5 Additional characteristics

5.1 Influence of a companion

The observed WR stars may be physical (typically WR + OB) or spurious (WR + anything on the line of sight) binaries. In both cases, the spectral lines of the WR stars are "diluted'' by the other star's light, and the corresponding location of these stars on our color diagrams lie closer to the origin of the axis than the corresponding uncontaminated spectral types (Sect. 4.3). This is indeed observed in Figs. 1 to 4 where the symbols corresponding to the binaries lie closer to the origin than those representing single objects.

  

Figure 5: Simulations for WN binaries. A downward arrow above another symbol indicates a binary. A plus sign above another symbol indicates a "+ abs'' star. The other symbols are explained in Table 3. The longer arrows indicate the displacement imposed to the representative points of some WN stars when their light is mixed with that of an O6V companion star (see text for details)

  

Figure 6: Simulations for WC binaries. A downward arrow above another symbol indicates a binary. A plus sign above another symbol indicates a "+ abs'' star. An upward arrow below another symbol indicates a weak lined WC star. The other symbols are explained in Table 3. The longer arrows indicate the displacement imposed to the representative points of some WC stars when their light is mixed with that of an O6V companion star. From left to right, the WR stars used for the simulations are WR 103 (WC9), WR 57 (WC7) and Brey 8 (WC5-6)

The first interesting fact to notice is that the binarity has no effect on the WN-WC separation as can be seen in Fig. 1. This means that, even when dealing with spatially unresolved double (or even multiple) objects, photometry alone still allows not only to detect the WR stars but also to separate them into WN and WC stars. But, in most cases, much more can still be achieved. As can be seen in Fig. 2, most of the binaries lie in their proper zone. It is true for the WN7 stars and it is also true for the WNEw stars with only one exception, WR 43, a WNEw star which is lying in the intermediate strip between the WN7 and the WN8-9 regions, with l( He II) $\approx$ 0.25. As a matter of fact, this star has some reasons to be mislocated on our diagram since Hofman et al. (1995) have shown that, within a circle of 5'' diameter around the star, diffraction-limited speckle masking observations allow to count 20 different objects. Among them, 4 are of the WN type, the others probably of type Of. It is interesting to see that even in such a difficult situation, photometry still allows to classify the object as a WN star.

The other "mislocated'' objects in Fig. 2 are Brey 40A (the diamond at l( He II) $\approx$ 0.2), a WN3+O6 (Niemela 1991), Brey 21 (the open circle at l( He II) $\approx$ 0.25), a WN5?+B1I (Smith et al. 1996) and WR 97 (the open circle at l( He II) $\approx$ 0.38), a WN5b + 07 (Smith et al. 1996). Due to their low intrinsic luminosity ($M_v \approx -3.$, van der Hucht 1992), it is expected that the contamination of a WN3 star by an O6 star ($M_{v} \approx -5.5$) shifts markedly the representative point towards the origin. It is more surprising for a WN5 star with a luminosity of the order of $M_{v} \approx -5.$ (van der Hucht 1992). To quantify this assertion, we have simulated the contamination of the WR stars of our sample by an O6 object.

As can be seen in Fig. 5, the contamination of a strong lined WN3 star (Brey 1, located at l( He II) $\approx$ 2.2) by a single O6 star brings its representative point close to the location of Brey 40A. But the contamination of a strong lined WN6 star (WR 110, located at l( He II) $\approx$ 2) leads to a markedly shorter displacement which is not sufficient to bring a representative point from the strong lined WNE region to the vicinity of the origin of the graph. This could indicate either that the WR component of Brey 21 and WR 97 is less luminous than a classical strong lined WNE star or that the WR light is diluted by more than one companion.

The results of a few other simulations are also shown on the graph. As expected, due to their high luminosities, the representative points of the late WN stars are less affected by the presence of a companion. It is also interesting to notice that the nature of the displacements is such that, basically, a contamination maintains roughly the representative points within their spectral region: a WN8-9 star will always remain in its zone, a WN7 star will roughly do the same, and it is only close to the origin that an early WN star could enter the WN7 + abs region. All the preceding conclusions are not affected if the simulations are made with another spectral type for the companion because the displacement of the representative point comes from the dilution of the WR spectral lines, not from the spectral characteristics of the companions.

Similar simulations have been carried out with WC objects (Fig. 6). Again, the displacement is towards the origin of the axes, i.e. such as to bring the representative points of the binaries in the binaries region, as defined in Figs. 4 and 6. The length of this displacement is a function of the relative luminosities of the WR star and its companion star. It is interesting to notice that the WC9 stars are so well separated from the other WC stars that most of their binaries probably remain in a distinct region of the plot.

5.2 Influence of reddening

An important fact is that these results are little affected by reddening (simulated up to A_v = 7.). Its effect is indeed weak and roughly the same for all WR and normal stars, so that differential reddening in a given part of the sky could blur a bit the separations between the different regions in our color diagrams, but not in a drastic way.

5.3 Radial velocity shifts

Observing external galaxies with narrow band filters raises the question of adapting the filters to the radial velocities of the targets. In order to estimate to which extent our results were sensitive to these velocities, we performed synthetic photometry with nominal and enlarged l( He II), l( C IV) and l( He I) filters (by a factor of 2) on blue- and red- shifted WR spectra.

Results are encouraging since, thanks to the width of the WR lines, blueshifts are no source of major problems and redshifts can be tolerated up to several hundreds of km/s. Of course, the useful part of our plots slightly shrinks as the redshift increases, i.e. the lines begin to get out of the filters and the stars slowly travel to the origin of the diagram, but sufficient discriminating power is left to operate till $\sim$ 500 km/s redshifts. The first property to suffer from further redshift is the WC9-WN separation. Apart from that, only slight adaptation of the borders of the different regions could be necessary in the l( C IV) vs. l( He I) plot at such redshifts.

Tests performed with wider band pass bring no surprise as they prove these to be less sensitive to redshift, but they also reduce discriminating power because the light of the line is more and more diluted by the continuum light as the filters widen.