4. Integrated spectroscopy

Integrated spectra of star clusters allow one to determine their basic parameters such as reddening, age and metallicity. Bica & Alloin (1986, 1987) have studied integrated spectra in the visible and near-infrared ranges of Galactic open and globular clusters, as well as Magellanic Cloud clusters. They investigated the behaviour of metallic and Balmer-line equivalent widths (W) as well as the continuum energy distribution from 3700 to 10000 Å. Once the age and metallicity are established from W measurements and comparisons with templates, the reddening can be derived by matching the observed cluster spectrum to that of the template which most resembles it.

4.1. Westerlund1

  
Figure 6: Integrated spectrum of Westerlund1 corrected for E(B-V)=4.4 (top), the template YA (middle) and the template YB (bottom)

We show in Fig. 6 (click here) the near-IR integrated spectrum of Westerlund1, corrected for E(B-V)=4.4 (see discussion below). Given the extremely high absorption for Westerlund1, its spectrum has no measurable flux for  Å. The prominent could arise either from stellar sources in clusters younger than 35 Myr (Bica et al. 1990), or extended gas (e.g. HII region). As already pointed out by W87 star No. 9 is a Be star with pronounced emission lines, and star No. 31 is possibly another Be. W87 also detected diffuse H emission in the cluster central region. We have investigated in our integrated spectrum the intensity variation of H across the cluster profile and found that the highest values occur within a small region approximately 5 pixel wide, in the central part of the cluster. The emission is fainter in the remaining zones of the central region as well as in the wings of the cluster profile. The pronounced inner H emission could be assigned to the Be star located in the central part of the cluster (see Fig. 1 (click here) in W87), while the weak emission could be due to extended nebular gas. Diffuse emission occurs in HII regions as old as 5 Myr, or even older clusters which have residual gas emission from photoionization and/or supernova remnants, as can be seen in the templates of Santos et al. (1995).

The presence of in emission, when caused by stellar sources such as Be stars, is a remarkable feature in the YB and YC templates of Bica et al. (1990), whose age ranges are 7-12 Myr and 12-35 Myr, respectively. The YB template reflects the full development of the red supergiant (RSG) phase, and consequently the near-IR spectrum presents strong TiO bands and CaII triplet lines as well as a flat continuum. We show in Fig. 6 (click here) the YB template, where the strong emission and the effects of the RSGs can be seen. The slightly older YC template (see Bica et al. 1990) is bluer, since RSGs do not contribute significantly.

The M2I star (No. 26 in W87) near the center of Westerlund1 produces moderate TiO bands around  Å and  Å in our reddening-corrected integrated spectrum (Fig. 6 (click here)). Notice also that the CaII triplet is well developed in the cluster, which sets a lower age limit of 5 Myr; the template YA in Bica et al. (1990), corresponding to this evolutionary stage (5 Myr age 7 Myr), is also shown in Fig. 6 (click here).

The above considerations about occurrence of diffuse and stellar emissions, RSG, and developed CaII triplet indicate that the most probable age for the cluster is  Myr, in very good agreement with the estimation of 7-9 Myr by W87.

A prominent absorption feature in the range  Å is present. Sanner et al. (1978) have studied several diffuse interstellar features in the near-IR, including four within this range. Based on integrated spectra of globular clusters, Armandroff & Zinn (1988) have studied an interstellar feature at a longer wavelength ( Å), clearly detectable for E(B-V) > 1.2. We suggest that the absorption feature found in the  Åregion is of interstellar origin, because it is also prominent in globular clusters more reddened than E(B-V)=2.0 (Bica et al. 1997). Sanner et al. (1978) have only detected a few interstellar lines in this region, probably due to the fact that their stars are not enough reddened, and also because their spectra were not corrected for the telluric A band.

The reddening of Westerlund1 was obtained by varying E(B-V) following Seaton's (1979) law to match the continuum of the YA template, as shown in Fig. 6 (click here). Notice the flux excess of the cluster spectrum with respect to that of the YA template for  Å, which we attribute to the contribution of the M2I star. An attempt to match the RSG-dominated continuum of the YB template yields E(B-V)=4.0. We favour the higher reddening value because the cluster spectrum resembles most that of the YA template. This spectroscopically derived reddening is consistent with that estimated from the CMD-fitting (Sect. 3). We finally adopted .

We measured W of the CaII triplet lines using as continuum points the regions around  Å and  Å (Bica & Alloin 1987). Table 2 (click here) presents the window limits and the Ws. The error assigned to each measurement was calculated by considering local high and low continuum tracings in order to take into account the spectral noise. Although the correlations between W of CaII triplet lines with metallicity are basically single-valued for integrated spectra of star clusters (Bica & Alloin 1987), there are sources of scatter which affect the determination. These uncertainties are related to the different stellar components contributing to the integrated light rather than to the errors arising from the quality of the spectrum itself. The contamination by TiO bands and Paschen lines affects the CaII windows. TiO may reinforce the metallicity dependence of the CaII triplet (especially for the 8498 and 8542 Å lines), while Paschen lines introduce a systematic increase for blue clusters whatever their metallicity. The Paschen absorptions affect more the CaII 8662 Å line.

 

Window Limits (Å) Westerlund1

CaII+TiO 8476-8520

CaII+TiO 8520-8564

CaII+P13 8564-8700

To estimate the metallicity of Westerlund1 we compared the W(CaII) values with those measured by Santos & Bica (1993) and Bica & Alloin (1987) for three open clusters with independent metallicity determinations from individual stars. The sum of the equivalent widths for the three CaII lines in Westerlund1 is . This value compares well with the average () of those for NGC 4755, NGC 6067, and NGC 6705, whose [Fe/H] values are 0.0 (Brown et al. 1986), -0.1 (Piatti et al. 1995) and +0.21 (Thogersen et al. 1993), respectively. We thus conclude that Westerlund1 has nearly solar metal content.

4.2. Westerlund2

The observed CCD integrated spectrum of Westerlund2 (Fig. 2 (click here)b) was obtained by scanning the slit  25 across the central region of the cluster in the north-south direction. The most relevant features in this spectrum are the nebular emission lines [OIII]4959, 5007 Å, H, and [SIII]9068 Å, and a continuum with a pronounced slope denoting important reddening. These nebular emission lines are expected since Westerlund2 is the core of the HII region RCW 49 (Belloni & Mereghetti 1994). We adopted as template spectrum that of the cluster NGC 3603, which is also the core of an HII region (Santos & Bica 1993 and references therein). We assume as the age of Westerlund2 that of NGC 3603, i.e., 2-3 Myr (Melnick et al. 1989; Santos & Bica 1993). This value is consistent with the presence of a WN7 star (MSP91) located away from the core of Westerlund2, not included in our integrated spectrum. Furthermore, MSP91 have spectroscopically recognized several O7V stars, which confirms the adopted age range.

  
Figure 7: Integrated spectrum of Westerlund2 corrected for E(B-V)=1.4 (top) compared to the NGC 3603 corrected for foreground (bottom)

We show in Fig. 7 (click here) the spectrum of Westerlund2 with a reddening correction E(B-V)=1.4 to match that of NGC 3603, which in turn is corrected for its foreground reddening , Santos & Bica 1993). Notice that the core of NGC 3603 presents several WR stars (Moffat & Niemela 1984), which are responsible for the observed WN features in the integrated spectrum. The CaII triplet has not yet developed in such young cluster, and consequently we have not measured equivalent widths.

Our spectroscopic reddening estimation should be considered as a foreground value. From 72 member stars of Westerlund2 in MSP91's Table 2 (click here), we derived a mean cluster reddening of . The difference between this value and the spectroscopic one suggests an internal reddening of in such a young object. For comparison, Santos & Bica (1993) estimated an internal reddening of in NGC 3603.

We have reexaminated MSP91's estimation of the cluster distance (7.9 kpc) by considering four cluster core O7V stars (Nos. 167, 183, 188 and 203). We derived from their V and M_v magnitudes a mean apparent distance modulus of . Subsequently, using the information available in MSP91, the mean E(B-V) colour excess for these stars was determined in two different ways. First, we compared the observed CCD (B-V) index with the intrinsic (B-V)₀ colour according to the MK type. The second method consists simply in averaging the E(B-V) values calculated by MSP91 from their UBV photometry. These procedures yield and , respectively, which in turn lead to a cluster distance of 5.5 and 5.9 kpc. The considerably larger distance obtained by MSP91 certainly comes from uncertainties in their ZAMS-fitting, owing to the almost vertical position of the observed CMD-sequence.