## Spectral analysis of early-type stars using a genetic algorithm based fitting method

M.R. Mokiem$^1$, A. de Koter$^1$, J. Puls$^2$, A. Herrero$^{3,4}$, F. Najarro$^5$ and M.R. Villamariz$^3$

1 - Astronomical Institute Anton Pannekoek, University of Amsterdam,
2 - Universit\"ats-Sternwarte M\"unchen
3 - Instituto de Astrof\'{\i}sica de Canarias
4 - Departamento de Astrof\'{\i}sica, Universidad de La Laguna,
5 - Instituto de Estructura de la Materia, Consejo Superior de Investigaciones Cient\'{\i}ficas

We present the first automated fitting method for the quantitative
spectroscopy of O- and early B-type stars with stellar winds. The
method combines the non-LTE stellar atmosphere code {\sc fastwind}
from Puls et al.\ (2005) with the genetic algorithm based optimizing
routine {\sc pikaia} from Charbonneau (1995), allowing for a
homogeneous analysis of upcoming large samples of early-type stars
(e.g.\ Evans et al.\ 2005). In this first implementation we use
continuum normalized optical hydrogen and helium lines to determine
photospheric and wind parameters. We have assigned weights to these
lines accounting for line blends with species not taken into account,
lacking physics, and/or possible or potential problems in the model
atmosphere code. We find the method to be robust, fast, and
accurate. Using our method we analysed seven O-type stars in the young
cluster Cyg OB2 and five other Galactic stars with high rotational
velocities and/or low mass loss rates (including 10~Lac, $\zeta$~Oph,
and $\tau$~Sco) that have been studied in detail with a previous version
of {\sc fastwind}. The fits are found to have a quality that is comparable
or even better than produced by the classical by eye'' method. We
define errorbars on the model parameters based on the maximum
variations of these parameters in the models that cluster around the
global optimum. Using this concept, for the investigated dataset we
are able to recover mass-loss rates down to $\sim$ $6 \times 10^{-8}$
$M_{\odot}{\rm yr}^{-1}$ to within an error of a factor of two,
ignoring possible systematic errors due to uncertainties in the
continuum normalization. Comparison of our derived spectroscopic
masses with those derived from stellar evolutionary models are in very
good agreement, i.e.\ based on the limited sample that we have studied
we do not find indications for a mass discrepancy. For three stars we
find significantly higher surface gravities than previously
reported. We identify this to be due to differences in the weighting
of Balmer line wings between our automated method and by eye''
fitting and/or an improved multidimensional optimization of the
parameters. The empirical modified wind momentum relation constructed
on the basis of the stars analysed here agrees to within the error
bars with the theoretical relation predicted by Vink et al.\ (2000),
including those cases for which the winds are weak (i.e.\ less than a
few times $10^{-7}$ $M_{\odot}{\rm yr}^{-1}$).

Reference: A\&A in press
Status: Manuscript has been accepted