Globular clusters contain many X-ray sources at lower luminosities, erg s-1. These sources were first discovered with the Einstein satellite (Hertz & Grindlay 1983), and many more were found with ROSAT (for a compilation, see Johnston & Verbunt 1996). The nature of these low-luminosity sources is the subject of debate, because various types of objects can emit X-rays at such luminosities, such as soft X-ray transients in quiescence, cataclysmic variables, RS CVn binaries, and recycled neutron stars (see e.g. Fig. 8 in Verbunt et al. 1997). The most compelling identification of a dim X-ray source with an object observed at other wavelengths is the recycled radio pulsar in M 28: the X-ray flux varies on the pulse period (Danner et al. 1994). Plausible identifications with cataclysmic variables have been suggested for dim X-ray sources in NGC 6397, NGC 6752, NGC 5904 and 47 Tuc (Cool et al. 1995b; Grindlay 1993; Hakala et al. 1997; Verbunt & Hasinger 1998). These identifications are based on the proximity of the X-ray position to that of a cataclysmic variable, and thus their probability depends on the accuracy of the X-ray position.
The accuracy of the ROSAT position of a detected X-ray source is determined by two factors: the statistical accuracy of the position of the source on the detector, and the accuracy with which the position of the detector as a whole is projected on the sky. For a sufficient number of photons the statistical error is less than an arcsecond, but the projection in general has a typical error of . Secure identification of a source in the detector field reduces the error in the projection to the statistical error of the identified source, provided that the optical (or radio) position has better accuracy. Only one identification is necessary, because the roll angle of the detector (i.e. the North-South direction) is accurately known; nonetheless, identification of more than one source is preferable to allow checks on internal consistency. In a globular cluster the surface density of possible counterparts is so high that chance coincidence usually cannot be excluded; a secure identification can usually be made only for X-ray sources detected well outside the cluster. This method has been used by Verbunt & Hasinger (1998) to improve the positional accuracy of the sources in the core of 47 Tuc from to , whereby the area in which the source is expected to lie is reduced sufficiently to exclude several proposed identifications, and increase the probability of others, including two possible cataclysmic variables.
In this paper we investigate three clusters known to contain multiple dim X-ray sources in their core, which have been observed in long exposures with the ROSAT HRI, and one cluster known to harbour a transient. We analyse hitherto unpublished observations and detect both previously published and new X-ray sources. All source positions are checked in the SIMBAD data base versus positions of other objects, and we find objects in the Hipparcos or Tycho Catalogues (ESA 1997; Perryman et al. 1997; Hog et al. 1997) with each cluster, i.e. counterparts with very accurate positions. In Sect. 2 we describe the observations and our data reduction procedures; Sects. 3 to 5 describe the results for Cen, NGC 6397, and NGC 6752, respectively. In Sect. 6 we discuss an observation of Liller 1. A discussion of our results is given in Sect. 7.
© European Southern Observatory (ESO) 2000
Online publication: June 20, 2000