3. X-ray data analysis
The all-sky survey performed by the Röntgensatellit (ROSAT ; Trümper 1983) provides an excellent opportunity to obtain X-ray observations for the entire Pop II sample. The ROSAT all-sky survey (RASS) was performed with the Position Sensitive Proportional Counter (PSPC; Pfeffermann et al., 1987) between July 1990 and January 1991. Parts of the sky missed during this period were observed in February and August 1991. The PSPC has an energy range of 0.1-2.4 keV, and a spectral resolution at 1 keV. In the survey mode, the sky was scanned along great circles perpendicular to the ecliptic plane, with the scan period being synchronized to the orbital period of . In the field of view of the PSPC, an X-ray source was visible for s per scan. The visibility interval for each Pop II binary was determined by its ecliptic latitude, but lasted for at least 2 d. The resulting exposure times amounted to 300-1700 s, with a mean value of 770 s.
For each Pop II binary in our sample, we extracted regions of the sky, centered at the respective optical position. All source searches were conducted in three energy bands: Total: PSPC channels 8-240 (0.1-2.4 keV), Soft: 8-41 (0.1-0.28 keV), and Hard: 52-201 (0.5-2.0 keV). The source detection was performed by means of a maximum likelihood (ML) algorithm (cf. Cruddace et al., 1989). This algorithm calculates the probability p that the observed photon distribution matches the distribution derived from the theoretical point-spread function. The source parameters (counts, position, extent) are varied to maximize the likelihood . All of the source detection analysis was performed by using the EXSAS software (Zimmermann et al. 1994).
For each of the 86 RASS data sets, the following source detection technique is applied to the X-ray images of the total, soft, and hard bandpasses:
(i) All obvious sources are found using the LDETECT source detection algorithm (Cruddace et al., 1989). These sources are extracted from the X-ray image. The remaining image is then fitted by a spline function to give a smoothed background image.
(ii) The source detection is performed with an ML algorithm. The search is started from the optical position. Since we expect differences between the optical and X-ray positions due to errors in the RASS attitude solution, the position is taken as free parameter. The X-ray peak closest to the optical position is found. The source counts within a radius around the X-ray position are determined. The extraction radius will contain of the total source counts.
(iii) A detection threshold of , corresponding to an existence probability , was chosen. Therefore, we would expect 0.08 false detections (i.e., practically none) within our 86 trials. Sources with are considered as detections. For sources with , upper limits on the source counts are computed. The upper limit gives the number of counts within the extraction radius needed above the background to detect a source with a certain confidence level. We adopted a confidence level of (i.e., ).
For the minority of times during the RASS when the attitude was calculated from the gyros instead of the star trackers, the attitude solution was not correct and the detected photons had an incorrect position, about off in ecliptic latitude. As a consequence, some RASS sources were shifted as a whole or split into two images. We did not find evidence for this positional offset in the X-ray images of our Pop II stars, although it is possible that we obtain only upper limits for sources which actually would have been detections. However, since only a few percent of the RASS sources are estimated to suffer from this attitude problem (Voges 1995), we should miss at most one detection if any.
© European Southern Observatory (ESO) 1997
Online publication: June 5, 1998