3.1. Source detections
The EGRET data analysis of the combined 7 weeks of Virgo data revealed a strong source consistent with the location of 3C 273 (Fig. 1). The overall detection significance at the position of the quasar for energies above 100 MeV is 10.4 assuming -statistics for a known source. The 7-week mean flux (E100 MeV) is (43.45.8) 10- 8 ph cm-2 sec-1, which is roughly 3 times the average flux listed in the third EGRET catalogue (Hartman et al. 1999). A flux level in excess of that - (48.311.8) 10-8 ph cm- 2 sec-1 in VP 308.6 - had been reported previously only once (Hartman et al. 1999). 3C 273 is identified with the -ray source on the basis of its sky location.
Simultaneously with these EGRET findings, COMPTEL observed significant emission from the sky position of 3C 273 in three (1-3 MeV, 3-10 MeV, 10-30 MeV) out of its 4 standard energy bands. The source is not detected at the lowest COMPTEL energies (0.75-1 MeV). The overall detection significance is 7.7. The average flux values in the individual COMPTEL bands (see Table 2) are among the largest ever measured in this energy window, showing that the source was active in -rays during this period.
Table 2. Fluxes and upper limits for 3C 273 for the different analysed periods in units of 10-5cm-2s-1 for the COMPTEL bands and 10-8cm-2s-1 for the EGRET band. The energy bands are given in MeV. The errors are 1. The upper limits are 2. An upper limit is given when the significance of an individual flux value is less than 1. The errors and upper limits are statistical only. The observational periods A-D are defined in Table 1 and described in the text.
3.2. Time variability
To check for time variability we have subdivided the total 7-week observation into slices of the 7 individual VPs covering typically one week each (Table 1). 3C 273 was not detected by EGRET during the first two VPs (detection-significance threshold 3.5. It then appeared just above the detection threshold and increased further to show the largest -ray flare observed during the CGRO era. This maximum flux level of (7720) 10- 8 ph cm-2 sec-1 is reached in VP 610. Thereafter the flux returned to an intermediate level, which is clearly detectable. Fits of the two EGRET light curves shown in Figs. 2 and 3 (upper panel) assuming a constant flux resulted in -values of 19.9 for the light curve containing 7 flux points (individual VPs) and 12.5 for the one containing 3 flux points. According to -statistics these values correspond to probabilities of 2.9 10-3 and 1.9 10-3, respectively, for a constant flux, which convert to 3.0 and 3.1 evidence for a time-variable 100 MeV flux. The largest change in flux occurred between VPs 610 and 610.5 when the flux dropped by a factor of 4.1 within 7 days. This is the shortest time variability at -ray energies ever reported for 3C 273. The significance that the two flux values are different is 2.7. During the two-week outburst (VPs 609 and 610) the source reached a flux level even slightly in excess to the COS-B flux reported by Swanenburg et al. (1978).
We checked for time variability in the COMPTEL bands by subdividing the data into the same time slices as chosen for EGRET. No obvious time variability is visible in either energy band, however, the statistics in the different COMPTEL bands became marginal, resulting in large error bars on the flux values. To improve the statistical significance we combined individual VPs. We defined three time intervals which were selected according to the EGRET light curve: a pre-flare period (VPs 606-608) covering 3 weeks which we shall call A in the following, a flare period (VPs 609 and 610) covering two weeks which we shall call B, and a post-flare period (VPs 610.5 and 611) which we shall call C (see also Table 1). The simultaneous EGRET and COMPTEL 3C 273 fluxes for different energy bands and various time periods are listed in Table 2 and are plotted in Fig. 3. In contrast to EGRET, COMPTEL observes no obvious time variability. The same -procedure as applied to the EGRET data showed that the different COMPTEL flux values are consistent with a constant level as is also obvious from Fig. 3. In particular the COMPTEL light curves do not show a hint of increased -ray emission during the EGRET flaring period. This result suggests that the observed flare is either solely a high-energy (30 MeV) phenomenon, or a time offset of at least 2 weeks between the EGRET and COMPTEL -ray bands is required.
3.3. Energy spectra
The EGRET spectral analysis followed the standard EGRET procedure (see Sect. 2). The energy range between 30 MeV and 10 GeV was subdivided into 10 energy intervals and the likelihood analysis was used to estimate the number of source photons in each energy bin. The data were fit to a single power-law model of the following form
with the parameters (photon spectral index) and (intensity at the normalization energy ). was chosen such, that the two free parameters are minimally correlated. We derived 1-errors on the parameters by adding 1.0 to the minimum -value (Mattox et al. 1996).
This approach was applied to the sum of all data as well as to selected subsets (see Fig. 3, Table 1) to check for a possible trend in time. In addition we summed the subsets A and C having roughly equal flux levels, which we shall call D, to check for a possible spectral trend with flux by using the improved event statistics. The results of the spectral fitting are given in Table 3. First of all the spectra are well fitted by simple power-law functions: the average spectral index in the EGRET band is = 2.400.14, which is comparable to previous results. For instance, von Montigny et al. (1997)found spectral indices in the range between 2.2 and 3.2 with a trend of spectral hardening with increased source flux. The EGRET fits show this trend as well: the spectrum is hardest during the flaring period. However, this is not significant because the error on the spectral slope is quite large.
Table 3. Results of the power-law fitting (I(E) = I0 () for the different time periods. The left part gives the results for fitting only the EGRET data (30 MeV - 10 GeV), while the right part shows the results of fitting the combined EGRET and COMPTEL (3 MeV - 10 GeV). The errors on the fit parameters are derived by the +1 contour level.
To derive the COMPTEL fluxes of 3C 273, we have applied the standard maximum-likelihood method as described in Sect. 2. Background-subtracted and deconvolved source fluxes in the 4 standard energy bands have been derived by taking into account the presence of further -ray sources (3C 279 at / = 194.1o/-5.8o, 4C -02.55 (1229-021) at / = 188.0o/-2.4o) and source candidates (at / = 193.5o/0.5o and / = 173.5o/8.5o), showing some evidence in the maps (Fig. 1). We note that inclusion of these further sources and source candidates has only a marginal effect on the flux of 3C 273. Their inclusion or exclusion changes the derived 3C 273 fluxes only within its error bars. Assuming a power-law shape, the quasar is fitted typically with a spectral index of 2 throughout the COMPTEL energy band. For example, the sum of all data (`ALL') which has the best statistics, yields = 1.920.13 between 0.75 and 30 MeV, which is significantly harder than = 2.410.14 as found in the EGRET range for the same observational period (Fig. 4). This result indicates a spectral hardening towards lower energies with the turnover starting at a few MeV. This spectral behaviour of 3C 273 is well known and has been reported previously (e.g. Lichti et al. 1995; von Montigny et al. 1997).
To take advantage of the contemporaneous observations of both instruments in neighbouring energy regimes, we combined the deconvolved EGRET and COMPTEL spectra for the different time periods. Fitting the whole energy range (0.75 MeV to 10 GeV) with a power-law model, we derive harder spectral slopes and increased -values compared to fitting solely the EGRET range. This effect weakens if we exclude the lowest-energy (0.75-1 MeV) spectral point, and disappears when we subsequently remove the next (1-3 MeV) COMPTEL flux point from the fitting procedure. This behaviour is easily explained by the spectral turnover at low energies, which seems to affect the fits only below 3 MeV. Fitting a broken power-law model
where I0 describes the differential source flux at the normalization energy E0, the high-energy spectral photon index, the break in spectral photon index towards lower energies ( = - ), and Eb the break energy, provides consistent results. The best-fit value for the break energy for the sum of all data is found to be at 5 MeV, which, however, is not well defined due to the small lever-arm towards lower energies. Considering these facts, we conclude, that the EGRET power-law spectrum extends into the COMPTEL band down to 3 MeV before it is substantially altered by the spectral turnover. This is illustrated in Fig. 4. By chance the quasar was observed simultaneously in the X-ray band by the BeppoSax satellite. Haardt et al. (1998) published the X-ray results, i.e. fluxes and spectra, of 3C 273 covering a monitoring period of 4 days (January 13, 15, 17, and 22, 1997). These X-ray observations are coincident in time with the transition from the -ray flaring period to the moderate post-flare -ray level (Fig. 2). Fig. 4 shows a broad-band high-energy spectrum of 3C 273, containing the COMPTEL and EGRET spectra for the sum of all 7 weeks and the best-fit power-law shape for the quasi-simultaneous X-ray measurements (covering only one day) provided by Haardt et al. (1998).
To investigate this high-energy component in more detail, we fitted the different observational subsets between 3 MeV and 10 GeV with single power-law models (Fig. 5). We take advantage of this enlarged (with respect to only EGRET) energy band, for which the spectral index can be determined more accurately. So, for the 3 MeV to 10 GeV band, the trend that during the flare the spectrum hardens as suggested by the EGRET analysis (see above), is observed more significantly. Especially, if we compare the periods B and D. The 1 statistical errors in spectral index during the flare and non-flare intervals do not overlap anymore. The fit results for the EGRET band only and this enlarged band are given in Table 3, and the latter ones are shown graphically in the Figs. 5 and 6. Along the 7-week observation, 3C 273 is observed to have a steep spectrum at the beginning, which hardens during the two-week flaring period, and turns back to roughly the same shape in the 2-week post flare period (Fig. 7). This result is consistent with the constant flux observed at COMPTEL energies. The flare occurs mainly at energies above 100 MeV, not affecting the COMPTEL points which results in a hardening of the overall -ray spectrum.
From this spectral analysis, we conclude too, that we either observed a phenomenon which occurs solely at high -ray energies, or that there are time delays between the different -ray bands. The first case would require an additional spectral component triggered by some mechanism which is only effective at EGRET energies. The second possibility would require that the COMPTEL energies are either delayed by two weeks or would be in advance by three weeks with respect to EGRET.
© European Southern Observatory (ESO) 2000
Online publication: February 9, 2000