7 Further classification of the cluster candidates

After this anticipation of the final results we return to the sequence of the REFLEX sample construction. Up to this point we have compiled a sample of cluster candidates relying only on machine based algorithms (except for the manual exclusion of obvious pathological cases like the bright diffraction spike stellar images and the multiple detections). The only two selection criteria are the X-ray flux limit for the X-ray sources and a signal of a galaxy overdensity in the optical data. This sample has still an estimated contamination of non-cluster X-ray sources of 30-40% (as discussed in the previous section). For the identification work that follows we treat each source individually, compile as much information as possible, and try to arrive at a safe classification in each case. The types of information used are the X-ray properties of the source, optical images (from the STScI scans of the POSS and UK Schmidt plates or CCD images if available), and literature information (including previous X-ray source identifications).

The basic X-ray source parameters used to assess the source properties are the probability for an extension of the X-ray emission (from the Kolmogorov-Smirnov test mentioned above) and the spectral hardness ratio. The hardness ratio and its photon statistical error is compared to the expected hardness ratio for a thermal cluster spectrum for a temperature of 5 keV and the absorbing interstellar column density at the source position (Dickey & Lockman 1990), for details see Böhringer et al. (2000). As a measure of the consistency of the observation with X-ray emission from the intracluster plasma of a cluster, we take the deviation of the predicted and observed hardness ratio in units of the statistical error (in $\sigma$ units).

In a first step we are discarding all sources that can be unambiguously classified as non-cluster objects. As a safe exclusion criterion we have either accepted a well documented previous identification or combined at least two quality criteria which exclude the identification as a cluster. Thus the following information leads to discarding a cluster candidate source:

i) Positive identification as a non-cluster X-ray source in the literature.

ii) The X-ray source is both well consistent with a point source and deviates by more than $3\sigma$ from the theoretically predicted hardness ratio of an X-ray cluster. In addition we find no indication of a cluster in the optical images.

iii) The X-ray source is point-like and there is an AGN spectrum observed for a galaxy or a point-like optical object at the X-ray source position.

iv) A point-like X-ray source coincides with a $b_{\rm J} < 12$ mag star (within a radius of about 30 arcsec) and there is no cluster visible in the optical image.

A large part of the contamination fraction can be removed from the source list by use of these criteria. The positive identifications are given simply by the observation of a clearly extended X-ray source and a galaxy cluster in the optical images. For all the X-ray sources for which no clear classifications can be obtained and also for all clusters that have not been identified previously and for which no redshift is available, further spectroscopic and if necessary imaging observations were conducted. In total 431 targets were observed by us within the ESO key programme for this project (including candidates with X-ray fluxes below the current flux limit of the REFLEX survey) as well as additional targets in related programmes (e.g. Cruddace et al. 2000 in preparation).

The identification strategy in the optical follow-up observations is similar to the scheme given above. We either try to establish the existence of a galaxy cluster as the counterpart of the X-ray source by securing several coincident galaxy redshifts in the X-ray source field or by arriving at an alternative identification of the X-ray source which is in general an AGN. AGN are found for about 10% of the sources for which spectra were taken. More details about the identification process and the different types of non-cluster sources found within the ESO key programme will be given in a subsequent paper which will also provide the object catalogue. Here we concentrate on discussing the statistics of some X-ray properties of the cluster and non-cluster X-ray sources in Sect. 10.