A Statistical and Multi-wavelength Study of Star Formation in Galaxies

Abstract

The main goal of this thesis is to characterize the processes that regulate large-scale star formation in galaxies. During the last fifteen years, the development of infrared astronomy through the satellites ISO, IRAS, Spitzer and Herschel has revolutionized our conception of galaxy evolution. By observing the light emitted by the interstellar dust, these observatories allow us to detect the energy and matter that remain elusive to the best optical telescopes, and have thereby discovered a substantial yet unexpected part of the star formation activity of galaxies. The work of my thesis hence rely heavily on the data acquired by the Herschel satellite, which allow for the first time the detection in the infrared of normal galaxies at great distances (z=2).Taking advantage of these new data, I perform a statistical study of several thousands of galaxies at different epochs of the Universe. In particular, I bring forward the best constraints available today on the properties of the of galaxies. The existence of this sequence (the correlation between the stellar mass, M*, and the star formation rate, SFR) turned out to be a incredibly useful tool to understand galaxy evolution. The small dispersion that is observed around this sequence suggests that the majority of galaxies are growing through long and steady episodes of star formation, rather than intense bursts like those triggered by the collision (of merger) of two galaxies. By developing a new image analysis technique, I show in particular that more than two thirds of the mass of stars present in the Universe today has been formed within Main Sequence galaxies, hence that this is the dominant mode of galaxy growth.Then I approach another aspect of the Main Sequence, that is the characterization of the evolution of its shape, i.e., the slope of the SFR-M* correlation. In agreement with other studies that were published independently, I find that this slope evolves and decreases with time, so that the most massive galaxies are forming relatively fewer stars per year today than they used to in the past. I study the various possible causes for this evolution, by quantifying for example the morphological evolution of these galaxies and the growth of bulges, as well as the evolution in their hydrogen gas content, which is the fuel for star formation. I deduce from these observations that the change of slope of the Main Sequence can be mainly attributed to a decrease of the star formation efficiency, rather than by a morphological process or a lack of gas.The various observations I have made throughout the work described above allow me to establish simple prescriptions to simulate the observable properties of galaxies, in particular their spectrum. I use these recipes to create a realistic simulation of a deep field, that I use to test my analysis methods and that reproduces consistently the cosmic infrared background.Lastly, I introduce some preliminary results on star formation in the young Universe (z=4) obtained thanks to new data acquired with the ALMA telescope. I describe in particular the resulting new constraints on the Main Sequence at this epoch, and study in more detail two extremely distant galaxies that I have discovered by chance in these data. These two galaxies are among the most distant known today, and are probably the most massive and most dusty ever detected in a Universe that is less than a billion years old.