MASSIVE STARS IN THE W33 GIANT MOLECULAR COMPLEX

Massive stars enrich the galactic interstellar medium via the feedback of radiative and mechanical energy, the deposition of chemically processed gas via their strong winds, and, latterly, solid state material during the post-main-sequence (post-MS) phase. Because of their luminosities, individual massive stars can be detected and resolved in external galaxies, providing direct measures of distances and spatially resolved metallicity gradients (e.g., Kudritzki et al. 2014). At and beyond the end of their lives they power a wide variety of highly energetic transient phenomena—first during their deaths in supernovae (SNe) or gamma-ray bursts and subsequently by accretion onto their stellar corpses in X-ray binaries (e.g., Güdel & Nazé 2009; Eldridge et al. 2013).

Considerable uncertainty remains regarding the mechanism(s) of formation of massive stars, although it is strongly suspected that this process is hierarchical: massive stars are found in apparently isolated young massive stellar clusters (e.g., the Arches and Quintuplet clusters; Figer et al. 2002), in loose associations (e.g., Cyg OB2; Negueruela et al. 2008; Wright et al. 2014, and references therein), and in large molecular complexes (e.g., 30 Doradus and G305; Walborn & Blades 1997; Clark & Porter 2004).

Massive stars are very often part of binary systems (typically, a fraction of 91% of OB stars are found to have companions; Sana et al. 2014). A population of apparently isolated massive stars also exists, although it is not clear whether these have genuinely formed in isolation (e.g., Bestenlehner et al. 2011) or, instead, were lost from a natal aggregate due to dynamical or SN-driven ejection (runaway stars; Povich et al. 2008; Oh et al. 2014).

Because of the high and variable interstellar extinction and uncertain distances of stars within the Galactic disk, it has long been suspected that our census of massive star-forming regions is incomplete. Fortunately, the plethora of modern infrared and radio surveys—e.g., MAGPIS, GLIMPSE, WISE, MSX, 2MASS, UKIDSS, and VVV,¹² —allow us to identify both the natal giant molecular clouds (GMCs) and the stars that form within them. Subsequent analysis of the physical properties of GMCs and associated stellar population(s)—in terms of the mass function of pre-stellar clumps/cores (proto-stars) and already formed massive stars, and their temporal and spatial distributions—enable constraints to be placed on the mode of star formation that occurred in the region in question (e.g., Messineo et al. 2014a).

One such massive star-forming region is the W33 complex, located in the Galactic plane at longitude $l=\sim 12\buildrel{\circ}\over{.} 8$; a parallactic distance of $2.40_{-0.15}^{+0.17}$ kpc was determined from observations of water masers, which suggests a location in the Scutum spiral arm (Immer et al. 2013). Subtending 15' (∼10 pc), it comprises a number of distinct molecular and/or dusty condensations (see Immer et al. 2014 for a census), with an integrated IR luminosity of $\sim 8\times {{10}^{5}}$ ${{L}_{\odot }}$ and a total mass of $\sim (0.8-8)\times {{10}^{5}}\;{{M}_{\odot }}$ (Immer et al. 2013). Radio observations of one component—W33 Main—revealed the presence of an obscured (proto-)cluster apparently comprising a number of stars with spectral types ranging from O7.5 to B1.5 (Haschick & Ho 1983). The presence of ongoing massive star formation is also signposted by the presence of OH, H₂O, and CH₃OH masers (Immer et al. 2013) and the direct identification of a bipolar outflow and massive dusty torus associated with the young stellar object W33A (Davies et al. 2010).

Independently of these studies, Messineo et al. (2008, 2011) serendipitously identified a hitherto-overlooked young massive cluster—Cl 1813–178—in the vicinity of W33. Analysis of the post-MS content of the cluster suggested a mass $\gtrsim {{10}^{4}}\;{{M}_{\odot }}$, making it among the most massive aggregates in the Galaxy (Clark et al. 2013). Given the unusual mix of spectral types present, Messineo et al. (2011) quoted 4–4.5 Myr but highlighted that several cluster members had low luminosities for that age; stellar luminosities would appear to demonstrate some degree of non-coevality. Intriguingly, Cl 1813–178 is found in the vicinity of the pulsar wind nebula HESS J1813–178 (Helfand et al. 2007; Messineo et al. 2008). With a spin-down measurement of 44.7 ms and a spin-down luminosity of $\dot{E}\sim 5.6\times {{10}^{37}}$ erg s⁻¹, PSR J1813–1749 is one of the youngest and most energetic pulsars in the Galaxy (Halpern et al. 2012). While the energetic young pulsar potentially lies beyond both regions ($\gt 4.8$ kpc; Halpern et al. 2012), evidently this line of sight samples numerous regions of massive star formation.

Here, we present a near-infrared spectroscopic survey of bright stars in selected regions of W33 (cl1, cl2, and Mercer1; Messineo et al. 2011) and in the nearby GLIMPSE bubble N10 (e.g., Churchwell et al. 2006). We aim to determine the massive stellar content of W33, its star formation history, and, hence, relation to the nearby cluster Cl 1813–178 and pulsar PSR J1813–1749. In Section 2 we present the spectroscopic observations, and in Section 3 the available infrared photometry. In Sections 4 and 5 we describe the spectral features and stellar properties. In Sections 6, 7, and 8 we briefly describe the spatial and temporal distributions of the detected stars, before summarizing our findings in Section 9.

Spectroscopic targets with 2MASS ${{K}_{s}}$ from 6 to 11 mag and $H-$ ${{K}_{s}}$ $\;\gt \;0.5$ mag were selected from the regions listed in Messineo et al. (2011) that exhibited overdensities of bright stars or nebulae in 2MASS and GLIMPSE images. The Mercer1 region,¹³ to the west of the radio source W33 Main, appears as a sparse aggregate of bright stars with a pronounced arc of IR and radio emission to the south that is suggestive of a wind-blown structure (Figure 1, top panel). The cl1 region, immediately to the east of the embedded proto-cluster in W33 Main, contains an isolated bright star surrounded by an arc of emission at IR and radio wavelengths (Figure 1 , middle panel). The cl2 region, coincident with W33 B1 (e.g., Immer et al. 2014), contains a small cluster of stars associated with diffuse IR and radio emission (Figure 1, bottom panel). These targets were supplemented with a few isolated bright targets selected on the basis of their GLIMPSE colors—e.g., indicative of free–free emitters (e.g., Hadfield et al. 2007; Messineo et al. 2012)—such as star #1 and the candidate luminous blue variable (LBV) #15 from Cl 1813–178. Additionally, we observed some stars to the north of W33, in the cluster BDS2003-115 embedded in the GLIMPSE mid-IR bubble N10 (Figure 2).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Data were acquired with the Spectrograph for INtegral Field Observations in the Near Infrared (SINFONI; Eisenhauer et al. 2003) on the Yepun Very Large Telescope, under the ESO programs 087.D-0265(A) and 089.D-0790(A). The K-grating was used along with the $0\buildrel{\prime\prime}\over{.} 25$ pixel⁻¹ scale to yield a resolving power of ∼4500. Integration times per exposure ranged from 30 to 300 s. Typically, each observation consisted of four exposures, two on target and two on sky. Data-cubes were generated with the version 3.9.0 of the ESO SINFONI pipeline (Schreiber et al. 2004; Modigliani et al. 2007), using flat-fields, bad-pixel masks, distortion maps, and arcs. From each cube, stellar traces with signal-to-noise ratio larger than 30–40 were analyzed. Corrections for instrumental and atmospheric responses were accomplished with standard stars of B types; stellar Br$_{\gamma }$ and He i lines at 2.112 μm were removed from the standard spectra with a linear interpolation. A sky subtraction was performed to eliminate possible nebular lines and residuals from OH subtraction. A total of 86 cubes were observed, and 94 stars were extracted.

We searched for counterparts of the spectroscopically observed stars in the 2MASS Catalog of Point Sources (Skrutskie et al. 2006), in the UKIDSS catalog (Lucas et al. 2008), and in the DENIS catalog (DENIS Consortium 2005) by taking the closest match within 1''. Mid-infrared data were retrieved from the MSX (Egan et al. 2003) with a search radius of 5'', from the GLIMPSE (Churchwell et al. 2009) and the WISE (Wright et al. 2010) with a search radius of 2''.

For 21% of the sources, R-band counterparts with magnitudes from 12.38 to 19.80 mag were found in the Naval Observatory Merged Astrometric Dataset (NOMAD; Zacharias et al. 2004).

Photometric measurements are listed in the Appendix. For the observed targets, there is no additional information from the SIMBAD database.

3.1. UKIDSS Photometry

4.1. Spectral Classification

We detected a total of 23 new early types, which comprise 1 Wolf–Rayet, 13 stars of spectral type O or B, and 9 stars with indeterminate, but early, spectral types (OBAF), and 70 late-type stars (see Tables 1, 2, and 3).

Table 1.  Summary of Observations of Early-type Stars

ID	R.A.[J2000]	Decl.[J2000]	Spec.	Period	Date-Obs
	(hh mm ss)	(deg mm ss)			(yyyy mm dd)
M15a	18 13 20.99	−17 49 46.9	cLBV	P89	2012 Jun 21
1	18 13 34.81	−18 05 41.5	WN6	P89	2012 Jun 21
2	18 13 47.53	−17 57 10.7	OBAF	P89	2012 Jun 27
3	18 13 48.07	−17 56 15.6	B0-5	P87	2011 Aug 19
4	18 13 53.10	−17 55 57.3	B0-5	P87	2011 May 29
5	18 13 55.36	−17 57 00.7	OBAF	P87	2011 May 29
6	18 13 58.19	−17 56 23.9	B0-5	P87	2011 May 29
7	18 13 58.20	−17 56 25.4	O4-6	P87	2011 May 29
8	18 13 59.69	−17 57 41.2	O4-6	P87	2011 Aug 19
9	18 13 59.83	−17 57 44.4	OBAF	P87	2011 Aug 20
10	18 14 03.00	−17 58 52.3	B0-5	P87	2011 Aug 20
11	18 14 05.69	−17 28 40.3	O4-6	P87	2011 Jun 24
12	18 14 06.12	−17 28 33.2	OBAF	P87	2011 Jun 24
13	18 14 06.63	−17 56 06.3	B0-5	P87	2011 Aug 26
14	18 14 06.89	−17 28 43.2	OBe	P89	2012 Sep 15
15	18 14 06.93	−17 28 45.5	OBAF	P89	2012 Sep 15
16	18 14 07.89	−17 28 40.9	OBAF	P89	2012 Sep 15
17	18 14 08.09	−18 00 11.6	B0-5	P89	2012 Jun 26
18	18 14 08.30	−18 00 23.3	B0-5	P87	2011 May 29
19	18 14 08.74	−18 00 15.2	OBAF	P89	2012 Jun 26
20	18 14 08.89	−18 00 27.8	OBAF	P89	2012 Sep 15
21	18 14 09.03	−18 00 23.8	OBAF	P89	2012 Sep 15
22	18 14 16.52	−17 56 03.2	Oe	P87	2011 May 29
23	18 14 20.48	−17 56 11.2	O6-7	P87	2011 May 29

Note.

^aStar [MFD2011] 15 was discovered by Messineo et al. (2011).

Download table as:  ASCIITypeset image

4.1.1. Early-type Stars

The spectra of the newly discovered early-type stars are characterized by hydrogen and helium lines, as well as transitions of heavier elements, such as C, N, O, and Fe. Their spectral classification was accomplished by comparison to the near-infrared atlases of Hanson et al. (1996), Hanson et al. (2005), and Figer et al. (1997).

Star #1 is a newly discovered Wolf–Rayet star; its spectrum is characterized by strong and broad emission lines: He i/ N iii centered at 2.1117 μm, He ii/Br$_{\gamma }$ line at 2.1636 μm, and He ii at 2.1891 μm. The equivalent width ratios between the 2.1891 μm line and the two lines at 2.1636 μm and 2.112 μm were estimated to be 1.8 ± 0.3 and 4.3 ± 0.7, respectively,¹⁴ by using multiple Gaussian fits (Figer et al. 1997). This spectrum resembles that of WR134, a WN6b star (Figer et al. 1997); the suffix b indicates broad emission lines (e.g., Hadfield et al. 2007); derived ratios are almost identical to those calculated for WR134; thereby, star #1 is a WN6b.

The spectra of stars #7, #8, #11, and #23 are characterized by emission in the C iv 2.0705 μm and 2.0796 μm and O iii/N iii at $\approx 2.115$ μm lines, He i line at 2.059 μm and He ii 2.189 μm in absorption, and Br$_{\gamma }$ mostly in absorption (see Table 2 and Figure 3). This combination of features is characteristic of stars of mid- to late O spectral type; unfortunately, for this temperature range the assignment of luminosity classes from K-band spectra alone is somewhat problematic. Nevertheless, in this regard we note the strong morphological resemblance of these objects to the O4–6 I stars within the Arches cluster (Martins et al. 2008). The spectra of stars #8, #11, and #23 display the Si iv emission line at 2.428 μm. We denote O-type stars with O iii/N iii and Si iv in emission by Of$_{K}^{+}$, as described in Messineo et al. (2014a), although we caution that this does not necessarily indicate a super/hyper-giant classification (de Jager 1988). Star #8 has Br$_{\gamma }$ filled in, indicating I+ nature. The spectrum of star #23 has the He i line at 2.112 μm in absorption and most likely has a later sub-type.

Zoom In Zoom Out Reset image size

Table 2.  List of Lines Detected in Early-type Stars

ID	Center	FWHM	EWa	S/Nb	${{V}_{{\rm LSR}}}$d
	(μm)	(Å)	(Å)		(km s−1)
M15	2.059024	19 ± 0	−52.1 ± 0.4	278	63 ± 10
M15	2.166333	26 ± 0	−33.6 ± 0.4	355	44 ± 10
1	2.045707	33 ± 0	12.7 ± 1.0	6	⋯
1	2.111480	229 ± 1	−43.1 ± 5.8	18	⋯
1	2.163296	199 ± 2	−106.3 ± 16.3	32	⋯
1	2.188899	195 ± 1	−186.2 ± 14.4	68	⋯
2	2.165926	57 ± 8	6.3 ± 6.0	7	⋯
3	2.113018	19 ± 1	1.8 ± 0.9	10	⋯
3	2.166116	56 ± 0	6.9 ± 1.1	27	⋯
4	2.112701	38 ± 1	1.8 ± 0.4	12	⋯
4	2.165608	67 ± 2	3.2 ± 1.3	10	⋯
${{5}^{{\rm c}}}$	2.165962	77 ± 187	2.3 ± 7.1	4	⋯
6	2.112678	17 ± 11	2.0 ± 1.7	4	⋯
6	2.165574	73 ± 8	4.1 ± 2.7	6	⋯
7	2.058728	21 ± 4	0.7 ± 1.1	5	⋯
7	2.069326	11 ± 4	−0.2 ± 0.5	4	⋯
7	2.079126	11 ± 1	−0.5 ± 0.5	11	⋯
7	2.115881	32 ± 0	−1.5 ± 0.8	8	⋯
7	2.165977	55 ± 2	3.5 ± 1.2	13	⋯
7	2.189267	16 ± 14	1.9 ± 0.5	11	⋯
${{8}^{{\rm c}}}$	2.069469	13 ± 21	−0.4 ± 0.7	4	⋯
8	2.079379	16 ± 0	−1.4 ± 0.6	12	⋯
8	2.115700	46 ± 0	−4.3 ± 0.7	15	⋯
8	2.189147	15 ± 8	0.6 ± 0.9	4	⋯
8	2.427893	38 ± 11	−2.1 ± 2.0	4	⋯
9	2.166162	65 ± 34	3.4 ± 3.9	6	⋯
10	2.112934	17 ± 1	0.5 ± 1.3	7	⋯
10	2.166113	92 ± 6	6.9 ± 1.8	14	⋯
11	2.058704	13 ± 4	0.4 ± 2.1	4	⋯
11	2.079061	20 ± 3	−1.4 ± 0.8	8	⋯
11	2.115609	38 ± 3	−2.3 ± 0.7	12	⋯
11	2.166673	34 ± 6	2.6 ± 1.5	11	⋯
11	2.189657	20 ± 11	1.2 ± 1.1	6	⋯
${{11}^{{\rm c}}}$	2.427892	17 ± 19	0.0 ± 2.7	2	⋯
12	2.165692	57 ± 16	8.9 ± 5.1	8	⋯
13	2.112964	21 ± 15	−0.3 ± 1.1	3	⋯
13	2.166270	54 ± 3	5.9 ± 1.4	15	⋯
14	2.166573	⋯	1.3 ± 3.6	6	⋯
15	2.166294	81 ± 30	7.6 ± 3.9	8	⋯
16	2.166464	61 ± 0	10.4 ± 2.4	18	⋯
17	2.112767	10 ± 7	0.3 ± 1.2	6	⋯
17	2.165805	78 ± 4	5.8 ± 2.1	11	⋯
18	2.112785	19 ± 1	0.3 ± 0.4	14	⋯
18	2.165775	84 ± 0	5.6 ± 1.4	13	⋯
19	2.166143	44 ± 15	−0.1 ± 4.6	5	⋯
20	2.166230	92 ± 24	11.6 ± 6.9	6	⋯
21	2.166324	71 ± 0	8.0 ± 2.2	14	⋯
22	2.166032	16 ± 0	−5.7 ± 2.5	21	⋯
23	2.058749	16 ± 0	0.8 ± 2.0	6	⋯
${{23}^{{\rm c}}}$	2.068678	7 ± 420	0.5 ± 1.1	3	⋯
23	2.079160	16 ± 11	−0.5 ± 0.5	6	⋯
23	2.115583	25 ± 3	⋯	10	⋯
23	2.165405	96 ± 23	3.5 ± 1.4	11	⋯
23	2.189342	8 ± 6	1.4 ± 0.6	10	⋯
23	2.427330	27 ± 8	−1.2 ± 1.5	5	⋯

Notes.

^aErrors on the EWs are calculated following Vollmann & Eversberg (2006). ^bS/N = flux(peak)/continuum noise. ^cMarked ID numbers indicate hints for lines (with a peak S/N $\gtrsim \;2$) with poor measurements. ^dThe used SINFONI setting allows for an absolute wave calibration within 10 km s⁻¹. For star M15, the average offset of detected OH lines from their rest wavelengths is 3 km s⁻¹, with $\sigma =8$ km s⁻¹. Quoted errors are sqrt(centererr²+10²).

Download table as:  ASCIITypeset image

The spectrum of star #22 is characterized by emission in He i 2.059 μm (weak), Fe ii 2.08958 μm, probable N iii 2.11467 μm, and Br$_{\gamma }$ (strong). O iii / N iii is a signature of massive stars from O2 to O8; usually, O4–O7 stars have additional C iv lines, although they are faint in dwarfs. Star #22 appears to still be partially enshrouded; the iron emission, which is indicative of shocks, is located at the position of the star, with diffuse H₂ emission in its surroundings (lines 1–0 (S1), 1–0 (S0), 2–1 (S1), 1–0 (Q1), 1–0 (Q2), and 1–0 (Q3); Black & van Dishoeck 1987; Gautier et al. 1976; Scoville et al. 1983).

The spectrum of star #14 shows only a Br$_{\gamma }$ line in absorption with a central emission peak. Stars #14 and #22 have spectral morphologies that are reminiscent of stars associated with ultracompact H ii regions (Bik et al. 2005, 2006) and, pre-empting Section 4.2, IR excesses that are suggestive of emission from natal circumstellar envelopes. We therefore conclude that both are likely to be very young, early-type stars.

The spectra of stars #3, #4, #6, #10, #13, #17, and #18 show both He i 2.112 μm and Br$_{\gamma }$ in absorption, indicative of spectral types B0 to B5. The spectra of stars #2, #5, #9, #12, #15, #16, #19, #20, and #21 are noisy; however, a Br$_{\gamma }$ line in absorption is clearly visible. They have spectral types earlier than G types.

4.1.2. The Candidate LBV [MFD2011] 15

During the spectroscopic campaign, we re-observed [MFD2011] 15—the LBV candidate #15 in the cluster Cl 1813–178 (Messineo et al. 2011)—in order to search for the spectroscopic variability characteristic of this phase of stellar evolution.

[MFD2011] 15 was first identified with NIRPEC observations (McLean et al. 1998) with R = 1900 by Messineo et al. (2011); our new SINFONI spectrum benefits from twice the spectral resolution and an improved signal-to-noise ratio and is shown in Figure 4. Comparison to the earlier spectrum shows an almost identical morphology, with a strong P Cygni line in He i at 2.059 μm, as well as single-peaked emission in He i/ N iii/ C iii at 2.11407 μm, Mg ii at 2.13764 μm and 2.14411 μm, Br$_{\gamma }$ at 2.16655 μm, and He i at 2.185 μm. The larger baseline of the SINFONI detector and higher resolving power allow detections of H i at 1.94552 μm, He i at 1.95556 μm, a line emission at 2.10224 μm (likely due to Si iii 8–7), and a forest of Pfund H lines; intriguingly, we also detected Si iv at 2.42724 μm. With the exception of the Si iv transition, the spectrum of [MFD2011] 15 bears a strong resemblance to that of the bona fide LBV P Cygni (e.g., Clark et al. 2011). The lack of He ii lines results in a degeneracy on the temperature estimate (Najarro et al. 1994, 1997; Martins et al. 2007). Nevertheless, the presence of Si iv in emission may point to a higher temperature than previously assumed from quantitative analysis (∼16 kK; Messineo et al. 2011) and, therefore, higher luminosity; this is of interest since the current estimate (log $(L/{{L}_{\odot }})\sim 5.3$) places the star among the faintest of known (candidate) LBVs (Clark et al. 2009a). [MFD2011] 15 is included in a sample of Galactic LBVs that is homogeneously remodeled (M. del Mar Rubio-Diez et al. 2015, in preparation).

Zoom In Zoom Out Reset image size

4.1.3. Late-type Stars

K-band spectra of late-type stars are characterized by CO bands in absorption with a band-head at 2.2935 μm. We corrected these spectra for interstellar extinction by using the extinction law by Messineo et al. (2005) and the J, H, and ${{K}_{s}}$ magnitudes given in Table 6. We measured the CO equivalent widths from 2.290 to 2.320 μm with a continuum from 2.285 to 2.290 μm, as in Figer et al. (2006). The initial assumption of an average intrinsic $J-$ ${{K}_{s}}$ = 1.05 mag (H–${{K}_{s}}$ = 0.23 mag) introduces an uncertainty in ${{A}_{{{K}_{s}}}}$ of only 0.075 mag, which is negligible for spectral classification. A shift by 10% on the K-band extinction produces a median shift in EW of 1.4%. We estimated the uncertainty due to the continuum region adopted by adding small shifts and remeasuring; the percentage uncertainty has a median value of 5%. The list of detected late types is provided in Table 3.

Table 3.  Detected G-, K-, and M-type Giants

ID	R.A.[J2000]	Decl.[J2000]	EW(CO)	Spa	Spa	Obs. Date	ID	R.A.[J2000]	Decl.[J2000]	EW(CO)	Spa	Spa	Obs. Date
				RSG	RGB						RSG	RGB
24	18 14 18.4	−17 56 18	22.4 ± 5.4	<K1	K1	2011 May 29	59	18 13 53.8	−17 57 19	64.9 ± 3.1	M3	⋯	2011 May 29
25	18 13 59.1	−17 27 31	48.6 ± 1.6	K5	M7	2011 Aug 12	60	18 13 53.8	−17 57 18	21.3 ± 1.9	<K1	K1	2011 May 29
26	18 14 07.8	−17 29 44	32.3 ± 1.7	K1	M0	2011 Aug 19	61	18 13 55.3	−17 57 03	57.2 ± 1.2	M1	⋯	2011 May 29
27	18 14 02.0	−17 27 24	26.0 ± 1.8	<K1	K3	2011 Aug 20	62	18 13 55.4	−17 57 41	52.9 ± 1.9	M0	⋯	2011 May 29
28	18 10 52.0	−17 42 23	44.5 ± 1.1	K4	M5	2011 Jun 24	63	18 13 52.3	−17 56 51	48.4 ± 1.4	K5	M7	2011 May 29
29	18 10 51.8	−17 42 20	25.9 ± 19.8	<K1	K4	2011 Jun 24	64	18 13 52.2	−17 57 56	36.5 ± 3.2	K2	M2	2011 May 29
30	18 10 58.3	−17 41 24	53.2 ± 0.8	M0	⋯	2011 Aug 19	65	18 13 52.3	−17 57 54	49.5 ± 1.0	K5	M7	2011 May 29
31	18 13 54.9	−17 53 49	45.4 ± 6.8	K5	M6	2011 Jun 29	66	18 13 52.0	−17 57 52	29.0 ± 8.1	K1	K5	2011 May 29
32	18 13 44.9	−17 57 12	33.8 ± 2.8	K2	M1	2011 Aug 19	67	18 14 00.1	−17 56 31	41.4 ± 2.0	K3	M4	2011 May 29
33	18 13 50.2	−17 54 26	50.3 ± 1.2	M0	M7	2011 May 31	68	18 14 00.8	−17 56 51	43.2 ± 1.0	K4	M5	2011 May 29
34	18 14 03.1	−17 58 52	18.8 ± 2.2	<K1	<K1	2011 Aug 20	69	18 14 01.0	−17 56 50	21.6 ± 3.9	<K1	K1	2011 May 29
35	18 14 03.2	−17 58 50	44.0 ± 3.8	K4	M6	2011 Aug 20	70	18 14 09.4	−18 00 32	37.6 ± 0.6	K2	M2	2011 May 29
36	18 14 03.4	−17 58 49	42.0 ± 3.5	K4	M5	2011 Aug 20	71	18 10 52.5	−17 41 11	51.2 ± 0.8	M0	M7	2011 Aug 19
37	18 14 02.7	−17 55 38	49.0 ± 2.2	K5	M7	2011 Jun 24	72	18 13 52.7	−17 58 03	39.7 ± 2.9	K3	M3	2011 Aug 12
38	18 13 49.9	−17 57 05	43.7 ± 2.0	K4	M5	2011 Aug 19	73	18 13 52.9	−17 58 02	27.1 ± 2.4	<K1	K3	2011 Aug 12
39	18 14 27.7	−17 57 05	30.9 ± 1.6	K1	K5	2011 Jun 24	74	18 14 08.5	−18 00 19	20.9 ± 7.3	<K1	K1	2012 Jun 26
40	18 14 06.3	−17 28 33	18.9 ± 1.7	<K1	<K1	2011 Jun 24	75	18 14 08.8	−18 00 17	21.2 ± 3.5	<K1	K1	2012 Jun 26
41	18 14 05.6	−17 28 50	49.8 ± 1.1	K5	M7	2011 Jun 24	76	18 14 09.4	−18 00 48	32.4 ± 1.8	K1	M0	2012 Aug 13
42	18 10 54.8	−17 39 56	40.9 ± 2.2	K3	M4	2011 Jun 07	77	18 13 49.1	−17 56 15	25.0 ± 1.3	<K1	K2	2012 Sep 15
43	18 10 55.1	−17 40 25	31.8 ± 2.7	K1	K5	2011 Jun 07	78	18 13 55.4	−17 54 31	24.8 ± 1.6	<K1	K2	2012 Sep 02
44	18 10 52.7	−17 40 19	47.1 ± 1.1	K5	M7	2011 Jun 07	79	18 13 53.4	−17 55 19	21.2 ± 2.6	<K1	K1	2012 Sep 02
45	18 10 52.7	−17 40 08	44.1 ± 1.4	K4	M5	2011 Jun 07	80	18 13 52.2	−17 56 22	42.4 ± 0.9	K4	M4	2012 Aug 08
46	18 10 50.5	−17 40 29	23.7 ± 2.4	<K1	K2	2011 Jun 07	81	18 13 57.2	−17 58 08	31.9 ± 2.6	K1	M0	2012 Aug 11
47	18 10 55.2	−17 41 20	62.5 ± 9.0	M2	⋯	2011 May 20	82	18 13 57.2	−17 58 06	22.8 ± 6.6	<K1	K2	2012 Aug 11
48	18 10 56.6	−17 41 54	44.4 ± 1.6	K4	M5	2011 Jun 07	83	18 13 38.1	−17 43 19	46.5 ± 0.8	K5	M6	2012 Jun 03
49	18 14 01.8	−17 54 43	46.7 ± 1.2	K5	M7	2011 Aug 12	84	18 13 20.8	−18 06 26	44.6 ± 0.6	K4	M5	2012 Jun 21
50	18 14 03.9	−17 55 06	52.2 ± 1.5	M0	⋯	2011 Aug 12	85	18 13 13.5	−17 48 07	54.6 ± 0.7	M0	⋯	2012 Jun 03
51	18 13 54.7	−17 54 57	56.1 ± 1.4	M1	⋯	2011 Jun 24	86	18 14 44.5	−18 07 38	50.2 ± 1.3	K5	M7	2012 Jun 21
52	18 13 53.4	−17 55 10	14.2 ± 2.5	<K1	<K1	2011 Jun 24	87	18 13 48.2	−17 50 42	57.5 ± 1.8	M1	⋯	2012 Jun 21
53	18 13 52.9	−17 55 00	53.4 ± 1.4	M0	⋯	2011 Jun 24	88	18 13 54.8	−18 06 56	49.9 ± 2.6	M0	M7	2012 Jun 21
54	18 13 52.5	−17 56 16	60.0 ± 1.8	M2	⋯	2011 May 29	89	18 14 07.8	−17 28 37	55.4 ± 2.8	M1	⋯	2012 Sep 15
55	18 13 51.9	−17 56 27	44.9 ± 1.1	K4	M6	2011 May 29	90	18 14 02.9	−17 29 01	40.0 ± 1.2	K3	M4	2012 Sep 15
56	18 13 55.5	−17 56 18	43.4 ± 0.4	K4	M5	2011 May 29	91	18 14 05.5	−17 29 25	30.9 ± 3.0	K1	K5	2012 Aug 19
57	18 13 54.6	−17 56 12	57.1 ± 7.0	M1	⋯	2011 May 29	92	18 14 10.1	−17 27 57	29.4 ± 2.1	K1	K4	2012 Sep 15
58	18 13 54.1	−17 57 22	40.8 ± 6.5	K3	M4	2011 May 29	93	18 13 47.4	−17 57 10	41.3 ± 1.6	K3	M4	2012 Jun 27

Note.

^aSpectral types are estimated by using the relation between spectral types and EW(CO)s of red giants (RGBs), as well as that between spectral types and EW(CO)s of RSGs.

Download table as:  ASCIITypeset image

Spectral types were obtained by comparison with template spectra of red giants and red supergiants (Kleinmann & Hall 1986). For each star, spectral types are provided for two possible luminosity classes—giants and supergiants—which follow differing relations between EW(CO)s and spectral types (e.g., Figer et al. 2006; Messineo et al. 2014a). Typically, the spectral types obtained via this methodology are accurate to within two spectral types. For 13 stars in Table 3 we found EW(CO)s larger than 52 Å, i.e., larger than those of an M7 III; the spectra of seven of them (#30, #47, #50, #54, #57, #87, and #89) show water absorption and are likely Mira asymptotic giant branch (AGB) stars (see, e.g., Messineo et al. 2014b); the spectra of stars #51, #53, and #85 are quite noisy; the spectra of stars #59, #61, and #62 do not show water absorption and could be semi-regular AGB stars (Messineo et al. 2014b).

4.2. Extinction in ${{K}_{s}}$ Band and Bolometric Corrections

For early-type stars, we assumed intrinsic colors and effective temperatures, ${{T}_{{\rm eff}}}$, as tabulated per spectral type in Messineo et al. (2011), and based on the works by Bibby et al. (2008), Crowther et al. (2006b), Johnson (1966), Koornneef (1983), Humphreys & McElroy (1984), Lejeune & Schaerer (2001), Martins et al. (2005), Martins & Plez (2006), and Wegner (1994). For the WN6, we used the ${{T}_{{\rm eff}}}$ values and average infrared colors listed by Crowther (2007) and Crowther et al. (2006a). For late-type stars, we adopted the intrinsic colors given by Koornneef (1983).

Total extinction in ${{K}_{s}}$ band was calculated by assuming these intrinsic colors and by adopting the power-law curve with an index of −1.9 by Messineo et al. (2005). Estimates for the IR excess in three different colors (E(J–H), E(H–${{K}_{s}}$), and E(J–${{K}_{s}}$)) are provided in Table 4. Since the ${{K}_{s}}$ band of mass-losing early-type stars (e.g., WR) may have significant excess due to free–free emission and dust (Cohen et al. 1975), it is preferable to use E(J–H). For late-type stars, the E(J-${{K}_{s}}$) is typically used.

Table 4.  Color Properties of Newly Detected Early-type Stars

ID	Sp. Type	${{(J-{{K}_{S}})}_{o}}$	${{(H-{{K}_{S}})}_{o}}$	${{A}_{{{K}_{S}}}}(JH)$	${{A}_{{{K}_{S}}}}(J{{K}_{S}})$	${{A}_{{{K}_{S}}}}(H{{K}_{S}})$	$Q1$	$Q2$
		(mag)	(mag)	(mag)	(mag)	(mag)	(mag)	(mag)
1	WN6	0.370	0.260	0.981	0.953	0.902	−0.272 ± 0.075	−2.784 ± 0.008
2	OBAF	−0.050	−0.040	1.123	1.093	1.041	0.149 ± 0.043	⋯
3	B0−5	−0.160	−0.080	1.470	1.353	1.144	0.442 ± 0.137	⋯
4	B0−5	−0.160	−0.080	1.267	1.215	1.123	0.224 ± 0.077	−1.017 ± 0.086
5	OBAF	−0.050	−0.040	0.337	0.375	0.444	−0.071 ± 0.045	⋯
6	B0−5	−0.160	−0.080	1.254	1.295	1.369	−0.086 ± 0.658	⋯
7	O4−6	−0.210	−0.100	0.819	0.803	0.775	0.116 ± 0.163	0.318 ± 0.010
8	O4−6	−0.210	−0.100	2.883	2.809	2.676	0.292 ± 0.138	0.270 ± 0.143
9	OBAF	−0.050	−0.040	2.722	2.682	2.612	0.168 ± 0.043	⋯
10	B0-5	−0.160	−0.080	1.978	1.972	1.961	0.065 ± 0.051	⋯
11	O4-6	−0.210	−0.100	2.280	2.240	2.169	0.182 ± 0.107	⋯
12	OBAF	−0.050	−0.040	0.313	0.243	0.117	0.296 ± 0.054	⋯
13	B0-5	−0.130	−0.030	1.067	1.029	0.962	0.071 ± 0.059	⋯
14	OBe	−0.120	−0.060	2.465	2.622	2.902	−0.502 ± 0.061	⋯
15	OBAF	−0.060	−0.010	0.658	0.560	0.386	0.289 ± 0.054	⋯
16	OBAF	−0.060	−0.010	0.580	0.518	0.408	0.169 ± 0.063	⋯
17	B0-5	−0.130	−0.030	1.832	1.831	1.828	−0.059 ± 0.045	⋯
18	B0−5	−0.130	−0.030	1.657	1.639	1.608	−0.003 ± 0.051	⋯
19	OBAF	−0.050	−0.040	0.607	0.511	0.340	0.378 ± 0.051	⋯
20	OBAF	−0.050	−0.040	0.529	0.491	0.423	0.184 ± 0.048	⋯
21	OBAF	−0.050	−0.040	0.602	0.591	0.570	0.095 ± 0.046	⋯
22	Oe	−0.210	−0.100	2.872	3.085	3.465	−0.670 ± 0.129	−6.532 ± 0.344
23	O6-7	−0.210	−0.100	1.199	1.165	1.104	0.173 ± 0.083	⋯

Notes. The Q1 and Q2 parameters are defined as in Messineo et al. (2012).

Download table as:  ASCIITypeset image

Apparent bolometric magnitudes are calculated with dereddened ${{K}_{s}}$ magnitudes and bolometric corrections, $B{{C}_{K}}$, as listed in Tables 8, 9, and 10 by Messineo et al. (2011, and references therein); for the WN6, the adopted $B{{C}_{K}}$ is taken from Crowther et al. (2006a); for late-type stars $B{{C}_{K}}$ values per spectral type are available from the work of Levesque et al. (2005). As shown in the J–H versus H–${{K}_{s}}$ diagram and in the ${{K}_{s}}$−8 versus H–${{K}_{s}}$ diagram of Figure 5, the bulk of the sources follow the direction expected for reddening by interstellar dust. The two early-type stars #14 (OBe) and #22 (Oe) show significant infrared excess, possibly due to the presence of circumstellar material.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In the following, we discuss the properties of the detected massive early-type stars and their association with W33. Recently, Immer et al. (2013) measured parallactic distances of several water masers in the direction of W33. The centroid LSR velocities of three out of four H₂O maser sites are between 29 and 37 km s⁻¹, while that of the remaining, W33B, is 59.3 km s⁻¹. Nevertheless, the trigonometric parallaxes yield similar values for their distances.

Their average distance is 2.64 kpc with a standard deviation $\sigma =0.25$ kpc. Following Immer et al., for the entire W33 complex we adopt the parallactic distance to the maser W33B of $2.4_{-0.15}^{+0.17}$ kpc ($DM=11.90$ ± 0.16 mag). Derived absolute ${{K}_{s}}$, ${{M}_{K}}$, and bolometric magnitudes ${{M}_{{\rm bol}}}$ are listed in Tables 4 and 5.

Table 5.  Physical Properties of Newly Detected Massive Stars

IDa	Sp.	Classb	${{T}_{{\rm eff}}}$	${{K}_{{{S}_{o}}}}$	${{A}_{{{K}_{s}}}}$	$B{{C}_{{{K}_{S}}}}$	${{M}_{{{K}_{S}}}}$(der.)	${\rm log}$(L/${{L}_{\odot }}$)	Region
			(K)	(mag)	(mag)	(mag)	(mag)	(mag)
1	WN6	I	70000 ± 5000	7.043 ± 0.033	0.981 ± 0.025	−3.50 ± 0.50	−4.86 ± 0.16	$5.24_{-0.21}^{+0.21}$	South of W33
3	B0-5	III	24000 ± 6700	9.437 ± 0.059	1.470 ± 0.040	−2.83 ± 0.87	−2.46 ± 0.17	$4.01_{-0.35}^{+0.35}$	Mercer1, W33
4	B0-5	III	24000 ± 6700	8.386 ± 0.035	1.267 ± 0.025	−2.83 ± 0.87	−3.52 ± 0.16	$4.43_{-0.35}^{+0.35}$	Mercer1, W33
6	B0-5	III	24000 ± 6700	8.212 ± 0.245	1.254 ± 0.208	−2.83 ± 0.87	−3.69 ± 0.29	$4.50_{-0.37}^{+0.37}$	Mercer1, W33
7	O4-6	I	38000 ± 2500	7.271 ± 0.057	0.819 ± 0.051	−4.40 ± 0.15	−4.63 ± 0.17	$5.51_{-0.09}^{+0.09}$	Mercer1, W33
8	O4-6	I	38000 ± 2500	7.009 ± 0.067	2.883 ± 0.058	−4.40 ± 0.15	−4.89 ± 0.17	$5.61_{-0.09}^{+0.09}$	Mercer1, W33
10	B0-5	III	24000 ± 6700	9.207 ± 0.020	1.978 ± 0.015	−2.83 ± 0.87	−2.69 ± 0.16	$4.11_{-0.35}^{+0.35}$	Mercer1, W33
13	B0-5	V	25000 ± 5900	9.827 ± 0.026	1.067 ± 0.016	−3.12 ± 0.66	−2.07 ± 0.16	$3.97_{-0.27}^{+0.27}$	Mercer1, W33
17	B0-5	V	25000 ± 5900	10.343 ± 0.021	1.832 ± 0.013	−3.12 ± 0.66	−1.56 ± 0.16	$3.77_{-0.27}^{+0.27}$	cl2, W33
18	B0-5	V	25000 ± 5900	8.946 ± 0.023	1.657 ± 0.015	−3.12 ± 0.66	−2.96 ± 0.16	$4.33_{-0.27}^{+0.27}$	cl2, W33
22	Oe	⋯	⋯	7.576 ± 0.075	2.872 ± 0.071	⋯	−4.33 ± 0.18	⋯	cl1, W33
23	O6-7	I	35000 ± 1200	6.904 ± 0.037	1.199 ± 0.027	−4.17 ± 0.08	−5.00 ± 0.16	$5.56_{-0.07}^{+0.07}$	cl1, W33
11	O4-6	I	38000 ± 2500	8.083 ± 0.039	2.280 ± 0.032	−4.40 ± 0.15	−5.08 ± 0.10	$5.69_{-0.07}^{+0.07}$	BDS2003–115
14	OBe	⋯	⋯	9.691 ± 0.028	2.465 ± 0.026	⋯	−3.47 ± 0.10	⋯	BDS2003–115

Notes. A distance of $2.4_{-0.15}^{+0.17}$ kpc (DM = +11.90 ± 0.16 mag) is used for W33 (Immer et al. 2013) and of $4.29_{-0.19}^{0.17}$ kpc (DM = 13.16 ± 0.09 mag) for N10.

^aOBAF detections are not included in this table. ^bClasses are photometrically estimated using ${{M}_{K}}$ values from Martins & Plez (2006), Bibby et al. (2008), Humphreys & McElroy (1984), and Lejeune & Schaerer (2001).

Download table as:  ASCIITypeset image

5.1. O-type Stars

5.2. A New WN6

Found in the southwest periphery of W33 (as shown in Figure 7), the WN6b star #1 has an ${{A}_{{{K}_{s}}}}$ = 1.0 mag, when assuming the average intrinsic near-infrared colors for late WN stars by Crowther et al. (2006a). With a distance of 2.4 kpc we measured ${{M}_{K}}=-4.86\pm 0.16$ mag, which fits well with the average ${{M}_{K}}=-5.13$ mag ($\sigma =0.07$ mag) of two other WN6b stars analyzed by Crowther et al. (2006a). By using $B{{C}_{K}}$ = −3.5 ± 0.5 mag, we obtain log(L/${{L}_{\odot }}$) = $5.24_{-0.21}^{+0.21}$ and a mass of about $27\pm 2.5$ M $_{\odot }$. The detection of a late WN implies the existence of highly luminous progenitor supergiants (Georgy et al. 2012), such as those we detected in Mercer1.

Zoom In Zoom Out Reset image size

The compilations of Galactic WR stars by van der Hucht (2001), Mauerhan et al. (2011), Shara et al. (2012), and Faherty et al. (2014), identify 43 WN6 stars from a total of 443 WRs of all flavors. Among the 12 WN6 stars listed in the latter three works, 40% have broad features, suggesting that only 4% of known WRs have a similar classification. Rapid rotation and the presence of a magnetic field have been suggested to explain the broadening of their spectral lines and the flattening of the line peaks (Shenar et al. 2014).

5.3. Spectral Types B0–5

5.4. Late-type Stars

Complementing Figure 1, in which we show the location of (candidate) massive stars in the putative individual clusters, Figure 7 delineates the nominal locations of these regions on a map of the 8 μm emission. Massive stars are found throughout W33, with the richest region being the Mercer 1¹⁷ aggregate to the west of W33 Main. The presence of two O4–6 (super-)giants #7 and #8 suggests a burst of star formation occurred ∼2–4 Myr ago. A further five early to mid- B-type stars (#3, #4, #6, #10, and #13) have ${{A}_{{{K}_{s}}}}$ consistent with those of the O-type stars and bolometric magnitudes typical of dwarfs and/or giants. The remaining three stars (#2, #5, and #9) are of generic early (OBAF) spectral-type.

Immediately to the west of Mercer 1 and the embedded protocluster forming within the radio source W33 Main, we find the O4–6 (super-)giant #23, which demonstrates similar physical properties to stars #7 and #8. In GLIMPSE images, star #23 is surrounded by a yellow curved filament, similar to the mid-infrared bow shocks identified by Povich et al. (2008). The likely young massive Oe star #22 is located in front of the apex of the bowshock between #23 and the radio source W33 Main, consistent with its elevated extinction (${{A}_{{{K}_{s}}}}$ = $2.87\;\pm \;0.07$ mag). The lack of further massive stars in this region leads us to conclude that the cl1 region does not delineate a bona fide cluster.

The massive protostar W33A (Davies et al. 2010) is located in the northeast of these regions, while the stellar aggregate cl2 is located to the south. A sequence of reddened stars is detected within the compact nebula of cl2 (Figures 1 and 8); these are found at J–${{K}_{s}}$ ≈ 1.6 mag in the color–magnitude diagram shown in Figure 8. The two brightest stars, #17 and #18, are spectroscopic B0–5 types, while the remaining three, #19, #20, and #21, are classified as spectral type OBAF. Radio continuum emission is found in the direction of the cl2 cluster, centered on star #18; we estimate a flux density of 0.57 Jy at 20 cm and 0.64 Jy at 90 cm by using MAGPIS data and an aperture of 35''. Under the assumption of optically thin thermal emission with an electron temperature, ${{T}_{e}}$ = 10,000 K, this implies a Lyman continuum photon flux, ${{N}_{{\rm lyc}}}$, of ${{10}^{47.2}}$ s⁻¹ (e.g., Martín-Hernández et al. 2003; Rubin 1968; Storey & Hummer 1995). For comparison, an O9.5 V emits a number of ${{N}_{{\rm lyc}}}$ of ${{10}^{47.9}}$ s⁻¹ and an O9.5 III of ${{10}^{48.4}}$ s⁻¹ (from the more recent work by Martins et al. 2005). By comparing the results from Martins et al. (2005), Vacca et al. (1996), and Panagia (1973), after having corrected for relative average shifts, we estimate ${{N}_{{\rm lyc}}}$ = 10 $^{47.2}$ s⁻¹, 10 $^{48.0}$ s⁻¹, 10 $^{44.4}$ s⁻¹, and 10 $^{45.1}$ s⁻¹ for a B0 V, a B0 III, a B2 V, and a B2 III star, respectively. We find typical uncertainties of 0.2 dex for every ${{N}_{{\rm lyc}}}$ value. Therefore, stars #17 and #18 may already account for the requisite ionizing flux. While early B-type stars with initial masses of 9–12 M $_{\odot }$ are present in stellar populations with ages ranging up to 30 Myr (Ekström et al. 2012), the nebular emission associated with cl2 suggests a much younger age, likely of only a few Myr.

Zoom In Zoom Out Reset image size

Finally, further to the southwest we find the broad-lined WN6 star #1. We unsuccessfully searched the W33 area for other possible bright stars (${{K}_{s}}$ $\lt 10$ mag) with properties of free–free emitters or candidate red supergiants by using the infrared photometric criteria of Messineo et al. (2012). Thus, it is unlikely that there are any further young massive stellar aggregates associated with the complex. Consideration of these findings emphasizes that the massive stellar population is distributed across the confines of the W33 and appears not to be concentrated in rich young clusters similar to, e.g., Danks 1 and Danks 2 within the G305 star-forming complex (Davies et al. 2012).

This behavior appears to mirror the current location of cold molecular material, within which future generations of stars may form. Immer et al. (2014) report the detection of six molecular clumps along the east side of W33, with masses of $(0.2-4.0)\times {{10}^{3}}$ M $_{\odot }$ coincident with the peak of the CO intensity map (see Figure 7). By contrast, the evolved H ii region G12.745-00.153 (Mercer1) resides on the west side of W33, where the molecular matter has already been swept out, a configuration that is at least suggestive of sequential star formation. Similarly, the dense clump W33B1 (Immer et al. 2014) is located ∼35'' southwest of the cl2 cluster on the periphery of the apparent wind-blown nebula associated with the latter; it is conceivable that its influence is contributing to this new protostellar core by heating and compressing it (Immer et al. 2014).

The cluster candidate BDS2003-115 is located north of W33 (Bica et al. 2003; Messineo et al. 2011). It appears as a group of bright near-infrared stars in the core of a mid-infrared bubble (bubble N10, and candidate SN remnant (SNR) G13.1875+0.0389, Churchwell et al. 2006; Helfand et al. 2006; Watson et al. 2008)¹⁸ . Radio line observations of the ¹³CO $J=3-2$ transition yielded a radial velocity ${{V}_{{\rm LSR}}}$ = 50.2 ± 4.1 km s⁻¹ (Beaumont & Williams 2010). We calculated the distance to N10 using the Galactic rotation curve parameters determined by Reid et al. (2009) from fitting trigonometric parallaxes of star-forming regions, i.e., ${{R}_{\odot }}=8.4\;\pm \;0.6$ kpc and ${{{\Theta}}_{0}}=254\;\pm \;16$ km s⁻¹. We find a distance of $4.29_{-0.19}^{+0.17}$ kpc. This could correspond to the high-velocity component seen in direction of W33; the low-velocity component of W33 (35 km s⁻¹) is also detected along the line of sight of N10 (Dame et al. 2001). Bubble N10 is among the 28% of bubbles currently interacting with a molecular condensation (Deharveng et al. 2010); this clump (hereafter BGPS1865) is located on the western edge of the bubble and has a mass $\gt 1600$ M $_{\odot }$ (Ma et al. 2013, and references therein).

We detected one O4–6 star (star #11, ${{A}_{{{K}_{s}}}}$ = 2.28 ± 0.03 mag) in the center of the bubble N10, as well as three early type-stars (spectral type OBAF; #12, #15, and #16) and the embedded massive star #14. For a distance of 2.4 kpc, we obtain ${{M}_{K}}$ = −4.07 ± 0.1 mag for #11 (which is a typical value for dwarfs of 29 ± 3 M $_{\odot }$; Martins & Plez 2006; Ekström et al. 2012); for a kinematic distance of 4.29 kpc this is revised upward to ${{M}_{K}}$ = −5.08 ± 0.1 mag ((super-)giant of 36 ± 4 M $_{\odot }$; Martins & Plez 2006; Ekström et al. 2012). There is a hint for Si iv in emission at 2.428 μm in the spectrum of star #11. Given the uncertainty in reddening and temperature of the remaining four stars, we are unable to determine their luminosities.

The young massive cluster Cl 1813–178 is projected onto the northwest periphery of W33 and is coincident with SNR G12.72–0.00 (e.g., Brogan et al. 2006; Helfand et al. 2006; Messineo et al. 2008, 2011). Messineo et al. also suggested a possible association of this cluster, with G12.83–0.02 and the pulsar and TeV gamma-ray source PSR J1813–1749/HESS J1813–178 (Brogan et al. 2005).

The cluster contains six spectroscopically detected late O-type stars, 12 early B-type stars, two WN7 stars, and three transitional objects (the O6O7If star #5, the O8O9If #16, and the cLBV #15 in Messineo et al. 2011). Messineo et al. (2011) estimated a spectrophotometric distance of 3.7 ± 1.7 kpc, consistent within errors with a stellar kinematic distance of 4.8 ± 0.2 kpc, which is based on the radial velocity of the RSG member (${{V}_{{\rm LSR}}}$= 62 ± 4 km s⁻¹; Messineo et al. 2008) and on the Galactic rotation parameters presented by Reid et al. (2009).

Given our current understanding of stellar evolution, it appears difficult to reconcile the distance of Cl 1813–178 with the new parallactic distance to the W33 complex (∼2.4 kpc). On the basis of the spectral features, luminosity classes can be inferred only for the transitional objects, the two WRs, and two other B0–B3 stars with He i at 2.058 μm in emission. While the magnitudes of O7–O9 stars and B0–B3 stars are consistent with both distances of 4.8 and 2.4 kpc, the shorter distance would lead to extremely low luminosities for the three transitional objects, as shown in Table 6. For example, for the O6-O7If star, ${{M}_{K}}$ = −5.87 ± 0.11 mag at 4.8 kpc or −4.36 ± 0.17 mag at 2.4 kpc; the latter value is not compatible with the supergiant class. The spectrum of the O6–O7If star resembles that of star F15 in the Arches cluster (Martins et al. 2008); for F15 we derive ${{M}_{K}}$ = −5.76 mag by using the photometry by Figer et al. (2002), the extinction law by Messineo et al. (2005), and a distance of 8.4 kpc. Therefore, the O6–O7If star must be located behind the W33 complex.

Table 6.  Average ${{M}_{K}}$, ${{A}_{{{K}_{s}}}}$, and Number of Stars per Spectral Group and Luminosity Class in Cl 1813–178 (Table 2, Messineo et al. 2011)

		Cluster Distancea 4.8 kpc			W33 Distanceb 2.4 kpc
Sp. Group	Classc	Mkd	${{A}_{{{K}_{s}}}}$d	Nstar	Mkd	${{A}_{{{K}_{s}}}}$d	Nstar
B0–B3	I	−6.32 ± 1.01	0.94 ± 0.25	7	−5.78 ± 0.33	0.87 ± 0.18	3
B0–B3	V/III	−4.24 ± 0.60	0.98 ± 0.34	5	−3.32 ± 0.90	0.97 ± 0.30	9
O7–O9	I	−5.48 ± 0.28	0.84 ± 0.09	5	⋯	⋯	0
O7–O9	V/III	−4.93 ± 0.11	0.86 ± 0.02	1	−3.88 ± 0.36	0.83 ± 0.09	6
WN7	I	−5.86 ± 0.43	0.75 ± 0.08	2	−4.35 ± 0.43	0.75 ± 0.08	2
O6O7If	I	−5.87 ± 0.11	1.03 ± 0.02	1	−4.36 ± 0.17	1.03 ± 0.02	1
O8O9If	I	−7.32 ± 0.10	1.24 ± 0.01	1	−5.81 ± 0.16	1.24 ± 0.01	1
cLBV	I	−7.53 ± 0.10	1.09 ± 0.02	1	−6.03 ± 0.16	1.09 ± 0.02	1

Notes.

^aCluster kinematic distance (Messineo et al. 2008; Reid et al. 2009). ^bW33 distance (Immer et al. 2013). ^cLuminosity classes are photometrically assigned: for B0–B3 supergiants ${{M}_{K}}$ $\lt -5.0$ mag, for B0–B3 giants or dwarfs ${{M}_{K}}$ $\gt -5.0$ mag. For O7–O9 supergiants ${{M}_{K}}$ $\lt -5.0$ mag, for O7–O9 giants or dwarfs ${{M}_{K}}$ $\gt -5.0$ mag. ^dWhen Nstar $\gt 1$, quoted errors are the standard deviations.

Download table as:  ASCIITypeset image

Furthermore, for a distance of 2.4 kpc, the cluster members of spectral type O or B would have initial masses below 25 M $_{\odot }$, i.e., below the theoretical and observed lower mass limit for the progenitors of late WN stars (Georgy et al. 2012; Messineo et al. 2011). As a consequence, it would be difficult to understand the presence of the two WN7 (#4, #7 in Messineo et al. 2011), one O8–9If/WN9h, and one O6–O7If star, given the resultant absence of a progenitor population. The average ${{M}_{K}}$= −5.86 mag ($\sigma =0.43$ mag) inferred for the two WN7 cluster members at 4.8 kpc is consistent with that of WN7 stars (−5.38 mag with a $\sigma =0.38$ mag) in Westerlund 1 (Crowther et al. 2006a).

While the SN that gave rise to the remnant G12.72–0.00 may have occurred in Cl 1813–178 (given the precise superposition), a physical association with SNR G12.82–0.02 and associated pulsar PSR J1813–1749 appears doubtful. Specifically, a comparison of the significant column density to the pulsar and SNR to the less extreme extinction inferred for cluster members led Halpern et al. (2012) to conclude that SNR G12.82–0.02 likely lies beyond both the cluster and the W33 complex ($d\sim 5-12$ kpc).

Finally, it is of interest that the distance estimate for the W33 complex was determined from parallax measurement of masers. If it were to be observed in the future, when such emission had ceased, it would be difficult to recognize it as a complex of discrete sources at the same distance and to further distinguish the distinct stellar population of Cl1813–178 that is projected on the edge of W33 (see Figure 7).

We performed a near-IR spectroscopic survey for massive stars, encompassing both W33 and the nearby mid-IR bubble/stellar cluster N10/BDS2003-115 to study their star formation history.

• 1.  
  We detected a total of 14 new early-type (OB and WR) stars and a further nine stars with spectra consistent with spectral types earlier than F. A large population of giants with spectral types G–M were uncovered, but no cool supergiants associated with W33 were identified.
• 2.  
  Following Clark et al. (2009b), the lack of RSGs precludes substantive star formation activity with W33 $\geqslant 6$ Myr, while the detected stellar population appears broadly consistent with an age of ∼2–4 Myr.
• 3.  
  The complex contains protostars (most notably the embedded protocluster W33 Main and the high-mass protostar W33A), massive evolved stars, and clear marks of sequential star formation and feedback. Star formation within W33 has not led to the formation of rich dense clusters, and the size of W33 (radius ≈5 pc) is typical for loose associations (Pfalzner 2009). When the GMC is exhausted and star formation has ceased, W33 will most likely resemble a loose, non-coeval stellar association similar to (but less massive than) Cyg OB2 (e.g., Negueruela et al. 2008).
• 4.  
  Given the spare nature of individual stellar "aggregates" and the limitations of the current data, we cannot infer integrated masses for the young populations within W33. W33 is probable less massive than other massive star-forming complexes of the Milky Way with known evolved stars, such as W43 (e.g., Blum et al. 1999; Chen et al. 2013), W51 (Clark et al. 2009b), Carina (e.g., Preibisch et al. 2011), and G305 (Davies et al. 2012), as suggested by their respective integrated radio and IR luminosities (e.g., Immer et al. 2013; Conti & Crowther 2004).
• 5.  
  The greater distance to the nearby young massive stellar aggregate Cl 1813–178 precludes a physical association with W33. The late O-type and B-type members of the cluster support a distinct older population than that observed in W33. Considering the extinction of Cl 1813–178 (${{A}_{{\rm V}}}$ = 9.1 mag), optical spectroscopy would yield precise spectral types and direct luminosity determinations for its constituent stars (e.g., Negueruela et al. 2010).
• 6.  
  Given the distances to W33 and to Cl 1813–178, an association with the energetic pulsar PSR J1813–1749 appears doubtful.

This work was partially funded by the ERC Advanced Investigator Grant GLOSTAR (247078). F.N. acknowledges funding from the Spanish Government Ministerio de Economia y Competitividad (MINECO) through grants AYA2010-21697-C05-01, FIS2012-39162-C06-01 and ESP2013-47809-C3-1 R. This publication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. This work is based on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under a contract with NASA. DENIS is a joint effort of several institutes, mostly located in Europe. It has been supported mainly by the French Institut National des Sciences de l'Univers, CNRS, and French Education Ministry, the European Southern Observatory, the State of Baden-Wuerttemberg, and the European Commission under networks of the SCIENCE and Human Capital and Mobility programs, the Landessternwarte, Heidelberg, and Institut d'Astrophysique de Paris. This research made use of data products from the Midcourse Space Experiment, the processing of which was funded by the Ballistic Missile Defence Organization with additional support from the NASA office of Space Science. This publication makes use of data products from WISE, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. This work is based in part on data obtained as part of the UKIRT Infrared Deep Sky Survey. This research made use of Montage, funded by the National Aeronautics and Space Administrationʼs Earth Science Technology Office, Computational Technnologies Project, under Cooperative Agreement Number NCC5-626 between NASA and the California Institute of Technology. The code is maintained by the NASA/IPAC Infrared Science Archive. This research has made use of the VizieR catalogue access tool, CDS, Strasbourg, France. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France. This research has made use of NASAʼs Astrophysics Data System Bibliographic Services. A special thanks goes to the great support offered by the European Southern Observatory. M.M. thanks Jos de Bruine and Timo Prusti for useful discussions and support while at ESA. We thank the referee Dr. Philip Dufton for his careful reading.

Figure 9 displays charts for the spectroscopically observed stars. For a few stars that are not easily identifiable in this figure additional SINFONI charts are provided in Figure 10.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size