Gamma-ray bursts (GRBs), short and intense bursts of ~ 100 keV - 1 MeV photons, were discovered accidentally in the late sixties by the Vela satellites . The mission of these satellites was to monitor the "Outer Space Treaty" that forbade nuclear explosions in space. A wonderful by-product of this effort was the discovery of GRBs.
The discovery of GRBs was announced in 1973 . It was confirmed quickly by Russian observations  and by observations on the IMP-6 satellite . Since then, several dedicated satellites have been launched to observe the bursts and numerous theories were put forward to explain their origin. Claims of observations of cyclotron spectral lines and of discovery of optical archival counterparts led in the mid eighties to a consensus that GRBs originate from Galactic neutron stars. This model was accepted quite generally and was even discussed in graduate textbooks [4, 5, 6] and encyclopedia articles [7, 8].
The BATSE detector on the COMPTON-GRO (Gamma-Ray Observatory) was launched in the spring of 1991. It has revolutionized GRB observations and consequently our basic ideas on their nature. BATSE observations of the isotropy of GRB directions, combined with the deficiency of faint GRBs, ruled out the galactic disk neutron star model 2 and make a convincing case for their extra-galactic origin at cosmological distances . This conclusion was recently confirmed by the discovery by BeppoSAX  of an X-ray transient counterparts to several GRBs. This was followed by a discovery of optical [11, 12] and radio transients . Absorption line with a redshift z = 0.835 were measured in the optical spectrum of the counterpart to GRB970508  providing the first redshift of the optical transient and the associated GRB. Latter, redshifted emission lines from galaxies associated with GRB971214  (with z = 3.418) and GRB980703  (with z = 0.966) were discovered. Galaxies has been discovered at the positions of other bursts. There is little doubt now that some, and most likely all GRBs are cosmological.
The cosmological origin of GRBs immediately implies that GRB sources are much more luminous than previously thought. They release ~ 1051 - 1053 ergs or more in a few seconds, the most (electromagnetically) luminous objects in the Universe. This also implies that GRBs are rare events. BATSE observes on average one burst per day. This corresponds, with the simplest model (assuming that the rate of GRBs does not change with cosmological time) to one burst per million years per galaxy. The average rate changes, of course, if we allow beaming or a cosmic evolution of the rate of GRBs.
In spite of those discoveries, the origin of GRBs is still mysterious. This makes GRBs a unique phenomenon in modern astronomy. While pulsars, quasars and X-ray sources were all explained within a few years, if not months, after their discovery, the origin of GRBs remains unknown after more than thirty years. The fact that GRBs are a short transient phenomenon which until recently did not have any known counterpart, is probably the main reason for this situation. Our inability to resolve this riddle also reflects the accidental and unexpected nature of this discovery which was not done by an astronomical mission. Theoretical astrophysics was not ripe to cope with GRBs when they were discovered.
A generic scheme of a cosmological GRB model has emerged in the last few years and most of this review is devoted to an exposition of this scheme. The recently observed X-ray, optical and radio counterparts were predicted by this picture [17, 18, 19, 20, 21]. This discovery can, to some extent, be considered as a confirmation of this model [22, 23, 24, 25]. According to this scheme the observed -rays are emitted when an ultra-relativistic energy flow is converted to radiation. Possible forms of the energy flow are kinetic energy of ultra-relativistic particles or electromagnetic Poynting flux. This energy is converted to radiation in an optically thin region, as the observed bursts are not thermal. It has been suggested that the energy conversion occurs either due to the interaction with an external medium, like the ISM  or due to internal process, such as internal shocks and collisions within the flow [28, 29, 30]. Recent work [20, 31] shows that the external shock scenario is quite unlikely, unless the energy flow is confined to an extremely narrow beam, or else the process is highly inefficient. The only alternative is that the burst is produced by internal shocks.
The "inner engine" that produces the relativistic energy flow is hidden from direct observations. However, the observed temporal structure reflects directly this "engine's" activity. This model requires a compact internal "engine" that produces a wind - a long energy flow (long compared to the size of the "engine" itself) - rather than an explosive "engine" that produces a fireball whose size is comparable to the size of the "engine". Not all the energy of the relativistic shell can be converted to radiation (or even to thermal energy) by internal shocks [32, 33, 34]. The remaining kinetic energy will most likely dissipate via external shocks that will produce an "afterglow" in different wavelength . This afterglow was recently discovered, confirming the fireball picture.
At present there is no agreement on the nature of the "engine" - even though binary neutron star mergers  are a promising candidate. All that can be said with some certainty is that whatever drives a GRB must satisfy the following general features: It produces an extremely relativistic energy flow containing 1051 - 1052 ergs. The flow is highly variable as most bursts have a variable temporal structure and it should last for the duration of the burst (typically a few dozen seconds). It may continue at a lower level on a time scale of a day or so . Finally, it should be a rare event occurring about once per million years in a galaxy. The rate is of course higher and the energy is lower if there is a significant beaming of the gamma-ray emission. In any case the overall GRB emission in -rays is ~ 1052 ergs / 106 years/galaxy.
We begin (section 2) with a brief review of GRB observation (see [37, 38, 39, 40, 41] for additional reviews and [42, 43, 44, 45] for a more extensive discussion). We then turn to an analysis of the observational constraints. We analyze the peak intensity distribution and show how the distance to GRBs can be estimated from this data. We also discuss the evidence for another cosmological effect: time-dilation (section 3). We then turn (section 4) to discuss the optical depth or the compactness problem. We argue that the only way to overcome this problem is if the sources are moving at an ultra-relativistic velocity towards us. An essential ingredient of this model is the notion of a fireball - an optically thick relativistic expanding electron-positron and photon plasma (for a different model see however ). We discuss fireball evolution in section 6. Kinematic considerations which determine the observed time scales from emission emerging from a relativistic flow provides important clues on the location of the energy conversion process. We discuss these constraints in section 7 and the energy conversion stage in section 8. We review the recent theories of afterglow formation in section 9. We examine the confrontation of these models with observations and we discuss some of the quantitative problems.
We then turn to the "inner engine" and review the recent suggestions for cosmological models (section 10). As this inner engine is hidden from direct observation, it is clear that there are only a few direct constraint that can be put on it. Among GRB models, binary neutron star merger  is unique. It is the only model that is based on an independently observed phenomenon , is capable of releasing the required amounts of energy  within a very short time scale and takes place at approximately the same rate [50, 51, 52] 3. At present it is not clear if this merger can actually channel the required energy into a relativistic flow or if it could produce the very high energy observed in GRB971214. However, in view of the special status of this model we discuss its features and the possible observational confirmation of this model in section 10.2.
GRBs might have important implications to other branches of astronomy. Relation of GRBs to other astronomical phenomena such as UCHERs, neutrinos and gravitational radiation are discussed in section 11. The universe and our Galaxy are optically thin to low energy -rays. Thus, GRBs constitute a unique cosmological population that is observed practically uniformly on the sky (there are small known biases due to CGRO's observation schedule). Most of these objects are located at z 1 or greater. Thus this population is farther than any other systematic sample (QSOs are at larger distances but they suffer from numerous selection effects and there is no all sky QSOs catalog). GRBs are, therefore, an ideal tool to explore the Universe. Already in 1986 Paczynski  proposed that GRBs might be gravitationally lensed. This has led to the suggestion to employ the statistics of lensed bursts to probe the nature of the lensing objects and the dark matter of the Universe . The fact that no lensed bursts where detected so far is sufficient to rule out a critical density of 106.5 M to 108.1 M black holes . Alternatively we may use the peak-flux distribution to estimate cosmological parameters such as and . The angular distribution of GRBs can be used to determine the very large scale structure of the Universe [57, 58]. The possible direct measurements of red-shift to some bursts enhances greatly the potential of these attempts. We conclude in section 12 by summarizing these suggestions.
Over the years several thousand papers concerning GRBs have appeared in the literature. With the growing interest in GRBs the number of GRB papers has been growing at an accelerated rate recently. It is, of course, impossible to summarize or even list all this papers here. I refer the interested reader to the complete GRB bibliography that was prepared by K. Hurley .
2 A few GRBs, now called soft gamma repeaters, compose a different phenomenon, are believed to form on galactic neutron stars. Back.
3 This is assuming that there is no strong cosmic evolution in the rate of GRB. Back.