DETECTION OF GRB 090618 WITH THE RT-2 EXPERIMENT ON BOARD THE CORONAS-PHOTON SATELLITE

During the GRB event, the RT-2 payload was completely in the "SHADOW" mode (away from the Sun) which started at 08:16:10.207 UT and ended at 08:37:35.465 UT. GRB 090618 was detected by the RT-2 instruments (Rao et al. 2009) with a large off-axis angle of 77°. Data from RT-2/S and RT-2/G are used for the present analysis. The scientific data from the detectors were stored in the memory of RT-2/E and then compressed using the on board software before being transferred to the satellite system (known as SSRNI) for a down-link to the ground station. During the "SHADOW" mode, the spectra are accumulated every 100 s (see the spectral analysis section) and eight channel count rates (for each detector) are accumulated every 1 s. These eight channels are divided equally between the 15–100 keV data from the 3 mm thick Na i (Tl) detector and the 26–1000 keV data from the 25 mm thick Cs i (Na) detector, operated in phoswich mode. For the present work, we use the latter four channel data, which have the ranges of 26–59 keV, 59–215 keV, 215–330 keV, and 330–1000 keV, respectively.

The "SHADOW" mode data of the RT-2/S and RT-2/G detectors of "GOOD" time span (away from high background regions) of ∼800 s from 08:23:27 to 08:36:47 UT were analyzed. During this time, the satellite was completely away from both the polar cap and high background South Atlantic Anomaly regions. GRB 090618 was detected by both the scintillator detectors. The light curves of 1 s time resolution were generated from the data obtained from both the detectors at different energy bands. Spectral data of 100 s, during the present GRB event, were also analyzed to study the evolution of the GRB spectrum.

BAT (Barthelmy et al. 2005) on board the Swift mission (Gehrels et al. 2004) is a highly sensitive, large field of view instrument, primarily designed to monitor the sky to detect gamma-ray events. It consists of an array of CdZnTe detectors, located behind a CAM. Swift also has an XRT (Burrows et al. 2005) and a UV/Optical Telescope (UVOT; Roming et al. 2005) which can make follow-up observations within a few hundred seconds after the trigger.

BAT registered GRB 090618, triggered at 08:28:29 UT. The BAT position of the burst with 3' uncertainty is R.A.(J2000), decl.(J2000) = 294.021, +78.353 (Schady et al. 2009a). The astrometrically corrected position using Swift-XRT data and Swift-UVOT data is R.A.(J2000), decl.(J2000) = 293.99465, +78.35677 (Evans et al. 2009; Schady et al. 2009c). We have analyzed the Swift BAT data to understand the prompt emission from this GRB.

We used heasoft6.5.1 for our analysis. First, we created the detector plane image using the task batbinevt and then the detector mask file was created using task batdetmask. The quality map file was created using task bathotpix. Mask weighting using new XRT and UVOT position to BAT data was applied using task batmaskwtevt. Then, mask-weighted light curves were generated using task batbinevt. The GRB light curve shows multi-peaked structure with a precursor and the main burst starts at T₀ + 50 s. We generated light curves for the full burst (T₀ to T₀ + 180) in four energy bands: 15–25 keV, 25–50 keV, 50–100 keV, and 100–200 keV (these are the energy ranges used in Ukwatta et al. 2010, for timing analysis).

The GRB was detected in the wide band of 26–1000 keV with both the RT-2 detectors. Since the GRB is incident at a large incident angle, meaningful light curves were available from the Cs i (Tl) detectors in different energy bands (26–59 keV, 59–215 keV, and 215–1000 keV ranges for RT-2/S, and 26–59 keV, 59–215 keV, 215–330 keV, and 330–1000 keV ranges for RT-2/G). The light curves, obtained from RT-2 and BAT instruments, are shown in Figure 1 with a bin time of 1 s and the time is given with respect to the BAT trigger time T₀ on 2009 June 18 08:28:29 UT. The BAT light curves are in units of mask-weighted counts, which are essentially background-subtracted counts per fully illuminated detector for an equivalent on-axis source. The RT-2 count rates are the combined light curves from the RT-2/S and RT-2/G detectors, and they are normalized to the nominal detector area of 100 cm² for each detector. The light curves are arranged with increasing energy (from the top) and wherever possible, light curves of comparable energy are plotted together (25–50 keV BAT data and 26–59 keV RT-2 data in panel (b), and 100–200 keV BAT data and 59–215 keV RT-2 data in panel (c)). For clarity, the 59–215 keV light curve from RT-2 is scaled down by a factor of 10 and vertically shifted up by 0.1 count s⁻¹.

Zoom In Zoom Out Reset image size

The burst profile was found to have a complex structure. The entire burst episode lasted for about 150 s with the brightest pulse detected at T₀ + 65 s. The other two weak emissions are registered at T₀ + 85 and T₀ + 115 s with low intensity. Another instrument, the Konus-RF on board the Coronas-Photon satellite, also detected GRB 090618 with an identical burst profile (Golenetskii et al. 2009). The burst profile is not clearly detected in the lowest energy band of RT-2 (26–59 keV), whereas in the high-energy X-rays, the burst is detected with significant emission. There is a very significant softening of the spectrum in the precursor with the 25–50 keV light curve peaking several seconds after the high-energy light curves. The main pulse at T₀ + 65 s shows a double structure, the 50–100 keV BAT profile showing a close similarity to the 59–215 keV RT-2 light curve. The pulse at T₀ + 85 s shows a single smooth structure in 100–200 keV (similar to the 59–215 keV RT-2 light curve), with an indication of multiple structures in the low energies.

We have attempted to model the pulse with the FRED profile developed by Kocevski et al. (2003). The empirical relation for the flux (photon counts s⁻¹) distribution is given by

$$\begin{eqnarray} &&F(t) = F_m\left(\frac{t}{t_m}\right)^r\left [\frac{d}{d+r} + \frac{r}{d+r} \left(\frac{t}{t_m}\right)^{(r+1)}\right ]^{-(r+d)/(r+1)},\quad \end{eqnarray} \tag{ 1 }$$

where F_m is the maximum flux at t_m, and r and d are the rising and decaying indices, respectively. First, we fit the total light curves (26–1000 keV for RT-2 and 15–200 keV for BAT). The start times for fitting (for each pulse) were kept fixed at t_m − t_st, where t_st were varied between 15 s and 25 s. Since there were negligible changes in the derived parameters, particularly for the measurement of width, we kept t_st fixed at 25 s. The reduced χ² for the RT-2 data is 1.38, for 75 degrees of freedom (dof). For the higher sensitivity BAT data, however, the FRED profile does not take into account the sub-structures in the pulses, and hence formally unacceptable fits (reduced χ² ⩾ 10) were obtained. Since our main emphasis is to find the broad pulse characteristics, we give below the parameters obtained from a FRED fitting. Furthermore, we find that the derived parameters for pulses 1 and 2 from the BAT data (these two pulses have a large overlap) are quite sensitive to the initial parameters used for fitting and hence the quoted formal errors could be underestimates.

The resultant derived parameters, along with the nominal 1σ errors (obtained from the criterion Δχ² = 2.7), are given in Table 1. There is reasonable agreement in the derived parameters, obtained from RT-2 and BAT data. While fitting, an upper bound of 1000 is kept for the parameter d (for d ≫ r, F(t) becomes independent of d). We have measured the pulse width as the FWHM value of the fitted light curve. The fitted light curves with burst profiles are shown in Figure 2 for the total energy band of 26–1000 keV for the RT-2 detectors (bottom panel) and for the total energy band of 15–200 keV for the BAT detector (top panel). The fitting procedure is repeated for light curves of different energies. The times of the maximum emission t_m, however, were kept fixed at those values obtained from the fitting for the full energy light curves of the respective detectors (RT-2 and BAT). The value of t_m could depend on energy up to about a second for these energies (see the delay analysis presented later), which might result in a further systematic error in the width measurement of <1 s. We have plotted the width as a function of mean energy in Figure 3. Mean energies are calculated as the mean energy of incident photons in the detector, based on the response function of the detectors, and they are derived to be 125.62 keV, 266.6 keV, and 427.5 keV for the 59–215 keV, 215–330 keV, and 330–1000 keV bands, respectively, for the RT-2 data. The mean energies for the four energy bands of BAT data are 20.82 keV, 35.45 keV, 68.07 keV, and 123.73 keV, respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The first two pulses have similar profiles in both data sets, though in the RT-2 data the first pulse is much stronger than the second one. There is an indication of further sub-structures in the BAT data which is not very apparent in the RT-2 data, possibly due to the lower sensitivity. The width decreased monotonically with energy. Also, there is a steepening trend for the latter pulses, particularly for the third pulse. For example, by defining a width index ξ such that width ∝ E^−ξ (Fenimore et al. 1995), we find ξ to be 0.18, 0.07, 0.14, and 0.05, respectively, for the four pulses, for a combined fit to the RT-2 and BAT data. These parameters are summarized in Table 3.

To investigate the softening of the spectra, we have performed cross-correlation analysis between the light curves of various energies using the BAT data. We have taken the 15–25 keV range as the base energy. Cross-correlation is done for the full light curves as well as in parts: the first part includes the precursor (T₀ to T₀ + 50), the second part includes pulses 1 and 2 (T₀ + 50 to T₀ + 77), and third and fourth parts include pulses 3 and 4, respectively. We have used the utility crosscor of the XRONOS software package (a part of the HEASOFT software package of HEASARC) to derive the cross-correlation function (CCF) with respect to lag. A Gaussian function was used to fit the CCF and measure the time lag, and the error is estimated using the criteria Δχ² = 4.0 (see Dasgupta & Rao 2006). The results of the cross-correlation analysis are given in Table 2. We found soft lags which show clear energy dependence. The measured lag increases with energy and they are shown in Figure 4. There is also a tendency for this steepening to be flatter for the latter pulses. Defining a delay index d_i such that delay = a + d_iln(E), where a is a constant, we find d_i to be very steep for the precursor (−3.71) and it increased from −0.50 for second part (pulses 1 and 2) to −0.13 for third part (pulse 3). For the fourth part (pulse 4), however, d_i is found to be −0.92. The values of ξ and delay index (d_i) are compiled in Table 3.

Zoom In Zoom Out Reset image size

Ukwatta et al. (2010) have derived delays for the light curve of T₀ + 46.01 to T₀ + 135.35 s as −0.171, −0.314, and −0.579 s, respectively, in three energy bands given in Table 3, which are close to the average values of delays for second and third parts (T₀ + 50 to T₀ + 100 s) reported in Table 3 (−0.247, −0.365, and −0.630 s, respectively, for the three energy bands). We have followed a method of cross-correlation analysis similar to that followed in Ukwatta et al. (2010) and we have confirmed that we get consistent results in the time span used in their work.

We have analyzed the spectral data during the GRB event. It showed the typical band spectrum with peak energy at about 164 keV and integrated 20 keV–1 MeV fluence of 2.8 × 10⁻⁴ erg cm⁻². We used 15–200 keV BAT data and the 100–650 keV RT-2 data of the GRB 090618 to perform joint spectral fitting.

The spectral information was available in the RT-2 data every 100 s. The output from each detector is passed through two amplifiers of different gains (G1 and G2) and spectral data were available for each of the amplifiers: 1024 channel spectra for G1 (covering the energy range of 26–215 keV for the Cs i detector) and 256 channel spectra for G2 (covering the energy range of 215–1000 keV). The number of channels are suitably rebinned for the spectral fitting.

The RT-2/S and RT-2/G detectors are essentially collimated phoswich detectors user for solar flare studies (Debnath et al. 2010). The shielding used for these detectors, however, are optimized to use them for hard X-ray spectroscopic studies of solar flares in the 15–100 keV region and as omni-directional hard X-ray/gamma-ray detectors between ∼50 keV and 1000 keV (Sarkar et al. 2010). GRB 090618 is incident at an angle of 77°, and we have used a Monte Carlo (MC) simulation technique using the GEANT4 toolkit to derive the spectral response of the RT-2 detectors for this large incident angle (see Sarkar et al. 2010 for details). We created a mass model of the detector including all parameters like detector sizes, collimator, shield materials, and mechanical support structures. Incident photons are used for the simulation in the energy range of 1–1000 keV in 50 equal bins in log scale and 10⁵ photons in each bin are considered. The number of photons registered in the detectors as a function of energy is normalized to the incident photons for a GRB model with the spectral parameters of α = −0.32, β = −5.0, and E₀ = 67.7 keV. The results are insensitive to model parameters (see later). This normalization is deemed as the effective area of the detectors. The derived effective areas for the Na i and Cs i detectors are shown in Figure 5. Since the effective area of Cs i is about an order of magnitude larger than the Na i detector, we have taken only the Cs i data (and the corresponding response function) for our spectral fitting. The response matrix for the RT-2 detectors are generated using the genresp tool of ftools. The channel–energy conversion is derived from the background spectral line at ∼58.0 keV (Nandi et al. 2009) and 122 keV line due to the on board calibration source (Co⁵⁷) and the energy resolution function is taken from the ground calibrations. Using this response matrix, a joint fit to the RT-2 and BAT data is performed. From the derived best-fit spectral parameters, MC simulation is again carried out and the RT-2 response matrix is again created. Convergent results were obtained in the first trial itself, indicating that direct blocking from the surrounding material and interaction with detector play a major part of the response and the second-order effects like scattering are unimportant for response matrix generation, at least to the level of sensitivity achieved for GRB 090618.

Zoom In Zoom Out Reset image size

Except for a 2 mm aluminum filter in RT-2/G, both the detectors are identical in performance. The setting for deriving the on board counts in the high-energy channels is lower in RT-2/S and hence we have a single energy bin above 215 keV in RT-2/S (compared to the two energy bins in RT-2/G). Though we get consistent spectral results for both the detectors, we report the spectral parameters derived from RT-2/S which has better spectral response above 100 keV. While making a simultaneous fit to the BAT data, we kept the relative area between RT-2 and BAT as a free parameter, and it is derived to be 1.30 for the combined fit to the full data. The relative area between RT-2 G1 and G2, however, is kept fixed at what is dictated by the response matrix. For the time-resolved spectral fit, the relative area between RT-2 and BAT is kept fixed at the value obtained for the full data (i.e., 1.30).

The spectrum can be well fitted with the model introduced by Band et al. (1993). In this model, two power laws are smoothly joined at (α − β)E₀, where α is the first power-law index, β is the second power-law index, and E₀ is the break energy. The best-fit spectral parameters are given in Table 3, along with the calculated peak energy E_p(=E₀(2 + α)), the best fit χ², and the degrees of freedom used for the fitting. Figure 6 shows the best-fit model along with the unfolded spectrum. We divided the BAT data into four parts and for each part we perform the spectral analysis. Since RT-2 does not have time-resolved spectral data, we have used the count rates as broad spectral data for this time-resolved spectral analysis. The results are shown in Table 3. Timing analysis parameters (width index ξ and delay index d_i) are also given in Table 3. We note that the spectral results agree quite well with that obtained from the Fermi data (Ghirlanda et al. 2010). The values of E_p, derived in this work, are 264 keV, 248 keV, 129 keV, and 33 keV, respectively, for the four parts which compares well with the value of 296 keV, 319 keV, 180 keV, and 81 keV derived for the peak of the light curves in these regions in the Fermi data (time ranges, after T₀, of 3–14 s, 63–67 s, 80–85 s, and 114–130 s, respectively).

Zoom In Zoom Out Reset image size