1 Introduction

Recent optical afterglow observations have convincingly shown that most GRBs are likely at cosmological distances (Djorgovski et al. 1997; Feroci et al. 1998). But their true nature remains a mystery. Though afterglow studies will shed important light on the energetics, progenitor and environment of GRBs, progress towards the origin of GRB itself still depends on the ultimate understanding of the gamma-ray emission mechanism. To address the physics of this early phase GRB emission we must decipher the complex information contained in the spectral and temporal evolution of the gamma-ray pulses.

Recent high-time-resolution studies of GRB spectra show that the typical hard-to-soft evolution can be characterized, using the Band et al. (1993) function, as the downward movement of the spectral break energy $E_{\rm pk}$ (Liang & Kargatis 1996), plus a softening of the low energy photon slope $\alpha$ (Crider et al. 1997; Crider et al. 1998). The Band et al. $\alpha$ often is very hard at the beginning of a pulse, with $\alpha_{\max}\gt$ in over 2/3 of pulses analyzed (Crider et al. 1997; Preece et al. 1998). This contradicts directly the upper limit on optically thin synchrotron $(\alpha_{\max}=-2/3)$ (Meszaros & Rees 1993; Katz 1994). Hence a synchrotron emission model requires additional absorption mechanism to turnover the low-energy spectral slope ("X-ray deficiency problem"). Also the decay of $E_{\rm pk}$ follows the Liang-Kargatis (1996) decay law in most pulses: $E_{\rm pk}=E_{\rm o\exp}(-\Phi/\Phi_{\rm o})$ where $\Phi$ is the running photon fluence, and $E_{\rm o}$ and $\Phi_{\rm o}$ are constants. This behavior must be satisfied by any emission and cooling model.