The physical properties and effective temperature scale of O-type stars as a function of metallicity. III. More results from the Magellanic Clouds

In order to better determine the physical properties of hot, massive stars as a function of metallicity, we obtained very high signal-to-noise ratio optical spectra of 26 O and early B stars in the Magellanic Clouds. These allow accurate modeling even in cases where the He I λ4471 line has an equivalent width of only a few tens of m Å. The spectra were modeled with FASTWIND, with good fits obtained for 18 stars; the remainder show signatures of being binaries. We include stars in common to recent studies to investigate possible systematic differences. The "automatic" FASTWIND modeling method of Mokiem and collaborators produced temperatures 1100 K hotter on average, presumably due to the different emphasis given to various temperature-sensitive lines. More significant, however, is that the automatic method always produced a "best" result for each star, even ones we identify as composite (binaries). The temperatures found by the TLUSTY/CMFGEN modeling of Bouret, Heap, and collaborators yielded temperatures 1000 K cooler than ours, on average. Significant outliers were due either to real differences in the data (some of the Bouret/Heap data were contaminated by moonlight continua) or the fact that we could detect the He I line needed to better constrain the temperature. Our new data agree well with the effective temperature scale we previously presented. We confirm that the "Of" emission characteristics do not track luminosity classes in exactly the same manner as in Milky Way stars. We revisit the issue of the "mass discrepancy," finding that some of the stars in our sample do have spectroscopic masses that are significantly smaller than those derived from stellar evolutionary models. We do not find that the size of the mass discrepancy is simply related to either effective temperature or surface gravity.