ${A}_{V}^{\mathrm{MS}}\gt 30$ mag and commensurately hard ${E}_{\mathrm{med}}\gt 3\,\mathrm{keV}$ (Figure 4). All nine of these stars have projected locations within or behind dense molecular clouds, so their extreme reddening is reasonable, but given the uncertainties in the extinction laws used for the selection, there is an increased chance that an obscured OB candidate could instead be an embedded, lower-mass protostar. Deep, NIR follow-up spectroscopy should reveal the true nature of these candidates.

Another outlying group of seven SED-fitting candidates (MOBc NGC 6357 11; Eagle 2, 6, and 7; M17 6; W3 2; and NGC 3576 4) has moderately high extinction ($15\,\mathrm{mag}\,\lt {A}_{V}^{\mathrm{MS}}\lt 20$ mag) but disproportionately hard median energy (${E}_{\mathrm{med}}\gt 3$ keV); these could be OB stars with intrinsically hard spectra or perhaps background contaminants (for example, X-ray-detected active galactic nuclei that were erroneously matched to IR-bright Galactic field stars). A similar outlier with intriguing, hard X-ray emission is the confirmed OB star NGC 6357 12, with ${E}_{\mathrm{med}}=2.97\,\mathrm{keV}$ and ${A}_{V}^{\mathrm{MS}}=5.3$ mag. The most extreme X-ray star plotted in Figure 4 is W3(OH) in the far upper-right corner, which, unlike the other stars in this plot, we fit using embedded YSO models (Table 2). Clearly, ${E}_{\mathrm{med}}\gt 3\,\mathrm{keV}$ is possible for certain massive stars, particularly when they are highly absorbed.

Recalling that our OB candidate selection criteria were designed to identify stars more luminous than the equivalent of a B1 V star, we can estimate the potential multiplicative factor ${F}_{\mathrm{inc}}$ by which the inferred massive stellar population of each region would increase due to the addition of newly confirmed OB candidates using

$$\begin{eqnarray*}&&{F}_{\mathrm{inc}}=\displaystyle \frac{{f}_{\mathrm{conf}}[\mathrm{MOB}1{\rm{c}}(\mathrm{SED})+\mathrm{MOB}1{\rm{c}}(\mathrm{NIR})]}{\mathrm{OB}1(\mathrm{pub})}+1,\end{eqnarray*}$$

where ${f}_{\mathrm{conf}}$ is the confirmation rate of candidate OB stars. The confirmation rate is a measure of the reliability of our OB candidate search (high ${f}_{\mathrm{conf}}$ implies few false-positive OB candidates that turn out to be late-type stars), and it will certainly vary from region to region.¹¹ In the (unrealistic) limit that all OB candidates in every region are eventually confirmed (${f}_{\mathrm{conf}}\to 1$), we would have ${F}_{\mathrm{inc}}$ ranging from zero for the three MYStIX regions with no OB candidates to $\gtrsim 3$ for regions such as NGC 6334, RCW 38, and Trifid, which have relatively few published OB stars compared to the number of their identified OB candidates.

We also consider completeness, the rate at which we identify true OB stars as candidates (as opposed to false negatives). In Table 2, all stars earlier than O9 were detected in X-rays, as well as all O giants and supergiants. However, a decreasing fraction of O9 to B1 main-sequence and B giant stars are detected by Chandra in the MYStIX fields. This is consistent with the long-established relationship between X-ray and bolometric luminosities in massive stars attributed to a trend in wind shock properties (Gagné et al. 2011). Hence, one estimate of completeness is the fraction of known O through B1 stars that have MYStIX X-ray detections, 67 of which satisfied our SED fitting search criteria (column 17 of Table 2), while virtually all of the remaining 55 satisfied our NIR search criteria (Figure 2). This fraction, ${f}_{X}=$ OB1X(pub)/OB1(pub), typically ranges between 0.5 and 1 among studied MYStIX regions (with the exception of NGC 1893, the most distant region, for which ${f}_{X}\sim 0.4$) and averages to ${\bar{f}}_{X}=119/172\,=70 \%$ across all regions (Table 5). The observed variation in f_X among the regions is caused primarily by differences in distance and depth of the X-ray observations.

Correcting for incompleteness in our X-ray search for new OB1 stars, we estimate the increase in the inferred total massive stellar populations in the MYStIX regions studied here as ${F}_{\mathrm{inc}}^{\prime }={F}_{\mathrm{inc}}/{f}_{X}$. The key unknown parameter governing the potential impact of our OB candidates on the inferred massive stellar population is, unsurprisingly, the confirmation rate.

We can begin to constrain ${f}_{\mathrm{conf}}$ by considering the five MYStIX regions (including Carina) that have recent spectroscopic follow-up of a subset of their OB candidates (Section 3.2):

• 1.  
  NGC 6357—All 12 OB candidates in Region 3 (see Figure 2(k)) followed up by NIR spectroscopy using VLT/KMOS were confirmed (A. Bik 2015, private communication). An additional two, brighter OB candidates (MOBc NGC 6357 19 and 21) are intermingled with these confirmed candidates and have a high probability of eventual confirmation as newly discovered, O-type stars (Appendix A.7). With 26 OB candidates in NGC 6357, we can report $0.5\gtrsim {f}_{\mathrm{conf}}\leqslant 1$. Adding the 14 confirmed and high-probability candidates to the 15 OB1(pub) stars (Table 5), we have already doubled (${F}_{\mathrm{inc}}\sim 2$) the known massive stellar content of this rich star-forming complex.
• 2.  
  M17—A subset of the candidates in M17 were observed with the long-slit spectrograph of WIRO. The types were determined using the red (5300 Å–6700 Å) portion of the spectra, since this includes several He i lines and one He ii line and because the blue portion is often not usable, given the high extinctions. Many of the candidates were still too obscured to detect—candidates with $J\gtrsim 10$ mag could not be classified, while all candidates brighter than this limit were assigned spectral types. Five relatively unobscured candidates were hence confirmed as OB stars, while four others were classified as late-type stars (Table 3). With 25 OB candidates in M17, we can place only a loose constraint on the confirmation rate, $0.2\gtrsim {f}_{\mathrm{conf}}\gtrsim 0.8$. With 36 pubOB1 stars (Table 5), we have already increased the cataloged massive stellar population by 10%, with potential for up to a 60% increase ($1.1\gtrsim {F}_{\mathrm{inc}}\gtrsim 1.6$). High-resolution, NIR spectroscopy is needed to classify the numerous obscured OB candidates located near the crowded NGC 6618 cluster and the M17 giant molecular cloud.
• 3.  
  Trifid—Three OB candidates in Trifid have been spectroscopically confirmed as early-type stars (Kuhn et al. 2016, in preparation), while the remaining five candidates have not yet been classified spectroscopically—hence $0.38\leqslant {f}_{\mathrm{conf}}\leqslant 1$.
• 4.  
  W3—Kiminki et al. (2015) independently identified and spectroscopically classified four of our candidates as OB stars, while one candidate was previously classified as a late-type star. With eight candidates in W3, we have $0.5\leqslant {f}_{\mathrm{conf}}\gtrsim 0.9$, similar to that for NGC 6357.
• 5.  
  Carina—Alexander et al. (2016) obtained visible-light spectra for 71 of the 94 OB candidates in the Carina Nebula identified by P11, classifying 21 as OB stars and 50 as late-type stars. This constrains the confirmation rate for the Carina OB candidates to $0.2\leqslant {f}_{\mathrm{conf}}\lt 0.5$. These relatively low confirmation and high rejection rates for Carina may not be representative of the MYStIX regions. The large CCCP X-ray mosaic surveyed by P11 included a much greater portion of the surrounding Galactic field than was typical for MYStIX X-ray observations (Feigelson et al. 2013), and the majority of the rejected OB candidates are relatively unobscured and located in the outlying regions, away from the principal star-forming clusters.

Taken together, the above results favor an average ${\bar{f}}_{\mathrm{conf}}\approx 0.5$ across all MYStIX regions, predicting fractional increases of ${\bar{F}}_{\mathrm{inc}}\approx 0.5(125/172)+1\,=\,1.4$ in the cataloged massive stellar populations and ${\bar{F}}_{\mathrm{inc}}^{\prime }={\bar{F}}_{\mathrm{inc}}/0.7=1.9$. We therefore conclude that even a conservative 50% confirmation rate for our OB candidates would double the inferred massive stellar content averaged across the 18 MYStIX star-forming complexes studied here.

P11 predicted that the most likely populations of contaminating sources in their CCCP sample of X-ray-selected candidate OB stars would be F or G stars, either field giants or (very nearby) foreground dwarfs. The same predictions can be extended to our MYStIX OB candidates, but the recent spectroscopic follow-up of CCCP OB candidates by Alexander et al. (2016) allows us to refine our understanding of the most likely types of contaminants. The 50 CCCP OB candidates classified as later-type stars consisted of 32 G through K5 (super)giants, 1 F supergiant, a single K dwarf, and 15 F or A stars for which no luminosity classification was made. None of the CCCP candidates was classified as a late K or M giant or dwarf, confirming the prediction of P11 that the coolest stars, which dominate the Galactic field populations, are not selected as X-ray-emitting OB candidates. These results indicate that the most common types of contaminating source in our MYStIX OB candidate sample are F through early K giants matched to X-ray sources. We note that some of these stars could be evolved massive stars with strong intrinsic X-ray emission.

Seven of the CCCP OB candidates classified as (super)giant stars by Alexander et al. (2016) had X-ray spectra of sufficient quality that P11 were able to fit one- or two-component thermal plasma models to measure their absorbing columns and plasma temperatures. All of these sources exhibited at least one soft plasma component with ${kT}\gtrsim 1\,\mathrm{keV}$ (OBc 1, 3, 10, 12, 18, 32, and 41 in Table 5 of P11), consistent with the soft X-ray emission from symbiotic binary systems in which an evolved giant transfers mass to a white dwarf companion (Mürset et al. 1997; Luna et al. 2013). Symbiotic binaries can produce nova outbursts accompanied by spectacular X-ray flares that decay rapidly as the ejecta interact with the extended wind region surrounding an evolved giant (Sokoloski et al. 2006). In light of the new G0–K0 I spectral classification, such an outburst now provides the best explanation for the unusual X-ray variability observed for CCCP OBc 41 (Townsley et al. 2011b).

The soft X-ray emission from quiescent, symbiotic binary systems originates in accretion shocks or colliding winds (Mürset et al. 1997), and it is generally not possible to distinguish these systems from OB stars by analysis of the relatively low-resolution MYStIX X-ray spectra. P11 compared ${A}_{V}^{\mathrm{MS}}$ from SED fitting to the absorbing column ${N}_{{\rm{H}}}^{\mathrm{Xfit}}$ from X-ray spectral fitting in an attempt to evaluate which of their OB candidates had X-ray spectra consistent with reddened OB stars. Among the nine candidates identified as having X-ray spectra consistent with IR SED fitting, the results were ambiguous, with three confirmed new OB stars (OBc 39, 55, and 57), three classified as late-type stars (OBc 2, 12, and 67), and three that were not detected by AAT optical spectroscopy (Alexander et al. 2016).

Young, yellow giants or exotic, symbiotic binary systems are far less common than red giants in the Galactic field; hence the contamination they may introduce to our samples should not be severe. We do not see systematically increased numbers of candidates in regions that are projected near the Galactic center, as might be expected if our candidate lists were dominated by contamination from IR-bright field giants. For example, the Lagoon Nebula, Trifid Nebula, NGC 6334, and NGC 6357 all have longitudes within 10° of the Galactic center, yet the former two regions have few new candidates, while the latter host numerous candidates. By contrast, the Carina Nebula is relatively far from the inner Galaxy yet contains nearly as many OB candidates as all 18 of our MYStIX regions combined (P11).

We have identified 125 candidate X-ray-emitting O or early B (MOBc) stars through a systematic search of the combined X-ray and IR point source catalogs for 18 of the 20 MYStIX star-forming complexes, supplementing the 328 previously identified by spectroscopy (Table 5). Follow-up or independent spectroscopy has already confirmed 24 of our candidates as newly discovered Galactic OB stars. We also modeled the 1–8 μm SEDs of 239 previously published OB stars, with the effective temperature fixed based on the spectral type, to measure their interstellar reddening and bolometric luminosities and to provide additional context for our candidate selection methodology. This analysis identified six candidate massive binary/multiple systems (the principal ionizing stars in NGC 2264 and the Lagoon Nebula, two stars in the Eagle Nebula, and two stars in NGC 1893) and five candidate O-type (super)giants with no published luminosity class (four in M17 and one in NGC 6334).

Perhaps the clearest trend in Table 5 is an anticorrelation with historical research. We find few to no new candidate OB stars in lightly obscured complexes that have been heavily studied in the past using visible-light photometry and spectroscopy. The stellar populations of regions such as NGC 2264, the Rosette Nebula, the Lagoon Nebula, the Eagle Nebula, and W4 have been surveyed for decades from northern hemisphere observatories. In contrast, nebulous and heavily obscured regions accessible only to southern hemisphere observatories (such as RCW 38, NGC 6334, NGC 6357, and NGC 3576) have been less carefully examined. The majority of previously cataloged MYStIX OB stars have ${A}_{V}\gtrsim 3$ mag, while the large majority of candidate OB stars have ${A}_{V}\gtrsim 5$ mag, assuming dereddening to the OB ZAMS (the CMDs in Figure 2). Previous work has also typically focused on the most prominent parts of large star-forming complexes, whether rich clusters or bright H ii regions, but now in several cases—notably NGC 6357, M17, Trifid, and Carina (P11; Alexander et al. 2016)—new spectroscopically confirmed OB stars have been discovered by surveying the less studied yet often lightly obscured surrounding regions.

Notable results for individual regions include the identification of the OB population of a recently discovered massive cluster in NGC 6357, an older OB association in the M17 complex, three new luminous O stars near the Trifid Nebula, doubling of the known massive stellar population in NGC 6357, and the potential to more than double the known OB populations in RCW 38, NGC 6334, and NGC 3576. In a few cases, we have probably identified the principal ionizing star of an H ii region. MOBc NGC 6334 11 is a high-luminosity, highly obscured (${A}_{V}^{\mathrm{MS}}\simeq 34$) star located at the center of the radio H ii region NGC 6334 D, with associated nebular emission and masers (Appendix A.6). MOBc NGC 6357 21, a very luminous candidate consistent with an O3 V star on the ZAMS, appears to be the dominant member of the NGC 6357 F subcluster, which contains the largest concentration of newly discovered OB stars in our sample and ionizes the G353.24+0.64 H ii region (Appendix A.7). MOBc Trifid 3, a newly confirmed O-type giant, likely ionizes the PAH-bright MIR bubble in the far north region of the Trifid Nebula complex (Appendix A.11). Additional discussion of individual regions is presented in the appendix.

The 18 MYStIX regions studied here represent the great diversity of the heliocentric distance, obscuration, and size of the young massive stellar population and the degree of previous study. For one-third of them, existing catalogs of massive stars appear largely complete. For the rest, the priority should be continued spectroscopic follow-up, preferably in the NIR, to assess the true nature of these OB candidates and move ever closer to completing the census of massive stars in these prominent Galactic star-forming complexes.

No new OB candidates were found in 5 of the 18 examined MYStIX star-forming regions:

• 1.  
  The Flame Nebula—The NIR CCD and CMD (Figure 2(a)) do not reveal any bright candidates lacking IR excess emission in this well-studied, very young star-forming region (Meyer et al. 2008). The most luminous star in the cluster, IRS 2b, is an extended, embedded IR source. The most luminous IR point source is NGC 2024 1 (B0.5 V), which reveals its photosphere at wavelengths shorter than 4.5 μm. We modeled this source with ${A}_{{\rm{V}}}=8.1$ mag (Table 2), close to the averaged cluster extinction ${A}_{V}\simeq 10$ mag reported by Skinner et al. (2003).
• 2.  
  W40—Of the three OB stars identified with IR spectroscopy (Shuping et al. 2012), two were IR point sources for which we were able to measure A_V and ${L}_{\mathrm{bol}}$ with SED fitting (Table 2), but the most luminous, IRS 1A South, is an embedded, extended MIR source (Figure 2(b)). The other bright stars appearing in the NIR CMD are IR excess sources and hence are not selected as candidate OB stars.
• 3.  
  RCW 36—This embedded H ii region has two spectroscopically identified OB stars, and the more luminous of the two is the more highly obscured (Figure 2(c)).
• 4.  
  NGC 2264—This very well-studied region (Dahm 2008a) contains no new OB candidates, which is reassuring. HD 261782 is a bright, evolved intermediate-mass giant star (G5 IIIp; Dahm et al. 2007) that would satisfy our secondary NIR selection criteria (Figure 2(d)), but fitting of its full SED ruled it out as an OB star. This demonstrates that yellow and red (super)giants can be erroneously identified as OB stars by our NIR photometry criteria if they happen to have an X-ray detection in the MYStIX source lists but lack sufficient MIR photometry for SED fitting. Our SED fitting for HD 47839, the principal ionizing star, returns about twice the luminosity expected for its O7 V((f)) spectral type (Figure 3). As mentioned in Section 3.1, this could indicate multiplicity.
• 5.  
  NGC 2362—Unsurprisingly, we found no OB candidates in the oldest MYStIX cluster (Dahm 2008b), which lacks any obscuring molecular material (Figure 2(g)). For all of the known OB stars, our SED fitting measures very low extinction ($\langle {A}_{V}^{\mathrm{SED}}\rangle =0.5$ mag; Table 5) as expected, except for the bright O giant HD 57061 (O9 II), for which SED fitting returned anomalously high ${A}_{V}^{\mathrm{SED}}=2.3$ mag and $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}/{L}_{\odot }=6.03$ (Table 2 and Section 3.1). The IRAC [3.6] photometry for HD 57061 is saturated and was not used for SED fitting. IR excess emission due to non-LTE effects in a strong stellar wind could explain the anomalously high ${A}_{V}^{\mathrm{SED}}$ for HD 57061, leading to a corresponding overestimate of ${L}_{\mathrm{bol}}^{\mathrm{SED}}$ (P11).

In the Rosette Nebula and its associated molecular cloud (Román-Zúñiga & Lada 2008, p. 928), five well-established early-type stars were independently identified by our SED fitting selection: HD 46150, HD 46056, HD 46149, HD 46106, and HD 259012 (Table 2). No new OB candidates were found via SED fitting, but one candidate, HD 259513, was selected using our secondary NIR criteria (Table 4). This star has a late spectral type of unknown provenance listed in SIMBAD.

The Lagoon Nebula (M8) is composed of an H ii region, an open cluster dominated by B-type stars, and a giant molecular cloud with a few O stars and dozens of B stars at a distance of around 1.3 kpc (Lada et al. 1976; Tothill et al. 2008). Relatively few of the 28 spectroscopically confirmed OB stars were recovered by our SED fitting analysis because only 15 were matched to X-ray point sources, some of which had poor JHK_s photometry.

SED fitting identified four new candidate OB stars with $4\lt {A}_{V}^{\mathrm{MS}}\lt 7$ (Table 3), including MOBc Lagoon 4 in the middle of the open cluster (Figure 2(f)), which has been followed up spectroscopically in the NIR using the VLT and confirmed as an early-type star (A. Bik 2015, private communication). MOBc Lagoon 5 is actually a K-type giant. Some of the OB candidates appeared in earlier optical, IR, and X-ray studies of M8 members without being identified as possible massive stars.

Our SED fitting for HD 164794, the principal ionizing star, returns nearly double the luminosity expected for its O4((f)) spectral type (Figure 3). As mentioned in Section 3.1, this could indicate undiscovered multiplicity, with the caveat that the presence of a strong wind in this early O-type emission-line star might have caused our LTE models to overestimate its ${L}_{\mathrm{bol}}$ (see Section 2.2).

SED fitting identified no candidate OB stars in the very young complex DR 21, which lies within the much larger Cygnus X complex (Reipurth & Schneider 2008). This is not surprising, given that no large H ii region is present in DR 21 (Figure 2(h)) and the most luminous IR point sources are protostars and pre-MS stars with dusty disks (Povich et al. 2013; Romine et al. 2016). Bright, extended MIR emission associated with embedded massive stars also foils the MYStIX MIR point-source photometry. Two OB candidates were identified via NIR photometry. MOBc DR 21 2 is associated with the extended IR source [DS96] I1 (Davis & Smith 1996) and may be the ionizing source for this (ultra)compact H ii region. If confirmed, DR 21 2 would be among the most highly obscured new OB stars in MYStIX, behind A_V = 33 mag of extinction (Table 4). The other OB candidate, DR 21 1, lies on the edge of the MYStIX X-ray field and is not actually associated with DR 21.

RCW 38 is a centrally concentrated rich cluster producing a very bright H ii region at an uncertain distance of around 1.7 kpc (Wolk et al. 2006). SED fitting identified two new candidates with luminosities consistent with early B stars (Table 3). Both are inferred to have high absorptions (A_V = 8–11 mag). These two stars were previously identified by Wolk et al. (2006) among their list of 31 candidate OB stars in RCW 38 (their sources #107 and #143 ), but with lower absorptions and thus lower inferred bolometric luminosities. We identified three additional candidates via NIR photometry, two of which had previously been noted as OB candidates by Wolk et al. (2006). An additional candidate, RCW 38 5, was classified as either a YSO or an OB candidate by Winston et al. (2011). The large majority of the 31 OB candidates identified by Wolk et al. (2006) have not been selected here. While their methodology is similar to our secondary NIR selection, they include late B-type stars and stars with disks, which (particularly the latter in a very young, embedded region such as RCW 38) are much more numerous than the O-type through early B-type stars targeted by our search.

NGC 6334, the Cat's Paw Nebula, is a complex of rich clusters and H ii regions in the Sagittarius–Carina spiral arm (Persi & Tapia 2008; Feigelson et al. 2009). Its stellar population has been poorly characterized in the IR and optical due to high and variable extinction, extremely high field star contamination (at a projected location less than 10^◦ from the Galactic center), and bright nebular emission. The most luminous known OB star is 2MASS J17203178-3551111 (O7 with no luminosity class specified; Table 2), with $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}/{L}_{\odot }=5.5$, consistent with an unresolved, O-type binary (or multiple) system, which seems a more plausible interpretation than a single, evolved star, given its location in the midst of very young, highly embedded clusters (Figure 2(j)).

Our analysis reveals a rich population of 21 candidate massive stars, mostly concentrated along the filamentary "backbone" of the giant molecular cloud that runs parallel to the Galactic plane (Figure 2(j)). The most luminous OB candidates, if they are on the main sequence, surpass the most luminous stars currently known in the region:

• 1.  
  MOBc NGC 6334 6 (${A}_{V}^{\mathrm{MS}}=10$, $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{MS}}/{L}_{\odot }=5.8;$ Table 3) lies north of the principal embedded clusters of NGC 6334 (Figure 2(j)), only $20^{\prime\prime}$ from CD $-35^\circ 11482$, a possible Herbig Be star with MIR excess emission detected (B0.5:Ve, Table 2). Both stars appear to be members of the absorbed massive subcluster NGC 6334 G (Povich et al. 2013; Kuhn et al. 2014).
• 2.  
  MOBc NGC 6334 11 is extremely obscured, with ${A}_{V}^{\mathrm{MS}}\simeq 34$ mag (Table 3). This star was undetected in J by 2MASS, which is consistent with its extremely high reddening; hence it does not appear on the NIR CCD or CMD in Figure 2(j), but SED modeling indicates a very high luminosity of $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{MS}}/{L}_{\odot }=5.6$, equivalent to an early O star on the ZAMS. The IR source was previously noted as FIR-II 24 by Straw et al. (1989), who classified it as a less absorbed, lower-mass pre-MS star. Its rising MIR SED may indeed indicate some reprocessing of stellar radiation by circumstellar dust, and this, along with its relatively low X-ray median energy (${E}_{\mathrm{med}}=2.4\,\mathrm{keV};$ Figure 4), may indicate that our SED fit with a normal stellar photospheric model overestimated the reddening and hence the bolometric luminosity (assuming that the matching between the X-ray and MIR source is correct). It lies near the center of the radio H ii region NGC 6334 D, which is about ${2}^{\prime }$ (1 pc) in diameter (Rodriguez et al. 1982) and at the base of a much larger MIR bubble blowing out to the east (dark, round nebulosity in Figure 2(j)). These nebular structures lie on the southeast edge of the large heavily absorbed subcluster NGC 6334 J (Kuhn et al. 2014). Two other OB candidates are located east of NGC 6334 11, inside the MIR bubble.
• 3.  
  MOBc NGC 6334 15 and 16 are secondary NIR candidates that would be brighter than any single O-type dwarf if dereddened to the colors of the ZAMS. While we could not measure ${A}_{V}^{\mathrm{MS}}$ for these stars, their locations on the CMD (Figure 2(j)) would indicate ${A}_{V}\sim 17$ and 25 mag, respectively. Of the two, NGC 6334 15 is the better candidate because it is located close to the massive, obscured cluster at the southwestern end of the complex.

Russeil et al. (2012) recently reported about 15 new candidate OB stars in the region of NGC 6334 covered by MYStIX, based on deep optical UBV CCDs and distance cuts. As that study did not provide clear identifications or coordinates to the individual stars, we cannot compare their results with our sample.

NGC 6357, the War and Peace Nebula, is a large star-forming complex in the Sagittarius–Carina spiral arm near NGC 6334, with three rich clusters. The historical Pismis 24 cluster is observed near the edge of the large annular H ii region G353.1+0.9 (which we will call Region 1; see Figure 2(k)), and two other, more obscured rich clusters are found in the nearby molecular cloud with associated H ii regions G353.1+0.6 (Region 2) and G353.24+0.64 (Region 3; Cappa et al. 2011; Townsley et al. 2014). Among the MYStIX regions examined in this study, our SED fitting procedure has found the largest number of candidate OB stars in NGC 6357, 25 (Table 3), with one additional candidate found by secondary NIR photometric analysis (Table 4). Wang et al. (2007) presented a list of 24 candidate OB stars in Region 1 based on high X-ray luminosity or bright NIR emission (${K}_{s}\lt 10$ mag), and we have independently identified eight of these plus one additional candidate (MOBc NGC 6357 5) that was not previously selected. Only three of the high-luminosity X-ray candidates have been recovered here, and all three had ${K}_{s}\lt 10.2$ mag. Conversely, our SED fitting and NIR photometry criteria have excluded five of the NIR-bright Wang et al. (2007) candidates. Simple, single-band bright magnitude cutoffs, even when combined with X-ray detection, are vulnerable to identifying YSOs with K_s excess emission from disks or late K through M (super)giants that are excluded by our selection criteria. NGC 6357 has also been surveyed using deep UBVR images combined with DENIS/2MASS IJHK photometry by Russeil et al. (2012). Candidate massive stars with spectral types O to B3 were identified from U − B and B − V colors after dereddening from multiband color relations (allowing for spatial variation in the R_V ratio, normally 3.1) and removal of K-excess objects. Russeil et al. report 458 candidate OB stars in their Zone 2, which is similar, but not identical, to the extent of the MYStIX region for NGC 6357. As they do not provide a catalog of their OB candidates, we cannot make a star-by-star comparison, but our OB candidate list is obviously more selective, and limited to more luminous stars.

Pismis 24 in Region 1 is centered on an unusual concentration of spectroscopically established extremely massive and luminous OB stars, including O3 (super)giants (Walborn et al. 2002). Cl Pismis 24 17, with spectral type O3.5 III(f*), is the earliest star in the entire MYStIX sample for which we were able to measure a bolometric luminosity ($\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}/{L}_{\odot }\,=5.92;$ Table 2), and given that theoretical luminosities for giants and dwarfs converge at the upper end of the ZAMS, our SED-based luminosity appears reasonable (in spite of the likely presence of non-LTE wind effects in this star that could bias our models toward higher luminosities). Cl Pismis 24 1 SW+NEab, a very massive binary system with O4 III(f+) and O3.5 If* components (Maíz Apellániz et al. 2007), was the highest-luminosity system in the entire MYStIX sample, as measured by SED fitting, with $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}/{L}_{\odot }=6.22$ (Table 2). The components of Cl Pismis 24 1 SW+NEab are unresolved by our NIR and MIR photometry. This confusion caused a failure of the automated matching of the 2MASS counterpart to the X-ray source (Naylor et al. 2013), so we substituted the 2MASS photometry included with the associated GLIMPSE Archive source. The IRAC [3.6] and [4.5] magnitudes for this star are above the GLIMPSE saturation limits, so we did not include them in our SED fitting.

Region 2 contains two published O-type stars, [N78] 49 and 51. [N78] 49, recently reclassified as an O5.5 IV(f) star (Maíz Apellániz et al. 2016), is the most promising candidate for the principal ionizing source of the G353.1+0.6 H ii region, and it lies at the center of the very compact, dense NGC 6357 D subcluster (Kuhn et al. 2014) (see lower-right inset image in the top panel of Figure 2(k)). The star was detected in all four IRAC bandpasses and saturated at [3.6]. The SED rises toward longer MIR wavelengths, which could be due to free–free emission in a strong, optically thick wind, consistent with the observed emission-line spectrum of this star (Maíz Apellániz et al. 2016). It was strongly saturated in all three UKIRT images, and confusion with neighboring stars in the subcluster created an offset in the 2MASS point source position, causing the MYStIX automated matching algorithm (Naylor et al. 2013) to report no 2MASS counterpart to the X-ray source. We retrieved the 2MASS photometry matched to the GLIMPSE Catalog source and were able to achieve good SED fits to [N78] 49. The extremely high luminosity ($\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}/{L}_{\odot }=5.87$; the fourth most luminous star for which we were able to fit an SED; see Table 2 and Figure 3) is inconsistent with a single O-type subgiant, suggesting that multiple O stars contribute to the IR source, the luminosity class was underestimated, or both.

There are two OB candidates in Region 2, both of them less luminous than [N78] 49. MOBc NGC 6357 12 has been confirmed as an OB star by follow-up VLT NIR spectroscopy (A. Bik 2016, private communication). NGC 6357 13 is optically bright, with V = 12.3 and a photometrically estimated spectral type of O9.5 (Neckel 1984); this star was too bright in the NIR for VLT follow-up. Three other OB candidates, 14, 15, and 17, located on the boundary between Regions 2 and 3 have all been spectroscopically confirmed as OB stars using the VLT (Figure 2(k) and Table 3).

Region 3 of NGC 6357 has proven to be the most fruitful hunting ground for new OB stars (see the upper-left inset image in the top panel of Figure 2(k)), with 10 candidates clustered within the MYStIX subcluster NGC 6357 F (Kuhn et al. 2014). All but two of these candidates have already been confirmed as OB stars by VLT follow-up, including MOBc NGC 6357 26, which was selected via our secondary NIR criteria. The two remaining unconfirmed OB candidates, NGC 6357 19 and 21, could not be observed spectroscopically using VLT because they are too bright in the NIR (${K}_{s}\leqslant 8;$ Table 3).¹² Most remarkable is the unstudied star MOBc NGC 6357 21, which appears very bright in the IR (K = 7.0, $[5.8]=6.7$) with $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{MS}}=5.8$, equivalent to an O3 V star on the main sequence observed through ${A}_{V}^{\mathrm{MS}}=6$ mag of extinction. This star is the best candidate for the principal ionizing source for the G353.24+0.64 H ii region. The median extinction in Region 3 is $\simeq 8$ mag, compared to 5–6 mag in Regions 1 and 2; one star has an inferred ${A}_{V}=14$, and another has ${A}_{V}=24$. We thus provide the best census available of candidate massive stars in this easternmost cluster of the NGC 6357 complex. The large population of probable massive stars paints a picture of a cluster that is nearly as rich as Pismis 24, but hidden behind higher extinction.

LS 4151, on the northern outskirts of NGC 6357, appears to have an uncertain spectral type (O6/7 III, $\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}/{L}_{\odot }\,=5.41;$ Table 2), and its SED fitting results place it near the ZAMS (Figure 3), suggesting a dwarf (or subgiant) rather than a true giant star.

On the far northwestern outskirts of NGC 6357, MOBc NGC 6357 1 is associated with an arcuate 24 μm nebula that is likely a bow shock caused by its ejection from the compact cluster core (Gvaramadze et al. 2011). These authors estimated a photometric spectral type of O9–O9.5, which is consistent with our SED fitting results (Table 3).

The Eagle Nebula (M16) lies near the Omega Nebula (M17), about 2 kpc from the Sun in the inner Galactic plane (Oliveira 2008). Field star contamination is very high. The ionizing cluster, NGC 6611, has a well-studied population of 13 O stars and over 50 B stars (Walker 1961); the earliest is HD 168076 with spectral type O4 V((f+)) (Table 2). SED fitting for HD 168076 and another star, NGC 6611 161 (O8.5 V), gives nearly double the luminosity expected for their spectral types (Figure 3). This could indicate undiscovered multiplicity (Section 3.1) or that the stars are actually O-type giants that have been misclassified as dwarfs. In the case of HD 168076 the presence of a strong wind might have caused our LTE models to overestimate its ${L}_{\mathrm{bol}}$ (see Section 2.2). In contrast, NGC 6611 421 and 283, both classified as Be stars, have significantly lower ${L}_{\mathrm{bol}}^{\mathrm{SED}}$ than expected for their spectral types. Both stars are located on the outskirts of the main NGC 6611 cluster (Figure 2(l)). Their lack of MIR excess emission suggests that they could be evolved, classical Be stars rather than Herbig Be stars with dust disks. They could be background stars unassociated with Eagle, or their spectral types are actually later than reported.

We report eight new candidate OB stars, seven from SED fitting and one extreme candidate from secondary NIR selection (Table 3 and 4). The most luminous SED-fitting candidate, MOBc Eagle 6, lies 16' east of HD 168076 but could be reddened by as much as ${A}_{V}^{\mathrm{MS}}=19$ mag of visual extinction. The luminous, secondary NIR candidate MOBc Eagle 8 is dubious, as its NIR colors are close to the reddened giant locus in the CCD, it is too bright in J for dereddening to the OB main sequence, and it is located on the far northern end of the MYStIX X-ray field, where the Chandra off-axis resolution is relatively poor. This increases the likelihood of a chance match between the X-ray point source and bright IR star in this very crowded field (Figure 2(l)).

Most of the NGC 6611 cluster stars are lightly absorbed with $\langle {A}_{V}^{\mathrm{SED}}\rangle \sim 3$ mag, while the new candidates appear to probe much more deeply into the cloud, with a mean $\langle {A}_{V}^{\mathrm{MS}}\rangle \sim 15.6$ mag (Table 5). The Eagle Nebula displays the greatest discrepancy in extinction between published and candidate OB stars (14 mag) of all the MYStIX regions except NGC 6334.¹³

The ionizing cluster of M17, NGC 6618, is extremely rich, dominated by a pair of O4+O4 spectroscopic binaries, a dozen other O stars, and many B stars (Townsley et al. 2003; Hoffmeister et al. 2008). Due to the combination of extremely bright MIR nebulosity and crowding of stars within the dense cluster, Spitzer/IRAC photometry is unavailable for a large fraction of the published OB stars in NGC 6618 (Table 2). We would therefore not expect our SED fitting selection, which depends on MIR photometry, to be very effective at identifying new candidates in NGC 6618 even if the cluster were not already very well studied. Nevertheless, our SED fitting did find three new OB candidates on the outskirts of NGC 6618, MOBc M17 4, 5, and 6 (Figure 2(m)), that may have gone unidentified as hot, massive stars due to very high extinction of ${A}_{V}^{\mathrm{MS}}=14$ to 40 mag (the latter value, for M17 4, was the maximum extinction allowed for the SED fitting). The brightest of these, M17 6, would be an early O star on the main sequence. Five of the seven OB candidates identified through secondary NIR selection in M17 are located near or within NGC 6618, the most notable of these being MOBc M17 23, with a projected location near the center of the cluster but behind an estimated extinction of ${A}_{V}^{\mathrm{MS}}=38$ mag (Table 4); this candidate could be another highly obscured mid-O star.

MOBc M17 3 is located within a subgroup of known OB stars associated with the M17 X subcluster immediately north of the main NGC 6618 cluster (Broos et al. 2007), which is dominated by the previously identified O6 star NGC 6618 B 260 (Hoffmeister et al. 2008).¹⁴ Two NIR-selected candidate OB stars, MOBc M17 24 and 25, are found within the highly obscured X-ray cluster M17 North (Broos et al. 2007), which is embedded within an area of MIR nebulosity extending north from the main H ii region between two IR dark clouds that run approximately north–south (Figure 2(m)). MOBc M17 24 lies behind as much as ${A}_{V}^{\mathrm{MS}}=37$ mag of extinction (Table 4).

Half of the OB candidates identified via SED fitting are found in a loose group far to the northeast of the main NGC 6618 cluster. Five of these, MOBc M17 7, 11, 12, 17, and 18, have already been spectroscopically followed up using WIRO and confirmed as OB stars (Table 3 and Figure 2(m)). These new OB stars are part of an older, more dispersed massive cluster or OB association, NGC 6618PG, which likely ionizes the extended H ii region and MIR bubble M17 EB described by Povich et al. (2009). Two massive, luminous members of NGC 6618PG—BD-16 4816 (O5.5 V((f))z) and LS 4957 (also known as BD-16 4831; B0 Ib)—have been previously classified spectroscopically; several others have been previously identified as OB candidates based on spectral types estimated from photometry only (Reed 2003; Povich et al. 2009). Four previously classified O-type stars with no published luminosity class (Cl* NGC 6618 B 137, Cl* NGC 6618 SLS 373, Cl* NGC 6618 SLS 17, and BD-16 4818) have high luminosity from SED fitting, consistent with evolved giants (Figure 3). In spite of their historical designations suggesting membership in the very young, embedded NGC 6618 cluster, these four stars may also be part of the earlier generation of star formation in M17. The most extreme of these, SLS 373, could be an O-type supergiant ($\mathrm{log}{L}_{\mathrm{bol}}^{\mathrm{SED}}/{L}_{\odot }=5.83;$ Table 2).

In contrast to the NGC 6618PG field, only three OB candidates and no known OB stars are found to the east of NGC 6618, where diffuse X-ray outflow from the H ii region is observed (Townsley et al. 2011a). The Chandra/ACIS total exposure on the eastern region was 85 ks, compared to 40 ks for the NGC 6618PG field (Townsley et al. 2014). Comparing the lack of X-ray detected massive stars in this deeper observation of a similarly unobscured field to the relative wealth of newly confirmed OB stars in the NGC 6618PG field reinforces our conclusion that NGC 6618PG is a distinct, loose OB cluster or association and not, for example, a uniformly distributed group of misclassified field stars or runaway OB stars from NGC 6618.

As mentioned in Section 3.1, four previously identified O stars in M17 have ${L}_{\mathrm{bol}}^{\mathrm{SED}}$ that is significantly lower than expected for their published spectral types. These stars (Cl* NGC 6618 B 272, B227, B213, and B 136) all have projected locations within the obscured H ii region, so they are very likely members of NGC 6618, but spectral classification is especially difficult in regions of high obscuration and nebular contamination; hence they may be B-type stars rather than O9/9.5 stars as reported by Hoffmeister et al. (2008). Two of the four lack MYStIX X-ray detections, which is unusual among the published O-type stars but common among B-type stars (Table 2).

The huge W3–W5 complex near Galactic longitude 135° has been extensively studied (Heyer & Terebey 1998; Megeath et al. 2008). W3 Main is a rich embedded cluster with many small H ii regions, W3(OH) is a clumpy group of nascent stars, and an isolated O star is found in W3 North (Feigelson & Townsley 2008). The lightly obscured and older IC 1795 cluster (Roccatagliata et al. 2011) lies between W3 Main and W3(OH) (Figure 2(n)). The W4 complex (Figure 2(o)) lies one degree east of W3. It has one well-studied, young massive cluster located near the center of a bipolar superbubble.

In W3, we report seven OB candidates in the vicinity of the W3 Main cluster, plus one candidate near W3(OH), MOBc W3 8, that has been spectroscopically classified as an early B-type binary system (Rivera-Ingraham et al. 2011) but omitted from the MYStIX compilation of OB stars (Broos et al. 2013). The most luminous OB candidate, MOBc W3 1, has been recently identified independently and spectroscopically classified as an O-type binary system by Kiminki et al. (2015), who also classified two of our other candidates, MOBc W3 3 and 5, as B-type stars (Table 3). The most highly obscured OB candidate, MOBc W3 6, comes from our secondary NIR selection with an estimated ${A}_{V}^{\mathrm{MS}}=27$ mag. This star is located among the known OB population and (ultra)compact H ii regions in W3 Main and was classified as a candidate YSO by Rivera-Ingraham et al. (2011). Our available NIR photometry does not indicate a K_s excess from circumstellar material (6 is the topmost star in the NIR CCD of Figure 2(n)), but we cannot rule out MIR excess emission for this star, which is bright at 3.6 μm. The absorptions to the other MYStIX candidate OB stars exhibit $4\lt {A}_{V}^{\mathrm{MS}}\lt 17$ mag, generally higher than the ${A}_{V}\sim 4$ mag for the optically bright members studied by Oey et al. (2005). It is possible that some are outlying members of the W3 Main cluster rather than the IC 1795 cluster.

The X-ray properties and visible-light spectra of the massive stellar population of W4 were recently studied in detail by Rauw & Naze (2016). Our SED fitting of HD 15570, a dominant ionizing supergiant star with spectral type O4 If+, returned an anomalously high ${A}_{V}^{\mathrm{SED}}=3.7$ mag compared to the average $\langle {A}_{V}\rangle =2.4$ mag for W4 (Table 5) and a correspondingly high $\mathrm{log}{L}_{\mathrm{bol}}/{L}_{\odot }=6.10$ (Table 2). The MIR photometry for HD 15570 (Kuhn et al. 2013a) does not appear heavily affected by saturation, but the rising SED toward [5.8] and [8.0] triggered a marginal IR excess flag (Povich et al. 2013), so these two longest-wavelength bandpasses were not used in our SED fitting for this star. This MIR excess could have been produced by an optically thick wind, which would have caused the static, LTE stellar atmosphere models used in our SED fitting to overestimate both luminosity and extinction (P11). Indeed, Bouret et al. (2012) derived a significantly lower $\mathrm{log}{L}_{\mathrm{bol}}/{L}_{\odot }=5.94$ for HD 15570 using a more appropriate, non-LTE wind model for the FUV–visible spectrum.

We identify only one candidate OB star in W4, at the extreme western corner of the ACIS-I field (Figure 2(o)). This candidate may be spurious, because while it satisfied our NIR–MIR SED fitting criteria, its NIR colors and magnitudes alone, while uncertain, are inconsistent with a reddened OB star. It could also be a mismatch of the X-ray source with the IR counterpart, given the reduction in positional accuracy for far off-axis Chandra point sources.

The Trifid Nebula (M20) H ii region in the inner Galactic plane (Rho et al. 2008) is illuminated by a young cluster dominated by the O7.5 Vz star HD 164492A, and the surrounding field suffers from extreme field star contamination (Figure 2(p)). Due to the combination of bright nebulosity and filamentary cloud obscuration, optical study of the cluster is poor; only a few B stars and no additional O stars are known in the cluster. The northern component of the nebula ∼10' north of HD 164492 appears as a blue reflection nebula in the optical band. The A7 Iab supergiant HD 164514 lies in the middle of the reflection nebula, but it is debated whether it is the source of radiation (Voshchinnikov 1975; Lynds & Oneil 1986). A third nebular component lying 5–10' further to the north consists of an MIR bubble (CN 95 in the Churchwell et al. 2007 catalog) not observed in visible light. Rho et al. (2006) suggested this third component is an independent H ii region with active star formation and an unidentified principal ionizing star.

Our SED fitting search uncovered four candidate massive stars, including the newly confirmed early-type stars MOBc Trifid 1, 2, and 3, located 1' apart inside the CN 95 bubble in the extreme northeast corner of the MYStIX X-ray field (Kuhn et al. 2016, in preparation). Trifid 2 and 3 are among the most luminous OB candidates in our sample, equivalent to O3 V stars on the main sequence. The spectral classification and SED fitting luminosity for Trifid 2 indicate an evolved O6–O8 (super)giant, and this star is likely the principal ionizing source of the northern portion of the Trifid Nebula/CN 95 bubble (Rho et al. 2006, Churchwell et al. 2007). These three stars are spatially associated with and have similar extinction (${A}_{V}\simeq 5$–6.5 mag) to Trifid D, a subcluster of lower-mass, pre-MS stars identified by Kuhn et al. (2014). We note the possibility that this cluster and its associated MIR bubble are only coincidentally associated with the Trifid Nebula to the south; however, this region is unlikely to lie in front of the blue reflection nebula seen in visible light.

Four candidate OB stars identified by our secondary NIR criteria and one other candidate identified via SED fitting are found elsewhere in the Trifid field, most of them near the main nebula (Figure 2(p)) and none, to the best of our knowledge, previously noted in the literature.

NGC 3576 is an embedded H ii region in the Sagittarius–Carina spiral arm, projected near the more distant NGC 3603 young rich cluster and H ii region. Its stellar content is relatively poorly studied; most of the ionizing OB stars appear to be highly obscured and are not yet identified (Figuerêdo et al. 2002). The MYStIX field includes the embedded cluster and a second Chandra/ACIS-I pointing to the north (Townsley et al. 2011a) that reveals an older cluster dominated by HD 97319 (O9.5 Ib) and EM Car/HD 97484 (O8+O8 V binary).

We report 10 new OB candidates in NGC 3576, all but one in the southern ACIS-I pointing containing the embedded H ii region (Figure 2(q)). The southern obscured stars have an estimated extinction of $5.5\lt {A}_{V}^{\mathrm{MS}}\lt 16$ mag. The most luminous candidate is MOBc NGC 3576 3, which would be an O7.5 V star if dereddened by ${A}_{V}^{\mathrm{MS}}=6.8$ mag to the ZAMS (Table 3), and it appears to ionize a second, smaller H ii region a few arcminutes southwest of the main embedded H ii region. The sole new candidate in the northern, lightly obscured cluster is TYC 8959-1106-1 (MOBc NGC 3576 7 with ${A}_{V}^{\mathrm{MS}}=3.4$ mag).

NGC 1893 is the most distant MYStIX star-forming region, located toward the Galactic anticenter with low obscuration. The low-mass stellar population has been studied mostly based on the deep Chandra observation used by MYStIX (Prisinzano et al. 2011). There are nine new candidate late O-type and B-type stars, although two of them (MOBc NGC 1893 4 and 5) may be foreground A-type stars (Table 3). None lie in dense cluster cores or are associated with optical nebulosity, and most appear to trace the extended OB population reported by Negueruela et al. (2007). The most luminous OB candidate is MOBc NGC 1893 9, located toward the edge of the field beside a prominent pillar that does not appear to "know about" this star (Figure 2(r)). This unremarkable location, coupled with the low extinction (${A}_{V}^{\mathrm{MS}}\simeq 3$ mag; Table 3), increases the odds that this candidate is actually an unassociated field star.

SED fitting for HD 242926 returns approximately double the luminosity expected for its O8((f)) spectral type (Figure 3). This could indicate undiscovered multiplicity (Section 3.1), with the caveat that the presence of a strong wind in this emission-line O star might have caused our LTE models to overestimate its ${L}_{\mathrm{bol}}$ (see Section 2.2).