Concepciones y modelos mentales acerca del aprendizaje y la evaluación en ciencias del

and future work

Using an improved model code, EUV spectral energy distributions (SEDs) have been computed for a lar ge grid of stellar models spanning the parameter range observed for O and early B stars. These SEDs have been incorporated into an evolutionary population synthesis code to investigate the time-depen- dence of the integrated SEDs from evolving clusters of massive stars. Purpose of these calculations is to provide a crucial ingredient for the simulations of the photoionized gas in star-for ming regions, which then yield information about the star-for mation history of observed clusters.

The new method used for computing the SEDs renders the influence of spectral lines on the EUV radia- tion field in identical quality as the high-resolution synthetic spectra used for comparison with observed UV spectra.

By means of exemplary UV analyses of individual O stars it has been shown that the models reproduce most features of the observed UV spectra. As the appearance of the observable UV spectrum depends strongly on the spectral shape of the EUV radiation field, this result represents strong evidence that the computed SEDs are on a realistic level, an essential requirement for their application in photoionization calculations. Some minor discrepancies still remain, however, to be resolved in future work.

The mass loss rates and terminal velocities from models with consistently calculated hydrodynamics have been shown to reproduce the theoretically predicted wind-momentum–luminosity relation, as well as the predicted metallicity dependence thereof, showing no distinct differences for dwarfs and supergiants. Comparison of the observed UV spectra of a sample of galactic O stars with the synthetic spectra of two sets of models, one based on selfconsistent hydrodynamics, the other on wind parameters derived from an analysis of optical lines, shows discrepancies that are consistent with the scenario of a fragmented stellar wind, although an in-depth investigation of other possible explanations, such as non-solar abun- dance patterns, remains to be perfor med. (This will require a detailed spectral analysis and comparative

study of a sample of Galactic, LMC, and SMC stars.) The different relations previously obtained for dwar fs and supergiants from analyses of the Hα line might therefore be the result of inadequate assump- tions made in modelling the optical lines.

Futur e work

The overall good agreement of the synthetic spectra from the current generation of stellar models with the observations indicates that phenomena pertaining to deviations from symmetry and stationarity, such as the time-dependent phenomena (driving instabilities) and the influence of rotation, are obviously sec- ond-order effects. Our short to medium term research plans in the field of hot stars will therefore instead focus on applications such as the following:

1. A detailed comparison of the stellar and wind parameters derived from fits to the UV spectra via models with consistent dynamics and those determined from analyses of the optical lines that is unhampered by the current uncertainties in the distances to the individual stars. Most probably this will involve a sample of stars in the LMC or SMC, where the relative error in the distances is much smaller than for stars in our galaxy.

2. Using the population synthesis methods to calculate integrated spectra for a large range of stellar metallicities from synthetic UV spectra (using the same technique as described in chapter 6, but focusing on the observable UV range), for the determination of stellar abundances and the physical properties of the most UV-luminous stars in star-for ming galaxies at high redshifts. A first tentative step in this direction has already been undertaken by Mehlert et al. (2001), comparing the observed spectrum of a high-redshift galaxy with the synthetic spectrum of a representative massive star of one-fifth solar metallicity.

3. Computing models for Population III stars, the very first generation of very massive, extremely low to zero metallicity stars, which are suspected to have played a significant role in the reionization of the universe. Questions to be addressed regard in particular the dynamics of stellar winds at such low metallicities.

Improvements and refinements planned in this regard for the atmosphere modelling comprise: 1. A more consistent treatment of shock emission.

Shocks have so far been mostly ignored in the computations of spectral energy distributions of stel- lar models used in photoionization calculations of HIIregions or planetary nebulae, due to the lack

of a consistent quantitative theory describing distribution and strength of the shock emission. As shocks, however, can increase the number of He++-ionizing photons by several orders of magnitude, at least an approximate inclusion of shock emission in the models is necessary. We are currently implementing an improved semi-empirical description of shock emission, a preliminary version of which has already in the case of ζ Puppis been shown to simultaneously reproduce the observed

ROSAT spectrum as well as the influence on the observed visible UV range (Pauldrach, Hoffmann,

and Lennon 2001). It is suspected that a similar simultaneous fit of both X-ray and UV spectra for a series of stars with existing ROSAT observations should enable us to correlate the shock emission with the fundamental stellar parameters and/or give us the necessary expertise to obtain reasonable estimates of the shock emission on basis of the visible UV spectra alone. This knowledge can then be incorporated into an improved systematic computation of stellar SEDs.

2. A comoving-frame treatment of important hydrogen and helium lines.

To compute the NLTE line transition rates, we presently use the Sobolev-with-continuum method. This is an excellent approximation for most lines and allows a fast numerical treatment of (currently) over ten thousand line transitions in the computation of the occupation numbers. However, for a few tens of lines (among which are some H and He lines) it is known to be an imperfect approxima- tion (e.g, Sellmaier et al. 1993). This has not been a major issue in the past, but for future planned analyses a more rigorous treatment of these particular lines will become necessary for two reasons:

7. Summary, conclusions, and future work 81

a. With decreasing metallicity, the hydrogen and helium lines gain in relative importance, and at very low metallicities it is essentially these comparatively few lines which drive the winds. Thus, to compute realistic mass loss rates at low metallicities, it is important that not only the occupa- tion numbers of the connecting levels be accurate, but also the treatment of the transfer of momentum from the radiation field to the wind material via these lines (which in our models is at present treated in Sobolev-approximation).

b. For diagnostic purposes, i.e., to determine the parameters of individual observed stars and to verify the models via a comparison to observations, we have in the past relied primarily on UV spectra. The lar ge number of resonance lines there is quite insensitive to minor approxima- tions, and the presence of lines from several different ionization stages makes them suitable as tracers of the ionization balance throughout the wind. In the centers of galaxies, however, absorption by dust severely attenuates the UV spectral range, and we must turn to optical and infrared spectra for diagnostics. The lines in the optical and infrared ranges are weaker, subor- dinate lines, and although these lines are of minor importance in establishing the ionization bal- ance, using them for diagnostic purposes requires a much more accurate treatment. In particu- lar, for several of these lines broadening effects must be considered and the line source func- tions must be computed in the comoving frame.

3. A more consistent calculation and treatment of the line force.

Mass loss in hot stars has a profound influence on the spectral energy distributions and spectra of the stars, and one of the main effects of metallicity is its influence on the mass loss rate. Knowledge of the metallicity-dependence of the mass loss rate as function of the stellar parameters is essential to obtain realistic synthetic spectral energy distributions and integrated cluster spectra derived from individual synthetic stellar spectra. Although (as shown in this thesis) the current approach already seems to accurately reproduce the theoretically predicted dependence of the wind momenta on metallicity, a more correct determination of the line force (using, e.g., detailed observer’s frame or comoving-frame calculations) must at least be investigated within the framework of our present mod- els. In particular for very low metallicities (such as expected for high-redshift systems) where obser- vational determinations of the mass loss rates do not (yet) exist and thus these parameters cannot be deduced from observations, this will be an important step in producing models with realistic spectral ener gy distributions.