The evidence for a jet outflow is based on the observed steepening of the light curve after ~ day , which is attributed to the transition between the afterglow early relativistic expansion, when the light-cone is narrower than the jet opening half-angle j and the late expansion, when the light-cone has become wider than the jet, -1 j, leading to a drop in the effective flux [36, 37, 38]. A jet opening half-angle j ~ 3-5 degrees is inferred, which reduces the total energy requirements to about 1051-1052 ergs. This, even allowing for substantial inefficiencies, is compatible with currently favored scenarios based on a stellar collapse or a compact merger, e.g.  and Section 1.
Observations with the Beppo-SAX and HETE-2 satellites indicated the existence of a sub-class of GRBs called X-ray flashes (XRFs), whose spectrum peaks at energies 30-80 keV instead of the 300 keV - 1 MeV of classical GRBs, and with wider jet opening angles, e.g. . The relative frequencies of XRFs versus GRBs led to considerations about a possible continuum distribution of angles, as well as about the jet angular shape, including departures from simple top-hat (abrupt cut-off) including an inverse power law or a Gaussian dependence on the angle [40, 41, 42, 43].
A problem with simple internal shock synchrotron models of the prompt MeV emission is that the low energy photon number spectral slope, which is expected to be -2/3, is found to be flatter in a fraction of BATSE bursts . In addition, the synchrotron cooling time can be typically shorter than the dynamical time, which would lead to slopes -3/2 . In either internal shock Fermi acceleration or in magnetic reconnection schemes, a number of effects can modify the simple synchrotron spectrum to satisfy these constraints. Another solution involves a photospheric component, discussed below.
A natural question is whether the clustering of spectral peak energies in the 0.1-0.5 MeV range is intrinsic or due to observational selection effects [46, 47]. A preferred peak energy may be attributed to a blackbody spectrum at the comoving pair recombination temperature in the fireball photosphere . A photospheric component can address also the above low-energy spectral slope issue with its steep Rayleigh-Jeans part of the spectrum, at the expense of the high energy power law. This was generalized  to a photospheric blackbody spectrum at low energies with a comptonized photospheric component and possibly an internal shock or other dissipation region outside it producing Fermi accelerated electrons and synchrotron photons at high energies. Photospheric models with moderate scattering depth can in fact lead to a Compton equilibrium which gives spectral peaks in the right energy range  and positive low energy slopes as well as high energy power law slopes (the positive low energy slopes can always be flattened through a distribution of peak energies). A high radiative efficiency can be a problem if the photosphere occurs beyond the saturation radius rsat ~ r0 , where r0 is the base of the outflow and = L / Mc2 is the asymptotic bulk Lorentz factor . However, a high radiation efficiency with low and high energy slopes can be obtained in all cases if significant dissipation (either magnetic reconnection or shocks) is present in the photosphere [51, 52]. This can also address the phenomenological Amati  and Ghirlanda  relations between spectral peak energy and burst fluence [51, 55].