*3.1. Observed Temporal and Spectral Poperties of GRB Prompt Emission*

The GRB observed duration distribution is bimodal, which leads to the division into short and long *duration* events (with an observer frame separation at ∼2 s). The average spectral properties of events in the two duration classes display some differences, with short GRBs featuring on average harder spectra with respect to long ones [155], the difference residing primarily in the low-energy part of the spectrum [156,157]. Short events are detected at a lower rate with respect to long events (with an instrument-dependent short/long ratio of 1:3 for *CGRO*/BATSE, 1:5 for *Fermi*/GBM and of 1:9 for *Swift*/BAT). However, several different instrumental and cosmological effects shape (and hence bias) these observed properties [53,158,159].

The prompt emission has no apparent periodicity [160] and features a power density spectrum consistent [161,162] with turbulent dissipation processes. Different methods were employed to measure the minimum *variability* timescale of prompt emission light curves, resulting in different distributions [163,164]. Ref. [164] reports similar rest-frame distributions for the minimum variability timescale in long and short GRBs, centered around 0.5 s and extending down to 10 ms in ∼10% of the events. Another interesting feature of the prompt emission is the presence of *spectral lags*. These consist in the fact that pulses observed in the lower energy bands of the gamma-ray instruments are seen to lag behind the corresponding pulses in the higher energy bands [165]. This feature seems to be more commonly present in long GRBs, while short events typically have lags consistent with zero [166].

Most attempts at identifying the fundamental building blocks of GRB prompt emission light curves adopted parametric functions to represent pulses (e.g., [165,167]). With the caveat that these methods are applied to large samples of light curves of GRBs with unknown redshift, the results show apparent (observer frame) differences between short and long GRBs [168].

On the longer timescales, periods of activity can at times be separated by quiescent phases. In a sizable fraction (∼ 15%) of long GRBs, a long quiescence phase (reaching, in some cases, >100 s) separates *precursor* activity from the main emission episode [169]. Precursors have also been identified in short GRBs [170]. Long apparent quiescences also separate the main event from late time pulses, or *flares*, often observed in the X-ray band by *Swift*/XRT. X-ray flares share some common properties with the prompt emission [171] and are thus often interpreted as linked to late-time central engine activity. No significant differences in the average spectral properties of precursors and main emission episodes have been identified [172], while flares appear clearly softer.

The prompt emission of GRBs is characterized by a *non-thermal* spectrum. The presence of thermal-like emission has been identified in a few cases either during the initial phases of the burst [173] or along its full duration [174], with no evidence of such emission component in short bursts [29]. A combination of multiple emission components (e.g., the sum of a power law with a high energy cutoff and a black body) was also adopted to interpret observed spectra [175,176]. Observationally, the spectral energy distribution of GRB prompt emission typically peaks at 0.1–1 MeV and the low (resp. high) energy spectrum, below (resp. above) the peak, is consistent with a power-law with photon index ∼−1 (resp. −2.5).
