Abstract
Fifty five years after the discovery of gamma-ray bursts (GRBs), the physical mechanisms responsible for the prompt emission (0.1-100 s) of their ultra-relativistic jet are still debated. This thesis deals with this short emission phase, very variable and very energetic, in particular at high energy (from keV to MeV). Schematically, one can distinguish the non-thermal emission of electrons accelerated by shocks in the jet or by magnetic reconnection, from the quasi-thermal emission near the photosphere. My work is based on a version of the internal-shock model, and on the parameterization of the synchrotron spectra that the numerical code predicts at keV-MeV energies. This spectral model (Internal Shock Synchrotron Model, ISSM) presents a continuous curvature resulting from the superposition of instantaneous synchrotron spectra which evolve rapidly in the shocked regions. In the first part of the thesis, I confronted this model to the prompt emission spectra of GRBs detected by the GBM instrument of the Fermi space mission, and I compared its performance to phenomenological models such as the ad-hoc Band function. The latter mathematical models are commonly used in the community to summarily fit GRB keV-MeV spectra. I used 460 GRBs detected by the GBM, which I selected for their high fluence in order to reach the spectral accuracy that is required to distinguish between models. I developed a complete analysis chain (data preparation, selection of GBM detectors on precise quantitative criteria, light curve segmentation, spectral fits, post-processing scripts). I showed that the ISSM physical model outperforms all phenomenological models, that it reproduces better the observed spectra on short timescales, and even better when these spectra are averaged. These results reinforce with a high level of confidence the hypothesis of an internal-shock synchrotron origin of the non-thermal prompt emission of GRBs. In a second part, I studied the few rare GRBs showing an additional quasi-thermal spectral component at low energy. This photospheric emission is expected in the so-called "fireball" scenario, in which the plasma of the ejecta expands under its own pressure and reach an ultra-relativistic speed. I showed that the detection of this spectral component depends strongly on the modeling of the non-thermal synchrotron component. In particular, the use of a realistic model like ISSM leaves little room for an additional component. These results thus seem to indicate an initial energy reservoir consisting partly of a Poynting flux and not only of the thermal energy of the adiabatically expanding fireball. The last part of the thesis concerns the study of the exceptional gamma-ray burst GRB 220101A with the GBM and LAT instruments of Fermi. This very energetic burst is the most distant ever detected by the LAT (z=4.618). It lies among the five rare GRBs which present a spectral break at high energy associated with a variable emission. Using a realistic gamma-gamma opacity calculation, this spectral attenuation allowed to measure the velocity of the ultra-relativistic jet (Lorentz factor ~ 110) and to locate the gamma emission regions, at a distance from the central black hole (a few 10^(14) cm) where internal shocks typically take place.