Astron. Astrophys. 320, 500-524 (1997)

5. Discussion

In this section we compare various properties of LMC WNL stars with their Galactic counterparts and discuss the evolutionary implications of our results.

5.1. Comparison of Galactic and LMC WNL stars

We consider first if the LMC WNL stars show any evidence for a lower metallicity in their observed spectra compared to the Galactic WNL stars. If we can demonstrate that the LMC stars in our sample are of lower metallicity, then we can investigate what effect this has on the parameters derived in Sect. 4. In Fig. 7, we show the equivalent width ratio of N III /N IV , which is purely an indicator of spectral excitation, plotted against the metallicity indicator N III /He II . It is apparent that the LMC stars (particularly the WN6-7 subtypes) have weaker N III /He II ratios compared to their Galactic counterparts and thus appear to have lower metallicities. This cannot be a result of the LMC stars having a different excitation (i.e. weaker N III) because they show the same range of N III /N IV ratios for a given subtype. Likewise, Fig. 1 shows that the LMC and Galactic stars have similar He II emission line strengths. This is in contrast to the study of L.J. Smith & Willis (1983) who found that He II in LMC WN stars was twice as strong as that in Galactic WN stars. We find that their equivalent widths for the five stars in common are 20-50% larger than those given in the Appendix (Table 7). This coupled with the re-classification of their four LMC WN7 stars to WN6, and more measurements of Galactic WN8 stars (they only used WR40 for comparison) appears to explain the discrepancy. The inference from Fig. 7 that the spectra of LMC WNL stars indicate a lower nitrogen content than Galactic stars is quantitatively confirmed in Paper I. We find that the average nitrogen mass fraction is 0.6% which is identical to that predicted by Schaerer et al. (1993) for a WNL star with . In addition, the UV Fe IV-V line spectrum of Brey 13 (Fig. 3) is significantly weaker than Galactic WN8 stars.

Fig. 7. The equivalent width ratio of N III /N IV plotted against N III /He II for Galactic (open) and LMC (filled-in) WN6-9 stars. Observational data for Galactic stars are from Paper III and our own measurements from Hamann et al. (1995b)

One interesting difference between the two galaxies is the relative populations of WN9-11 and their evolutionary successors, the WN8 stars. In the LMC we find only three WN8 stars compared to nine WN9-11 stars, whereas in the Galaxy, there are 12 WN8s, one normal WN9 star (WR105, NS4) and currently one WN11 star (He 3-519). We note however that the distinction between WN8 and WN9 spectral classifications is often fairly subjective, as shown in Fig. 1a. For example, several Galactic WN8 stars (e.g. WR130, WR156) show N III 4640 a factor of thirty times stronger than N IV 4058 (recall Fig. 7), but fail a WN9 classification because the strength of N IV 4058 is non-negligible. WN9 classifications may therefore be more suitable for these objects. Another difficulty with accurately classifying Galactic WN8 stars is that strong interstellar extinction often prevents reliable N IV 4058 measurements, in contrast to the lightly reddened LMC stars.

We now examine if the terminal velocities of the LMC stars are lower than the Galactic WNL stars. Haser et al. (1994) and Walborn et al. (1995a) have found that O stars in the LMC have marginally slower winds than their Galactic counterparts, as anticipated from theory. In Fig. 8 we plot against stellar temperature . There is a definite trend (with a fair amount of scatter) for a decreasing with decreasing . Stars with a high for their are anomalous in some way. The Carina WNLa stars appear to be on a different evolutionary track and are more like extreme O stars (Paper III). The four 30 Dor members, Brey 75, 80, 89 and 90 have very high luminosities, and Brey 26 may be a close binary. Omitting these stars, there is no evidence that the WN6-8 stars in the LMC have lower velocities than their Galactic counterparts. The cool WN9-11 stars with log ( /K) 4.5 may have winds that are % slower but with only three Galactic WN stars in this group, this must be regarded as uncertain.

Fig. 8. A comparison of terminal wind velocities (km s-1) versus stellar temperature for Galactic (open symbols) and LMC (filled-in symbols) WNL and LBVs. The adopted terminal velocity for WR66 (HD 134877, WN8(h)) is 650 km s-1 based on previously unpublished optical spectroscopy, instead of 1500 km s-1 as adopted by Hamann et al. (1995a)

Fig. 9. a Observed H-R diagram for Galactic (open symbols) and LMC (filled-in symbols) WNL and LBVs; b Derived wind performance numbers, , for WNL stars versus surface hydrogen content (by mass). Physical parameters are taken preferentially from Papers I-III, L.J. Smith et al. (1994), L.J. Smith et al. (1995), and supplemented by the analyses of Hamann et al. (1995a)

We now compare the luminosities of WNL stars in the LMC and Galaxy by showing their positions in the H-R diagram in Fig. 9a. With the exception of the highly luminous 30 Dor WN6 stars Brey 89 and 90, we see no separation between Galactic and LMC WNL stars, in contrast to the claim of Koesterke et al. (1991) who found that the LMC stars were less luminous. If we were to restrict our comparison to WN6-8 stars, however, excluding 30 Dor members because of potential multiplicity, it is possible that the remaining LMC stars might possess lower luminosities, although the distances to many Galactic WR stars, and hence their luminosities, remain uncertain.

Since He II 4686 line luminosity determinations are important for studies of WR starburst regions, we note that the average He II 4686 line luminosity is 415 , (range 90-1 700 ) based on 14 LMC WN6-9 stars. For comparison, from 11 Galactic WN6-9 stars with known distances, the average He II 4686 line luminosity is 215 (range 50-440 ). Although the most luminous Galactic WNL stars show progressively weaker emission lines, this is not true for LMC WNL stars (in contrast to Morris et al. 1993b).

Current radiatively-driven wind theory (Kudritzki et al. 1989) predicts that mass-loss should scale as the square-root of the metallicity. This suggests typical LMC WNL stars ought to have mass-loss rates 0.2 dex lower than Galactic examples if the WR winds are radiatively-driven, as has recently been found, albeit marginally, for LMC O supergiants relative to Galactic stars by Puls et al. (1996). Further, assuming that WR winds are driven by multiple scattering, we might expect their momentum rates to scale with the number of optically thick lines, and so be proportional to metallicity (e.g. Gayley 1995). In Fig. 9b, the wind performance number ² is plotted against the atmospheric hydrogen content. The two parameters show a reasonable correlation in the sense that the wind performance number increases as the star becomes more evolved i.e. the amount of helium in the wind increases. Stars with a hydrogen content of % are all WN6-8 stars and have high values of from 3-20, and 50 in the case of the extreme 30 Dor star Brey 80. There appears to be no difference between Galactic and LMC WN6-8 stars. Conversely, the region with % is occupied by LMC WN9-11 stars and two Galactic LBVs. All these stars have and can be considered to have radiatively-driven winds. It is also clear that the Galactic WNLa stars form a distinct group with high performance numbers and hydrogen contents. This diagram can be interpreted as an evolutionary sequence with the wind performance number increasing with the helium content. Hamann et al. (1995a) suggested a relation between hydrogen mass fraction and for Galactic WN stars (their Fig. 7). Inclusion of LMC WN stars broadly confirms their derived relation, though with significant scatter.

5.2. Stellar Evolution

There are a number of well known problems in comparing current single star evolutionary models with atmospheric analyses (see e.g. Hamann 1994). In addition, it is now recognised that rotational mixing has a significant effect on evolutionary models (Fliegner & Langer 1995; Langer, priv. comm.), allowing for instance, surface chemical enrichment at a relatively early post-main sequence phase for sufficiently massive stars. Nevertheless, inspection of the results from the latest evolutionary models (Meynet et al. 1994) suggests that at low metallicities, WR formation should be restricted to higher initial mass, higher luminosity progenitors, because of lower mass-loss rates, with the WR lifetime also reduced.

We have previously proposed that high initial mass stars ( 40 ) advance through a LBV stage, with WN9-11 probably dormant LBVs during their hot phase, before progressing on to WN8 stars (Smith et al. 1994; Papers I, III). Considering first the relationship between WN9-11 and LBV stars, we find that the stellar luminosities ( 5.5-5.8) and helium contents (55-75% by mass) of the WN9-11h stars (incorporating results from Paper I) are in close agreement with the LBV R71 in the LMC (Lennon et al. 1994). We note that BE294 has been identified as a LBV (Bohannan 1989), while R127 (HDE 269858f) was the prototype of the Ofpe/WN9 subclass in 1977 (probably WN11h at that epoch) before its discovery as a LBV (Stahl et al. 1983). The association of circumstellar ejecta nebulae with several LMC WNL stars allows further support of this evolutionary scenario. For example, the WN8 star Brey 13 has an ejecta-type nebula (Garnett & Chu 1994), while LBV-type nebulae associated with several WN9-11 stars (e.g. S119, S61) provide further evidence for a close link with LBVs (e.g. R127).

In Fig. 10a we illustrate this evolutionary sequence by showing the smooth progression of spectral morphology with increasing excitation and wind velocity (and decreasing hydrogen content) between B-supergiant LBVs and WN11-8 stars. Observations of P Cygni were obtained at the William Herschel Telescope in 1993 October using the Utrecht echelle spectrograph. We note that WN8-11 stars are found in the field in the LMC (with the exception of the hydrogen-poor WN8 star Brey 81), suggesting that their initial masses are lower than the WN6-7 stars located in 30 Dor.

Fig. 10. Morphological sequences amongst post-main sequence massive stars between 4000-4900. a Sequence from the B supergiant LBV P Cygni through WN9-11 stars to the classical WN8 star Brey 36. b Sequence from the O3 supergiant HD 93129A, through Sk-  22 (O3 If/WN6) and WR25 (HD 93162, WN6ha) to the WN6h star Brey 89. LMC spectra are radial velocity corrected (for the purpose of comparison). H/He ratios are taken additionally from Papers I-II, Langer et al. (1994) and Kudritzki et al. (1991). All data are plotted to the same scale, shifted vertically for clarity

We find that the luminosities of LMC WNL stars span a similar range to Galactic WNL stars, suggesting a similar range of initial masses. While stellar luminosities of most WN6-7 stars are unexceptional, the 30 Doradus WN6 stars Brey 89 and Brey 90 have extreme luminosities (log () 6.2). The spectral synthesis study of 30 Dor by Vacca et al. (1995) has shown that the age of the starburst is at most 3 Myr. Examination of the evolutionary models of Meynet et al. (1994) implies that only stars with initial masses of 90-110 will be WNL stars by this time. In Paper III we found that the Carina WNLha stars, with high stellar luminosities, were closely related to, and directly descended from, massive Of stars, with WN6-7 sucessors. This result has recently been supported by the direct mass determination of 72 by Rauw et al. (1996) for HD 92740 (WR22, WN7ha) and the presence of spectroscopically similar stars in the young cluster NGC3603 (Drissen et al. 1995). We find similar evidence here that (at least some) WN6-7 stars are descended from very massive progenitors, probably involving direct evolution from O3 If/WN6 stars as proposed in Paper III. The O3 If/WN6 stars, also located in and around 30 Dor (Walborn 1994), probably represent LMC equivalents of the Galactic WNLha stars, due to their similar spectral morphologies and extreme stellar properties (Pauldrach et al. 1994).

We illustrate the evolutionary sequence for the most massive stars in Fig. 10b, which shows the progression of spectral morphologies with increasing emission strength, width and helium content between early Of stars and WN6-7 stars. Observations of HD 93129A were obtained at the MSO 1.9m in 1995 June using the coudé spectrograph, while those of Sk-  22 were taken at the AAT in 1994 December using the RGO spectrograph. Although the chemical content of Sk-  22 is uncertain, other O3 If/WN6 stars show negligible helium enrichment (Pauldrach et al. 1994; de Koter et al. 1994).

Overall, we find that while there is observational evidence for a lower metal content in LMC WNL stars, this does not seem to affect the stellar parameters in comparison to Galactic WNLs. In particular the predicted dependence of metallicity on wind strength is not seen although the high performance numbers of the WN6-8 stars question the application of radiation-driven mass-loss theory to Wolf-Rayet stars. We find tentative evidence that the radiatively-driven winds of the WN9-11 stars have lower terminal velocities than the few Galactic examples. The observed properties of LMC WNL stars support our previous evolutionary schemes involving direct evolution to WN6-7 stars from the most massive O stars, located in starburst regions. In contrast, less massive O stars advance through an intermediate LBV phase, with Ofpe/WN9 stars, now revised to WN9-11, representing a hot dormant LBV stage, and subsequently evolve to a classical WN8 star.

© European Southern Observatory (ESO) 1997

Online publication: June 30, 1998
helpdesk.link@springer.de