The dynamics of the expected relativistic fireball expansion were investigated by [16, 17]. The fact that photons of over 100 MeV are detected provides compelling evidence for ultra-relativistic expansion. To avoid degradation of the spectrum via photon-photon interactions to energies below the electron-positron formation threshold me c2 = 0.511 MeV the outward flow must have a bulk Lorentz factor high enough so that the relative angle at which the photons collide is less than -1, thus diminishing the pair production threshold [18, 19].
Since each baryon in the outflow must be given an energy exceeding 100 times its rest mass, a key requirement of the central engine is that it must concentrate a lot of its energy into a very small fraction of its total mass. This favours models were magnetic fields and Poynting flux are important.
The observed spectrum extends to high energies, generally in a broken power law shape, i.e., highly nonthermal. Two initial problems [9, 20] with the first expanding fireball models were that (a) they are initially optically thick and the photon spectrum escaping from the Thompson scattering photosphere would be expected to be an approximate blackbody, and (b) most of the initial fireball energy would be converted into kinetic energy of expansion, with a concomitantly reduced energy in the observed photons, i.e. a very low radiative efficiency.
Figure 1. Schematic GRB jet from a collapsing star.
A simple way to achieve a high efficiency and a nonthermal spectrum, which is currently the most widely invoked explanation, is by reconverting the kinetic energy of the flow into random energy via shocks, after the flow has become optically thin . Two different types of shocks may be expected. There will be an external shock, when the expanding fireball runs into the external interstellar medium or a pre-ejected stellar wind, and a reverse shock propagating back into the ejecta. As in supernova remnants, Fermi acceleration of electrons into a relativistic power distribution in the turbulent magnetic fields boosted in the shock leads to synchrotron emission [21, 22] resulting in a broken power law spectrum, where the high energy photon spectral slope fits easily the observations, and the single electron low energy photon slope -2/3 can, with a distribution of minimum energy electrons min, reproduce the observed average low energy photon slope values of -1 (see also [23, 24]). The reverse shock would lead to optical photons, while inverse Compton emission in the forward blast wave would produce photons in the GeV-TeV range .
Figure 2. A diversity of gamma-ray light curves from the BATSE instrument on the Compton Gamma Ray Observatory.
There could, additionally, be dissipation and acceleration within the outflowing jet itself. If the jet is unsteady, internal shocks [26, 27] can form as faster portions of the flow catch up with slower portions. And if magnetic stresses are important within the jets (i.e. they are Poynting dominated outflows , instead of the usual baryonic inertia dominated outflows) then magnetic reconnection can provide efficient mechanical conversion of bulk into random energy  (see also [30, 31]). Any of these models provide a generic scenario for explaining the radiation spectrum, largely independent of the specific nature of the progenitor.
Internal shocks continue to be the model most widely used by observers to interpret the prompt MeV emission, while the external shock model is the favored interpretation for the long-term afterglows starting at high energies and phasing into gradually longer wavelengths over periods of days to months. Coincidentally, the detection of the afterglows was preceded, a few weeks earlier, by the publication of quantitative predictions of the power law spectral and time dependence of X-ray, optical and radio afterglows , in general agreement with observations. Prompt optical afterglows were first detected in 1999 , while multi-GeV emission was reported by CGRO-EGRET , and more recently and in greater detail by Fermi (see Section 5).