... detector[*]
This detector is characterised by an rms read-out noise of 3.3 electrons per pixel (2.7 ADU with 1.21 electrons per ADU). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... IRAF[*]
The IRAF package is distributed by the National Optical Astronomy Observatories, which is operated by the Association of Universities for Research in Astronomy, Inc., under contract with the National Sciences Foundation. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...(Crowther 2004)[*]
For a hypergiant such as HD 190603 this value might be even higher. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... number[*]
According to latest results (Asplund et al. 2005), the actual solar value is slightly lower, log (Si/H) = -4.49, but such a small difference has no effect on the quality of the line-profile fits. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... fits[*]
Note that in their FT procedure Simon-Diaz & Herrero (2007) have used a Gaussian profile (with EW equal to that of the observed profile) as ``intrinsic profile''. Deviations from this shape due to, e.g., natural/collisional broadening are not accounted for, thus allowing only rough estimates of $v_{\rm mac}$ to be derived, which have to be adjusted during the fit procedure. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...$\lambda $4116[*]
Si IV $\lambda $4089 is unavailable in our spectra. Given that in early B-SGs this line is strongly blended by O II which cannot be synthesised by FASTWIND with our present model atoms, this fact should not affect the outcome of our analysis. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... temperatures[*]
Note that the Balmer lines remain almost unaffected by $v_{\rm mic}$ so that a direct effect of $v_{\rm mic}$ on the derived $\log g$ is negligible. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... below[*]
Though He II is rather weak at these temperatures, due to the good quality of our spectra its strongest features can be well resolved down to B2. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... later)[*]
At these temperatures, only He I is present, and no conclusions can be drawn from the ionisation balance. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...$M_\odot $[*]
A mass of 7 $M_\odot $ as derived for HD 198478 (second entry) seems to be rather low for a SG, suggesting that the B-V colour adopted from SIMBAD is probably underestimated. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...Q[*]
We do not directly derive the mass-loss rate by means of H$_\alpha $ , but rather the corresponding optical depth invariant Q, see Markova et al. (2004); Repolust et al. (2004). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... substantially[*]
And might introduce a certain ambiguity between $\beta $ and the run of the clumping factor, if the latter quantity is radially stratified. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... misclassified[*]
Note that already Lennon et al. (1992) suggested that HD 191243 is likely a bright giant, but their argumentation was based more on qualitative rather than on quantitative evidence. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... et al.[*]
Four B1 stars from the Lefever et al. sample have the same $T{_{\rm eff}}$ and thus appear as one data point in Figs. 8 (right) and 9. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...Azzopardi & Vigneau (1975)[*]
Using low quality objective prism spectra in combination with MK classification criteria, both of which contribute to the uncertainty. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...analysis[*]
These authors have employed the unblanketed version of FASTWIND (Santolaya-Rey et al. 1997) to determine wind parameters/gravities while effective temperatures were adopted using the unblanketed, plane-parallel temperature scale of McErlean et al. (1999). 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... respectively[*]
Slope of regression should correspond to 4/$\alpha'$, if the relations were unique. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... stars[*]
HD 41117, with $\dot M$ from H$_\alpha $ being 0.37 dex lower than from the radio excess. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... globally.[*]
In contrast to Q', $\eta $ is not completely radius-independent, but includes a dependence $\propto$ $R_\star^{-0.5}$, both if $\dot M$ is measured by H$_\alpha $ and by the radio excess. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... ones[*]
From Eq. (14) with ratios of (5/2.5)2 and (5/0.4)2. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... O-dwarfs[*]
A detailed UV-analysis by Martins et al. (2004) showed the mass-loss rates of young late-O dwarfs in N81 (SMC) to be significantly smaller (factors 10 to 100) than theory predicts (see also Bouret et al. 2003). In the Galaxy, the same dilemma applies to the O9V 10 Lac (Herrero et al. 2002) and maybe also for $\tau$ Sco (B0.2V), which show very low mass-loss rates. 
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Copyright ESO 2008