Time-averaged Analysis

We obtained 3 240-channel energy spectra for this event of initial and final times with respect to the trigger time as follows: -60-67 s (interval A), 68-195 s (interval B) and 196-323 s (interval C). We fitted to the three spectra the phenomenologically established spectral law for SGRs: an optically thin thermal bremsstrahlung (OTTB). A 10% systematic error has been added to the statistical error in order to take into account the calibration uncertainties at the large off-axis location of the event.

The time integrated spectrum in interval A cannot be fitted with a simple OTTB law, showing an excess at high energies (also reported for the March 5th event. This is likely due to the strong spectral evolution as it appears from the 1-s ratemeters (see below). In fact, from the ratemeter data we note that during the first seconds the emitted spectrum is very hard and it becomes much softer soon after the peak. Assuming different spectral components for these two fractions of interval A, we fitted the energy spectrum with the sum of an OTTB and of a single power law (PL, $I(E) \propto E^{-\alpha})$.The best fit parameters are kT=$(31.2\pm2.5)$ keV and photon index $\alpha = 1.47\pm0.16$ (reduced $\chi^{2}$=0.705 with 186 d.o.f.). We note, however, that the spectrum can equally be fit with the sum of a PL and a blackbody (BB). In this case the photon index is $\alpha = 1.71\pm0.15$, the BB temperature is kT=$(16.4\pm0.8)$ keV, and reduced $\chi^{2}$=0.74. We have also tried to fit the spectrum two OTTB laws, two power laws and a broken power law (that usually fits time-averaged spectra of classical gamma-ray bursts) but the fits are unacceptable.

In interval B from the ratemeter data we do not deduce strong spectral variation, therefore we first used a simple OTTB law. However, the best fit to the spectrum for interval B has a kT=$(34.2\pm1.2)$ keV and the reduced $\chi^{2}$ is 1.996 (75 d.o.f.). We note the existence of a hard excess, that is likely biasing upward the determination of the effective temperature. Thence we added a PL component to the fit, and the best parameters are kT=$(27.6\pm1.9)$ keV and $\alpha = 4.5\pm0.2$ (reduced $\chi^{2}$=1.360, 73 d.o.f.). In this case substituting the OTTB with a BB gives a only slightly worse fit (reduced $\chi^{2}$=1.424, 73 d.o.f.) with kT=$(15.5\pm1.0)$ keV and $\alpha = 4.5\pm0.1$.

In the case of interval C we obtain a satisfactory OTTB fit with kT=$(28.9\pm1.4)$ keV (reduced $\chi^{2}$ 1.06 with 71 d.o.f.).

Fig. 5.21 shows the energy spectra and best fit models for the time intervals A, B and C. For A and B we reported the OTTB+PL fit and for C the OTTB.