3 In-flight data analysis and calibration

The GRBM is performing very well since the satellite launch. The background level along the orbit, outside the region near the South Atlantic Geomagnetic Anomaly (SAGA), remains very stable, with an average variation of $\sim$10%. Because of background stability and source intensity, background subtraction for GRB data analysis is in general not critical, except for very weak events, events occurring close to the crossing of the SAGA and, for spectral evolution analysis, during the very beginning and tails of some events. For the ratemeters data, the background during the event is estimated by interpolating the data in $\sim$250 s before and after the event with a polynomial, the order of which is chosen basing on the background variation level. In the case of the 240-channels spectra, polynomials are used to interpolate the background trend in different energy ranges using 3 packets (each one covering 128 s) before and after the packet(s) containing the event spectrum. After background subtraction and proper grouping, GRBM data are analyzed using the response matrices described in previous section, which have also been converted to FITS format to be used with standard spectral analysis software packages like XSPEC. The systematics due to uncertainties in the knowledge of the response functions are on average of the order of 10%. They are mainly due to the uncertainty on calibration sources fluxes and to the statistical errors on calibrations data. The systematic error term is added in quadrature to that on the data before performing the fits.

The goodness of the response matrices has been verified in flight with Crab Nebula flux and spectrum measurements performed exploiting source occultation by the Earth. The Crab flux and spectrum derived from both the 2-channels ratemeters and the 240-channels spectra are consistent with values found by other experiments in the hard X-rays and soft gamma-rays, i.e. a photon index of $\sim$2.2 and a 100 keV flux density of $\sim 60\ 10^{-5}$ photons cm^-2 s^-1 keV^-1, with reduced $\chi$² values of $\sim$1.3 for $\sim$ 13 d.o.f. In particular, by fixing the photon index at the commonly adopted X-rays value of 2.1, we obtain a normalization of $9.64 \pm 0.49$ photons cm^-2 s^-1 keV^-1, fully consistent with the standard value of 9.7. The 240-channels spectra fits have been performed ignoring the energy channels from $\sim$40 to $\sim$70 keV. Indeed, as can be seen from Fig. 1, the calibration-based response matrix is performing very well in the whole detector energy band except for this energy range, in which the detector response is not well described, with a discrepancy between expected and measured counts/channel of $\sim$40%. This is observed also in LS3 and the reason is under investigation by means of Monte Carlo simulations. Up to now, the 40-70 keV region has been ignored when fitting the 240-channels spectra. The efficiency in the lower (40-100 keV) energy channel of the 2-channels response matrices has been corrected to account for this effect.

In addition to Crab Nebula measurements, cross-checks with BATSE results on a sample of 33 GRBs detected by both experiments and arriving from directions between $\pm 30\rm ^{\circ}$ with respect to LS1 and LS3 axis have been performed, showing that the fluences and peak fluxes in the range 50-300 keV agree within 10%.

  

Figure 2: GRB 970111 1s spectral evolution measured with LS3