Evolved massive stars in the local group. II. A new survey for Wolf-Rayet stars in M33 and its implications for massive star evolution: Evidence of the "Conti scenario" in action

We expect the evolution of massive stars to be strongly influenced by mass loss and hence to be sensitive to metallicity. It should be possible to tes t this "Conti scenario" be comparing the populations of evolved massive sta rs among the Local Group galaxies, but such investigations have been hamper ed by incompleteness. In a previous paper, we presented results of a new su rvey for red supergiants (RSGs) in selected regions of the Local Group gala xies M33, M31, and NGC 6822. In the present paper, we survey eight fields i n M33 for Wolf-Rayet stars (WRs), using interference-filter imaging with a CCD to select candidates. Follow-up spectroscopy is used to confirm 22 newl y found WR stars, 21 of WN type. We establish that our survey would readily detect WRs as weak-lined as any known, and we conclude that our survey is essentially complete. This survey confirms suspicions that the previous pho tographic surveys were only 50% complete for WN-type WRs and allows us to c ombine the data with equally complete samples on other Local Group galaxies . We find that the relative number of WC- and WN-type WRs correlates extrem ely well with metallicity, varying by a factor of 3 with galactocentric dis tance within the plane of M33, and continuing the trend to lower and higher metallicity galaxies. The WC/WN ratio within 3 kpc of the sun is slightly above this trend, and we argue that WN stars are underrepresented in this s ample. The WC/WN ratio is anomalously high in IC 10, given its low metallic ity, and we demonstrate that this is not because of selection effects but i s likely caused by IC 10's current status as a starburst system. We examine the spectral properties of WC stars within these galaxies, confirming the previously reported trends that the spectral lines are stronger and broader in regions of lower metallicity. We suggest that the different WC spectral subclasses do not primarily indicate different physical properties for the se stars but rather are simply a reflection of the effect that the initial metal abundances has had on the stellar wind structure. Finally, we compare the luminous RSGs with WRs in these galaxies. We find that there is a very strong correlation of the relative numbers of RSGs and WRs with metallicit y, in the sense predicted by Maeder, Lequeux, & Azzopardi: at lower metalli cities the fraction of luminous (M-bol < -7) RSGs is higher, with a factor of 6 change within the disk of M33 [Delta log (O/H) = 0.35 dex], and a fact or of similar to 10 change from M31 (or the inner portions of M33) to NGC 6 822 CA log (O/H)= 0.5 dex]. This is easily explained by the Conti scenario in terms of massive stars spending proportionately less of their He-burning lifetimes as RSGs rather than WRs at higher metallicities and hence higher mass-loss rates. Finally, we note that the presence of luminous RSGs and W Rs stars is extremely well correlated for the OB associations in M31 and M3 3: where one finds one, one finds the other. To the extent that an associat ion is strictly coeval, this argues that some stars of 15 M-. and above ind eed do go through both RSG and WR stages. The presence of WR stars of both WN and WC types in the same associations as luminous RSGs further suggests that some WCs, at least, have gone through the RSG phase. We include an App endix providing a complete catalog of confirmed WR stars in Local Group gal axies beyond the Magellanic Clouds.