In this thesis I present work using the MOPED algorithm to extract in a non-parametric fashion star formation histories and galaxy masses from the spectra of galaxies in the Sloan Digital Sky Survey. The recovered parameters for all galaxies are combined to give insight into the processes of star and galaxy formation on both individual galaxy and cosmic scales. The MOPED algorithm allows use of the entire spectral range, rather than concentrating on specific features, and can be used to estimate the complete star formation history without prior assumptions about its form. By combining the star formation histories of 96,545 galaxies in the redshift range 0 < z < 0:34 the cosmic star formation rate is determined from the present day to z ~ 6. The results show that the peak of star formation occurred at z ~ 0:6, and that 26% of the mass of stars in the present-day Universe was formed at z ~ 2. The average metallicity rises from Z/Z= 0:44 at high redshift to a peak of 0:8 at z ~ 1 before declining to a level around 0.25 atthe present day. Although the peak in star formation is more recent than previously thought, the sample used includes galaxies with a range of masses not accessible to traditional studies, down to a limit of L ~ 2 x 10-3L*. By cutting the sample into ranges of mass it can be seen that the redshift at which starformation activity peaks is an essentially monotonically increasing function of final stellar mass. The time of the peak in star formation ranges from z > 2 for the highest mass galaxies (MS < 1012M) to z ~ 0:2 for the lowest (MS < 1010M). A typical L* galaxy appears to have its peak at around z » 0:8. These differences in star formation with mass reconcile the redshift of the peak found in this work with the previous estimates, generally deep surveys only probe the SFR of galaxies with MS < ML*. The stellar mass calculated using the reconstructed spectra eliminates contamination from either emission lines or AGN components. Using these masses it is possible to construct the mass function for the stellar mass component of galaxies which give excellent agreement with previous works, but extend their range by more than two decades in mass to 10 7.5 < Ms/h-2M < 1012. I present both a standard Schechter fit and a fit modified to include an extra, high-mass contribution, possibly from cluster cD galaxies. The Schechter fit parameters are phi* = (7:8 +/- 0:1) £ 10-3h3Mpc-3, M* = (7.64 +/- 0.09) x 10*10h-2M and alpha = -1.159 +/- 0.008. The sample also yields an estimate for the contribution from baryons in stars to the critical density of omega b*h = (2.39+/-0.08)x10-3, in good agreement with other indicators. No evolution of the mass function in the redshift range 0:05 < z < 0:34 is apparent, indicating that almost all stars were already formed at z » 0:34 with little or no star formation activity since then and that the evolution seen in the luminosity function must be largely due to stellar fading. The star formation history can be interpreted as a measure of how gas was transformed into stars as a function of time and stellar mass: the Baryonic Conversion Tree (BCT). There is a clear correlation between early star formation activity and present-day stellar mass: the more massive galaxies have formed about 80% of their stars at z > 1, while for the less massive ones the value is only about 20%. Comparing the BCT to the dark matter merger tree indicates that star formation efficiency at z > 1 had to be high (as much as 10%) in galaxies with present-day stellar mass larger than 2 x 10*11M, if this early star formation occurred in the main progenitor. The LCDM paradigm can accommodate a large number of red objects; it is the high efficiency in the conversion from gas to stars that needs to be explained. On the other hand, in galaxies with present-day stellar mass less than 10*11M, efficient star formation seems to have been triggered at z ~ 0:2. This work shows that there is a characteristic mass (M » 10*10M) for feedback efficiency (or lack of star formation). For galaxies with masses lower than this, feedback (or star formation suppression) is very efficient while for higher masses it is not. The BCT, determined here for the first time, should be an important observable with which to confront theoretical models of galaxy formation.

Le décalage vers le rouge est la principale mesure par laquelle les astronomes peuvent cartographier l’Univers dans la direction radiale. Afin de tester les hypothèses d’homogénéité et d’isotropie, il est nécessaire de mesurer avec precision le décalage vers le rouge d’un grand nombre de galaxies. De plus, différents modèles cosmologiques ne peuvent être distingués qu’au travers d’une analyse précise des structures à grandes échelles tracées par ces galaxies. Pour obtenir un grand nombre de ces mesures, il est nécessaire de mener d'importantes campagnes d’observations pour établir des relevés couvrant une large portion du ciel. Ces mesures trouvent aussi d’autres applications en astronomie comme par exemple l’analyse du cisaillement gravitationnel, la calibration des mesures photométriques, l’étude des halos de matière noire, de la morphologie des galaxies, des structures à grandes échelles et de la distribution des galaxies.Dans tous les relevés de galaxies, les mesures les plus problématiques sont pour les objets de plus faible luminosité, où le bruit instrumental devient gênant, et qui se trouvent être en général les objets les plus lointains. Pour ces objets, les mesures de décalages vers le rouge peuvent souvent devenir imprécise et, la plupart du temps, elles sont simplement exclues de l’analyse en appliquant des coupures en magnitudes ou en rapport signal à bruit. Cette procédure est une méthode brutale pour séparer les mesures probablement imprécises des mesures fiables.Dans cette thèse, nous développons un algorithme permettant la mesure du décalage vers le rouge des spectres de galaxies dans le regime de faible rapport signal à bruit. La première partie de cette thèse présente les différents concepts relatifs à l’estimation du décalage vers le rouge et au débruitrage de signaux par transformation en ondelettes et filtrage par taux de fausse détection (False Detection Rate, FDR en anglais). La seconde partie détaille comment ces concepts sont mis à contribution dans l’élaboration de l’algorithme Darth Fader (Denoised and Automatic Redshifts THresholded with a False DEtection Rate). Enfin, la dernière partie présente l’application de cet algorithme à des données synthétiques générées à partir du COSMOS Mock Catalogue, mais aussi sur des données réelles tirées du relevé WiggleZ.Nous montrons que Darth Fader fonctionne efficacement à bas rapport signal à bruit étant donné un choix approprié du taux de fausse détection et d’un critère de comptage de traits caractéristiques judicieux. Nous montrons aussi que Darth Fader permet d’éliminer le continuum des spectres à bas rapport signal à bruit, ce qui rend possible l’estimation du décalage vers le rouge par corrélation croisée. Enfin, nous montrons sur des spectres de test issues du relevé WiggleZ que notre algorithme est capable d’analyser une part importante du relevé de façon autonome avec une haute précision, sans nécessiter d'inspection visuelle (alors que les données WiggleZ ont à l’origine été entièrement soumises à l’inspection visuelle). En conclusion, Darth Fader est un algorithme prometteur pour l’analyse des grands relevés de galaxies, en particulier pour exploiter les objets à faible rapport signal à bruit qui sont habituellement simplement ignorés. / Summary: Redshift is the primary measure by which astronomers can map the Universe in the radial direction. In order to test the assumptions of homogeneity and isotropy, accurate redshifts of galaxies are needed, and for a great many of them. Additionally different cosmological models can only be distinguished by careful observations of the large scale structure traced by these galaxies. Large sky surveys are the only mechanism by which redshifts for a large number of galaxies can be obtained. Accurate redshift estimation is additionally required for many other fields of astronomy including but not limited to: weak lensing, studies of dark matter haloes, galaxy morphology studies, chemical evolution studies, photometric calibration, and studies of large scale structure and galaxy clustering.Problems exist in all surveys at the dim limit of observation, which usually corresponds to the higher redshift objects in the survey, where noise becomes problematic. Magnitude or signal-to-noise ratio cuts are often employed in order to eliminate potentially troublesome objects; such a procedure is a blunt tool for separating good redshift candidates from ones likely to be inaccurate.In this thesis we develop an algorithm to tackle redshift estimation of galaxy spectra in the low signal-to-noise regime. The first part of this thesis introduces the concepts of denoising, particularly False Detection Rate denoising, wavelet transforms and redshift estimation algorithms. The second part details how these concepts are united into the Darth Fader (Denoised and Automatic Redshifts THresholded with a FAlse DEtection Rate) algorithm. The final parts of this thesis apply the algorithm both to idealised synthetic data generated from the COSMOS Mock Catalogue, and to a subset of real data from the WiggleZ survey.We show that Darth Fader can operate effectively at low signal-to-noise given an appropriate choice of FDR parameter for denoising, and an appropriate feature-counting criterion. We also show that Darth Fader can remove the continua of spectra effectively at low signal-to-noise for the purposes of redshift estimation by cross-correlation. Additionally we show from tests on spectra from the WiggleZ survey that our algorithm has the ability to process a substantial subset of that data without the need for visual inspection (to which the entire WiggleZ spectral survey has been subjected), and to a high degree of accuracy. We conclude that the Darth Fader algorithm has potential to be used in large-sky survey pipelines, particularly where signal-to-noise is expected to be poor.

Due to the finite speed of light and a vast, expanding universe, telescopes are just now receiving the light emitted by galaxies as they were forming in the very early universe. The light from these galaxies has been redshifted (stretched to longer, redder wavelengths) as a result of its journey through expanding space. Using sophisticated techniques and exceptional multi-wavelength optical and infrared data, we isolate a population of 378 galaxies in the process of formation when the Universe was only two billion years old. By matching the distinctive properties of the light spectra of these galaxies to models, the redshift, age, dust content, star formation rate and total stellar mass of each galaxy are determined. Comparing our results to similar surveys of galaxy populations at other redshifts, a picture emerges of the growth and evolution of massive, star-forming galaxies over the course of billions of years.

Due to the finite speed of light and a vast, expanding universe, telescopes are just now receiving the light emitted by galaxies as they were forming in the very early universe. The light from these galaxies has been redshifted (stretched to longer, redder wavelengths) as a result of its journey through expanding space. Using sophisticated techniques and exceptional multi-wavelength optical and infrared data, we isolate a population of 378 galaxies in the process of formation when the Universe was only two billion years old. By matching the distinctive properties of the light spectra of these galaxies to models, the redshift, age, dust content, star formation rate and total stellar mass of each galaxy are determined. Comparing our results to similar surveys of galaxy populations at other redshifts, a picture emerges of the growth and evolution of massive, star-forming galaxies over the course of billions of years.

Due to the finite speed of light and a vast, expanding universe, telescopes are just now receiving the light emitted by galaxies as they were forming in the very early universe. The light from these galaxies has been redshifted (stretched to longer, redder wavelengths) as a result of its journey through expanding space. Using sophisticated techniques and exceptional multi-wavelength optical and infrared data, we isolate a population of 378 galaxies in the process of formation when the Universe was only two billion years old. By matching the distinctive properties of the light spectra of these galaxies to models, the redshift, age, dust content, star formation rate and total stellar mass of each galaxy are determined. Comparing our results to similar surveys of galaxy populations at other redshifts, a picture emerges of the growth and evolution of massive, star-forming galaxies over the course of billions of years.

Due to the finite speed of light and a vast, expanding universe, telescopes are just now receiving the light emitted by galaxies as they were forming in the very early universe. The light from these galaxies has been redshifted (stretched to longer, redder wavelengths) as a result of its journey through expanding space. Using sophisticated techniques and exceptional multi-wavelength optical and infrared data, we isolate a population of 378 galaxies in the process of formation when the Universe was only two billion years old. By matching the distinctive properties of the light spectra of these galaxies to models, the redshift, age, dust content, star formation rate and total stellar mass of each galaxy are determined. Comparing our results to similar surveys of galaxy populations at other redshifts, a picture emerges of the growth and evolution of massive, star-forming galaxies over the course of billions of years.