4. The catalogue

A total of 95 GRBs have been included in the catalogue. The information on these events is presented in Table 1. In the first column of the table the names of the bursts are given, which were formed using a common terminology: the first letter "W" indicates the instrument's name, WATCH, then follows the year, month and day of the detection of the event. The burst may have a letter ("b" or "c") appended to its name, if it is the second or third burst detected in a day. The second column yields the time (UT) of the trigger. In the subsequent columns the basic characteristics of the bursts are given. As a measure of the burst duration we use the quantity T₉₀ which is the length of the interval during which 90% (from 5% to 95%) of the total counts from a burst was accumulated. The fluence and the peak energy flux of the bursts were calculated in the two energy bands: 8-20 keV and 20-60 keV (spectral shape similar to that of the Crab Nebula was assumed). When calculating these quantities we made a correction for the aspect of the burst source, which is known for the majority of the presented events from localizations with either WATCH or CGRO/BATSE (Meegan et al. 1996). However, for 18 bursts such information is not available, hence we accepted for them as an estimate of the detector geometrical efficiency (see Fig. 1 (click here)) its expectation value of 0.7. The quoted errors for the fluxes are purely statistical ones, the uncertainties due to unknown spectral shapes and source aspects have not been considered. The peak flux was calculated using the count rate data with the best time resolution available for a given event. Also, given in the table is the ratio of the fluences in the 20-60 keV and 8-20 keV energy bands, which characterizes the hardness of the burst spectrum. Finally, the last column of the table contains information on detections of the bursts by other experiments that were in orbit during the period examined: KONUS/GRANAT, PHEBUS/GRANAT, SIGMA/GRANAT, GINGA, BATSE/CGRO, COMPTEL/CGRO, OSSE/CGRO, DMS, Mars Observer, PVO, ULYSSES, WATCH/EURECA and YOHKOH (Golenetskii et al. 1991; Terekhov et al. 1994; Terekhov et al. 1995; Sunyaev et al. 1993; Ogasaka et al. 1991; Meegan et al. 1996; Hanlon et al. 1994; Hurley et al. 1994; Brandt et al. 1994).

4.1. Burst time histories

In Fig. 2 the time histories of the detector count rate during the bursts in the two WATCH energy bands are shown. For the longer bursts, the best time resolution available is either 7 s or 14 s. For the shorter events, time histories of 1 s resolution are presented. Finally, for the two shortest events W900404 and W930106, which lasted less than 1 s, light curves obtained by integrating the count rate over 20 ms intervals are presented. We note that these two time histories are necessarily distorted to some degree by the modulation effect caused by the rotation of the collimator. The background was in most cases estimated by averaging the count rate over time intervals immediately before and after the burst. When the background was strongly variable during the burst, we used approximation of the count rate by polynomials of first or higher orders.

        

       
Figure 2: Time histories of the gamma-ray bursts detected by WATCH in the two energy ranges: 8-20 keV (upper panels) and 20-60 keV (lower panels). Time runs relative to the burst start (see Table 1). The background count rate is indicated by the dashed line

4.2. Burst durations

Figure 3 (click here) shows the duration (T₉₀) distribution for the bursts detected. It was obtained using those 89 out of 95 events whose durations could be determined reliably. At least 5 of the 6 events excluded from the analysis are apparently short (shorter than a few seconds), but they were too weak to force the on-board burst logic to trigger. Hence no count rate data with a good time resolution needed for determining their durations is available. The mean burst duration is 66 s, and the maximum of the distribution lies in the interval 10 to 100 s. Our sample contains 2 events shorter than 2 s and 7 events longer than 200 s. In other experiments, two classes of bursts were identified: of duration shorter and longer than  s (Kouveliotou et al. 1993). It was noticed that the energy spectra of the shorter bursts of type 1 were generally harder than those of the longer bursts of type 2 (Kouveliotou et al. 1993; Lestrade et al. 1993). The negligibly small number of type 1 events in our catalogue compared to the previous results (Meegan et al. 1996; Terekhov et al. 1995) can be accounted for by at least two selectional effects: 1) all bursts with detection significance less than included in our duration sample are longer than 7 s, because this value is the time resolution of the data, 2) WATCH is sensitive to X-ray photons with energies significantly below the effective energy of the photons emitted in type 1 bursts.

  
Figure 3: Histogram of durations (T₉₀) for the WATCH bursts

4.3. Burst spectral hardness and evolution

The spectral information on the emission produced during GRBs provided by WATCH is limited to the count rates in the two energy ranges 8-20 keV and 20-60 keV. Hence to describe the spectra of the observed bursts we calculated the ratios of the burst energy fluxes in the higher and lower bands. As follows from the cross-correlation diagram presented in Fig. 4 (click here), there is no clear dependence of burst overall hardness ratio (the 20-60 keV fluence divided by the 8-20 keV fluence) on burst duration. This graph gives further support to the above statement that most of the bursts in the WATCH catalogue are of the same type.

  
Figure 4: Burst hardness ratios (fluence over the 20-60 keV band divided by that over the 8-20 keV band) vs. burst durations

The WATCH observations illustrate that the energy spectra are usually not constant but evolve throughout bursts, the typical situation being that the spectra during the burst rise and decay phases are softer than that at the peak of the event. This is demonstrated by Fig. 5 (click here), where burst peak hardness ratio, i.e. the ratio of the peak energy fluxes in 20-60 keV and 8-20 keV (in some bursts the two flux maxima are not coincident in time), is shown as a function of burst overall hardness ratio. It can be seen that the former is larger than the latter for the majority of events, reflecting the fact that bursts generally have "sharper'' profiles in the harder energy band. In 13 bursts (Table 2 (click here)) this spectral evolution reveals itself especially distinctly as a significant activity observed only at 8-20 keV either preceding or following the hard X-ray event (see the corresponding time histories in Fig. 2, for discussion on part of these events with occurences before October 1992 see also Castro-Tirado (1994) and Castro-Tirado et al. (1994)). Similar X-ray precursor and tail activities have been observed before in a number of bursts by a few space-flown GRB instruments sensitive to medium or soft X rays (Murakami et al. 1991). In the observations carried out with the GRB detector on board the GINGA satellite, which had a low-energy cut-off at as low as 1.5 keV, such X-ray-active events accounted to about one third of the total bursts detected (Murakami et al. 1992). The photon spectra measured with GINGA at 1 to 10 keV during both the burst X-ray precursor and tail could be approximated by a black-body model with temperatures between 1 and 2 keV, indicating that the emission mechanism at these burst phases may be thermal at variance with the apparently non-thermal emission during the main gamma-ray event.

  
Figure 5: Burst peak hardness ratios vs. burst overall hardness ratios. The dashed line indicates the case of equality of these two quantities

 

Burst name Type of X-ray activity

W900222 precursor and tail

W900708 precursor and tail

W900901 tail

W910817 tail

W911209 precursor

W920718 precursor

W920723b tail

W920903 tail

W920903b precursor

W920925 precursor

W921013b tail

W921022 tail

W930705 tail

Table 2: Gamma-ray bursts with precursor or tail X-ray activity

 

4.4. Distribution of bursters in space

For describing the distribution of GRB sources in space, the statistics (Schmidt et al. 1988) is widely used. In this statistics, for each burst detected according to the set criterium one calculates the quantity by the formula: , where is the minimum count rate to satisfy the burst detection criterion, and is the maximum count rate during the burst. If the bursters are homogeneously distributed in space, will be uniformly distributed in the interval (0, 1), and its expectation will be 0.5.

We have carried out a test on the WATCH GRB catalogue. Calculating the values of the parameters and , in order to have a homogeneous sample, we used the count rate data with 14 s integration time, for data of such resolution exist for all of the events. Of crucial importance in our case is the definition of , since the inclusion of many bursts into the catalogue was dependent on observations of the event by other experiments. We have chosen the criterion that the peak count rate must exceed the background by at least (this limit very roughly corresponds to a flux threshold of  erg cm^-2 s for the WATCH sensitivity energy range and a typical trigger time scale of 8 s), for it proves that virtually all cosmic bursts meeting it can be registered with WATCH independently, through localization. The average value of for 43 so selected events is . This implies that the burst sources are homogeneously distributed within the WATCH sampling distance. The WATCH instrument is not sensitive enough to make it possible for us to analyze the distribution of more distant bursters, a deficit of which has been observed in a number of burst experiments (Meegan et al. 1996; Terekhov et al. 1995).

4.5. Burster locations in the sky

The location of the source of a burst can be found from the WATCH data if the following three requirements are met: 1) the burst is strong enough, 2) it lasts longer than  1 rotation of the modulation collimator, and 3) the modulation pattern used for localization is not significantly distorted by the presence in the light curve of bright details on time scales shorter than the rotation period of the collimator. The presence of other bright sources in the field of view can possibly make the procedure of deriving the source position unreliable even when the above conditions are satisfied. Therefore we considered a source localized only if the same position was resulted from two statistically independent modulation patterns. These patterns may belong to different time intervals, different energy ranges, or, in rare cases, different phase intervals of the same modulation pattern. For several weak bursts an additional independent verification was obtained through comparing the positions provided by WATCH and BATSE.

We have succeeded in localizing the sources of 47 bursts (Table 3 (click here)). The statistical uncertainty of position determination is inversely proportional to the significance of source detection and varies between 7 arcmin and 1.5 deg for the localized bursts (the radius of a circle with an area equal to the area of the confidence region), the localization region being an ellipse somewhat contracted along the source off-axis angle . In preparation of this catalogue special efforts were made to decrease the influence of various systematic effects on burst localizations. Significant progress in this direction has now been achieved, in the first instance due to the use of information provided by the star tracker of the SIGMA telescope for determination of the attitude of the spacecraft at the times of bursts. Besides, on the basis of an ample archive of WATCH data on localizations of bright persistent X-ray sources, we have found more accurate values for some of the parameters relevant to the instrument, including the mounting angles defining the orientation of the WATCH detectors with respect to the SIGMA star tracker. Unfortunately, the star tracker was at times off during WATCH observations, and for 16 bursts we thus were bound to use other information resources to calculate the attitude, namely readings of the navigational instruments of the spacecraft and the knowledge of the celestial positions of bright X-ray sources that are always present in the field of view and can be localized with the WATCH detectors. This leaves an irremovable uncertainty resulted from the rapid ( min period) and virtually unpredictable wobbling of the spacecraft attitude. For the 31 positions that were calculated using precise navigational information we conservatively estimate the remaining systematic error at 0.2. This uncertainty is mainly due to the not complete accounting for various physical phenomena in the instrument mathematical model currently used, in particular the dependence of the instrument's positional response on the energy spectrum of the incident radiation. The cumulative localization uncertainties given in the last column of Table 3 (click here) were calculated by summing (in quadrature) the statistical () and estimate systematic errors. Although we have already arrived at the state that for most of the localized burst sources it is the statistical error that mostly contributes to the location uncertainty, a cooperative effort between IKI and DSRI is now in progress to further improve the modelling of the instrument, which we expect will eventually enable further reducing of the localization regions of several stronger bursts.

 

Burst l b Total error name

W900118 174.68 -44.32 289.46   16.64 0.54 0.73

W900123b 357.19 -38.56 347.89 -72.61 0.60 0.78

W900126 131.15 -37.79 258.76    3.09 0.24 0.32

W900222 336.73   34.83   92.09 -19.25 0.79 0.81

W900708 185.91   30.62 181.40   82.99 0.22 0.30

W900708b 252.79   16.20   34.92   33.74 0.43 0.48

W900901 276.31 -45.18 349.28 -14.59 0.45 0.67

W900925 133.14 -36.72 258.92    5.00 0.77 0.80

W900929 169.68  -6.48 265.83   49.59 0.44 0.49

W901009 348.61   30.40   99.28 -27.98 0.51 0.72

W901116   39.93   24.97 151.95 -31.73 0.48 0.52

W901121   30.39   72.40 128.24   10.26 0.55 0.59

W901219 348.62 -54.00 329.84 -57.77 0.55 0.59

W910122 297.48 -71.23 324.06 -30.26 0.66 0.69

W910219 212.94   58.54 104.43   55.61 0.93 0.95

W910310 184.10    6.38 279.48   67.64 0.24 0.55

W910627 199.60  -2.60 316.32   59.57 1.08 1.09

W910821 353.08 -72.01 311.26 -43.81 0.38 0.43

W910927   49.70 -42.72 250.24 -56.37 0.80 0.94

W911016 297.37  -4.71   35.35 -15.05 0.78 0.92

W911202 171.97 -22.59 278.81   36.34 0.80 0.94

W911209 261.92 -44.19 345.18  -5.15 0.60 0.78

W920210 154.15   47.89 167.78   53.51 1.08 1.19

W920311 132.25 -36.39 258.20    4.65 0.26 0.33

W920404 323.07   22.53   73.92 -20.85 0.63 0.66

W920714 221.43 -30.75 330.15   26.01 0.48 0.52

W920718   21.37  -3.36 143.31 -64.88 0.75 0.78

W920718b 296.17 -55.95 341.69 -29.51 0.62 0.65

W920720 145.67 -11.20 246.46   30.31 1.07 1.09

W920723b 287.08   27.33   59.26    8.69 0.19 0.28

W920814 259.83 -45.17 343.53  -4.47 1.13 1.24

W920902 279.08 -22.81   10.87  -7.04 0.41 0.46

W920903 295.87   35.46   70.02    5.81 0.67 0.70

W920903b 301.54   22.59   61.53  -5.05 0.22 0.30

W920925 201.11   42.20 100.96   73.49 0.73 0.76

W920925c 330.80   25.48   81.64 -23.60 0.34 0.39

W921013   87.97    1.93 204.27 -12.31 1.50 1.51

W921013b 117.71   33.41 186.99   26.20 0.25 0.32

W921022 254.43  -9.64  10.21   19.97 0.52 0.72

W921029   35.82  -0.42 166.28 -55.37 1.23 1.25

W930612 109.24 -71.20 282.48 -23.56 0.68 0.70

W930703 311.06    8.04   54.10 -20.65 0.59 0.77

W930706 281.42 -20.18   14.23  -7.83 0.42 0.47

W940419 358.82 -48.19 326.63 -66.27 1.00 1.02

W940701 145.67  -6.15 241.88   33.54 1.54 1.56

W940703 133.20   28.11 197.08   37.69 0.12 0.24

W940907 161.42 -31.81 273.93   23.88 0.84 0.98

 

When analyzing a celestial distribution of sources, it is necessary to know how long different regions of the sky have been monitored. We have compiled from the WATCH data an exposure map, which is shown in Galactic coordinates in Fig. 6 (click here). Calculating this map, in cases when a sky region had been observed simultaneously by two or three WATCH detectors, the corresponding time interval was appended only once. We considered an area of the sky being in sight of the instrument if it was not more than 65 off axis. The maximum of the exposure map of 510 days is located in the vicinity of the Galactic center (l= -7, b= 3), and its minimum of 218 days has coordinates l=87, b=29. The exposure time averages 372 days over the celestial sphere, which corresponds to an all-sky monitoring efficiency of 21%.

In Fig. 7 (click here), a celestial map of the positions of the burst sources in Galactic coordinates is presented. In order to check the angular distribution of the bursts for possible large-scale anisotropies, we have calculated its dipole and quadrupole moments relative to the Galactic center and the Galactic plane, respectively. Upon correction for exposure time (terms are summed with weights inversely proportional to the position exposure) the dipole moment ( is the source angular distance from the Galactic center), the quadrupole moment (b is the source galactic latitude). Therefore our observations are consistent with an isotropic distribution of the burst sources on the sky, in agreement with the corresponding BATSE result (Meegan et al. 1996). We carried out similar calculations for samples of bursts selected out by various attributes. We have not found any significant deviations from isotropy, in particular, the distribution of the sources of the stronger bursts is apparently isotropic. Note that the dipole and quadrupole moments calculated using the 32 events of the first WATCH catalogue of GRBs are: , , respectively (Castro-Tirado et al. 1994). Thus, the tendency for bursters to concentrate towards the Galactic center, evident at a significance of in the first catalogue, is not confirmed by the new data obtained since October 1992.

  
Figure 6: Sky exposure map in Galactic coordinates constructed from the data of WATCH observations in 1989-1994. Exposure time is measured in days

  
Figure 7: Positions of 47 burst sources in the sky in Galactic coordinates. The circles shown are the twice-zoomed real localization circles