Astron. Astrophys. 327, L5-L8 (1997)

3. Data sample and analysis

The Mkn 501 data sample comprises data from 14 nights from March 15/16 to April 13/14, 1997 with a total observation time of 26.7 hours. Bad weather conditions and the rising moon prevented continuous observation. All observations were carried out in a mode where Mkn 501 was displaced in declination by from the optical axis of the telescopes, with the sign of the displacement changing every 20 min. A region displaced symmetrically by the same amount in the opposite direction was used to provide a control sample.

The image analysis and the reconstruction of the shower axis from the images is described elsewhere (Daum et al., 1997). In the present analysis, improved corrections for the telescope pointing were applied, and an algorithm to estimate the shower energy was added. Monte-Carlo simulations were used to determine the relation between the light yield measured in a camera as the sum of pixel amplitudes, , the energy E of the shower, and the distance r to the shower core. In addition, the fluctuation of the light yield, , was determined, taking into account the error in the measurement of r. The shower energy is then obtained as a weighted average over telescopes.

The system is expected to provide a -ray energy threshold of 500 GeV, an energy resolution of 20%, an angular resolution of about , and a determination of the shower impact point of about 15 m in each coordinate. The angular resolution was verified by observations of -rays from the Crab Nebula (Daum et al. 1997).

Already in the raw data, before selection cuts, a clear signal of Mkn 501 is visible. Fig. 1 shows the distribution of the reconstructed shower directions for all events which triggered at least two telescopes, and provided two images with 40 or more photoelectrons and at least two pixels with more than 10 photoelectrons. The position of Mkn 501, as reconstructed from such distributions (after cuts on the image shape, to reduce background), is consistent with its nominal position within the statistical error of .

Fig. 1. Distribution of the reconstructed shower directions relative to the direction to Mkn 501, for events where at least two telescopes triggered, before shape cuts.

For a quantitative analysis, we plot the distribution in the angle between the shower axis and the source location; shown in Fig. 2 (a) is . For the uniform background from charged cosmic rays one expects a flat distribution in . A -ray point source causes an excess around . The observed distribution shows these features. An estimate for the background under the signal is obtained by plotting the distribution of shower axis relative to a virtual source displaced by the same amount from the telescope axis as the real source, but in the opposite direction. This backgound is shown as a shaded histogram; it is flat in . In the region up to around the source, 3574 excess events are counted, corresponding to an average rate of 134 events/h.

Fig. 2. Line: distribution of events in the square of the angle relative to the direction to the source. The shaded histogram shows the background, see text for details. (a) before cuts, (b) after loose shape cuts, and (c) after tight shape cuts.

The shapes of Cherenkov images can be used to suppress cosmic-ray background relative to -ray showers; -rays generate narrower and more compact images. Therefore the width of each image in a given event is scaled to the Monte-Carlo expected width of -ray images as a function of image amplitude and distance to the shower core. As selection parameter the mean scaled width is calculated for all telescopes participating in an event. To maintain high efficiency and to minimize corrections, a very loose cut is applied by selecting events with a mean scaled width below 1.3. Fig. 2 (b) shows the angular distribution of events after this loose cut. The background is reduced by a factor of about 3, while the number of events in the peak is nearly unchanged. We verified that the high selection efficiency is maintained for all shower energies. At the expense of signal statistics, the background can be reduced further. Fig. 2 (c) illustrates the effect of tight cuts (Daum et al. 1997), which almost completely eliminate the background.

© European Southern Observatory (ESO) 1997

Online publication: April 6, 1998
helpdesk.link@springer.de