More than 80% of the stars in the Universe are expected to have initial masses below eight to ten times the mass of our sun. These low mass stars, including our sun, become cool red giants during one of the final evolutionary stages of their life: the Asymptotic Giant Branch (or AGB) phase. AGB stars are among the main producers of carbon and heavy (s-process) elements in the Universe. These elements are synthesized inside the star and mixed to the stellar atmosphere where stellar winds are responsible for the loss of more than 50% of the stellar mass, hence, AGB stars are strong polluters of the interstellar medium. The ejected material can clump together into dusty particles which may serve as ingredients for the birth of new stars and planets. When most of the AGB stellar envelope is lost, the AGB star stops releasing nuclear energy from interior processes and swaps its giant face for a planetary nebulae look, whereafter it fades away as a white dwarf.<p><p>The dredge-up of carbon and s-process elements into the AGB atmosphere causes an important chemical anomaly among them: initial oxygen-rich stars (M stars) are transformed into carbon-rich stars (C stars). As a consequence, a group of oxygen-rich AGB stars exists which makes the transition between M and C stars. These transition stars are classified as S.<p><p>Although AGB stars are identified as producers of heavy elements, their nucleosynthesis and mixing processes are weakly constrained due to large uncertainties on their estimated temperature, gravity and chemical composition. Stronger constraints on the atmospheric parameter space, hence interior processes, of AGB stars can be obtained by investigating the atmosphere of S stars. Since they are transition objects on the AGB, they trace the rise of the s-process. S stars are less numerous than C stars, but their optical spectra are brighter making it easier to identify atomic and molecular lines. Therefore, S stars belong to the most interesting objects along the AGB to perform this task.<p><p><p><p>From a practical point of view, the spectra of S stars are extremely difficult to study since they are dominated by different, overlapping molecular bands, and the spectral shape may vary strongly from star to star due to their transition status. Therefore, tailored model atmospheres for S stars are of utmost importance to understand the spectroscopic, and even photometric, changes in terms of variations in the atmospheric parameters. A comparison between the models and observations aims not only at constraining the atmospheric parameter space of S stars, it will also test the reliability of 1D state-of-the-art model atmospheres for such complex stars.<p><p><p><p>From an evolutionary point of view, the S-star family is contaminated with stars who gained their atmospheric enrichment in heavy elements from a companion star. Evidences were found that these binary S stars are not at all located on the AGB, hence, they are labelled as extrinsic S stars while S stars on the AGB are labelled as intrinsic. The difference in evolutionary stages between intrinsic and extrinsic S stars was already found 20 years ago, however, a separation in terms of surface temperature, gravity and chemical composition is not well-established due to the lack of S-star model atmospheres. Such a distinction in atmospheric parameters will facilitate the discovery of these intruders and even help to calibrate stellar evolutionary models of single and binary stars.<p>To achieve these goals, the first step consists in the construction of a grid of model atmospheres for S stars. The grid will be used to quantify the influence of atmospheric parameters on the model structure and emergent flux. These results will be analyzed to derive precise atmospheric parameters of observed S stars, using a set of well-defined photometric and spectroscopic indices. Once the best model atmosphere has been selected for all observed S stars, their atmospheric parameters will be discussed in view of their evolutionary stage. The best-fitting model atmosphere will also be used to derive abundances from spectral syntheses. The abundance profiles are compared with stellar evolution model prediction to constrain nucleosynthesis and mixing processes inside S stars. Derived abundances of unstable elements will be used to estimate, for the first time, the age of AGB stars. Finally, their abundance profile will be discussed as a function of their time spent on the AGB. / Doctorat en Sciences / info:eu-repo/semantics/nonPublished

Surface brightness asymmetries are a very common feature of stars. Among other effects they cause a difference between the projected centre of mass and the photocentre. The evolution of those surface features makes this difference time-dependent. In some cases the displacement can be a non-negligible fraction of the star radius R, and if R>1 AU, of the parallax. We investigate the impact of surface brightness asymmetries on the Gaia astrometric solution and on the data processing flow. In particular we derive analytical expressions for the change in the derived astrometric parameters for a single-star, with respect to the parameters for a uniformly-bright star, as a function of the characteristics of the surface brightness asymmetries. These predictions are confirmed by the results of the processing of simulated astrometric Gaia data where a photocentre motion caused by surface brightness asymmetries has been added using a Gaussian Markovian model.<p>In the case of a red supergiant star, the average photocentre shift is about 0.1 AU. Such a photocentric noise translates in a 10% inaccuracy on the parallax (independently of the distance), which becomes larger than the statistical error on the parallax derived from the data reduction for stars that are up to about 4 kpc away. For the most nearby stars, we derive an inaccuracy on the parallax that can be 10 times its statistical error. Finally we estimate that up to about 4000 stars among red supergiants and bright giants may have astrometric parameters that are inaccurate at levels bigger than expected because of the surface brightness asymmetries. In the determination of this number, a crucial role is played by the Gaia observable magnitude range. The fact that Gaia will not observe stars brighter than 5.6 in the Gaia G band means that the closest stars will not be observed. Yet, the impact of the surface brightness asymmetries is proportional to their angular size, meaning that the stars whose astrometric accuracy would be most affected are not observed.<p>Various non-Gaussian spot models (as applicable in the case of magnetic spots) have been implemented and analytical predictions for the effects of such magnetic spots are computed for the most representative classes of magnetic stars.<p>Another effect of the presence of surface brightness asymmetries is their impact on Gaia data processing flow. The quality of the fit of the data is evaluated with the F2 parameter that is a transformation of χ2 such that it has a unit normal distribution when the model is adequate and it is independent of the number of measurements. If the goodness-of-fit F2 of the single-star solution is not good enough (F2>3), a chain of solution of growing complexity is tried until a satisfactory one (with F2<3) is obtained. If no good solution is found, a so-called stochastic solution is computed where a "cosmic" error is added to the data in order to obtain a single-star solution with F2=0. We show that the photocentre noise induces an increase in the goodness-of-fit parameter, causing this chain of solutions to be entered. Depending on the characteristics of the photocentre noise, a variable fraction of the stars in our simulations end up with a non-single-star solution. Yet, we show that these (orbital) solutions are not acceptable because non-significant or non-physical. Finally, an important fraction of stars is assigned a stochastic solution with a cosmic noise matching well the photocentric noise.<p><p>/<p><p>Les asymétries de brillance de surface sont une caractéristique commune des étoiles. Parmi d'autres effets, elles provoquent une différence entre la projection du centre de masse et le photocentre. L'évolution de ces structures de surface rend cette différence variable avec le temps. Dans certains cas, le déplacement du photocentre peut être une fraction non négligeable du rayon de l'étoile R et, si R>1 UA, de la parallaxe. Nous examinons l'impact des asymétries de brillance de surface sur la solution astrométrique de Gaia et sur le processus de traitement des données. En particulier nous dérivons des expressions analytiques pour le changement des paramètres astrométriques déerivées pour une étoile simple, par rapport aux paramètres pour une étoile uniformément lumineuse, en fonction des caractéristiques des asymétries de brillance de surface. Ces prévisions sont confirmées par les résultats de simulations du traitement des données astrométriques de Gaia, auxquelles des mouvements du photocentre causés par des asymétries de brillance de surface ont été ajoutés en utilisant un modèle gaussien markovien.<p><p>Dans le cas d'une étoile super-géante rouge, le décalage moyen du photocentre est d'environ 0.1 UA. Un bruit photocentrique de cette amplitude se traduit dans une imprécision de 10% sur la parallaxe (indépendamment de la distance), qui peut devenir plus grande que l'erreur statistique sur la parallaxe déerivée par la réduction des données, pour les étoiles plus proches d'environ 4 kpc. Pour les étoiles les plus proches, nous évaluons une imprécision sur la parallaxe qui peut être 10 fois leur erreur statistique. Finalement, nous estimons que jusqu'à environ 4000 étoiles parmi les super-géantes rouges et géantes brillantes peuvent avoir des paramètres astrométriques inexactes à des niveaux plus grands que prévu en raison des asymétries de brillance de surface. Dans la détermination de ce nombre, la gamme de magnitudes observables par Gaia joue un rôle crucial. Le fait que Gaia n'observera pas les étoiles plus brillantes que 5.6 mag (en bande Gaia) signifie que les étoiles les plus proches ne seront pas observées. Pourtant, l'impact des asymétries de brillance de surface est proportionnel à leur taille angulaire, ce qui signifie que les étoiles dont la précision astrométrique seraient la plus affecté ne seront pas observées.<p>Différents modèles de taches ont été réalisés et des prédictions analytiques pour les effets de ces taches magnétiques sont calculés pour les classes les plus représentatives des étoiles magnétiques. <p>Un autre effet de la présence des asymétries de brillance de surface est leur impact sur le traitement des données de Gaia. La qualité de l'ajustement des données est évaluée avec le paramètre F2 qui est une transformation de χ2 telle qu'il ait une distribution normale lorsque le modèle est adéquat. Si la qualité de l'ajustement F2 de la solution étoile-simple n'est pas acceptable (F2>3), une chaîne de solutions de complexité croissante est essayée jusqu'à ce qu'une solution satisfaisante (avec F2<3) soit obtenue. Si aucune solution satisfaisante n'est trouvée, une solution dite stochastique est calculée où une erreur "cosmique" est ajoutée aux données afin d'obtenir une solution étoile-simple avec F2=0. Nous montrons que le bruit du photocentre induit une augmentation de F2, ce qui provoque l'activation de cette chaîne de solutions. Selon les caractéristiques du bruit du photocentre, une solution étoile-non-simple est obtenue pour une fraction variable des étoiles dans nos simulations. Nous montrons que ces solutions (orbitales) ainsi obtenues ne sont pas acceptables car non significatives ou non-physiques. Enfin, une fraction importante d'étoiles se voient attribuer une solution stochastique avec un bruit cosmique correspondant au bruit photocentrique. / Doctorat en Sciences / info:eu-repo/semantics/nonPublished