3.1. Detected cluster candidates
Using the various odd/even catalogs containing the objects extracted from single frames (150 sec exposures), lists of cluster candidates have been produced for each patch and band using the cluster identification pipeline based on the matched-filter method described in detail in Paper II. The cluster identification has been applied to the I-band data from patch B, using the same parameters to describe the cluster radial profile and luminosity function (kpc, Mpc and , ), the same SExtractor detection parameters ( and corresponding to the area of a circle with radius ), and the same selection criteria (, and ) as given in Paper II. Over the analyzed area of 1.1 square degrees, 19 new cluster candidates were found, which gives a density of 17.2 cluster candidates per square degree, consistent with the results discussed in Paper II and with the value estimated by Postman et al. (1996). Out of the 19 detections 12 are "good" candidates, i.e. , those detected at in at least one catalog, or at in both I-band catalogs. In addition, there are 7 candidates detected at 3 in only one catalog, nearly all at large redshift (). In Paper II the noise properties of the cluster-finding procedure were investigated and the frequency of noise peaks was estimated to be 4.2 per square degree. Using this noise frequency it is found that the number of questionable candidates is consistent with that expected for noise peaks, about 5 in the area analyzed in Patch B.
The properties of the detected cluster candidates are summarized in Table 1 which gives: in Column (1) the cluster id; in Columns (2) and (3) the right ascension and declination (J2000); in Column (4) the estimated redshift; in Column (5) the measure of the cluster richness derived from the maximum likelihood (see Paper II, Eq. (1)); Column (6) gives the number of galaxies within a magnitude interval of 2 magnitudes delimited at the bright end by the magnitude of third brightest cluster galaxy; in Columns (7) and (8) the significance of the detections in the even and odd catalogs, respectively; in Column (9) the significance of the detection using the galaxy catalogs extracted from the V images (see below); and in Column (10) other identifications. The upper part of the table lists the "good" candidates, while the remaining candidates are listed in the lower part of the table. Note that the detection with the largest significance is the cluster easily seen near the center of the patch. This cluster is Abell S84 at (Abell et al. 1989, Strubble & Rood 1987).
Table 1. Preliminary cluster candidates for EIS Patch B.
For simultaneous display of the cluster candidates in all available passbands a new facility for extracting image postage stamps from the coadded images was implemented in the ESO Science Archive. Using this facility all cluster candidates detected in I-band were visually inspected and most were found to be promising. Even though candidates in the lower part of the table are less conspicuous, most do seem to be possible clusters. Out of the seven candidates listed in the bottom part of the table, at least 4 lie in a region where the limiting isophote varies significantly between the odd and even frames which may explain the fact that they were detected only in the odd catalogs. Note that 2 of them were detected in the V catalogs. From this examination one also encounters cases where superposition of clusters at significantly different redshifts may occur, as judged from the available color information.
The cluster-finding pipeline was also used to detect cluster candidates in the available band catalogs for patches A and B. For this purpose the same radial profile as for the band data, and the same slope of the Schechter function were adopted, while the characteristic magnitude was taken from Postman et al. (1996) , corrected to the Johnson system according to the transformation given by those authors. Furthermore the limiting magnitude was chosen to fit the 80% completeness limit for the band catalogs of . Even/odd band candidate catalogs were produced, using the same peak finding criteria given above, but in contrast to the I-band procedure no selection criteria on detection significance, persistency or cluster richness (see Paper II) were imposed.
These detections were cross-identified with the I-band ones listed in Table 1, and those listed in Paper II for which band data are available (16 candidates). The latter are listed in Table 2, which reproduces part of Table 2 in Paper II, adding also the V-band information. The matching between the I- and V-band catalogs was done based on position only, using a search radius of 1 arcmin, centered on the nominal band detection. The choice of 1 arcmin was dictated by the estimated uncertainty in the position derived for the cluster candidates. It should be noted that in most cases where a match is found, the estimated redshifts agree to within . Only in 5 cases the discrepancy was larger. The significance of the detections in the V-band for the case of patch B cluster candidates is given in Column (9) of Table 1 and for patch A in Column (9) of Table 2. In case the candidate is detected both in the even and in the odd catalogs, the listed V-band significance is the highest of the two.
Table 2. Cluster candidates for EIS Patch A that are found in the region where both I and V-band data are available. In addition to the information already presented in Table 2 of Paper II, here also the V-band detection significance is listed. The format of the table is the same as for Table 1.
In total there are 35 candidate clusters where color information is available over a 2 square degree area. The regions where color information is available are typical regions with respect to the data quality as characterized by the seeing and limiting isophotes. The distribution of estimated redshifts for this sample is shown in Fig. 1. The median estimated redshift for the whole set presented in this figure is . The shaded portion of the histogram shows the distribution of estimated redshifts (as obtained from the I-band detection) for the cluster candidates also detected in V-band (combining patches A and B). Out of 19 candidates with , 18 ( 95%) are also detected in V-band; for clusters at , 5 out of 16 are detected in V-band. This result is not surprising since the ability to detect clusters varies with redshift, with the redshift range of the candidates detected in V-band being smaller than that of those detected in I-band. Using the rule of thumb that the data should reach at least one magnitude fainter than at the redshift of a given cluster to allow for its detection, one can translate the galaxy catalog limiting magnitudes adopted in the cluster search into limiting redshifts for cluster detection. The I-band limit of then translates into a limiting redshift between (no-evolution model) and (passive evolution model). These values are in good agreement with the results presented in Paper II, and by Postman et al. (1996), where cluster detections are limited to . Using the same argument, the V-band limit of translates into a limiting redshift between (no-evolution) and (passive evolution). This argument is consistent with the above findings, providing strong support for the reality of the detected candidates. Note that the two matched detections with estimated redshift are among the cases where a significant discrepancy between redshifts estimated from the V- and I-band data is present, with the V-band estimate being significantly lower than the I-band one.
The probability of detecting a cluster also depends on its richness. Despite the small number statistics, nearby candidates not detected in V data tend to be poor, with an estimated richness close to the lower limit adopted for the inclusion of a candidate in the catalog (). At high-redshifts, only very rich clusters, probably with a large fraction of ellipticals, are detected in the two passbands. There are two such cases in the above table, but note that neither would have been included as cluster candidates based on the V detection alone. Their appearance on the images strongly suggests that both are likely to be clusters at high redshifts. However, since some galaxies are seen in the V images, either their matched filter redshifts are overestimated or there are foreground concentrations leading to their detection in the V data. Only spectroscopic follow-up will be able to resolve such cases.
3.2. Color-magnitude diagrams
The availability of data in two passbands can, in principle, provide an alternative way of confirming cluster candidates and their estimated redshifts, based on the detection of the sequence of cluster early-type galaxies in a CM diagram. As described in paper III a color catalog was constructed for regions where multi-band data were available. The color catalog was built by merging of the and band catalogs, associating objects based on their extension (see Paper III for further details). Using this color catalog a color-magnitude diagram, based on magnitudes measured within a 5.3" aperture, was constructed for each cluster candidate. This diagram includes all galaxies within a radius of 0.5 h-1 Mpc (H0 = 75 km s-1/Mpc) from the nominal cluster position that are detected in both bands. Fig. 2 shows these diagrams for all cluster candidates with . Also indicated in the plot are the values of and the color of a typical elliptical (no-evolution) at the estimated redshift of the cluster, as derived from the matched filter. At low redshift, the sequence of early-type galaxies is clearly visible, but at the evidence for a CM relation is, in most cases, less compelling. In no cases with a CM sequence was visible, because of the relatively shallow V-band data.
Considering the combined patch A and B sample, one finds that out of 35 clusters in the region of overlap of the V- and I-band images, there are 11 with evidence for a CM relation, with redshifts extending out to . Furthermore, the redshift estimates based on color and the matched filter seem to agree, in most cases, within 0.1. In addition there are 6 cases in that redshift range where there is a weaker suggestion of a color sequence.
Comparison between the cases detected in both I and V and the cases where a color sequence is found shows that in most cases (16 out of 17) where the CM-diagram supports the detection, also the candidate is detected in both bands. There are 4 cases where a detection is found in both V and I-band and no color sequence is seen. All of these cases have significances below 4, and two of them correspond to the matched detections with shown in Fig. 1. In one case evidence for a color sequence is found but there is no detection in the V-band. Altogether this shows that the V- and I-band data provide a consistent picture, giving further support to the reality of the detections based on I-band data only.
© European Southern Observatory (ESO) 1999
Online publication: April 19, 1999