*Astron. Astrophys. Suppl. Ser.* **141**, 23-64

**J. Puls - U. Springmann - M. Lennon**

**Send offprint request: **J. Puls

e-mail: uh101aw@usm.uni-muenchen.de

Universitäts-Sternwarte München, Scheinerstr. 1, D-81679 München, Germany

Received March 18; accepted September 28, 1999

This paper analyzes the inter-relation between line-statistics and radiative
driving in massive stars with winds (excluding Wolf-Rayets) and provides
insight into the qualitative behaviour of the well-known
force-multiplier parameters
and ,
with special
emphasis on .

After recapitulating some basic properties of radiative line driving, the correspondence of the local exponent of (almost) arbitrary line-strength distribution functions and , which is the ratio of optically thick to total line-force, is discussed. Both quantities are found to be roughly equal as long as the local exponent is not too steep.

We compare the (conventional) parameterization applied in this paper
with the so-called -formalism introduced by Gayley ([1995])
and conclude that the latter can be applied alternatively in its *most
general* form. Its "strongest form'', however (requiring the Ansatz
to be valid, with
the line-strength of the strongest line), is
justified only under specific conditions, typically for Supergiants with
K.

The central part of this paper considers the question concerning the shape of the line-strength distribution function, with line-strength as approximate depth independent ratio of line and Thomson opacity. Since depends on the product of oscillator strength, excitation- and ionization fraction as well as on elemental abundance, all of these factors have their own, specific influence on the final result.

At first, we investigate the case of hydrogenic ions, which can be treated
analytically. We find that the exponent of the
differential distribution is -4/3 corresponding to
,
as
consequence of the underlying oscillator strength distribution.
Furthermore, it is shown that for *trace* ions one stage below the major
one (e.g., H I in hot winds) the equality
is valid throughout the wind.

For the majority of non-hydrogenic ions, we follow the statistical approach suggested by Allen ([1966]), refined in a number of ways which allow, as a useful by-product, the validity of the underlying data bases to be checked. Per ion, it turns out that the typical line-strength distribution consists of two parts, where the first, steeper one is dominated by excitation effects and the second one follows the oscillator strength distribution of the specific ion.

By summing up the contributions of all participating ions, this *direct*
influence of the oscillator strength distribution almost vanishes. It turns
out, however, that there is a second, indirect influence controlling the
absolute line numbers and thus
.
From the actual numbers, we find an
average exponent of order
,
similar to the value for
hydrogen.

Most important for the shape of the *total* distribution is the
difference in line-statistics between iron group and light ions as well as
their different (mean) abundance. Since the former group comprises a large
number of meta-stable levels, the line number from iron group elements is
much higher, especially at intermediate and weak line-strengths.
Additionally, this number increases significantly with decreasing
temperature (more lines from lower ionization stages). In contrast, the
line-strength distribution of light ions remains rather constant as function
of temperature.

Since the line-strength depends linearly on the elemental abundance, this quantity controls the relative influence of the specific distributions on the total one and the overall shape. For solar composition, a much more constant slope is found, compared to the case if all abundances were equal.

In result, we find (for solar abundances) that iron group elements dominate the distribution at low and intermediate values of line-strength (corresponding to the acceleration in the inner wind part), whereas light ions (including hydrogen under A-star conditions) dominate the high end (outer wind). Typically, this part of the distribution is steeper than the rest, due to excitation effects.

Finally, the influence of *global* metallicity *z* is discussed.
We extend already known scaling relations (regarding
mass-loss, terminal velocity and wind-momentum rate) with respect to
this quantity. In particular, we demonstrate that, besides the
well-known direct effect (
), the curvature
of the line-strength distribution at its upper end induces a decrease of
for low metallicity and/or low wind density.

Summarizing the different processes investigated, the force-multiplier
parameter
becomes a decreasing function of decreasing
,
increasing
and decreasing *global* metallicity *z*,
consistent with the findings of earlier and present empirical results and
observations.

**Key words: **atomic data -- stars: atmospheres -- stars: early type --
stars: mass-loss

- 1 Introduction
- 2 The line-force from an ensemble of lines and the line-strength distribution function
- 3 Line-strength distribution functions for stellar wind conditions and the validity of Gayley's Ansatz
- 4 What determines the slope?
- 5 Metallicity effects, thin winds and scaling relations
- 6 Summary and discussion
- Appendix A: Line force and local slope
- Appendix B: Line-strength distribution with exponential cutoff.
- Appendix C: On the difference of and
- Appendix D: LTE line-strength distribution for hydrogenic ions
- Appendix E: Frequency-integration of the line intensity distribution function
- References

Copyright The European Southern Observatory (ESO)