MASSIVE STARS IN THE Cl 1813-178 CLUSTER: AN EPISODE OF MASSIVE STAR FORMATION IN THE W33 COMPLEX^*

An understanding of the mechanisms of formation, evolution, and end state of massive stars is fundamental for the studies of galaxies at all redshifts. Massive stars contribute to the chemical enrichment of the interstellar medium with their strong winds and by exploding as supernovae. Massive stars are the most luminous stars, can easily be detected in external galaxies, and provide distance estimates. They are the sources of the most energetic phenomena in the universe, gamma-ray bursts (e.g., Woosley & Bloom 2006).

The availability of large surveys of the Galactic plane at radio and infrared wavelengths opens a golden epoch for studying the formation, evolution, and environments of massive stars. More than 1500 new candidate stellar clusters have been discovered, and among them several young clusters rich in massive stars may be hidden (Messineo et al. 2009).

In Messineo et al. (2008, hereafter referred to as Paper I), we presented the serendipitous discovery of a young massive cluster, Cl 1813-178, in the Galactic disk at l = 12°, with a spectroscopically identified population of massive stars, a red supergiant (RSG) star, two blue supergiants (BSGs), and one Wolf-Rayet (WR) star. Here, we present a follow-up study of the cluster. Near-infrared photometry and spectroscopy, and X-ray observations, reveal a large number of massive stars.

The cluster is located in the W33 complex and is associated with two supernova remnants (SNRs), SNR G12.72−0.00 and G12.82−0.02, and the highly magnetized pulsar associated with the TeV γ-ray source HESS J1813−178. Interestingly, the W33 complex appears to contain several other candidate stellar clusters and several SNRs. Clusters do form in large complexes (e.g., Beuther et al. 2007), and their spatial distribution varies from cloud to cloud, indicating that several external and internal triggers can be at work (e.g., Clark et al. 2009). W33 is an ideal laboratory to investigate various issues about massive stars and multi-seeded star formation, and to verify the presence of triggered sequential star formation, which is suggested by the presence of SNRs. The association of the stellar cluster with two SNRs can shed light on the initial masses of the supernova progenitors and on the fate of massive stars.

We describe the observations and data reduction in Section 2, and analyze the spectra and the cluster color–magnitude diagram (CMD) in Section 3. A discussion on the massive members and spectrophotometric distances is presented in Sections 4 and 5. Cluster age and mass are derived in Sections 6 and 7. An overview of candidate clusters in the direction of the W33 complex is given in Section 8. Finally, our findings are summarized in Section 9.

2.1. IR Photometry

Photometric measurements from the near-infrared Two Micron All Sky Survey (2MASS) catalog (Skrutskie et al. 2006), the Galactic Legacy Infrared Mid-Plane Survey Extraordinaire (GLIMPSE) catalog (Benjamin et al. 2003) as well as from the Naval Observatory Merged Astrometric Data set (NOMAD; Zacharias et al. 2004) were used.

Images of the central cluster region with the J, H, and K_s filters were obtained during two nights of observation, 2008 June 23–24,¹¹ using the near-IR camera SofI mounted on the ESO NTT (Moorwood et al. 1998). We used SofI in large-field mode with a pixel size of 0288 and a total field of view of 49 × 49 (Figure 1). For each filter, we performed a random dithering pattern of 14 images and reached a total exposure time of 840 s, 1176 s, and 1680 s in J, H, and K_s, respectively. Each image is a combination of 50, 70, and 100 exposures, each one 1.2 s long, in J, H, and K_s bands, respectively.

Zoom In Zoom Out Reset image size

Data reduction was performed using standard IRAF routine. For each filter, we obtained a sky image by median-combining the dithered frames and subtracted the sky image from each frame. Flat fielding was performed by using the SpecialDomeFlat template, which applies the appropriate illumination correction, as described in the SofI User Manual. Finally, the dithered frames were averaged into a single image. Standard crowded-field photometry, including point-spread function modeling, was carried out on each image using DAOPHOT II/ALLSTAR (Stetson 1987). The internal photometric accuracy was estimated from the rms frame-to-frame scatter of multiple star measurements (0.03 mag $<{\sigma }_J \sim {\sigma }_H \sim {\sigma }_{K_{{\rm s}}}<0.06$ mag for 8 < J ∼ H ∼ K_s < 18). The instrumental magnitudes were converted into the 2MASS photometric system.¹² The 2MASS catalog was used as an astrometric reference frame.¹³ For stars that were saturated on the SofI images, magnitudes were taken from the 2MASS catalog.

2.2. Near-IR Spectroscopy

2.3. X-ray Data

3.1. Spectral Classification

Spectral classification was performed by comparing the spectra with spectral atlases (e.g., Hanson et al. 1996, 2005; Figer et al. 1997; Martins et al. 2007; Kleinmann & Hall 1986; Wallace & Hinkle 1996).

We observed a total of 60 stars and detected 24 early-type stars and 36 late-type stars. In addition, we considered the previous observations reported in Paper I, which added an extra early-type star (no. 11) to our new sample. The observed stars were divided into candidate cluster members and non-members and are listed in Tables 1 and 7. We first listed the sample of stars given in Table 1 of Paper I, using the same identification numbers, then we appended the new targets ordered in K_s magnitude. The spectra of the newly detected early-type stars are shown in Figures 2 and 3.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In the following, we describe the spectra by grouping the stars into late types and early types. Early types divided into four subgroups (WR stars, candidate luminous blue variables (cLBVs), O-type, and B-type stars).

3.1.1. Red Supergiants

In the K band, the spectra of late-type stars have CO band heads at 2.29 μm. By measuring the equivalent widths of the CO bands, it is possible to distinguish between giants and supergiants (see the Appendix B).

Among the detected late-type stars, four stars, nos. 1, 32, 38, and 39, are found to have equivalent widths of the CO bands, EW(CO), typical of RSGs.

Star 1 is a K2-K5I cluster member, as discussed in Paper I on the basis of high-resolution data.

Stars 32, 38, and 39 have EW(CO)s typical of RSGs with spectral types M2.5, M3.5, and M1, respectively. Their photometric properties indicate that they are unrelated to the stellar cluster. We further discuss these stars in the following sections.

3.1.2. Wolf-Rayet Stars

Star 4 is a known WR star of type WN7 (number 8 in Hadfield et al. 2007). Star 7 has similar emission lines at 2.114 μm, 2.166 μm, and 2.189 μm. By comparing it with the spectral atlas of Figer et al. (1997), star 7 appears to be another WN7 star (see Figure 4).

Zoom In Zoom Out Reset image size

3.1.3. Candidate Luminous Blue Variables

The spectrum of star 15 is dominated by He i lines at 2.058 μm and 2.112 μm, and Br γ line in emission. Faint Mg ii lines are detected at 2.138 μm and 2.144 μm. The type of detected lines and the line profiles indicate that star 15 is a P-Cygni B supergiant. Its spectrum is similar to that of the cLBV star in the W51 region (Clark et al. 2009). More information on the stellar photometric variability is needed to firmly confirm that it is another LBV star.

3.1.4. O-type Stars

Spectra of O and B stars have hydrogen lines, helium lines, and other atomic lines (e.g., C iv, N iii, C iii, and Mg ii). Hydrogen and helium lines are usually seen in absorption; they may be in emission in supergiants. The Br γ line is the only hydrogen line detectable in K band. The He i transitions at 2.058 μm and 2.112 μm are usually detected in late-O and early-B stars. The He ii line at 2.189 μm appears only in O-type stars. With our resolution, the He ii line at 2.189 μm may be present down to O9.5I or O9V.

The spectrum of star 6 has a strong He i line at 2.058 μm in absorption, C iv lines at 2.069 μm and 2.079 μm in emission, strong emission line at 2.11 μm (probably due to He i/N iii/C iii), and an He ii absorption line at 2.189 μm. N iii at 2.25 μm is seen in emission. There is an indication for a faint C iv line at 2.083 μm and a N iii line at 2.10 μm. The simultaneous presence of He i, He ii, and C iv lines is typical of an O-type star. The strengths of the He i and He ii absorption lines, as well as the detection of N iii at 2.25 μm, indicate a O6–O7f+.

The spectrum of star 16 has Br γ line in emission, an He i line at 2.058 μm with P-Cygni profile, and an He i/N iii line at 2.112 μm. An He ii line at 2.189 μm is seen in absorption. A faint line is seen in emission at 2.105 μm. Only an indication for a C iv line at 2.08 μm is visible. The spectrum resembles that of an O8–O9If star, similar to HD151804 (Morris et al. 1996).

Stars 18, 19, and 20 have He i lines at 2.058 and 2.112 μm in absorption, a N iii line at 2.115 μm in emission, and Br γ line in absorption. An He ii line at 2.189 μm clearly appears in absorption. The absence of C iv lines, and the presence of N iii and He ii lines, suggest a late O-type (O7-09).

The spectra of stars 5, 8, and 9 have He i lines at 2.058 μm and 2.112 μm in absorption, and Br γ line in absorption. There is an indication for an N iii line at 2.115 μm in emission, and an He ii line at 2.189 μm in absorption. Star 5 has a clear detection of an N iii line. The absence of C iv lines, and the presence (or indication) of N iii and He ii lines, suggest that this is a late O-type (O7–O9).

3.1.5. B-type Stars

Here, we list all detected early-type stars, which do not have He ii lines. These may be B-stars or O9 dwarfs. With our resolution, the He ii line at 2.189 μm may be present down to O9.5I or O9V. Once the He ii line disappears, assigning a spectral type and luminosity class can turn into a quite degenerate issue. This degeneracy may be partly broken if the absorption components of the Br γ line and of the He i lines are sufficiently filled in by the stellar wind contribution. Typically, an He i line at 2.112 μm in absorption is observed down to B8Ia stars, while the same line is not detected in dwarfs/giants later than B3V. Such matters are highly dependent on the quality of the spectra.

From a comparison with the atlas of Hanson et al. (1996), stars 2, 3, 11, 12, 13, 14, 17, 21, 22, 23, 24, 25, and 26 appear to be likely stars with types from B0 to B2.

The spectra of stars 2 and 13 have an He i line at 2.058 μm in emission, an He i line at 2.112 μm, and Br γ line in absorption. From the relative strengths of the Br γ and He i lines (Hanson et al. 1996), the presence of He i in emission at 2.058 μm, and the lack of He ii lines, these stars appear to be B0–B3 supergiants.

The spectra of stars 3, 22, 24, 25, and 26 have an He i line at 2.112 μm in absorption as well as Br γ line in absorption. Star 17 has Br γ line in absorption and a trace of He i at 2.112 μm. These are likely B0–B3 stars or O9–B3 stars (depending on the luminosity class).

The spectra of stars 11, 21, and 23 have Br γ line in absorption and He i lines at 2.058 and 2.112 μm in absorption. These spectra are typical of early B-type stars.

Star 12 has Br γ line in absorption. The lack of He i at 2.112 μm in absorption suggests that star 12 has a spectral type later than a B8I or B3–B4V star.

The spectrum of star 14 has Br γ in absorption, an He i line at 2.112 μm in absorption, an indication for an He i line in emission at 2.058 μm, and a faint He i line at 2.113 μm in emission. The lack of He ii at 2.189 μm and the presence of He i lines suggest that star 14 is a BSG star with spectral type between B0 and B3.

The spectrum of star 11 is shown in Paper I. Stars 2 and 3 were also observed with NIRSPEC in 2008 April (Paper I), and we obtained the same spectral classification.

3.2. Photometric Results

3.2.1. Cluster Size

We assume as a cluster center the flux-weighted centroid of a 2MASS K_s-band-smoothed image (R.A. = 18:13:24.15, decl. = −17:53:29.64). We take as a cluster radius the average distance from the cluster center where the surface brightness becomes equal to the average brightness of the surrounding field (35). The 61 spectroscopically observed stars are located within 65 from the cluster center, but predominantly (50) within the cluster radius. Within 35 from the cluster center, we observed 41 stars out of the 48 stars with K_s < 9.5 mag, and 30 stars out of the 31 stars with K_s < 9.0 mag.

A fraction of 48% of stars brighter than K_s = 9.5 mag, within the cluster radius, have early types and are likely to be members. The high degree of completeness of the spectroscopic observations of the bright sample allows for a precise numbering of stars in various post-main-sequence phases (WR, cLBV, RSG, and OB stars).

Table 2. Derived Properties of Stellar Members

ID	Ks	J − Ks	H − Ks	EJ − Ks	EH − Ks	AK	MK	Lum	$\log (\hbox{$T_{\rm eff}$}[{\rm K}])$	Sp. Type	Class
1	3.79	1.67	0.44	1.02	0.31	0.54 ± 0.13	−10.16	4.95 0.24 − 0.25	3.60 0.01 − 0.02	RSG	I
2	7.22	1.42	0.50	1.50	0.54	0.80 ± 0.02	−6.99	5.70 0.34 − 0.34	4.31 0.12 − 0.12	B0–B3	I
3	7.79	1.41	0.46	1.49	0.50	0.79 ± 0.02	−6.41	5.46 0.34 − 0.34	4.31 0.12 − 0.12	B0–B3	I
4	7.94	1.68	0.66	1.31	0.39	0.69 ± 0.05	−6.16	5.76 0.16 − 0.17	4.70 0.04 − 0.05	WN7	I
5	8.56	1.40	0.50	1.61	0.60	0.87 ± 0.02	−5.72	5.75 0.12 − 0.13	4.50 0.03 − 0.03	O7–O9	I
6	8.57	1.72	0.58	1.93	0.68	1.03 ± 0.02	−5.87	5.92 0.11 − 0.12	4.54 0.02 − 0.02	O6–O7If	I
7	8.66	1.64	0.61	1.51	0.50	0.80 ± 0.05	−5.55	5.68 0.16 − 0.17	4.70 0.04 − 0.05	WN7	I
8	8.75	1.11	0.37	1.32	0.47	0.71 ± 0.01	−5.36	5.61 0.12 − 0.13	4.50 0.03 − 0.03	O7–O9	I
9	9.34	1.40	0.46	1.61	0.56	0.86 ± 0.02	−4.93	5.47 0.13 − 0.14	4.51 0.03 − 0.03	O7–O9	III
11	9.59	2.15	1.26	2.23	1.30	1.37 ± 0.02	−5.19	4.98 0.34 − 0.34	4.31 0.12 − 0.12	B0–B3	III
12	6.75	2.32	0.75	2.23	0.74	1.19 ± 0.02	−7.85	5.07 0.13 − 0.14	3.98 0.03 − 0.03	B9–A2	I
13	6.85	1.94	0.71	2.02	0.75	1.08 ± 0.03	−7.64	5.96 0.34 − 0.34	4.31 0.12 − 0.12	B0–B3	I
14	6.91	1.30	0.47	1.38	0.51	0.74 ± 0.02	−7.24	5.79 0.34 − 0.34	4.31 0.12 − 0.12	B0–B3	I
15	6.96	1.97	0.82	1.99	0.84	1.09 ± 0.02	−7.53	5.38 0.22 − 0.23	4.20 0.01 − 0.01	LBV	I
16	7.33	2.10	0.77	2.31	0.87	1.24 ± 0.01	−7.32	6.36 0.11 − 0.12	4.48 0.01 − 0.01	O8–O9If	I
17	8.52	1.18	0.45	1.26	0.49	0.68 ± 0.03	−5.56	5.13 0.34 − 0.34	4.31 0.12 − 0.12	B0–B3	I
18	8.61	1.44	0.44	1.65	0.54	0.88 ± 0.05	−5.67	5.74 0.13 − 0.14	4.50 0.03 − 0.03	O7–O9	I
19	8.72	1.53	0.53	1.74	0.63	0.94 ± 0.02	−5.62	5.72 0.12 − 0.13	4.50 0.03 − 0.03	O7–O9	I
20	9.14	1.46	0.47	1.62	0.51	0.86 ± 0.03	−5.13	5.58 0.11 − 0.12	4.54 0.01 − 0.01	O7–O8	I
21	9.29	1.44	0.49	1.60	0.57	0.86 ± 0.03	−4.98	5.04 0.38 − 0.38	4.36 0.13 − 0.13	O9–B3	III
22	9.34	2.08	0.76	2.24	0.84	1.22 ± 0.03	−5.28	5.16 0.38 − 0.39	4.36 0.13 − 0.13	O9–B3	III
23	9.44	1.21	0.39	1.34	0.42	0.71 ± 0.02	−4.67	5.03 0.30 − 0.30	4.39 0.11 − 0.11	O9–B3	III
24	10.31	1.37	0.44	1.50	0.47	0.79 ± 0.02	−3.89	4.72 0.30 − 0.30	4.39 0.11 − 0.11	O9–B3	V
25	10.82	1.51	0.52	1.64	0.55	0.88 ± 0.04	−3.46	4.55 0.30 − 0.30	4.39 0.11 − 0.11	O9–B3	V
26	10.84	2.80	0.90	2.93	0.93	1.55 ± 0.02	−4.12	4.81 0.30 − 0.30	4.39 0.11 − 0.11	O9–B3	V

Download table as:  ASCIITypeset image

3.2.2. Color–Magnitude Diagram and Cluster Reddening

In the top panels of Figure 5, we display a (J − K_s) versus K_s diagram of 2MASS data points within the cluster radius of 35 as well as a diagram of a comparison field. The same CMD with deeper SofI data, covering the central 49 × 49, is shown in the bottom panel of Figure 5.

Zoom In Zoom Out Reset image size

Several distinct populations of stars are seen in the CMD. There is a sequence of stars at J − K_s = 0.5 mag, which we attribute to field stars in the closer Sagittarius–Carina spiral arm. There is a broad sequence with J − K_s from 3 to 6 mag, which is populated mainly by field giant stars in the background of the cluster. Three candidate RSGs were detected from this sequence. Finally, there is a sequence of bright stars with J − K_s ≃ 1.5 mag and with K_s from 3 to 14 mag (bottom panel of Figure 5). This sequence is seen only in the cluster field, and this is the brightest extension of the cluster sequence.

To confirm and isolate the cluster sequence, we performed a statistical decontamination. 2MASS data were used in order to have a comparison field. We counted the number of stars in a grid of 0.5 mag in (J − K_s) and of 1.0 mag in K_s. From each bin of the cluster CMD, we subtracted a number of stars equal to that of the corresponding field bin. First, we subtracted all contaminating giant stars (spectroscopically detected), then we continued with a random subtraction. The decontaminated diagram is displayed in the right panel of Figure 5, and different symbols indicate known spectral types. The cluster sequence appears populated by massive stars. An average interstellar extinction of $A_{{K}_{\rm s}}=0.83\pm 0.2$ mag (A_v = 9.1 mag) was estimated by matching the colors of the observed cluster sequence with a theoretical isochrone of solar composition from the Geneva group (Lejeune & Schaerer 2001) and by assuming a power-law extinction curve A_λ ∝λ^−1.9 (Messineo et al. 2005). This measurement is independent of age since the isochrones are almost vertical sequences in this plane. An isochrone of 4.5 Myr and solar metallicity, taken from the non-rotating models of the Geneva group, is overplotted on the diagram (Lejeune & Schaerer 2001).

In the following, we analyze the properties (colors, magnitudes, and luminosities) of massive stars in the Cl 1813-178 cluster, which we list in Table 2. A kinematic distance of 4.8 kpc is used (see section 4.1). For each star, we assumed an intrinsic color consistent with its stellar spectral type (Koornneef 1983; Martins & Plez 2006; Wegner 1994); we used the extinction law by Messineo et al. (2005) and the kinematic distance, and we estimated values of $A_{{K}_{\rm s}}$ and M_K (see Table 2). The M_K values were transformed into bolometric magnitudes, Mbol, by adding the bolometric corrections. The adopted effective temperatures and bolometric corrections per spectral type are given in the Appendix.

4.1. Red Supergiants

Star 1 is the brightest star within the cluster area, K_s = 3.8 mag, and is consistent with being a K2–K5 I cluster member (Paper I). From the radial velocity (V_LSR = 62 ± 4) of star 1 (Paper I) and the Galactic rotation curve by Reid et al. (2009), we obtained a revised near kinematic distance of 4.8^+0.25 _{− 0.28} kpc. With intrinsic colors from Koornneef (1983) and bolometric correction, BC_K, taken from Levesque et al. (2005), we estimated a luminosity L_*=(9 ± 7) × 10⁴ L_☉.

Stars 32, 38, and 39 have EW(CO)s larger than the typical values for red giants. Stars 32 and 38 were observed with NIRSPEC, and their K-band spectral coverage allows us to study both the CO band heads and the shape of the continuum. The large values of EW(CO)s and the absence of water absorption suggest a supergiant luminosity class for stars 32 and 38 (Comerón et al. 2004). Star 39 was observed with UIST on UKIRT, with the K-long filter only, and we do not have information on the shape of its continuum. We derived spectral types M2.5I, M3.5I, and M1I, respectively, by assuming a supergiant luminosity class for all three stars.

Stars 32, 38, and 39 are among the brightest stars with red J − K_s (from 3 to 9 mag), much redder than the cluster sequence as shown in Figures 5 and 6. By assuming intrinsic H − K_s and J − K_s colors for a given spectral type (see the Appendix), we obtained $A_{{K}_{\rm s}}$ = 2.3 mag, 3.6 mag, and 2.3 mag, respectively. These values of interstellar extinction are the combined contribution of interstellar and possible circumstellar extinction (Messineo et al. 2005). By analyzing the distribution of detected stars in the K_s − [8.0] versus J − K_s plane (Figure 6), star 38 appears to show an excess in K_s − [8.0], which is suggestive of a circumstellar envelope. Stars 32 and 39 have K_s − [8.0] and J − K_s colors in agreement within errors with those of naked late-type stars (Messineo et al. 2004, 2005). We corrected for the total extinction by using the intrinsic colors for given spectral types and the extinction law by Messineo et al. (2005). We then artificially reddened the stars to emulate the average cluster condition. By assuming that these stars were moved along the reddening vector to the average cluster extinction ($A_{{K}_{\rm s}}$ = 0.8 mag), we obtained K_s = 6.7 mag, 5.8 mag, and 7.2 mag, respectively.

Zoom In Zoom Out Reset image size

If these RSGs were at the same distance as the stellar cluster, they should be the brightest stars at near-infrared wavelengths, and their K_s magnitudes would increase with later spectral types. Their spectral types are later than the bright (K_s = 3.8 mag) K2I star, and their de-reddened magnitudes are similar to those of the brightest OB stars (K_s ≃ 7 mag). The spectral types and magnitudes of these three RSGs are not consistent with cluster membership; they are too faint to be RSGs and part of the cluster. This reasoning is valid also for stars with moderate mass loss, e.g., no. 38, because bolometric corrections BC_K are almost constant for colors K_s − [8.0] < 3 mag (chapter 5, Figures A1 and A2, Messineo 2004).

We estimated a density of non-member RSGs equal to 0.16 in the central 35. These RSGs could represent an outburst farther away, perhaps related to the Galactic bar.

4.2. Wolf-Rayet Stars

4.3. Luminous Blue Variables

4.4. OB Stars

4.5. X-ray Emitters

Massive evolved stars in a young stellar cluster span a range of masses (see Figure 7), and therefore of luminosities. Spectrophotometric distances from OB supergiants are less accurate than those from dwarfs and giants. When considering the dwarfs and the M_K magnitudes for O9, B0, B1, and B2 dwarfs (Martins & Plez 2006; Humphreys & McElroy 1984; Koornneef 1983), we estimated distances of 3.8 ± 0.6 kpc, 3.2 ± 0.5 kpc, 2.6 ± 0.4 kpc, and 1.9 ± 0.3, respectively.

Zoom In Zoom Out Reset image size

From the photometric properties of the candidate giants (nos. 9, 21, 22, and 23) and M_K values for classes II and III from Humphreys & McElroy (1984) and Wegner (1994), we derived an average spectrophotometric distance of 3.0 ± 0.7 kpc for class III or 4.5 ± 0.6 for class II.

For the two WN7 stars, we adopted intrinsic magnitudes from Crowther et al. (2006a) and the extinction law by Messineo et al. (2005), and obtained a distance of 2.6 kpc for star 4 and 5.8 kpc for star 7. The average distance for the two WN7 stars is 4.2 ± 1.6 kpc.

The derived spectrophotometric distances are listed in Table 4. While the distance estimates from giants and WRs within errors are consistent with the kinematic distance, the distance estimates from dwarfs are only consistent with the kinematic distance if the dwarfs are all late-O stars. Higher-resolution spectra are needed to refine the spectral types.

Observations of radio hydrogen recombination lines of the W33 complex reveal two velocity components (Bieging et al. 1978; Goss et al. 1978). By using the radial velocity component at 35 km s⁻¹ and the rotation curve of Reid et al. (2009), we calculated a near kinematic distance of 3.53^+0.40 _{− 0.46} kpc, while we obtained a distance of 4.82^+0.25 _{− 0.28} kpc with the radial velocity component at 62 km s⁻¹. A radio monitoring program of methanol masers in the direction of the W33 complex is currently ongoing. It will yield parallactic distances of the masers (e.g. Reid et al. 2009; Brunthaler et al. 2011).

The Cl 1813-178 cluster contains a large number of evolved massive stars (BSGs, a cLBV, two WN7 stars, and one RSG star), which are listed in Table 1. We plot the inferred stellar luminosities versus stellar effective temperatures in Figure 8, together with theoretical stellar models from the Geneva group (Meynet & Maeder 2000). By comparison with the models, we estimated initial masses from 20 to 100 M_☉.

Zoom In Zoom Out Reset image size

Models predict a large span of stellar masses in post-main-sequence phase, e.g., a population of 4.0 Myr would have a mass of 25–35 M_☉ at a turn-off (TO), but it would still contain stars of about 50 M_☉, or even 100 if rotating models are considered. This is well illustrated in Figure 7, where we show the predicted maximum initial masses and TO masses as a function of cluster age. We used non-rotating models by Schaller et al. (1992) and Schaerer et al. (1993), models with high mass loss by Meynet et al. (1994), and the rotating models by Meynet & Maeder (2000). Metallicity ranges from solar to twice solar. For the same age, different models predict variations in the initial masses of up to 50%.

The mixture of evolved stars indicates progenitor masses larger than 20 M_☉, and more likely of 25–60 M_☉, with a few exceptions.

The Of stars are among the most luminous stars. The luminosity of the O8–O9If star suggests an extremely massive star (100 M_☉; Meynet & Maeder 2003). The O6–O7If star has an estimated L_*=(8.1 ± 2.2) × 10⁵ L_☉, which is predicted for 30–70 M_☉.

WR stars of WN7 type have been found only in stellar clusters with masses at the TO larger than 35–40 M_☉ (Massey et al. 2001; Clark et al. 2005a). The luminosities of the WR stars in the Cl 1813-178 cluster (L_*=(5.2  ±  2.3) × 10⁵ and (4.5  ±  1.9) × 10⁵ L_☉) are similar to those inferred for WN7 stars in the Westerlund 1 cluster (L_* from 3.1 × 10⁵ to 5.0 × 10⁵ L_☉). The stellar luminosities suggest initial masses from 40 to 70 M_☉.

The cLBV and RSG have luminosities expected for less massive stars (∼20 M_☉). However, the 2MASS magnitudes of the RSG star have errors of about 0.3 mag, due to saturation, and the resulting luminosity could be underestimated.

Models of young simple stellar populations by Meynet & Maeder (2003) predict that stars with masses between 9 and 25–35 M_☉ have an RSG phase. Single WR stars have initial masses greater than 26–30 M_☉, while binary WR stars have initial masses greater than 20–25 M_☉ (Eldridge et al. 2008). By assuming coevality between the RSG and the WR stars, the Cl 1813-178 cluster would be between 4 and 6 Myr old. A cluster with stars exceeding 100 M_☉ (like the O8–O9If) would require an age of 3–4 Myr, and a TO at 35 M_☉. Therefore, the simultaneous presence of RSGs, WRs, and Of stars would further narrow the possible age range to 4–4.5 Myr. However, either the luminosities of the cLBV, of the RSG star, and/or of the late-B supergiant are underestimated by ∼0.7 dex, or the "cluster" is more of an "association" with some degree of non-coevality.

The cLBV star appears to be rather faint. The estimated luminosity of 2.4 × 10⁵ L_☉ is similar to that of the HD168607 and HD316285 LBVs (Clark et al. 2005a). Other known LBVs in clusters are typically among the most luminous members, e.g., qF362 and the pistol star in Quintuplet (e.g., Mauerhan et al. 2010). In Westerlund 1, the W243 LBV has an initial mass of about 40 M_☉, which is consistently similar to the masses of the cluster WRs (40–50 M_☉), and larger than the estimated mass at the TO (30 M_☉) (Ritchie et al. 2009). The cLBV in Cl1813-178 has a mass smaller than that of the two detected WR stars. Its mass is consistent with the mass of the RSG and would require a mass at the TO of about 20 M_☉ and an age of 5–7 Myr. Further observations are recommended to explore the degree of coevality in Cl1813-178. Some non-coevality would easily explain the discrepant masses. A population with an age of 4.0–4.5 Myr, and a spread in age of 1 Myr, could explain the observed range of stellar masses.

The stellar mass at the TO for a population with an age of 4.0–4.5 Myr is likely between 25 and 35 M_☉. This is consistent with the observations of dwarfs and giants. Three OB dwarfs with K_s < 10.0 mag were detected with O9–B3 spectral types, while the giants with O7–B3 types have K_s = 9.3–9.5 mag. An O7V star (∼32 M_☉) at an average extinction of $A_{{K}_{\rm s}}$ = 0.8 mag and a distance of 4.8 kpc is expected to have a K_s ≃ 10.4 mag (Martins & Plez 2006), while an O9V (∼22 M_☉) star has a K_s = 10.9 mag.

A further spectroscopic survey of fainter stars is needed to sample the TO region. Moreover, high-resolution spectra would allow for a more precise spectral classification. By narrowing the errors shown in Figure 8, and increasing the sample, we will be able to better constrain the age and verify the degree of coevality.

By adding the masses of the spectroscopically identified massive stars, we estimated a minimum cluster mass of 990 M_☉. We considered all 34 stars with J − K_s between 1 and 3 mag and K_s < 10 mag. By assuming that they all have masses greater than 35 M_☉, and using a Salpeter mass function integrated from 0.8 to 120 M_☉, we estimated a cluster mass of 13, 000 ± 3000 M_☉, where the error is the Poisson error of the number of massive stars. By assuming that these 34 stars have masses greater than 25 M_☉, we obtained a cluster mass of 8700 ± 2000 M_☉. The two calculations take into account the uncertainties of the mass at the TO (25–35 M_☉), predicted for a coeval population with an age of 4–4.5 Myr.

The Cl 1813-178 cluster is, therefore, a new addition to the list of 13 known young massive clusters (⩾10⁴ M_☉) in the Milky Way (Messineo et al. 2009; Negueruela et al. 2010).

A 24 μm image from MIPSGAL, the survey of the inner Galactic plane using the Multiband Infrared Photometer for Spitzer aboard the Spitzer Space Telescope (Carey et al. 2009), of the whole W33 region is shown in Figure 9, together with contours of radio continuum emission at 20 cm (White et al. 2005). The complex extends over an area of roughly 25' × 20' (Bieging et al. 1978). Radio observations show that W33 is made up of a number of discrete sources. Some of these radio sources have been classified as candidate SNRs by Brogan et al. (2006) and Helfand et al. (2006) on the basis of morphology and spectral indices, with MAGPIS data (White et al. 2005). See Table 6.

Zoom In Zoom Out Reset image size

The Cl 1813-178 cluster appears located on the edge of the W33 complex. The radial velocity of the K2I star in the Cl 1813-178 cluster is 62 ± 4 km s⁻¹ (Paper I) and well agrees with the high-velocity gas component of W33 (Dame et al. 2001). The spatial coincidence of the Cl 1813-178 cluster with the SNR G12.82-0.02 and G12.72-0.00 has already been reported in Paper I and supports the association of the cluster with the complex. The filamentary shape of the Cl 1813-178 stellar cluster and its location on the edge of W33 suggest a secondary episode of star formation, perhaps triggered by an expanding shell.

We searched for other possible candidate clusters and associations with SNRs in the direction of the W33 complex. A number of candidate stellar clusters have been identified in GLIMPSE and 2MASS images in the direction of the W33 region by Mercer et al. (2005) and Bica et al. (2003), which we list in Table 5. The W33 MYSO (Davies et al. 2010) is projected into SNR6 (G13.1875+0.0389; Helfand et al. 2006). The spatial coincidence of the SNR and the MYSO suggests a physical association. The W33A MYSO could be an episode of triggered star formation induced by a supernova explosion. Mercer1 candidate cluster is the object number 1 in the list of Mercer et al. (2005). It was identified as a stellar overdensity in the GLIMPSE catalog with an automatic algorithm. It is located at the center of the molecular complex, and it appears as a spread overdensity. Two other candidates are reported in literature in the surrounding of the W33 complex, but without a clear connection with the W33 complex. The BDS2003-115 candidate cluster is about 20' North of the main W33 complex and is associated with SNR5 (G12.83-0.02; Helfand et al. 2006). The BDS2003-7 candidate cluster appears as a small group of stars without associated radio emission (Bica et al. 2003).

We visually inspected the 2MASS images and located two other clumps of stars (cl1 and cl2). The cl1 candidate appears as a group of point sources on bright nebular emission in the 2MASS K_s image. Inspection of the GLIMPSE and MAGPIS images reveals the presence of an H ii region, suggesting the presence of massive stars (see Table 5 and Figures 10 and 11). The cl2 candidate is another small concentration of bright stars (K_s = 8–10 mag) in another H ii region. Nothing is reported in previous literature about both, cl1 and cl2, clumps. In addition, we searched for stellar overdensities in the direction of the W33 complex using both 2MASS and GLIMPSE star counts. Detections are hampered by strong variations of the background level, which are due to variations of interstellar extinction and nebular emission. A spectroscopic and photometric follow-up study of these regions with SINFONI and UKIDSS data is ongoing, in order to confirm the presence of massive stars.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Near-infrared spectroscopic follow-up observations are needed to characterize these sources and to confirm their association with the W33 complex. However, the associations of Cl 1813-178, cl1, and cl2 with H ii regions and/or SNRs suggest that these are other condensations of massive stars in the W33 complex.

A near-infrared spectroscopic survey of the brightest stars in the direction of the Cl 1813-178 cluster is presented. Among the 61 observed stars, 25 massive stars were detected. Two WR stars of type WN7, a cLBV, and 21 OB stars were identified. Among the OB stars, an O8–O9If star and an O6–O7If star were discovered. Eight of these evolved stars also have X-ray emission, as detected by the Chandra and XMM satellites. The hardness of the X-ray emission from the two WN7 stars strongly suggests binary systems.

A spectrophotometric analysis of the OB stars reveals 14 supergiants, 4 giants, and 3 dwarfs. From the giants, dwarfs, and WRs, we derived average spectrophotometric distances of 3.8  ±  1.0 kpc, 2.9  ±  0.8 kpc, and 4.2  ±  1.6 kpc. The distances from giants and WRs are in agreement with the kinematic distance. The distance estimates from dwarfs are only consistent with the kinematic distance if the dwarfs are late-O stars.

The mixture of evolved massive stars is reminiscent of other Galactic young massive clusters, such as Westerlund 1, Quintuplet, Galactic center, and Cl 1806-20. We estimated stellar luminosities, and therefore masses, by comparing the luminosities with evolutionary tracks from the Geneva group. By assuming a Salpeter mass function, we obtained a cluster mass of (1.0 ± 0.2) × 10⁴ M_☉. A likely cluster age of 4–4.5 Myr is derived; however, a spread in age of about 1 Myr cannot be excluded. In order to better constrain the degree of coevality, further spectroscopic observations are required.

The Cl 1813-178 cluster is located on the Western edge of the W33 complex. We have located several other candidate stellar clusters that could belong to the same complex.

This work was partially funded by the ERC Advanced Investigator Grant GLOSTAR (247078). The material in this work is partly supported by NASA under award NNG 05-GC37G, through the Long-Term Space Astrophysics program. Part of this research was performed in the Rochester Imaging Detector Laboratory with support from a NYSTAR Faculty Development Program grant. We also acknowledge support from the US National Science Foundation under grant AST-0709479 (R.M.R.). R.P.K. acknowledges support by the Alexander-von-Humboldt Foundation. F.N. acknowledges financial support from the Spanish Ministerio de Ciencia e Innovacion under the project AYA2008-06166- C03-02. 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 NSF. This research has made use of Spitzer's GLIMPSE survey data, the Simbad and Vizier database. We thank the staff of the Joint Astronomy Centre for their great support during the UKIRT run. We thank the staff astronomers of the European Southern Observatory for their excellent support during the SofI observations. We thank Dr. Thomas Dame for kindly providing us with a CO map of the W33 complex. A special thank to Dr. Andreas Brunthaler for discussion on parallactic distances to the W33 complex using methanol masers. We acknowledge helpful comments and suggestions by an anonymous referee.

The EW(CO)s can be used for classification in subclasses of late-type stars. There is a linear correlation between stellar effective temperatures and EW(CO)s of the CO bands. Furthermore, since giants and supergiants follow two different relations, information on the luminosity class can also be obtained (see, e.g., Figer et al. 2006; Davies et al. 2007).

Among the observed 60 stars, 37 are found to be late-type stars. For 36 of them, spectra were taken with the K-long grism, covering the region of the CO band head at 2.29 μm. CO bands at 2.29 μm in absorption are detected in all spectra. Star 55 was observed only with the K-short grism; however, the presence of Na lines (at 2.2075 and 2.2077 μm) and Mg i at 2.11 μm suggests a spectral type later than G0.

The equivalent widths of 32 of these late-type stars are compatible with that of giant stars with spectral type from K0III to M7III (Table 7). For star 58 we do not report any spectral type, since its spectrum has a poor signal to noise.

Star 32, 38, and 39 have EW(CO)s typical of RSG stars, with spectral types M2.5, M3.5, and M1, respectively. However, their photometric properties indicate that they are unrelated to the stellar cluster. Star 1 is consistent with being a K2–K5I cluster member (see Paper I).

Young massive clusters with ages from 4 to 30 Myr may contain yellow supergiants (YSG) stars, e.g., the Westerlund 1 (Clark et al. 2005a) and RSGC1 clusters (Figer et al. 2006). YSG are rare F- or G-type supergiants in transition toward the RSG phase or, back from the RSG locus, evolving blueward. In a coeval population, YSGs or RSGs are expected to be brighter in K_s than early-type massive stars. In the Cl 1813-178 cluster, all detected stars with CO band head at 2.29 μm appear fainter than the brightest OB stars, and the gap between the brightest early-type star (K_s = 6.75 mag) and the K2I star (K_s = 3.79 mag) is devoid of stars. Furthermore, the CO band heads of the five late-type stars with K_s between 7 and 8 mag indicate late-M giants (M2–M5), or late-K supergiants (K2–K4). Their EW(CO)s are not consistent with being F or G supergiants.

In order to estimate stellar luminosities, estimates of effective temperatures and bolometric corrections as a function of spectral type need to be known. Since an homogeneous calibration extending from O stars down to early A stars is missing, we summarize all the adopted values in Tables 8–11.

For O-type stars, we used the T_eff from Martins et al. (2005). For early-B supergiants, we adopted the T_eff values given by Crowther et al. (2006b), while, for late-B and A supergiants, those by Humphreys & McElroy (1984). For B giants, we adopted the T_eff by Humphreys & McElroy (1984). For B and A dwarfs, we used T_eff estimated by Humphreys & McElroy (1984) and Johnson (1966).

For O-type stars, we used the bolometric corrections in K band (BC_K) by Martins & Plez (2006). For early-B-type stars, we used those provided by Bibby et al. (2008). We estimated the BC_Ks of late-B supergiants and giants by assuming bolometric corrections in the V band (BC_V) from Humphreys & McElroy (1984), and intrinsic V − K colors from Koornneef (1983) and Wegner (1994). For dwarf stars, a set of homogeneous BC_K was obtained by interpolating an isochrone of 0.5 Myr and solar metallicity from Lejeune & Schaerer (2001) at the assumed effective temperatures (Humphreys & McElroy 1984; Johnson 1966).