In most of the scenarios currently considered, the progenitor is a black hole recently formed, surrounded by an accretion disk. This black hole can result from the collapse of a very massive star in fast rotation (hypernova model) or from the collapse of two neutron stars (merger model). The matter is ejected in both cases with an ultra relativistic speed (Lorentz factor larger than 100) in a cone whose opening is of a few degrees, the observer being placed by chance in a direction included in this cone. Various matter shells are ejected with different speeds. When a shell meets another one, there is an internal shock. The electrons accelerated by this shock radiate, probably by synchrotron radiation in the existing magnetic fields, and produce gamma rays (emission which has given the name to the phenomenon), but also X rays and visible light (prompt emission). Several internal shocks can take place, which accounts for the complexity and the variety of the GRB phenomenon. Finally the shells collide with the surrounding matter, which can be the interstellar environment or the material expelled previously by the progenitor. A radiation, probably due to the synchrotron emission, is then produced, covering all the electromagnetic spectrum from X rays to the radio waves (afterglow emission).

These scenarios have been confirmed by several observations:

• In several cases, long GRBS have been firmly associated to type Ib/Ic supernova (GRB980425/SN1998bw, GRB030329/SN2003df, ?). In several other cases, the study of the light curves has shown a rebrightening after a few days, which is also compatible with the appearance of a supernova. The long GRBs seem to be linked to the most massive stars.
• the short GRBs have been observed in galaxies with a weak star formation activity. They appear to be linked to the coalescence of two neutron stars.

This property makes the GRBs exceptional beacons to study the early Universe. In particular, they offer a unique possibility to study the epoch of the first stars formation and of the reionisation of the Universe, the so-called Dark Ages. By definition, there were no quasars at these epochs since it takes several hundred million of years to build a supermassive black hole. The first stars, however, are expected to be supermassive (100 times solar masses or more) and a fraction of these are expected to produce detectable GRBs.

Furthermore, GRBs can be used as probes of cosmological lines-of-sight in a manner similar to that of bright quasars. Indeed, when looking for absorption systems in the spectrum of distant quasars or GRBs, we select galaxies with no luminosity (or star formation) bias. Absorbers are thus complementary to galaxies selected in emission (e.g. Lyman break galaxies). While quasars, which are bright and do not fade away, will remain important probes of the weak lines of the Ly? forest at redshifts 2 < z < 5, the advantage of GRBs over quasars is that they will allow the study of the more rare high column density absorption systems such as the Damped Ly? absorbers, the metal systems (MgII, CIV, ?) or specifically interesting species (D, H2 , C*, CII, ?).

Their advantages as background sources are many fold. First, it will allow to probe galaxies selected thanks to their metal absorption up to the highest redshifts where there might be no other observational techniques to find faint bound objects. At intermediate redshifts, the link between galaxies and CIV in particular has clearly been established. Assuming this also holds at higher redshifts (z>7), CIV could then be used as tracers of the very first galaxies of the Universe. Second, GRB being bright and transient background sources, they will be the ideal tool to observe extreme environments in absorbers. Whilst many high column density system have been found at the GRB position, including extremely high N(HI), the first truly intervening Damped Ly? system has only been recently discovered (GRB 050730). It is expected that large N(HI) and/or dusty systems missing from current quasar absorbers survey (because of the dust extinguishing the quasar) will be probed with GRBs thanks to their extremely high intrinsic luminosity. Third, the central question of the relation of all these absorbers to galaxies remains. Indeed, the steadily bright quasars make it extremely difficult to search for the galaxy counterparts to the intervening absorbers. In GRBs, the afterglows will disappear and the search for emission from the absorbing galaxy will only be affected by the light from the much fainter GRB host galaxy. Absorbers detected towards GRBs will therefore be the perfect tool to bridge systems detected in absorption with galaxies detected in emission.

Moreover the GRBs host-galaxies themselves describe a sample of galaxies selected in a new way (GRB-selected), spanning the whole history of the universe. The interpretation of their properties in the global cosmic and galactic evolution framework will be possible once the nature of GRBs and their host is better understood.

Due to the link between the long GRBs and the massive stars, it has been also considered that they may provide a unique opportunity to trace the star formation process in the early Universe. However infrared observations with the Spitzer satellite have shown very recently that the host galaxies are not a good tracer of the stellar formation history as their properties seem to be not correlated with the stellar formation of nearby and distant galaxies. This application may be, at the end, not so straightforward.

By using the characteristics of the gamma ray spectrum and of the optical light curve, it may be possible to standardize the GRBs luminosity, as it has been already done with type Ia supernovae. This method, which is still the object of a very intense debate, could offer remarkable potentialities by allowing the measurement of the cosmological parameters up to redshifts of 10-15, against <2 for type Ia supernovae. Systematic errors are in addition different between these two methods, which make them perfectly complementary.