Quantitative Spectroscopy of Supergiants

Blue supergiants of spectral types B and A are the visually brightest stars in spiral and irregular galaxies, with their most luminous members (at M_V≃-10) outshining entire dwarf galaxies. This characteristic allows us to use them as probes to study the local universe in great detail. In principle, already existing large telescopes and instrumentation facilitate quantitative spectroscopy of these objects as far as the Virgo and Fornax clusters of galaxies. Beyond their challenging stellar atmospheres and opportunities for testing sophisticated non‐LTE physics, they offer numerous applications to modern astrophysics. Quantitative spectroscopy of supergiants will contribute to improving our understanding of massive‐star evolution. Galactic abundance gradients and abundance patterns, as will be obtained from studies of large ensembles of supergiants in our own and other galaxies, will foster the understanding of galactochemical evolution. Finally, they are promising independent indicators for calibrating the extragalactic distance scale, by application of the wind momentum–luminosity and the flux‐weighted gravity–luminosity relationships (R. P. Kudritzki et al. 1999, A&A, 350, 970; R. P. Kudritzki, F. Bresolin, & N. Przybilla 2003, ApJ, 582, L83).

In view of this large potential, the objective of this thesis is to improve the status of quantitative spectroscopy of BA‐type supergiants and to provide first applications on a sample of Galactic and extragalactic targets. It is shown that among the model atmospheres available at present, the best suited for analyses of supergiants are line‐blanketed classical LTE models. An investigation of the impact of various parameters such as helium abundance and line blanketing on the atmospheric structure shows that for the most luminous objects, an accurate treatment of these parameters is essential for a quantitative analysis, whereas the less luminous supergiants react less sensitively. Spectrum synthesis is used to model the line spectra. It is the only technique capable of providing analyses of spectra of different qualities from low to high resolution and able to cope with heavy line blending at a broad range of signal‐to‐noise ratios. The stellar parameters are determined from purely spectroscopic indicators, from temperature‐ and gravity‐sensitive ionization equilibria and the Balmer line wings. Elemental abundances are derived by modeling individual spectral features. Several tens of thousands spectral lines from 28 chemical species are included in the line formation, allowing almost the entire observed spectra to be reproduced. Non‐LTE effects become important in blue supergiants, where a strong radiation field at low particle densities favors deviations from LTE. Comprehensive model atoms are therefore constructed for C i/ii, N i/ii, O i, and Mg i/ii in order to determine non‐LTE level populations (N. Przybilla, K. Butler, & R. P. Kudritzki 2001, A&A, 369, 936; N. Przybilla & K. Butler 2001, A&A, 369, 955; N. Przybilla et al. 2000, A&A, 359, 1085; N. Przybilla et al. 2001, A&A, 369, 1009). Highly accurate radiative and collisional atomic data recently determined for astrophysical and fusion research using the R‐matrix method in the close‐coupling approximation are incorporated. In addition, model atoms for H, He i, O ii, S ii/iii, Ti ii, and Fe ii are adopted from the literature, the atomic data being updated to more modern values in some cases. Thus, an improved treatment for the main elements of astrophysical interest is achieved.

Extensive testing of the atomic data is performed for the nearby bright main‐sequence standard Vega (A0 V), at well‐determined stellar parameters and atmospheric structure. A high‐resolution and low‐noise spectrum with large wavelength coverage from the visual to the near‐IR is used for this purpose. Further tests are performed for the Galactic supergiants η Leo, HD 111613, HD 92207, and β Ori, with similar high‐quality spectra in order to study the non‐LTE effects across the parameter space. Accurate and consistent stellar parameters are derived for these. Non‐LTE ionization equilibria of several elements—typically N i/ii, O i/ii, Mg i/ii, and S ii/iii—agree simultaneously, provided that a realistic treatment of line blocking is used. These parameters also constitute important input data for further studies of the stellar winds of these objects. Accounting for non‐LTE reduces the random errors and removes systematic trends in the analyses. In particular, the improved treatment of electron collisions largely removes longstanding discrepancies in analyses of lines from different spin systems of a given ion. The computed non‐LTE line profiles fit the observations well for the different species at a given elemental abundance. In the parameter range covered, all lines from He i, C i/ii, N i/ii, O i/ii, and S ii/iii are significantly strengthened by non‐LTE effects; Mg ii remains almost unaffected, except for the strongest lines, and non‐LTE weakening is found for the lines of Mg i, Ti ii, and Fe ii in supergiants. The nature of the non‐LTE effects is investigated: non‐LTE strengthening generally occurs in conjunction with a strong overpopulation of metastable energy levels, while non‐LTE weakening is due to overionization of minor ionic species. In extreme cases, as for several strong lines of N i and O i, non‐LTE abundance corrections up to a factor of 50–100 are found, whereas typical mean non‐LTE abundance corrections for the diagnostic lines are within a factor of up to 3. Estimates of the systematic uncertainties in the non‐LTE abundance analysis of CNO and Mg are provided. Accounting for these and random errors, it is shown that absolute elemental abundances can be derived with accuracies of ∼0.1–0.25 dex (1 σ uncertainties), depending on the element, in contrast to ∼0.2–0.3 dex (only random contribution) typically achieved in previous studies. The statistical significance of the analyses is also largely improved because of the large wavelength coverage of the spectra, with many lines per element being available. In addition, hitherto unaccounted effects on metal line strengths are found from the consistent treatment of microturbulence in the non‐LTE computations and line formation.

The abundance analysis for Vega confirms its status as a mild λ Bootis star: the light elements, CNO, show an underabundance of ∼0.25 dex when compared to the solar composition, while the heavier elements are depleted by ∼0.55 dex. All four Galactic supergiants have metallicities close to solar. The non‐LTE abundances for individual heavier elements in a star cluster around a mean offset to the solar composition, whereas in a classical LTE analysis misleading "abundance patterns" are seen. In particular, LTE analyses tend to overestimate the abundances of the α‐process elements and to underestimate the iron group abundances; this systematic effect strengthens with increasing luminosity. It might be suspected that other elements with an atomic structure comparable to the species investigated are susceptible to similar non‐LTE mechanisms; these need to be quantified in future studies. The increasing scatter of individual abundances around a mean value with increasing stellar effective temperature and luminosity can be interpreted in terms of neglected effects in the atmospheric modeling. Non‐LTE effects on the atmospheric structure become more pronounced at higher temperature, while sphericity and hydrodynamical outflow velocity fields are noticeable only for very luminous objects.

A similar analysis is performed on high‐resolution spectra of supergiants in nearby Local Group galaxies, although at lower signal‐to‐noise ratios. VLT/UVES and Keck/HIRES spectra are available for this (K. A. Venn et al. 2000, ApJ, 541, 610; K. A. Venn et al. 2001, ApJ, 547, 765). One early A‐type supergiant near the center of the dwarf irregular galaxy NGC 6822 is found to be metal poor by ∼0.55 dex, confirming similar values from the literature, from both stellar and H ii region studies. The non‐LTE overionization of Ti ii shows a distinctive strengthening at this low metallicity. Two objects of early‐A spectral type are analyzed in M31, both located at a galactocentric distance of ∼12 kpc. The objects have abundances compatible with and slightly above solar, also in good agreement with measurements from close‐by H ii regions in M31. The non‐LTE effects follow the same trends as in the Galactic counterparts.

While indications are found that the abundances of the heavier elements in the supergiants cluster around a mean value, distinctive abundance patterns for the light elements are derived, with enriched He and N, depleted C, and O being compatible with the heavier element abundances. The predictions for chemical mixing from recent stellar evolution models, accounting for mass loss and rotation, can be verified; good agreement is found for the He enrichment and the N/C ratios in the Galactic supergiants, whereas the observed N/O ratios are lower than those from the models. With the presently achieved accuracy in the abundance determination, different stages of stellar evolution—first crossing of the Hertzsprung‐Russell diagram from the main sequence to the red versus a blue‐loop scenario with first dredge‐up abundances—can be distinguished with confidence. From the Galactic sample, η Leo is apparently undergoing a blue loop, while the other stars have directly evolved from the main sequence. In the NGC 6822 and M31 supergiants, similar He enrichments and unchanged O abundances are derived. However, as the diagnostic lines of the main indicators for mixing, C and N, are not covered by the available spectra, further conclusions on their evolutionary status cannot be drawn at present.

Low‐resolution spectra of supergiants in the Sculptor Group spiral galaxy NGC 300 and in NGC 3621 in the field, at distances of 2.0 and 6.6 Mpc—well beyond the Local Group—were recently obtained with FORS1 on the VLT (F. Bresolin et al. 2002, ApJ, 567, 277; F. Bresolin et al. 2001, ApJ, 548, L159). In order to perform quantitative analyses of these stars, the applicability of the spectrum synthesis technique is tested on the most luminous Galactic sample star, HD 92207, its spectrum being artificially degraded to FORS1 resolution. It is shown that the stellar parameters can be determined with sufficient accuracy from the spectral classification and the Balmer line strengths to constrain the metallicity to ±0.2 dex. Also, individual abundances for a few key elements of astrophysics can be determined at this resolution. Two early A‐type supergiants in NGC 3621 and two similar objects in NGC 300 are studied, each among the most luminous stars in these galaxies. The objects have metallicities of slightly subsolar to ∼0.2 times solar, typically, in accordance with expectations from their position in their host galaxies. The stellar metallicities agree with literature data on abundance gradients for these galaxies, as derived from H ii regions; the reddening of the objects is compatible with being due to the Galactic foreground. Analyses of stars at such distances are performed for the first time.