L'objectif de mon travail de thèse a consisté à tenter de comprendre quels sont les mécanismes principaux à l'origine de la formation d'étoiles dans les galaxies au cours des derniers dix milliards d'années. Alors qu'il avait été proposé dans le passé que le rôle des fusions de galaxies était dominant pour expliquer l'allumage de la formation d'étoiles dans les galaxies lointaines formant leurs étoiles à de très grands taux, des études plus récentes ont au contraire mis en évidence des lois d'échelles reliant le taux de formation d'étoiles des galaxies à leur masse stellaire ou masse de gaz. La faible dispersion de ces lois semblait être en contradiction avec l'idée d'épisodes stochastiques violents de formation stellaire liés à des interactions, mais plutôt en accord avec une nouvelle vision de l'histoire des galaxies où celles-ci sont nourries de manière continue en gaz intergalactique.Nous nous sommes particulièrement intéressés à l'une de ces lois d'échelles, la relation entre le taux de formation d'étoiles (SFR) et la masse stellaire des galaxies, appelées communément la séquence principale des galaxies à formation d'étoiles. Nous avons étudié cette séquence principale, SFR-M*, en fonction de la morphologie et d'autres paramètres physiques comme le rayon, la couleur, la clumpiness. Le but étant de comprendre l'origine de la dispersion de cette relation en lien avec les processus physiques responsables de cette séquence afin d'identifier le mode principal de formation d'étoile gouvernant cette séquence. Ce travail a nécessité une approche multi-longueurs d'ondes ainsi que l'utilisation de simulations de profils de galaxies pour distinguer les différents types morphologiques de galaxies impliqués dans la séquence principale.

This thesis addresses the challenge of understanding the typical star formation histories of red sequence galaxies, using linestrength indices, mass-to-light ratios and redshift evolution as complementary constraints on their stellar age distribution. We first construct simple parametric models of the star formation history that bracket a range of scenarios, and fit these models to the linestrength indices of low-redshift cluster red-sequence galaxies. For giant galaxies, we con firm the downsizing trend, i.e. the stellar populations are younger, on average, for lower σ galaxies. We find, however, that this trend flattens or reverses at σ < 70km/s. We then compare predicted stellar mass-to-light ratios with dynamical mass-to-light ratios derived from the Fundamental Plane, or by the SAURON group. For galaxies with σ ~ 70 km/s, models with a late 'frosting' of young stars and models with exponential star formation histories have stellar mass-to-light ratios that are larger than observed dynamical mass-to-light ratios by factors of 1.7 and 1.4, respectively, and so are rejected. The single stellar population (SSP) model is consistent with the Fundamental Plane, and requires a modest amount of dark matter (between 20% to 30%) to account for the difference between stellar and dynamical mass-to-light ratios. A model in which star formation was 'quenched' at intermediate ages is also consistent with the observations, although in this case less dark matter is required for low mass galaxies. We also find that the contribution of stellar populations to the 'tilt' of the Fundamental Plane is highly dependent on the assumed star-formation history: for the SSP model, the tilt of the FP is driven primarily by stellar-population effects. For a quenched model, two-thirds of the tilt is due to stellar populations and only one third is due to dark matter or non-homology. Comparing to high redshift cluster data, we find again the SSP and quenched models, as well as a model where strangulation begins at intermediate ages after a period of constant star formation, are preferred. They predict the recent faint-end build up of the red sequence, along with observed dwarf-to-giant ratios. Only the SSP model appears to predict the observed M/L evolution, but only if selection effects are carefully modeled.

This thesis addresses the challenge of understanding the typical star formation histories of red sequence galaxies, using linestrength indices, mass-to-light ratios and redshift evolution as complementary constraints on their stellar age distribution. We first construct simple parametric models of the star formation history that bracket a range of scenarios, and fit these models to the linestrength indices of low-redshift cluster red-sequence galaxies. For giant galaxies, we con firm the downsizing trend, i.e. the stellar populations are younger, on average, for lower σ galaxies. We find, however, that this trend flattens or reverses at σ < 70km/s. We then compare predicted stellar mass-to-light ratios with dynamical mass-to-light ratios derived from the Fundamental Plane, or by the SAURON group. For galaxies with σ ~ 70 km/s, models with a late 'frosting' of young stars and models with exponential star formation histories have stellar mass-to-light ratios that are larger than observed dynamical mass-to-light ratios by factors of 1.7 and 1.4, respectively, and so are rejected. The single stellar population (SSP) model is consistent with the Fundamental Plane, and requires a modest amount of dark matter (between 20% to 30%) to account for the difference between stellar and dynamical mass-to-light ratios. A model in which star formation was 'quenched' at intermediate ages is also consistent with the observations, although in this case less dark matter is required for low mass galaxies. We also find that the contribution of stellar populations to the 'tilt' of the Fundamental Plane is highly dependent on the assumed star-formation history: for the SSP model, the tilt of the FP is driven primarily by stellar-population effects. For a quenched model, two-thirds of the tilt is due to stellar populations and only one third is due to dark matter or non-homology. Comparing to high redshift cluster data, we find again the SSP and quenched models, as well as a model where strangulation begins at intermediate ages after a period of constant star formation, are preferred. They predict the recent faint-end build up of the red sequence, along with observed dwarf-to-giant ratios. Only the SSP model appears to predict the observed M/L evolution, but only if selection effects are carefully modeled.

Proefschrift--Universiteit Leiden, 2005. / Résumé en néerlandais. Notes bibliogr.

Dissertation--Department of astronomy--Stockholm--Stockholm university, 1998. / Notes bibliogr.

Dissertation--Department of astronomy--Stockholm--Stockholm university, 1998. / Notes bibliogr.

Proefschrift--Rijksuniversiteit Groningen, 2000. / Résumé en néerlandais. Notes bibliogr.

Diss.--Bonn Universität, 2000. N°: D 98. / Bibliogr. p. 117-123.

Proefschift--Universiteit Leiden, 2004.

Thesis (Ph. D.)--University of Oregon, 2000. / Typescript. Includes vita and abstract. Includes bibliographical references (leaves 94-96). Also available for download via the World Wide Web; free to University of Oregon users.