2. The classification parameters
The new classification by SSM96 uses:
We first examine the numerical relationships between the classification parameters, their stellar model equivalents and other closely related quantities.
2.1. FWHM 4686 vs. other measures of the terminal velocity
Fig. 1 shows FWHM 4686 (from SSM96) vs. other measures of the
terminal velocity . Values from Hamann et al.
(1995) were estimated from the width of helium lines in the visible.
Values from Eenens & Williams (1994) were derived by fitting to
the P Cygni profiles of HeI 1.083 µ and 2.058
µ. Values from Prinja et al. (1990) were derived from the
short wavelength edge of the saturated CIV 1548 violet edge.
Correlation is good with the two determinations that depend on helium
lines but less good with the UV values. The solid line in the diagram
is the regression line fitted to the Eenens & Williams data
(filled points) and represents the data with a standard deviation of
174 km/s. It is:
() = 0.28 () + 7.7
The regression line fitted to the HKW data is slightly steeper (dot-dash line, slope 0.32, intercept 6.4), and the line to the Prinja data (ignoring the outliers discussed below) is slightly flatter (dotted line, slope 0.24, intercept 9.8). The differences are only marginally significant.
The stars with derived terminal velocities well above the line in Fig. 1 (both the HKW and Prinja et al. data sets) are identified in Fig. 1 by their numbers in the Sixth Catalogue (van der Hucht et al. 1981). They are distinguished from the rest of the stars in the diagram by having relatively weak lines. WR 2 (WN 2b) and WR 3 and 46 (WN3b) have lines between 2 and 9 times weaker than WN 3 and 4 stars of comparable line width (see SSM96). WR 22, 24, 25 and 89 are ha stars. WR 128 is the only known WN 4h star and has the weakest lines of any star in this ionisation subclass. The apparent implication is that these stars, with the weakest lines, have a different wind velocity structure from the majority of WN stars.
2.2. Hydrogen abundance
Fig. 2 shows the plot of (by number) from visual inspection of the Pickering series vs. H/He (by number) from the model fits. For WN 8 stars, overestimates H/He by about a factor of 2, reflecting the presence of significant amounts of . For WN7 stars, is a good estimator of H/He. For WN 3-6 stars, underestimates H/He by about a factor of 2, possibly due to blending of NIII with the HeII when the lines become broad.
The sequence of decreasing hydrogen contribution to the spectrum is indicated by the hydrogen subclasses: ha, h, (h) and o. Detection by visual inspection becomes unreliable below = 0.5, the boundary between h and (h). There are only two stars with either or H/He below 0.45 ( = 0.1, by mass), indicating that the two methods suffer a detection limit at about the same value. The number of stars in any range of H/He clearly increases to lower values, so there must be a significant number of stars with H/He 0.5 where detection is difficult.
We note the detection by HKW95 of hydrogen in WR 136, 138 and 157, stars for which hydrogen is not detected from visual inspection of the spectra. We do not change the classification of the stars in what follows, adopting the policy that the spectral classification should reflect only what is available from visual inspection of the spectra. By this policy, the controversial (h) on the classification of WR 136 is inappropriate since the Pickering series for this star is, to visual inspection, completely smooth.
Fig. 3a shows the derived by HKW95 from model fits to spectra of galactic WR stars vs. with the symbols indicating the broad, b, and hydrogen subclasses: ha, h and o; (h) is included with h. We notice that there is a clear relationship between and the broad, b, and hydrogen subclasses of WN stars; the b-stars separate strikingly from the narrow-line stars and the hydrogen subclasses also separate.1
Schmutz (private communication) has pointed out that, without an extended atmosphere, all WN stars are hot enough to have = -0.3 and that the colour derived from a model depends directly on the extension of the atmosphere, i.e. optical thickness of the wind. Hence, the extreme values in Fig. 3a are not surprising: models that fit b stars have optically thick extended atmospheres and are redder; models that fit ha stars have no extension and are at the blue limit. However, Fig. 3a suggests that the h subclass alone is sufficient to predict . If true, this could provide a simple method for deriving the colour excess of any individual star. But is it true for the real stars as distinct from models fitted to the helium lines of those stars?
For galactic WN stars, the only independent values are those of Lundström & Stenholm (1984). Schmutz & Vacca (1991) have shown that, the derived using models agree well, indicating that the models also predict the correct continuum shape - at least for that subset.
For LMC WN stars, Fig. 3b shows observed colours that have been dereddened by nulling the 2150 Å feature (Morris et al. 1993, Vacca & Torres-Dodgen 1990). Scatter in the data is much larger than in Fig. 3a, but the trend is similar. Average values for the galactic models and for the LMC reddened stars are given in Table 1.
Table 1. Average of WN stars in the Galaxy and the LMC
It is not clear whether the scatter in the LMC data (Fig. 3b) and the redder colours of the LMC o and h stars is due to the effect of crowding on the LMC photometric data (cf. Schmutz & Vacca 1991), or whether there is more than one parameter needed to define the atmospheric extent and intrinsic colour of a WN star.
2.3. Ionisation subclass
Fig. 4 shows the ionisation subclass parameter HeII 5411/HeI 5875 vs. , the effective temperature at optical depth = 10 (HKW95) from model fits to the helium lines. The symbols indicate the broad, b, and hydrogen subclasses: ha, h and o. The label c is used for the WN/C stars. The ionisation subclass boundaries are indicated on the HeII/I axis. The error bar on of dex is the standard deviation of differences between the determinations by HKW95 and by Crowther et al. (1995a), both using pure helium atmospheres. The error bar on HeII/I of is the standard deviation of differences between various determinations, as tabulated by SSM96 (Table 9).
HKW95 have emphasised that the previously defined ionisation subclasses are not well separated by . With ionisation subclasses in the new system defined primarily by the HeII/I ratio, the same criteria as used to define , we might expect some improvement. The clear separation of b-stars from narrow-weak line stars discovered by Schmutz et al. (1989) and demonstrated further by HKW93 is obvious. Within each group, there is a correlation between the HeII/I ratio and . (A possible reason for the difference in at the same HeII/I will be suggested in Sect. 4.2.)
For the b-stars, the ionisation subclasses are approximately separated by boundaries indicating that, in this group, is probably the primary parameter controlling the ionisation subclass. However, the scatter is considerable and three stars in particular, WR 36, 75 and 110, lie at low values of HeII/I for their values of . Clean separation of the ionisation subclasses requires sloped (dashed) lines. While this could be an artefact of the observational scatter, the possible relevance of a second parameter is addressed in Sects. 4.2 and 6.
For the narrow-weak line stars, the scatter around a mean line is less than two standard deviations in (as derived above), however the small difference in average between the ionisation subclasses suggests that, for this group, another parameter may be dominant. There appears to be no significant separation of the hydrogen subclasses. (A plot against the effective temperature at = 2/3 from HKW95, is very similar but with less separation between the two groups, less scatter in the b-stars but more in the narrow-line stars.)
We note the presence of two stars in the gap between the two groups: WR 149, which is included by HKW95 as an s-star but which fails the b-star criteria, and the WN3(h) star WR 152.
Fig. 5 shows ionisation subclass vs. , with different symbols indicating the broad, b, and hydrogen subclasses: ha, h and o. The data are (with exceptions listed in Table 2) from HKW (93 and 95) who dereddened the stars using the values from the models. For binaries, we have corrected the using the methods of line diminution of the Balmer and HeI lines of the O stars (Smith & Maeder, 1989) and of the HeII 5411 emission line of the WR star using the calibration from SSM96. New or revised distance modulae are used for WR 6, 78, 108, 78, 157 and 138. All revised data are given in Table 2. (WR 46 is included in Table 2 but not plotted because the values are too uncertain.)
Table 2. Absolute magnitudes of WN stars other than given by HKW
The new elements in this often-produced diagram (which looks rather different from the recent version by HKW95) are: revised ionisation classifications; improved values for WN stars in binaries and inclusion of information about the amount of hydrogen. We note the following points:
The last two points may prove of interest with regard to the evolutionary sequences of WN stars.
© European Southern Observatory (ESO) 1998
Online publication: June 2, 1998