The primary objective of this work is to investigate the potential of utilizing Hadoop and Amazon Redshift in the Amazon Web Services ("AWS") cloud, in order to design and implement a data warehouse, the efficacy of which will be tested afterwards. Contributions of this work include: documenting the technologies in the AWS cloud in Czech, demonstration of the design and performance tests of the data warehouse and the ETL part. Another considerable benefit is the added value to the company for whom the project was designed, and which is currently using the output of the project.

We present a measurement of the abundance of carbon monoxide in the early universe, utilizing the final results from the CO Power Spectrum Survey (COPSS). Between 2013 and 2015, we performed observations with the Sunyaev-Zel'dovich Array to measure aggregate CO emission from z similar to 3 galaxies with the intensity mapping technique. Data were collected on 19 fields, covering an area of 0.7 square degrees, over the frequency range 27-35 GHz. With these data, along with data analyzed in COPSS I, we are able to observe the CO(1-0) transition within the redshift range z = 2.3-3.3 for spatial frequencies between k = 0.5-10 h Mpc(-1), spanning a comoving volume of 4.9 x 10(6) h(-3) Mpc(3). We present estimates of contributions from continuum sources and ground illumination within our measurement. We constrain the amplitude of the CO power spectrum to P-CO = 3.0(-1.3)(+1.3) x 10(3) mu K-2(h(-1) Mpc)(3), or Delta(2)(CO)(k=1 h Mpc(-1)) = 1.5(-0.7)(+0.7) x 10(3) mu K-2, at 68% confidence, and PCO > 0 at 98.9% confidence. These results are a factor of 10 improvement in sensitivity compared to those of COPSS I. With this measurement, we constrain on the CO(1-0) galaxy luminosity function at z similar to 3. Assuming that CO emission is proportional to halo mass and using theoretical estimates of the scatter in this relationship, we constrain the ratio of CO(1-0) luminosity to halo mass to A(CO) = 6.3(-2.1)(+1.4) x 10 (7) L circle dot M circle dot-1. Assuming a Milky Way-like linear relationship between CO luminosity and molecular gas mass, we estimate a mass fraction of molecular gas of f(H2) = 5.5(-2.2)(+3.4) x 10(-2) for halos with masses of similar to 10(12)M(circle dot). Using theoretical estimates for the scaling of molecular gas mass fraction and halo mass, we estimate the cosmic molecular gas density to be rho(z similar to 3) (H-2) = 1.1(-0.4)(+0.7) x 10(8) M(circle dot)Mpc(-3).

An extensive survey of extragalactic objects with accurate distances is difficult to perform. Distance at that scale is commonly found with redshift and most easily identified with spectroscopic observations, which are time intensive. This is especially a concern when surveying for the elusive, possibly non-existent dwarf galaxies in the centers of voids, whose light would be so faint that a spectroscopic survey to find them would be unreasonably time consuming. Photometric methods to calculate redshift could provide a much quicker way to find distance, but the photometric methods used thus far are characterized by large errors that make them unusable for research on nearby voids. I have been testing two photometric redshift methods that have been recently developed: the Gaussian filter method and the ramp filters method. The Gaussian filter method isolates objects that have emission within the bandpass of the Gaussian filters. Spectroscopic follow-up observations of objects observed with the Gaussian filter method confirmed that 41.9% of the objects identified by the method actually had emission lines within the bandpass of the filters, which is more than a random sampling where only 7% of galaxies had emission within that wavelength range. Assuming that all the lines identified were Hα, the photometric redshift calculated from the Gaussian filter observations had an error of 945 km/s (or 430 km/s by excluding one outlier). Ramp filter observations of objects with SDSS redshifts provided a way to identify redshift with an error of 641.8 km/s. These errors are about an order of magnitude less than other redshift methods. These methods show promise for uses in observational cosmology, specifically extragalactic survey work and searching for void galaxies.

Galaxy clusters, the largest gravitationally-bound structures in the universe, are composed of 50-1000s of galaxies, hot X-ray emitting gas, and dark matter. They grow in size over time through cluster and group mergers. The merger history of a cluster can be imprinted on the hot gas, known as the intracluster medium (ICM). Merger signatures include shocks, cold fronts, and sloshing of the ICM, which can form spiral structures. Some clusters host double-lobed radio sources driven by active galactic nuclei (AGN). First, I will present a study of the galaxy cluster Abell 2029, which is very relaxed on large scales and has one of the largest continuous sloshing spirals yet observed in the X-ray, extending outward approximately 400 kpc. The sloshing gas interacts with the southern lobe of the radio galaxy, causing it to bend. Energy injection from the AGN is insufficient to offset cooling. The sloshing spiral may be an important additional mechanism in preventing large amounts of gas from cooling to very low temperatures. Next, I will present a study of Abell 98, a triple system currently undergoing a merger. I will discuss the merger history, and show that it is causing a shock. The central subcluster hosts a double-lobed AGN, which is evacuating a cavity in the ICM. Understanding the physical processes that affect the ICM is important for determining the mass of clusters, which in turn affects our calculations of cosmological parameters. To further constrain these parameters, as well as models of galaxy evolution, it is important to use a large sample of galaxy clusters over a range of masses and redshifts. Bent, double-lobed radio sources can potentially act as tracers of galaxy clusters over wide ranges of these parameters. I examine how efficient bent radio sources are at tracing high-redshift (z>0.7) clusters. Out of 646 sources in our high-redshift Clusters Occupied by Bent Radio AGN (COBRA) sample, 282 are candidate new, distant clusters of galaxies based on measurements of excess galaxy counts surrounding the radio sources in Spitzer infrared images.

No description available.

Les modèles et les observations actuels de l’Univers sont en accords aux grandes échelles. Néanmoins, cela ne permet pas d’expliquer la formation des structures baryoniques dans l’Univers. Afin de contraindre la formation des structures dans l’Univers, des observations sont nécessaires à différents redshifts pour observer les structures à différentes étapes d’évolutions. Le satellite Planck donne accès à des objets dont les redshifts sont compris entre z=2 et z=4. Ainsi environ 1200 objets extragalactiques susceptibles d’être à haut redshift sont détectés dans 26% du ciel, autour des pôles galactiques. Un sous échantillon de 228 sources est observé à plus grande sensibilité et résolution angulaire avec le satellite Herschel. Grâce à ces observations je peux résoudre les sources ponctuelles de Planck en plusieurs sources Herschel. J’identifie alors trois possibilités de type de sources, soit c’est une candidate de source amplifiée par lentillage gravitationnel, soit c’est un candidat d’amas de galaxies à haut redshift, soit c’est un alignement de sources fortuit le long de la ligne de visé. Après analyse, j’ai identifié dans cet échantillon 11 sources amplifiées par lentillage gravitationnel parmi les plus brillantes du ciel sub-millimétrique. Ces sources ont des redshifts compris entre z=2.2 et z=3.6 (Canameras et al., 2015) et permettent de sonder la formation stellaire dans l’Univers lointain. Il reste 217 surdensités de sources que j’ai étudié. En utilisant la couleur des sources Herschel, j’ai calculé une estimation de la densité des sources. 50% des champs Herschel de l’échantillon on une surdensité supérieur à 10sigma lorsque l’on sélectionne les sources rouges (S250/S350 < 1.4 et S500/S350 > 0.6). Cela montre que l’échantillon est principalement composé de surdensité de source rouge. En ajustant un corps noir modifié à 35K sur les données photométrique Herschel-SPIRE je montre que la distribution de redshift photométrique des sources piquent à z=2. Notre échantillon se compose donc de 11 sources lentillées et 217 potentiels amas de galaxies à redshift z~2. Ces potentiels amas de galaxies se compose d’en moyenne 9 sources dans 4.5’ de diamètre. Les sources individuelles ont des luminosités infrarouges d’environ 4.1012 Lsol ce qui correspond d’après Bell et al. 2003 à un taux de formation stellaire (SFR) de 700 Msol.yr-1. Cela mène pour les potentielles structures à une luminosité de 4.1013 Msol et un SFR de 7000 Msol.yr-1. Notre échantillon trace donc les structures denses du ciel à fort taux de formation stellaire. Cet échantillon sera donc clé dans la compréhension de la formation des structures et des étoiles à des redshifts autour de z=2. / Actual models and observations of the Universe agreed at large scale. But the formation of baryonic structures remains unknown. To constrain structure formation in the Universe, observatoins are needed at different redshift to see different evolution steps. Planck satellite gives an acces to objetcs in the redshift range z=2 to z=4. Thus we detect about 1200 extragalactic objects in 26% of the sky near galactic poles, candidates to be at high redshift. A sample of 228 sources is observed at higher sensibility and resolution than Planck with Herschel satellite. This lead to resolve ponctual Planck sources in several Herschel sources. So I identify three possible types of object : candidates sources of gravitationally amplified lenses, galaxy cluster candidates and alignement of sources along the line of sight. I find in the sample 11 sources amplified by gravitationnal lensing, some of the brightest of the submillimeter sky. Those have redshift between z=2.2 and z=3.6 (Canameras et al., 2015) and provide a zoom in the stellar formation of the far Universe. It let 217 overdensity of sources that I study. With their colour in Herschel data, I find an estimate of their density. 50% of Herschel fields have an overdensity greater than 10sigma using red sources colour selection (S250/S350 < 1.4 et S500/S350 > 0.6). This show that our ample is mainly composed of red overdensity. Fitting Herschel-SPIRE photometry with a modify black body at 35K, I show that the photmetric redshift distribution of Herschel sources peaks around z=2. Our sample is now composed of 11 lensed sources and 217 galaxy cluster candidates at z~2. These galaxy cluster candidates contain an average of 9 SPIRE sources in 4.5’ diameter beam. Individual sources have a luminosity distribution peaking at 4.1012 Lsun which leads to a star formation rate (SFR) of 700 Msun.yr-1 (Bell et al. 2003). This gives an estimate structure luminosity of 4.1013 Msun and an SFR of 7000 Msun.yr-1 assuming that sources are members of the same structure. Our sample traces dense structure at high star formation rate in the full sky. This sample will be a key in the comprehension of structure formation and star formation at redshift about z=2.

It is now believed that most, if not all, galaxies contain a supermassive black hole (SMBH) and that these play a crucial role in their host galaxies' evolution. Whilst accreting material, a SMBH (known as an active galactic nucleus, AGN, during this growth phase) releases energy which may have the effect of quenching star formation and constraining the growth of the galaxy. It is believed that AGNs can be divided into two broad fundamental categories, each with its own feedback mechanism. The radiative-mode of feedback occurs in gas-rich galaxies when substantial star formation is occurring and their young AGNs are growing rapidly through efficient accretion of cold gas. A fraction of the energy released by an AGN is transferred into the surrounding gas, creating a thermal "energy-driven" wind or pressure "momentum-driven" wind. Gas and dust may be expelled from the galaxy, so halting star formation but also cutting off the fuel supply to the AGN itself. The jet-mode occurs thereafter. The SMBH has now attained a large mass, but is accreting at a comparatively low level as gas slowly cools and falls back into the galaxy. The accretion process generates two-sided jets that generate shock fronts, so heating the gas surrounding the galaxy and partially offsetting the radiative cooling. This restricts the inflow of gas into the galaxy, so slowing the growth of the galaxy and SMBH. There are several convincing theoretical arguments to support the existence of these feedback mechanisms, although observational evidence has been hard to obtain. A new radio telescope - the Low Frequency Array (LOFAR) - recently started operations. LOFAR is especially suitable for investigating AGN feedback. It has been designed to allow exploration of low radio frequencies, between 10 and 240 MHz, which are particularly relevant for research into AGN activity. Also, with its large field-of-view and multi-beam capability, LOFAR is ideal for conducting extensive radio surveys. A project to image deeply the ELAIS-N1 field was started in May 2013. This thesis uses a number of surveys at different wavelengths, but particularly the low-frequency radio observations of the ELAIS-N1 field, to improve our knowledge of jet-mode AGN feedback and hence of the interplay between the complicated processes involved in galaxy formation and evolution. The more important pieces of research within the thesis are as follows: - A sample of 576 AGNs in the nearby universe was assembled and used to find a relationship between radio luminosity, X-ray luminosity and black hole mass. Moreover, the relationship is valid over at least 15 orders of magnitude in X-ray luminosity, strongly suggesting that the process responsible for the launching of radio jets is scale-invariant. - The established "Likelihood Ratio" technique was refined to incorporate colour information in order to optimally match the radio sources in the ELAIS-N1 field with their host galaxies. - The resulting catalogue was used to investigate ways in which radio sources can be matched automatically with their host galaxies (and so avoiding laborious visual examination of each source). The conclusions have helped the design of a pipeline for an extensive wide-area survey currently being conducted by the LOFAR telescope. - The catalogue was also used to investigate the evolution of jet-mode AGNs. This involved: deriving source counts; obtaining redshifts for each object; classifying the radio sources into the different populations of radiative-mode AGNs, jet-mode AGNs and star-forming galaxies; and using the above preparatory work in order to derive a luminosity function for jet-mode AGNs. - Key conclusions are that (1) feedback from jet-mode AGNs peaks at around a redshift of 0.75, (2) the space density of jet-mode AGNs declines steadily with redshift and (3) the typical luminosity of a jet-mode AGN increases steadily with redshift.

In recent years, advances in deep optical, and especially deep near-infrared imaging with the Hubble Space Telescope (HST) and wide-field ground-based telescopes such as VISTA, have revolutionized our understanding of the cosmological evolution of galaxies and supermassive black-holes (as manifest through active galactic nuclei; AGN). In particular, the dynamic range provided by the survey `wedding cake' of available HST+ground-based optical/IR data offers new opportunities to push the meaningful statistical study of galaxy and AGN evolution out to high redshifts. Much recent attention has focused, unsurprisingly, on using these new data to push studies of galaxy formation back to within a billion years of the Big Bang, and exploring the role of young galaxies in driving cosmic hydrogen reionization during the crucial era corresponding to redshifts z ≃ 6-10. However, these data have not been as thoroughly exploited at intermediate redshifts, and have only recently been used to explore black-hole/AGN evolution. In this thesis I utilise the latest deep optical/near-infrared imaging and spectroscopy to explore three key facets of cosmological evolution. First, I present a new, robust measurement of the evolving rest-frame ultraviolet (UV) galaxy luminosity function (LF) over the key redshift range from z ≃ 2 to z ≃ 4. My results are based on the high dynamic range provided by combining the Hubble Ultra Deep Field (HUDF), CANDELS/GOODS-South, and UltraVISTA/COSMOS surveys. I utilise the unparalleled multi-frequency photometry available in this survey `wedding cake' to compile complete galaxy samples at z ≃ 2; 3; 4 via photometric redshifts (calibrated against the latest spectroscopy). This study is important as the peak of star-formation is shown to happen within a redshift range z = 2 - 4 and determining the exact epoch that the galaxies were forming most of their stars depends significantly on the UV luminosity density which requires robust measurements of the galaxy UV luminosity function and its accurate parameterization. My new determinations of the UV LF extend from M1500 ≃ -22 (AB mag) down to M1500 =-14.5, -15.5 and -16 at z ≃2, 3 and 4 respectively (thus reaching ≃ 3-4 magnitudes fainter than previous blank-field studies at z ≃ 2 - 3). At z ≃ 2 - 3 I find a much shallower faint-end slope (α = -1:32 ± 0:03) than the steeper values (α ≃ -1:7) reported in the literature, and show that this new measurement is robust. By z ≃ 4 the faint-end slope has steepened slightly, to α = -1:43 ± 0:04, and I show that these measurements are consistent with the overall evolutionary trend from z = 0 to z = 8. I then calculate the UV luminosity density (and hence unobscured star-formation density) and show that it peaks at z ≃ 2:5 - 3, when the Universe was ≃ 2:5 Gyr old. Second, I have used these data to revisit the possibility that X-ray AGN played a significant role in cosmic hydrogen reionization which is one of the major processes in the formation of the Universe we see today. Hence, it is really important to understand this phenomenon thoroughly by studying the properties of sources capable of ionising photons, such as star-forming galaxies and high redshift AGNs. Although most recent studies have suggested that the emerging population of young star-forming galaxies can bathe the Universe in sufficient high-energy photons to complete reionization by z ≃ 6, some authors have reasserted the potentially important role of high-redshift AGN in the hydrogen reionization process. In an effort to clarify this situation, I reinvestigate a claimed sample of 22 X-ray detected active galactic nuclei (AGN) at redshifts z > 4, which has reignited the debate as to whether young galaxies or AGN reionized the Universe. These sources lie within the GOODS-S/CANDELS field, and I examine both the robustness of the claimed X-ray detections (within the Chandra 4Ms imaging) and perform an independent analysis of the photometric redshifts of the optical/infrared counterparts. I confirm the reality of only 15 of the 22 reported X-ray detections, and moreover find that only 12 of the 22 optical/infrared counterpart galaxies actually lie robustly at z > 4. I recalculate the evolving far-UV (1500Å) luminosity density produced by AGN at high redshift, and find that it declines rapidly from z ≃ 4 to z ≃ 6, in agreement with several other recent studies of the evolving AGN luminosity function. The associated rapid decline in inferred hydrogen-ionizing emissivity contributed by AGN falls an order-of-magnitude short of the level required to maintain hydrogen ionization at z ≃ 6. I conclude that AGNs make a very minor contribution to cosmic hydrogen reionization. Finally, I have utilized the deep optical/near-infrared survey data to explore the prevalence of quenched/passive galaxies at high redshift. Applying a robust method to isolate passive galaxies from star-forming galaxies is the key to improving our understanding of the quenching process. Focusing primarily on the deep HUDF data-set, I have revisited the effectiveness of simple colour-colour (UVJ) selection techniques in isolating robust samples of quenched galaxies, and find that dust plays a more important role in this selection process than has been previously appreciated. Through careful SED fitting I successfully isolate a sample of apparently dust-free quiescent galaxies in the redshift range 0:5 < z < 4:5 but (at least in the HUDF) fail to find any galaxy which has remained truly quiescent for > 1 Gyr. I conclude by focusing on the properties of a refined/robust sample of apparently quenched galaxies at z > 3, and in particular establishing the contribution of quenched galaxies to stellar-mass density at early times. I conclude with a summary of my findings, and a brief discussion of the most promising avenues for future advances with the next generation of facilities, such as the James Webb Space Telescope (JWST).

J'ai consacré mon travail de thèse à la recherche de quasars à haut redshift dans le sondage Canada France High-z Quasar Survey in the Near Infrared (CFHQSIR). L'objectif principal de ce programme est de découvrir des quasars de redshift de l'ordre de z $\sim$ 7 dans les champs larges du CFHTLS (CFHTLS Wide) en utilisant des images réalisées dans le proche infrarouge avec la caméra WIRCam installée au foyer du CFHT. J'ai tout d'abord effectué la calibration photométrique de l'ensemble des données de CFHQSIR. J'ai ensuite étudié la qualité des images ainsi que leur propriété de bruit et leur profondeur. J'ai dans un deuxième temps calculé le nombre de quasars qu'il serait possible de détecter avec CFHQSIR et en ai déduit les contraintes envisageables sur la fonction de luminosité des quasars à z $\sim$ 7. J'ai ensuite procédé à l'identification de candidats quasars parmi les dizaines de milliers de sources que comptent les 130 degrés carrés couvert par CFHQSIR. Des observations de suivi photométriques des candidats ont finalement permis de révéler une cinquantaine d’objets, dont une quinzaine a été sélectionnée pour des observations spectroscopiques au Very Large Telescope (VLT). Celles-ci n’ayant, pour la plupart, pas encore été realisées, la nature exacte de ces sources ne pourra être connue que dans les prochains mois. Pour finir, la dernière partie de ma thèse s’est focalisée sur l’étude de modèles statistiques bayésiens afin de compléter ma méthode de sélection de candidats quasars. Cette étude m'a finalement permis de vérifier que la majorité des objets retenus pour des observations spectroscopiques étaient effectivement les candidats les plus probables. / My PhD work is focused on the search for high-redshift quasars in the Canada France High-z Quasar Survey in the Near Infrared (CFHQSIR). The main scientific objective of this CFHT Large Program is to search for quasars at redshift z $\sim$ 7 with near-infrared images of the CFHTLS Wide fields acquired with the CFHT WIRCam camera. Firstly, I carried out the photometric calibration of the CFHQSIR images. I performed a detailed analysis of the CFHQSIR data by studying their quality, as well as their noise properties and their depths.Secondly, I computed the number of high-redshift quasars that could be detected with CFHQSIR and the corresponding constraints which could be put on the z $\sim$ 7 quasar luminosity function. Then, I proceeded to the identification of quasar candidates among the many thousands of sources in the 130 square degrees covered by CFHQSIR. Photometric follow-up observations of the candidates revealed about fifty objects, of which fifteen were chosen to be spectroscopically observed with the Very Large Telescope (VLT). Given that this spectroscopic follow-up is not yet completed, the exact nature of these sources will only be known in the coming months. To finish, I applied Bayesian model comparison to my sample in order to complete and consolidate my selection procedure. My candidates were finally classified according to their probability to be a high-redshift quasar. I verified that the majority of the most likely candidates were selected for spectroscopic observations.

Les futurs sondages à grande échelle, comme la mission Euclid, produiront un large set de données qui nécessitera la mise en place de chaînes de traitement complètement automatisées pour analyser les données, extraire l'information et s'assurer que l'ensemble des exigences est respecté. À partir des données collectées, le redshift est une quantité essentielle à mesurer. Des méthodes distinctes pour l'estimation de redshift existent dans la littérature, mais il n'y a pas à l'heure actuelle une définition d'un critère de qualité automatisé pour juger de la fiabilité d'une mesure de redshift spectroscopique. Dans ce travail, nous avons présenté les techniques usuelles de l'analyse spectrale, comme le filtrage et l'extraction du continu, qui peuvent être réalisées en amont de l'estimation du redshift. Puis, nous nous sommes en grande partie intéressés à la problématique de prise de décision quant à l'évaluation de la fiabilité d'une mesure de redshift en développant une nouvelle méthodologie qui exploite les algorithmes de Machine Learning (ML) et les caractéristiques de la fonction de densité en probabilité du redshift (zPDF) définie dans un cadre bayésien.Notre idée consiste à quantifier, via le ML et des descripteurs des zPDFs, la fiabilité d'une mesure de redshift au sein de partitions distinctes qui décrivent différents niveaux de confiance. Par ex. une estimation de redshift obtenue à partir d'une zPDF multimodale ne dépeint pas le même niveau de confiance que celle obtenue à partir d'une zPDF unimodale. Nous estimons que cette nouvelle méthodologie pourra être exploitée dans le cadre de futurs sondages spectroscopiques à grande échelle comme Euclid. / Large-scale surveys, as Euclid, will produce a large set of data that will require the development of fully automated data-processing pipelines to analyze the data, extract crucial information and ensure that all requirements are met. From a survey, the redshift is an essential quantity to measure. Distinct methods to estimate redshifts exist in the literature but there is no fully-automated definition of a reliability criterion for redshift measurements. In this work, we first explored common techniques of spectral analysis, as filtering and continuum extraction, that could be used as preprocessing to improve the accuracy of spectral features measurements, then focused on developing a new methodology to automate the reliability assessment of spectroscopic redshift measurements by exploiting Machine Learning (ML) algorithms and features of the posterior redshift probability distribution function (PDF). Our idea consists in quantifying, through ML and zPDFs descriptors, the reliability of a redshift measurement into distinct partitions that describe different levels of confidence. For example, a multimodal zPDF refers to multiple (plausible) redshift solutions possibly with similar probabilities, while a strong unimodal zPDF with a low dispersion and a unique and prominent peak depicts of a more "reliable" redshift estimate. We assess that this new methodology could be very promising for next-generation large spectroscopic surveys on the ground and space such as Euclid and WFIRST.