3. Stellar spectra
3.1. Optical classification
For spectral classifications of WN6-8 stars we follow the recent L.F. Smith et al. (1996) scheme in preference to earlier classifications since (i) the overall scheme based primarily on He I-II line strengths is more quantitative and objective. For instance it avoids the metallicity effect of using the N III 4634-41 to He II 4686 ratio as a criterion for WN8 subtypes; (ii) it is based on modern CCD observations, including our own; and (iii) it includes additional properties such as the presence or absence of hydrogen. Relative to previous schemes (e.g. Breysacher 1981) several stars are re-classified to WN6 from WN7 while Brey 47, previously mis-classified as WN8 is now given a WN6 classification (L.F. Smith et al. 1996).
Turning to the Ofpe/WN9 stars (and AB18), we have previously argued that four such stars should be re-classified as WN9 or WN10 stars, based on the relative strengths of their nitrogen lines (Paper I). Similarly, the spectral classification of the Galactic LBV candidate He 3-519 was revised from Ofpe/WN9 to WN11 by L.J. Smith et al. (1994). The purpose of this re-classification is as follows:
1. These objects form a natural, smooth extension to the WN morphological sequence at lower excitation, using the Wolf-Rayet classification criteria from Conti (1973).
2. The surface chemistries of the WN9-11 stars are quantitatively indistinguishable from stars of earlier WN spectral type and so should also be included in the WR population when comparing, e.g. theoretically predicted WR:O lifetimes with observation.
3. The (unintentional) implication that Ofpe/WN9 stars represent objects at a phase in their evolution directly between Ofpe and WN9 is misleading. For example, BE381 (previously Ofpe/WN9) is apparently progressing between a LBV and a classical WN8 phase (Paper I). Similarly, He 3-519 (WN11, previously Ofpe/WN9) is probably a dormant LBV, showing a visual excitation more representative of a B supergiant than a late Of star (L.J. Smith et al. 1994). In contrast, HD 152408 (O8 :Iafpe) appears to be advancing directly between an Of and a Wolf-Rayet stage and shows a spectral morphology quite distinct from LMC WN9-11 stars (Crowther & Bohannan 1997).
4. An extension of the WN sequence to lower excitation also mirrors the existing WC sequence. The Planetary Nebula central star CPD- 8032, classified [WC10], has almost identical He I-II line strengths to the WN10 star S9 (Crowther et al. 1996).
One indirect effect of our re-classification is that objects with substantial hydrogen ( 20-40%) at their surfaces may not be consistent with the usual core helium burning definition of a Wolf-Rayet star (Smith 1973). At present, however, the critical link between the observed H/He ratio (when it is greater than zero), and the state of the core (hydrogen or helium burning) is not known. Therefore, while still representing the final phase in massive star evolution, Wolf-Rayet spectral classifications may not necessarily imply core helium burning. We note that Galactic hydrogen-rich WN6-7ha (or WN6-7 abs) stars have been proposed as core hydrogen burning objects (e.g. Rauw et al. 1996), yet the existence of WR 123, a WN8 star with very little atmospheric hydrogen (H/He 0.10; Paper II), suggests that (at least some) WN8 stars must be helium burning.
Although our initial classification scheme for WN9-11 stars was limited to the relative strengths of N II-IV emission lines (L.J. Smith et al. 1994), we have since discovered that use of He I-II emission line strengths provides a fully consistent classification criterion (L.J. Smith et al. 1995). The advantage of the latter is that for lower quality data and weaker-lined objects, relative helium equivalent widths are more readily obtained than those of nitrogen, and this forms a natural extension to the WN classification scheme of L.F. Smith et al. (1996). In Fig. 1a we present comparisons of measured emission equivalent widths for He I 5876 versus He II 4686 for our LMC programme stars. We also show values for the Galactic WN6-11 counterparts, preferentially taken from Paper III and L.J. Smith et al. (1994), or our own measurements from Hamann et al. (1995b), LMC WN9-10 stars from Paper I, and several Of stars including the prototype LMC O3 If/WN6 star Sk- 22. Using the relative strengths of the N II -N IV emission lines as the primary classification criterion (Paper I), we are able to draw approximate boundaries to the spectral classes as shown in Fig. 1a, and then use this diagram for classification purposes. We note that the boundary lines are not vertical because while the strength of He II is primarily an excitation diagnostic, the strength of He I also strongly depends on wind density. It is thus possible to have WN6 and WN8 stars with the same He II line strength but vastly different He I line strengths. In Fig. 1b we present He II 4686 equivalent width versus FWHM measurements for Galactic and LMC WNL and Of stars. While WNL stars within both galaxies obey a tight relation between He II 4686 line strength and width (Bohannan 1990), this line alone is not a useful discriminator between WN9-11 and Of stars (L.J. Smith et al. 1995; Crowther & Bohannan 1997) although it is useful for identifying candidate WNL binaries.
From Fig. 1a, we re-classify Brey 75 as WN6, Brey 81 as WN8, AB18 (marginally) as WN9, BE294 as WN10 and the three Henize objects S119, S61, S142 as WN11. Following L.F. Smith et al. (1996) we further refine spectral classifications of these stars to incorporate hydrogen contents, as summarised in Table 1. No additional WN9 stars are found here from the sample of Bohannan & Walborn (1989) since those objects showing the strongest He II emission (i.e. highest excitation) were all discussed in Paper I. Relative to the Ofpe/WN9 subgroup structure of Walborn (1982), all Group A members are identified as WN9 stars, with Group B stars assigned either WN10 or WN11 classifications. The spectral appearance of the sole Group C member R99 from Walborn (1982) is found to be inconsistent with a Wolf-Rayet classification since it lacks the He I P Cygni profiles characteristic of WNL stars and is discussed separately (Sect. 3.6). Wolf-Rayet catalogue numbers for these new additions are Brey 37a (BE294), Brey 45a (S119), Brey 97a (S142) and Brey 99a (S61). We note that BE294 has been identified as a LBV (Bohannan 1989), while S119 has been proposed as a LBV by Nota et al. (1994) because of the similarity of its circumstellar nebula with established LBVs (R127, AG Car). This is consistent with the findings from Paper I and L.J. Smith et al. (1994) that WN9-11h stars generally represent either dormant LBVs or the immediate progenitors of classical Wolf-Rayet (WN8) stars.
3.2. Optical and ultraviolet spectroscopy
In Fig. 2 we present new blue and yellow spectroscopy of selected LMC programme WN6-11 stars, sorted by spectral type. We show BE381 (Brey 64) from Paper I in preference to AB18 since its line strengths are more representative of WN9. The progression to narrower, lower excitation spectral features at later spectral type is clearly demonstrated; compare He II 5412 to He I 5876 or N IV 4058 to N III 4634-41.
Fig. 3 shows representative flux calibrated ultraviolet spectra of LMC WN6-10 stars, including HST /FOS observations of the WN9h star Sk- 249c (HDE 269927c, Paper I). As in the optical we see a smooth progression to lower excitation, narrower and weaker emission lines in the ultraviolet from WN6 to WN10. In particular, highly excited transitions such as N V 1238-42, C IV 1548-51 and He II 1640 rapidly decrease in strength at later spectral type, while N III ] 1747-54 and Si IV 1394-1402 increase in strength, and Al III 1855-63 becomes visible. Overall the ultraviolet spectral appearance of Brey 89 (WN6h) is similar to the Galactic WN6ha star WN24 (HD93131, Paper III), while that of Brey 13 (WN8h) resembles the Galactic WN8 star WR40 (HD 96548, Paper III), except for weaker Fe IV-V lines, and stronger N IV ] 1486 and He II 1640 emission.
The ultraviolet spectra of some LMC WN9-10 stars have previously been discussed by Shore & Sanduleak (1984) and in Paper I, but at lower spectral resolution. From Fig. 3 the broad depression between 1500-1700 due to Fe IV 1 is readily apparent. Shore & Sanduleak (1984) gave a UV classification of B1 I for S61, although as for R84 and S9 (Paper I), He II 1640 emission is present. Since the observed optical excitation of very late WN stars is closer to early B than late O supergiants, this UV classification is anticipated.
3.3. Nebular emission lines
Many LMC WN9-11 stars show nebular lines superimposed on their stellar spectra. Circumstellar nebulae have been discovered around R127 (Walborn 1982; Clampin et al. 1993), S61 (Walborn 1982; Stahl 1987), S119 (Nota et al. 1994) and BE381 (Nota et al. 1996). In addition, R84, R99 and Sk c show nebular lines but further observations are required to determine if they arise in circumstellar nebulae or H II regions (Walborn 1982; Nota et al. 1996). While the presence of nebulae is of much interest for studying the evolutionary connections between WN9-11, LBV and WN8 stars, and the properties of the central stars (e.g. L.J. Smith 1995), they pose a severe problem for the quantitative analysis of the stellar spectrum, particularly for determining the H/He ratio. Since the nebulae are barely spatially resolved, it is very difficult to subtract the nebular emission during data reduction. The best approach appears to be to resolve spectrally the nebular lines and to subtract them using Gaussian fitting techniques. High resolution echelle observations of the WN11 star (and LBV candidate) S119 in the region of H are shown in Fig. 4. As described by Nota et al. (1994, 1996), the front and back of an expanding shell are clearly seen expanding at km s-1. To remove the nebular emission, we have used the Gaussian fitting package ELF within DIPSO ; Fig. 4a shows the three component Gaussian fit and Fig. 4b shows the underlying H emission with the nebular lines removed. In the Appendix (Sect. A.14) the H/He ratio for S119 is determined using intermediate and high resolution spectra. We find that the intermediate resolution spectra (nebular lines included) give an erroneously high value of H/He 4, while the true ratio is H/He 1.5.
We have examined the spectra of other WN9-11 stars in our sample with the aim of assessing if nebular lines are present, and if so, how much of a problem they pose for stellar analysis. In Paper I, we analysed the WN9h star Sk c which has nebular lines (Walborn 1982; Nota et al. 1996). Unpublished high resolution spectra obtained at the AAT with UCLES show that the nebula contributes only % of the total flux at H . For most of the stars in our present sample with nebulae, we have MSO H observations at a spectral resolution of 1 Å which is just sufficient to separate the stellar and nebular components. The H region is shown in Fig. 5 for S142, S119 and S61. We detect weak [N II ] nebular lines in the spectrum of S142, indicating that any nebular contamination of the stellar Balmer lines will be negligible. Further observations are required to determine if the nebular emission arises in a circumstellar nebula or an underlying H II region. S119, as discussed previously, has very strong nebular lines. In Fig. 5 the H line is just resolved, allowing us to separate the two contributions. S61 is known to have an associated nebula (Walborn 1982; Stahl 1987). From the Gaussian fits to H shown in Fig. 5, we find that 75% of the flux is nebular. We find that the true H/He ratio for S61 is 1.2 whereas assuming that the Balmer lines are purely stellar gives H/He 6 (Appendix, Sect. A.16). For the two remaining stars with known nebular lines R84 and BE381, the high resolution spectra of Nota et al. (1996) show that the nebular emission is negligible. Finally we note that IR analyses of LMC WN9-11 stars based on line ratios involving a hydrogen line (e.g. Br ) should be treated with caution if the star has strong nebular lines (e.g. S119 and S61).
3.4. Line measurements and wind velocities
Determinations of wind terminal velocities in hot luminous stars are best suited to observations of UV metal resonance lines (Prinja et al. 1990) or infrared He I P Cygni profiles (Eenens & Williams 1994). For the few stars with high resolution UV observations, we can measure terminal velocities from the bluemost absorption edge of Si IV 1393-1402 or C IV 1548-1551. For the majority of our programme stars, for which we only have optical observations, we use instead the He I 5876 profile to estimate wind velocities. In the Appendix (Table 6) we present measurements of wind velocities for our programme stars. Although wind velocities are not readily measured from low resolution IUE data, measurements resulting from the approximate method of Prinja (1994) appear to be broadly consistent with those resulting from optical profiles. BE294 shows several forbidden transitions ([Fe III ] 4658, 5270, [N II ] 5755) in its visual spectrum (Fig. 2). The measured half width at zero intensity of these profiles is in excellent agreement with the velocities from He I P Cygni profiles, as was previously found for S9 (Sk- 40, also WN10) in Paper I. For comparison in Table 6, we include measurements previously tabulated by Koesterke et al. (1991) and Rochowicz & Niedzielski (1995). In Sect. 4 minor adjustments are made to these wind velocities in order to optimise line profile fits.
In the Appendix (Table 7) we present equivalent width and FWHM measurements for selected optical emission lines for all programme stars (including those from Paper I) based on our AAT and MSO observations. Koesterke et al. (1991) have previously presented intermediate dispersion He I 5876 and He II 5412 profiles for four WN6-8 stars in common with our sample, while Nota et al. (1996) have recently presented high dispersion optical spectra for two WN10-11 programme stars. We find good agreement with the published equivalent widths of Koesterke et al. (1991), while we obtain higher equivalent widths than Nota et al. (1996), especially for strong optical lines.
3.5. Photometry and interstellar reddenings
Since only eight programme stars have narrow band optical photometry (Torres-Dodgen & Massey 1988) we utilise wide band B and V photometry taken from a variety of sources. These are typically within 0.05 mag of narrow band measurements for WNL stars. Optical and infrared photometry is presented in the Appendix (Table 8) for our programme stars. As in Paper I we have derived interstellar reddenings by combining theoretical energy distributions with UV, optical and IR photometry. For this a uniform Galactic foreground extinction of 0.05 mag was assumed, which was combined with the standard LMC extinction curve (Howarth 1983) and an extinction ratio of R = / =3.2 (Olson 1975). Comparison between the interstellar reddenings derived here and the arithmetic mean values resulting from the previous determinations of L.J. Smith & Willis (1983), Schmutz & Vacca (1991) and Morris et al. (1993a) shown in the Appendix (Table 8) is generally within 0.02 mag.
For consistency with Paper I, we adopt the standard LMC distance of 51.2 kpc from Panagia et al. (1991), corresponding to a distance modulus of 18.55 mag. Use of the latest LMC distance estimate of 47.3 kpc by Gould (1995) would result in uniform reduction in by 0.18 mag.
3.6. The peculiar emission line star R99
Before we move onto our theoretical model results, we turn briefly to the related LMC object R99. This star has received considerable attention because of its emission line characteristics in the ultraviolet (Shore & Sanduleak 1984), optical (Stahl et al. 1984; Stahl 1986) and infrared (McGregor et al. 1988). R99 was recognised by Walborn (1977, 1982) as being unique in its spectral morphology due to the near-absence of optical absorption features and thus is essentially 'unclassifiable'. Nevertheless, R99 was later included in the LMC Ofpe/WN9 subgroup by Bohannan & Walborn (1989). In Fig. 6, we compare our optical observations of R99 with the WN10h star BE294. While the He I-II and H I emission of R99 is consistent with a WN10h classification, the observed He I profiles are peculiar in that the usual sharp P Cygni absorption components are absent, but shallow, violet absorption extending to 1 050 km s-1 is observed. AAT-UCLES observations of R99 prohibit a sharp, nebular origin for He I-II or H I and confirm an asymmetric He II 4686 feature, which can be reproduced with a double Gaussian comprising a narrow He II 4686 feature with FWHM 300 km s-1 plus a broad component (FWHM 600 km s-1) centered at +340 km s-1.
HST -FOS observations of R99 are also presented in Fig. 6 and reveal strong, broad absorption profiles (e.g. C IV 1548-51), corresponding to an unusually high terminal wind velocity of 1 050 km s-1. The broad 'depression' between 1500-1700 due to Fe IV observed in WN9-11 stars (see Fig. 3) is clearly absent in R99. Shore & Sanduleak (1984) commented on its probable UV emission line variability and assigned an ultraviolet B0.5 supergiant classification. From its ultraviolet properties, R99 therefore appears to be of higher excitation, but lower wind density than typical WN9-11 stars. We will defer any further discussion to a future paper, but we note an overall optical and near-infrared spectral similarity of R99 to the peculiar LBV candidate HD 5980 (Sk 78) in the SMC during 1994 December.
© European Southern Observatory (ESO) 1997
Online publication: June 30, 1998