Abstract
It is well known that clusters of massive stars are influenced by the presence of strong winds, that they are sources of diffuse X-rays from shocked gas, and that this gas can be vented into the surrounding region or the halo through the champagne effect. However, the details of how these different environments interact and evolve are far from complete. This paper attributes the broad C ivλλ1500 absorption features (extending to −1900 km s−1) that are seen in the spectra of main sequence B stars in NGC 6231 to gas in the cluster environment and not the B stars themselves. It is shown that the presence of a WC star, WR 79, in the cluster makes this gas detectable because its wind enriches the cluster gas with carbon. Given the available data, it is not clear whether the absorbing gas is simply the far wind of WR 79 or a collective cluster wind enriched by carbon from the wind of WR 79. If it is simply due to the wind, then this wind must flow, unimpeded for more than 2 pc, suggesting that the inner region of the cluster is nearly devoid of obstructing material. If it is actually a collective wind from the cluster, then we could be witnessing an important stage of galactic feedback. In either case, the observations provide a unique and significant piece to the puzzle of how massive, open clusters evolve.
1 INTRODUCTION
Wolf Rayet stars have powerful, highly ionized stellar winds, and the signature of these winds defines their spectra class. Their winds have been imaged on scales of hundreds of AU (e.g. Tuthill et al. 2008) and their power is evident from the wind-blown bubbles they can produce, which can be several pc in diameter (e.g. Cazzolato & Pineault 2000; Nazé et al. 2003). In a cluster, these winds may carve out large cavities around the star or may interact with the winds of other cluster members. This paper shows that because WR 79 (HD 152270) resides in the young, compact open cluster NGC 6231, it is possible to use the main sequence B stars (BVs thereafter) in the cluster to probe the influence of its carbon rich, massive wind at distances more than a parsec from the star. This wind is identified as the source of strong, high-speed C ivλλ1550 absorption seen in IUE spectra of BVs in the core of NGC 6231. The following sections provide: an overview of NGC 6231 and WR 79, including why a previous explanation for the high-speed absorption is no longer tenable (Section 2); an explanation of why the wind of WR 79 can be observed in NGC 6231 (Section 3); a derivation of the absorption profiles expected for BVs embedded in a wind (Section 4); fits of the model profiles to the two BVs in NGC 6231 that have high-dispersion IUE spectra (Section 5), and a discussion of the results (Section 6).
2 NGC 6231
NGC 6231 is a compact open cluster at the core of Sco OB1, at a distance of 1.64 kpc (Balona & Laney 1995). The core of the cluster contains numerous BVs, 10 O stars, a B supergiant, and the WC+O star WR 79. Its stellar content has been well studied (e.g. Schild, Hiltner & Sanduleak 1969; Baume, Vázquez & Feinstein 1999), and it is a site of ongoing star formation (Sana et al. 2007). Polarimetric observations by Feinstein et al. (2003) suggest there has been a supernova explosion in the cluster, but they noted that their observations are also consistent with an expanding bubble of the sort attributed to the WC stars WR 101 and WR 113 by Cappa, Goss & Pineault (2002). Red DSS2 images of the region show that NGC 6231 is within a large, ∼1.7° × 3.5° (∼50 pc × 100 pc), cavity centred near ζ1 Sco, and gives the impression that the cluster might be venting gas from its H II region. However, such gas has not been detected. WR 79 is one of the most luminous objects in the cluster. It is a colliding wind binary system whose wind was modelled by Hill et al. (2000), who showed that the WC component dominates the overall wind flow.
Some years ago, Massa, Savage & Cassinelli (1984) noticed that the low-dispersion IUE spectra of several BVs in NGC 6231 had uniquely peculiar spectra, unlike any others seen before or since. The low-resolution IUE spectra of these stars have abnormally strong C ivλλ1550 absorption. They also obtained an IUE high-dispersion spectrum of the B1 V star, CPD−41o7719. Later, another high-dispersion spectrum of the B0.5 V star, CPD−41o7724, was also obtained. The spectra of these stars are compared to the normal B0.5 IV star, λ Lep, in Fig. 1. Because λ Lep was used as a flux standard by the IUE project, the spectrum shown is a mean of 78 spectra. All of the spectra were binned to 0.15 Å. Because λ Lep is a slow rotator, vsin i = 25 km s−1 (Abt, Levato & Grosso 2002), it was necessary to broaden its spectrum to obtain good matches to the NGC 6231 stars. The λ Lep spectrum was convolved with a rotational broadening function (e.g. Gray 1976) of 50 km s−1 to match CPD−41o7719 and 90 km s−1 to match CPD−41o7724. The spectra were also shifted in radial velocity to align the Si iii 1300 Å triplets. The agreement of the important Si iii 1300 Å triplets and the 1417 Å singlet, which are sensitive to the stellar parameters (Massa 1989), shows that λ Lep is a good spectral match for both stars, given the relatively poor S/N of the NGC 6231 spectra.
Comparisons of the NGC 6231 BV spectra (red) with that of the normal B0.5 IV star λ Lep (black). Top three panels: λ Lep and CPD−41o7719 (B1 V). Bottom three panels: λ Lep and CPD−41o7724 (B0.5 V). All spectra are binned to 0.15 Å. Aside from the stronger interstellar spectrum and the pronounced high-velocity C iv absorption in the NGC 6231 stars, the agreement with λ Lep is very good.
The one obvious difference between the NGC 6231 BVs and λ Lep is the high-velocity C iv absorption trough in the NGC 6231 stars, which extends to ∼− 2100 km s−1. In addition, there is no hint of excess emission near v ∼ 0, as one would expect from a normal wind. These profiles are very different from any seen in OB stars of any kind.1 Massa et al. (1984) interpreted the peculiar absorption in terms of an abnormal wind. They were able to achieve a good fit to the high-resolution profile but, in doing so, it was necessary to stretch the bounds of credibility in order to model. Modelling profiles that show very strong high-speed blue absorption with virtually no red emission with a spherically symmetric wind require a wind that accelerates extremely rapidly, so that nearly all of the wind material capable of producing low-speed emission is hidden behind the star. Further, this explanation is only plausible if one assumes that carbon is overabundant in the BVs. However, later analysis showed that the NGC 6231 BVs have fairly normal C abundance (e.g. Kilian, Montenbruck & Nissen 1994).
This study included four BVs Massa et al. (1984) identified as peculiar (including CPD−41o7719), and all were found to have normal C abundances. In addition, over the years since Massa et al. (1984), no other examples of BVs with similar profiles have emerged (see, Walborn, Parker & Nichols 1995). As a result, an alternative explanation is needed.
This paper proposes that the high-speed absorption in the BVs actually results because they are immersed in a radially expanding flow originating elsewhere. Fig. 2 shows the stars in the central region of NGC 6231. These are the stars listed in table 4 of Baume et al. (1999) and CPD−41o7719, and includes 10 O stars, a B supergiant, and WR 79. The symbol sizes are proportional to the V magnitudes and the coordinate system is in arcmin centred on WR 79. The red points indicate the positions of BVs observed in low dispersion by IUE. This includes the BVs HD 326328, HD 326332, HD 326333, CPD−41o7715, and Cl* NGC 6231 SBL, which were observed after the data used by Massa et al. (1984). Of these, the spectra of CPD−41o7715 and Cl* NGC 6231 SBL are peculiar by the Massa et al. (1984) criteria. Filled red points are BVs whose low-dispersion spectra have peculiar C iv absorption and open red points are BVs with normal spectra. The stars labelled 1 and 2 are CPD−41o7719 and CPD−41o7724, respectively, which have been observed in high dispersion by IUE. It is important to notice that the stars with peculiar spectra are those closest to WR 79.
Stars in the central region of NGC 6231. This includes the stars listed in table 4 of Baume et al. (1999) and CPD−41o7719. The size of 1 pc is shown for a distance of 1.64 kpc. The symbol sizes are proportional to V magnitudes and the coordinate system is centred on WR 79. Red points represent BVs observed by IUE. Filled red points are for stars whose low-dispersion spectra contain C iv peculiarities and open red points are for stars with normal spectra. The stars labelled 1 and 2 are CPD−41o7719 and CPD−41o7724, respectively, which were observed at high resolution with IUE.
It is emphasized that such peculiar absorption can only be observed in BVs for two reasons. First, they are strong UV sources at 1540–1550 Å, so the absorption can be detected. Secondly, unlike more luminous B stars and O stars, they do not have strong, high-speed absorption from their own winds, so it is possible to identify absorption due to a neighbouring source.
3 TOTAL COLUMN THROUGH THE WIND
This section develops a simple model to help understand why it is so unusual to observe the absorption due to a stellar or cluster wind in background stars at great distances from the wind source, and why it is possible to see such a wind in the case of NGC 6231.
To begin, consider the absorption caused by a spherically symmetric wind when viewed by a second star within the far wind (termed the embedded star), which has no wind (or a very weak one), so that it can be used as a probe of the strong wind. Fig. 3 shows the coordinate frame where the star with a massive wind is positioned at the origin and the embedded star is located at (x1, y1, z1) and the observer is at z = ∞. The position of the star is labelled by (p1, z1), where p1 is the impact parameter, defined as p = (x2 + y2)1/2, so that the radial distance between the source and the embedded star is r = (z2 + p2)1/2.
Schematic of a wind emerging from a star at the origin and a star embedded in the wind. The embedded star is located at (x1, y1, z1) and labelled by (p1, z1), where p = (x2 + y2)1/2 is the impact parameter. The dashed line points to the observer. It is clear that the range of velocities which absorb along the line of sight is dictated by the position of the star in the wind.
4 MODEL PROFILES
Fig. 4 shows examples of r(λ) for constant qi(vz). It demonstrates how the profiles respond to different values of v1 and vt. Note that for the wind of an isolated star, turbulent velocities ∼0.05–0.10v∞ are typically invoked. Given the chaotic nature of the intracluster environment and the colliding winds of the WR 79 binary, even larger values are not unreasonable.
A set of model profiles of the C iv absorption expected for embedded stars. All of profiles have v∞ = 2100 km s−1 and the arrows indicate the rest wavelengths of the C iv doublet. The black profile has a turbulent velocity of 200 km s−1 and v1 = −1500 km s−1, implying that the embedded star is well in front of the wind source. The red profiles have a turbulent velocity of 100 km s−1 and v1 = −1000 km s−1(solid), 0 km s−1(dotted) and +2100 km s−1(dashed). The profiles with v1 = 0 and +2100 km s−1 show how broad, featureless absorption can be present and that even red absorption can exist in stars well beyond the wind source.
5 FITS TO THE DATA
In this section, the IUE high-resolution spectra of CPD−41o7719 and CPD−41o7724 are fit to the model developed in Section 4. The ionization was assumed constant for the fitting procedure, i.e. qi(vz) ≡ qi. This was necessary since its functional form is unknown and would be difficult to disentangle from the velocity parameters which determine the shape of the profile with the limited data available.
In addition to the absorption profile, a template is needed for the unabsorbed flux from the BV. As shown in Fig. 1, the B0.5 IV star λ Lep is a good match to both stars, and even the low-velocity portion of the C iv doublets are in reasonable agreement. Consequently, λ Lep was used as a template for both stars.
Although NGC 6231 has a large number of β Cep stars (e.g. Meingast, Handler & Shobbrook 2013), neither of the program stars have been identified as such. However, during 1996, HST GHRS G140L spectra were obtained for CPD−41o7719 during three consecutive orbits (spanning roughly 4 h). Fig. 5 compares the three GHRS spectra with the IUE high-dispersion spectrum degraded to the GHRS G140L spectral resolution of 135 km s−1. Variability is clearly present in C iv, and most likely Si iv as well. However, the level is much smaller than typically observed in β Cep stars. This may indicate that CPD−41o7719 is a very low amplitude β Cep, perhaps with such a small optical amplitude that it has escaped detection in ground based searches. In any case, Fig. 5 demonstrates that exact agreement with the low-velocity portion of C iv, cannot be expected.
Comparison of three GHRS G140L spectra of CPD−41o7719 (solid, dashed and dotted curves) taken on consecutive orbits and an IUE high-resolution spectrum binned to the GHRS G140L resolution (points).
Fig. 6 shows non-linear least-squares fit to CPD−41o7719 and CPD−41o7724 by the absorption model using λ Lep as the underlying photosphere. The fits used the IUE errors for the normalized fluxes and the points crossed out are contaminated by low-velocity interstellar C iv absorption along the line of sight. A single v∞ was used for both fits. For a given fixed value of v∞, the values of vt, v1, and u0 were determined by a Levenberg Marquardt non-linear least-squares routine. The best-fitting value of v∞ was determined by a simple grid search. In both cases, χ2 was a minimum for v∞ ≃ 1900 ± 100 km s−1. The parameters derived from the fits are listed in Table 1 along with their 1σ errors. Considering the data quality and the simplicity of the model, the fits are considered quite good.
Fits to CPD−41o7719 (left) and CPD−41o7724 (right) using the parameters listed in Table 1. The black spectra are the observed spectra and the crosses represent points excluded from the fits due to interstellar contamination. The dotted spectrum is λ Lep (which was used as a template for both stars) and the red spectra are the fits, which are the dotted spectrum multiplied by the model absorption profile.
Fit parameters.
| Star | Sp Ty | p1 | v∞ | vt | v1 | u0 | Wλ |
|---|---|---|---|---|---|---|---|
| arcmin | km s−1 | km s−1 | km s−1 | 10−6 M⊙ yr−1 pc−1 | Å | ||
| CPD−41o7719 | B1 V | 2.8 | 1900 ± 100 | 310 ± 83 | 1319 ± 130 | 0.32 ± 0.05 | 4.25 ± 0.57 |
| CPD−41o7724 | B0.5 V | 2.3 | 1900 ± 100 | 507 ± 105 | 1021 ± 78 | 0.31 ± 0.05 | 5.37 ± 0.58 |
| Star | Sp Ty | p1 | v∞ | vt | v1 | u0 | Wλ |
|---|---|---|---|---|---|---|---|
| arcmin | km s−1 | km s−1 | km s−1 | 10−6 M⊙ yr−1 pc−1 | Å | ||
| CPD−41o7719 | B1 V | 2.8 | 1900 ± 100 | 310 ± 83 | 1319 ± 130 | 0.32 ± 0.05 | 4.25 ± 0.57 |
| CPD−41o7724 | B0.5 V | 2.3 | 1900 ± 100 | 507 ± 105 | 1021 ± 78 | 0.31 ± 0.05 | 5.37 ± 0.58 |
Fit parameters.
| Star | Sp Ty | p1 | v∞ | vt | v1 | u0 | Wλ |
|---|---|---|---|---|---|---|---|
| arcmin | km s−1 | km s−1 | km s−1 | 10−6 M⊙ yr−1 pc−1 | Å | ||
| CPD−41o7719 | B1 V | 2.8 | 1900 ± 100 | 310 ± 83 | 1319 ± 130 | 0.32 ± 0.05 | 4.25 ± 0.57 |
| CPD−41o7724 | B0.5 V | 2.3 | 1900 ± 100 | 507 ± 105 | 1021 ± 78 | 0.31 ± 0.05 | 5.37 ± 0.58 |
| Star | Sp Ty | p1 | v∞ | vt | v1 | u0 | Wλ |
|---|---|---|---|---|---|---|---|
| arcmin | km s−1 | km s−1 | km s−1 | 10−6 M⊙ yr−1 pc−1 | Å | ||
| CPD−41o7719 | B1 V | 2.8 | 1900 ± 100 | 310 ± 83 | 1319 ± 130 | 0.32 ± 0.05 | 4.25 ± 0.57 |
| CPD−41o7724 | B0.5 V | 2.3 | 1900 ± 100 | 507 ± 105 | 1021 ± 78 | 0.31 ± 0.05 | 5.37 ± 0.58 |
The v1 parameter can be used to infer the z1 distance for each star. Assuming a spherical wind from WR 79, and letting α be the angle between the line of sight and the flow at z1, one has v1 = v∞cos α and p1 = z1tan α, so |$z_1 = p_1 \cot [\cos ^{-1}(v_1/v_\infty )]$|. Using d = 1.64 kpc to convert the Table 1 data from arcmin to parsecs, results in p1 = 1.3 and z1 = 2.20 pc for CPD−41o7719 and 1.1 and 2.23 pc for CPD−41o7724.
Table 1 also lists the resulting Wλ values and their errors, which were calculated using the standard propagation of errors equation with numerical derivatives for the free parameters of the fit. Using equation (3), and the values of z1 derived above, these can be rescaled to z1 = 0 and ppc = 1. The results are 7.66 and 7.61 Å for CPD−41o7719 and CPD−41o7724, respectively, which are close to the expectations given in Section 3 for qi|$\dot{M}$|−6 ≃ 10.
6 DISCUSSION
This discussion examines the evidence favouring the proposed interpretation of the C iv absorption seen in the BVs and describes how the model could be used to constrain the kinematics of the flow if more data were available.
To assess the validity of the model, both direct and indirect evidence is examined. Beginning with the direct information, I compare the fit parameters (v∞, vt, v1, u0) given in Table 1 to known quantities and processes relevant to WR stars and H ii regions. The best-fitting v∞ (1900 km s−1) differs by only 16 per cent from 2270 km s−1, the v∞ determined for WR 79 by Prinja, Barlow & Howarth (1990). The large errors for vt show that they are poorly constrained by the data. Nevertheless, values of 300 km s−1 are common for WC stars (e.g. Hillier 1989), so considering the large errors and the chaotic nature of the NGC 6231 environment due to the presence of several O stars, the derived values for vt are considered reasonable. While v1 is not directly measurable, it was shown in the previous section that it can be used to derive the distance from the star to the wind source. It was found that z1 = 2.20 and 2.23 pc for CPD−41o7719 and CPD−41o7724, respectively. These values are only about twice their impact parameters and therefore consistent with the observed size of the cluster. They also indicate that the two BVs are in front of WR 79. Combining the z1 and impact parameters gives the radial distance of each star from WR 79. These are 2.57 and 2.46 pc for CPD−41o7719 and CPD−41o7724, respectively.
Knowing that the BVs are 2 pc or more from WR 79, it is possible to estimate two important properties about the gas along their lines of sight. First, it takes the absorbing gas ∼1000 yr to travel from the stellar surface to the point where it begins to absorb. Secondly, the mean number density of the wind 2 pc from the star is 0.0032 M−6/(μrpcv3) ≃ 1.57 × 10−4 M−6 cm−3. Thus, the absorbing material exists in a very rarefied environment.
It is also possible that the absorption is not due to just the wind of WR 79, but a collective wind whose carbon abundance has been enhanced by WR 79. To address this issue, consider the contributions to the wind flow of the 10 O stars (11 when both components of HD 152248 are included) and one B supergiant shown in Fig. 2. Their mass-loss rates were estimated by beginning with the spectral types given by Baume et al. (1999). These were translated to physical parameters using the Martins, Schaerer & Hillier (2005) calibration for Galactic O stars (as modified by Weidner & Vink 2010) which were then used to determine mass-loss rates through the Vink, de Koter & Lamers (2000) relations. The combined |$\dot{M}$| for all of these stars is 1.2 × 10−5 M⊙ yr−1. While these winds contribute very little to the C iv absorption, they could influence the wind flow from WR 79. The two components of the binary HD 152248 account for roughly half of the O star contribution. Together, their mass-loss is expected to be 6.1× 10−6 M⊙ yr−1. Further, the system is much closer (on the sky) to the two BVs (it is the bright star that touches CPD−41o7724 in Fig. 2). So although its wind should contribute little to the C iv absorption, its momentum could have a strong influence on the flow that is traced by the C iv from WR 79, possibly channelling it towards the line of sight to the cluster and affecting the ionization. Unfortunately, given the few sightlines available, it is currently not possible to pursue this possibility further.
In terms of indirect support for the current model, consider the statement made in Section 2 that BVs with anomalous spectra like those in NGC 6231 have not been observed before or since in more detail. Specifically, beginning with the WC stars listed in the van der Hucht (2001) catalogue with V ≤ 9 mag, the IUE archive was searched for BVs (with an SWP spectrum of any kind) within 30 arcmin of each. WR 79 is the only WC near BVs. There is a reason for this. Normal, faint BVs were not typically observed with IUE. Normal BVs are 2–4 mag fainter than O, B supergiant or WR stars. Consequently, there was very little motivation to observe them, when bright nearby BVs were available. However, one reason to observe faint BVs in open clusters was to use them as standard candles for extinction studies – which is how the BVs with peculiar spectra in NGC 6231 were uncovered. These extinction studies observed numerous BVs in several young, open clusters (e.g. fig. 1 in Fitzpatrick & Massa 2007), but only NGC 6231 contained a WC star and only NGC 6231 contained BVs with peculiar spectra. Since it is now known that the NGC 6231 BVs have normal abundances (Kilian et al. 1994), the only thing that makes them unique is that they cohabit a cluster with a WC star.
To summarize, it was argued that the high-speed C iv absorption is probably not intrinsic to the NGC 6231 BVs but related to the wind of WR 79 for the following reasons:
The BVs in NGC 6231 have normal abundances, so there is no reason to expect their winds to be abnormal.
Of the BVs observed in young clusters with IUE, only NGC 6231 contains BVs with peculiar C iv absorption, and only NGC 6231 contains a WC star.
The NGC 6231 BVs with peculiar C iv are located near WR 79, and those further away have normal spectra (Section 2).
The BV absorption profiles are well modelled by the profiles expected for stars embedded in a spherically expanding flow.
The values of v∞, v1, |$\dot{M}$|−6, AE, and qi determined from the profile fits all lie within the range of expected values.
While not conclusive, these facts imply that it is highly probable that WR 79 is responsible for the high-speed absorption. Unfortunately, the available data do not make it possible to determine whether the flow is dominated by the wind of WR 79, or simply enriched in carbon by it. This is because the two available lines of sight are close together and roughly the same distance from WR 79. As a result, they do not provide much independent information for distinguishing whether the high-speed gas is coming from WR 79 or a more extended source. Only additional sightlines with different locations in the cluster can break this degeneracy.
Acknowledgments
This paper benefited from the comments of an anonymous referee, and discussions with W.-R. Hamann and L. Oskinova. DM acknowledges partial support under NASA Grants NNX11AB19G and HST GO-13760 to SSI. The data presented in this paper were obtained from the Mikulski Archive for Space Telescopes (MAST). STScI is operated by the AURA, Inc., under NASA contract NAS5-26555. Support for MAST for non-HST data is provided by the NASA Office of Space Science via grant NNX09AF08G and by other grants and contracts. The DSS2 was produced at the STScI, under grant NAG W-2166.
Disc winds are also capable of creating absorption with little accompanying emission. However, these winds are typically characterized by dense, low-speed flows near the star (e.g. Bjorkman et al. 1994), and never show high-speed absorption without strong low-speed absorption. Thus, it is the presence of high-speed absorption together with a lack of emission and low-speed absorption that make the profiles of the NGC 6231 BVs unique.
