Astron. Astrophys. 324, 683-689 (1997)

2. Observations and analysis apparatus

2.1. The instrument

The COMPTEL telescope was designed to detect gamma-ray photons in the 0.75-30 MeV range with an energy resolution of 6-10% (FWHM). Within a field of view of about 1 steradian, it is able to locate gamma-ray sources with a spatial accuracy of typically. A detailed description of the instrument and the detection principle are presented by Schönfelder et al. (1993). Ideally, incoming photons are first Compton scattered in an upper detector layer, then completely absorbed in a lower detector layer. The energy deposits and locations in these layers determine the scatter direction, scatter angle, and total energy of each photon. Events are sorted in a 3D dataspace defined by the scatter direction and Compton scatter angle .

The CGRO observation programme consists of several successive phases. The present work involves all observations of the Galaxy from 16 May 1991 to 4 October 1994 (Phases I to III as listed by Gehrels et al. 1994). We specifically combined the 75 viewing periods with the telescope pointing within from the Galactic plane, of durations of typically s each. The 3D dataspace spans a total of in , respectively, and the events are binned in cells.

2.2. Data reduction and background models

The detection of celestial gamma-ray emission involves the identification of source signatures in the 3D dataspace. We have generated a map of the Galactic plane at by applying a maximum-likelihood method to the data (de Boer et al. 1992). This algorithm is very appropriate to a search for ⁴⁴ Ti sources. Testing indeed for the presence of point sources throughout the map, it gives flux estimates and statistical significances of such sources (Sect. 2.3). For comparison, we also generated maximum-likelihood maps for two neighbouring energy bands. We thus prepared 3D datasets for the following energy intervals: 0.89-1.07 MeV, 1.07-1.25 MeV, and 1.25-1.43 MeV, where the middle one is centered on the ⁴⁴ Ti line. This narrow bandwidth takes full advantage of the energy resolution of the instrument ( keV at ). Selections in time of flight (channels 115-130) and pulse shape (channels 0-110) are applied for rejection of backward-scattered photons and neutron events, respectively (Schönfelder et al. 1993).

Proper description of the background is critical, as the signal is not expected to exceed 1% of the total number of events. We have achieved background modelling within the 3D dataspace in two different ways:

(i) data from neighbouring energy bands are interpolated, taking into account the specific variations of the different dataspace variables with energy (Knödlseder et al. 1996);

(ii) data from the same observations and energy range are filtered with a smoothing technique similar to that described by Bloemen et al. (1994).

In principle, method (i) includes any continuum contribution in the background so that mainly line emission is imaged. In contrast, method (ii) makes no discrimination and images the total (line + continuum) celestial emission.

2.3. The significance of source detections

For each pixel of the sky image, the maximum-likelihood ratio is calculated as the ratio between the likelihood of the best fit to the data by the background model alone to that of the best fit including a point source at this position: . The quantity measures the statistical significance of the presence of a point source at the pixel location.

In the process of identifying known celestial objects as possible sources, follows a distribution with one degree of freedom: the source flux. For example, a value of would indicate that the object is detected in the gamma-ray band at the significance level.

A higher threshold is required in a search for previously unknown sources since now obeys a distribution with three degrees of freedom: source flux and coordinates. A confidence level implies . However this figure stems from a local optimisation with no account of the large sky area scanned by COMPTEL (Schönfelder et al. 1993). We require () in order to claim a serendipitous source detection: at this level, spurious excesses are expected to show up in our map.

© European Southern Observatory (ESO) 1997

Online publication: May 26, 1998

helpdesk.link@springer.de