O-STAR MASS-LOSS AND WIND MOMENTUM RATES IN THE GALAXY AND THE MAGELLANIC-CLOUDS

A new, very fast approximate method is presented to determine mass-los s rates of O-stars from H-alpha line profiles. The method uses H and H eII departure coefficients from unified model atmospheres parametrized in a simple way as function of wind velocity together with photospher ic NLTE line profiles as the inner boundary condition for a numericall y exact radiative transfer solution to derive a wind contaminated H-al pha -profile. The method is also applied to H-gamma to determine stell ar gravities corrected for wind emission. A detailed analytical discus sion of H-alpha line formation in O-star winds is given and it is demo nstrated that former very simple approaches considering only optically thin wind emission lead to significant systematic errors. Scaling rel ations and generalized curves of growth are presented that connect mas s-loss rate, terminal velocity, stellar parameters and H-alpha equival ent width. The method is applied to samples of O-stars in the Galaxy, LMC and SMC and mass-loss rates are derived from H-alpha in combinatio n with terminal velocities measured from IUE and HST spectra. The resu lts reveal that a tight empirical relation exists between the radius m odified stellar wind momentum rate M v(infinity)R(0.5) and the stella r luminosity. The variations of this relationship between the Galaxy, LMC and SMC are explained in terms of different abundances. Furthermor e, for almost all objects with dense winds (mostly supergiants), the c ommonly used velocity field exponent beta could be derived, indicating a typical value of beta approximate to 1. A comparison with the impro ved theory of radiation driven winds (as presented recently by Pauldra ch et al. 1994) shows that the observed wind momentum-luminosity relat ionship can be understood qualitatively in terms of the theory. Howeve r, there exist significant systematic discrepancies as a function of e ffective temperature, luminosity class and wind performance number eta = M v(infinity)c/L. We stress that these discrepancies would not have been detected with previous simplified H-alpha-approaches. The defici encies of the theory are discussed and suggestions for future improvem ents are made.