next previous
Up: The ESO Nearby

3. Comparison with the COSMOS catalogue

  At the start of the ENACS programme, the large-scale galaxy catalogues that are presently available in several flavours, like the COSMOS (e.g. Wallin et al. 1994) and the APM surveys (Maddox et al. 1990), and that are based on automatic scanning of photographic survey plates and subsequent computer processing, were not yet available. Therefore, we had to produce our own galaxy catalogues in the direction of the target ACO clusters, in preparation for the Optopus multi-object spectroscopy. The most important requirements that these cluster galaxy catalogues had to meet were that they should have very good positional quality (better than 1 arcsec), and that good photometry was available so that galaxy samples complete in apparent magnitude could be selected.

We prepared our cluster galaxy catalogues with the Leiden Observatory Astroscan automatic plate measuring machine. The positional quality, required for very good relative positioning of the Optopus fibres as well as to secure excellent positional consistency between galaxy and guide star fibres, could easily be met. The photometric requirement could not be met in an absolute sense, but the relative photometry of the Astroscan was known to be quite good. This allowed the definition of galaxy samples complete to a well-defined magnitude limit, the absolute value of which still had to be calibrated by photometric CCD imaging of well-chosen galaxy subsamples.

At the completion of the observational part of the ENACS programme, we have compared the ENACS positional and photometric systems with those of the COSMOS catalogue. Note that the latter contains a high-quality subset, the Edinburgh-Durham Southern Galaxy Catalogue, or EDSGC (Heydon-Dumbleton et al. (1989), for which the calibration is of higher quality than for the rest of the COSMOS catalogue, while its completeness has been studied in detail.

The comparison between the ENACS and COSMOS catalogues was made for only 77 of the 107 ENACS clusters, but this should not affect the general validity of the results. The 77 clusters in question are listed in Table 3 (click here). We have marked with an asterisk the 15 clusters within the EDSGC.

 

A0013 A0524 A2502 A2911* A3144 A3733
A0087 A0543 A2569 A2915 A3151 A3744
A0118* A0978 A2715* A2923* A3158 A3764
A0119 A1069 A2717* A2933 A3194 A3781
A0151 A2353 A2734* A3009 A3202 A3806
A0168 A2354 A2755* A3093 A3223 A3809
A0229 A2361 A2764* A3094* A3264 A3822
A0295 A2362 A2765 A3108 A3341 A3825
A0303 A2383 A2778* A3111 A3354 A3827
A0367 A2401 A2799* A3112 A3528 A3864
A0380* A2426 A2800* A3122* A3559 A3897
A0420 A2436 A2854 A3128 A3703 A3921
A0514 A2480 A2871* A3141 A3705
Table 3: ENACS clusters with COSMOS data

 

3.1. The cross-correlation between the ENACS and COSMOS catalogues

  First, we cross-correlated the galaxies in the ENACS catalogue with the galaxy catalogues for the relevant sections of the COSMOS catalogue (kindly provided by H.T. MacGillivray). Since it was not clear, a priori, to what extent the positional systems in both catalogues are indeed identical, we first determined the optimum maximum difference in position required for the cross-identification. It appears that a maximum position difference of 7 arcsec must be allowed in order not to miss plausible identifications. However, the surface densities of ENACS and COSMOS galaxies are such that the number of galaxies that is cross-identified does not change significantly if one increases the maximum allowed position difference to tex2html_wrap_inline1719 50 arcsec. In other words: chance coincidences start to be important only for position differences larger than tex2html_wrap_inline141750 arcsec.

When preparing the galaxy catalogues for the Optopus observations we limited the selection to the 30-50 brightest galaxies within the Optopus apertures on survey plates which, for the large majority of the ENACS clusters, are identical to the plates that were scanned for the COSMOS catalogue. Therefore, it is not very meaningful to ask which fraction of all galaxies in the COSMOS catalogue appears in the ENACS. The reason is that this fraction will depend on the effective magnitude limit of our galaxy samples which varies significantly between clusters (the magnitude distributions of the COSMOS and ENACS galaxies, which describe the completeness of the ENACS samples are discussed in Sect. 4.2 (click here)). Furthermore, the success-rate of the Optopus spectroscopy decreases towards fainter magnitudes (see Fig. 4 (click here)).

However, it is interesting to ask the complementary question: viz. what fraction of the ENACS galaxies do not appear in the COSMOS catalogue. That some ENACS galaxies (which through the spectroscopy have been "proven" to be galaxies) will not be found in the COSMOS catalogue is to be expected. Heydon-Dumbleton et al. (1989) have estimated that the EDSGC is >95% complete at bj = 20.0, and by determining the fraction of ENACS galaxies not found in the COSMOS catalogue we can provide independent evidence about the completeness of the COSMOS catalogue and its EDSGC subset; or more precisely: at least of those areas that contain rich clusters.

Two of the 77 clusters for which we have COSMOS data, viz. A2502 and A3144, have less than 4 ENACS galaxies, and we have not used those in the following analysis. That leaves 75 clusters which contain, in the areas of overlap between COSMOS and ENACS (which is not always the entire Optopus area) a total of 3896 ENACS galaxies. All these ENACS galaxies are well above the magnitude limit of the COSMOS catalogue. For 357 ENACS galaxies, there is no COSMOS counterpart within 7 arcsec; of these 357 galaxies, 226 do have a nearest neighbour between 7 and 100 arcsec (almost exclusively at more than 50 arcsec), while the remaining 131 have a nearest neighbour at more than 100 arcsec. Taken at face value, these numbers would seem to indicate that the COSMOS catalogue is 91% rather than >95% complete, at the magnitude limit of the ENACS samples which is generally 0.5 to 1.0 mag brighter than bj = 20.0.

This result is somewhat unexpected, but it must be realized that the two completeness estimates refer to slightly different parts of the COSMOS catalogue. Whereas the estimate by Heydon-Dumbleton et al. is the average for the entire EDSGC, our estimate refers to areas with high surface density in the COSMOS catalogue, where it may be more difficult to obtain the same completeness level as in the field. In addition, it is likely that the completeness of the COSMOS catalogue depends somewhat on galactic latitude, local galaxy surface density etc. Therefore, the two completeness estimates need not be really discordant.

We have checked whether the EDSGC is more complete than the total COSMOS catalogue. This appears not to be the case. Of the 357 ENACS galaxies without COSMOS counterpart 92 are in the 15 clusters with EDSGC data, which contain 805 ENACS galaxies in total. Consequently, 265 ENACS galaxies in the other 60 clusters (with a total of 3091 ENACS galaxies) have no COSMOS counterpart. In other words: the completeness estimates for the EDSGC and the rest of the COSMOS catalogue are essentially the same, viz. 89 and 91%.

One might naively expect that the relatively high surface density of, especially, bright and very extended galaxies could be the main reason why some ENACS galaxies do not appear in the COSMOS catalogue. In constructing the ENACS catalogues, plates of all clusters were inspected visually to ensure that the latter were included as much as possible. It is conceivable that the pattern recognition software used in constructing the COSMOS catalogue had some problems in recognizing, especially the brighter, late-type galaxies (we have seen several examples of this, e.g. in A3822). Yet, this effect does not seem to be the main cause for the apparent incompleteness of the COSMOS catalogue. The magnitude distribution of the 357 ENACS galaxies without COSMOS counterparts is virtually the same as that of the other 5258 ENACS galaxies (see Fig. 1 (click here)), and there is at most a small "excess" of brightest galaxies among the 357 ENACS galaxies without a counterpart in the COSMOS catalogue.

  figure318
Figure 1: The normalized distribution wrt apparent magnitude (R25) for the 357 ENACS galaxies without a counterpart in the COSMOS catalogue (solid histogram). For comparison, the normalized R25 distribution for the other 5258 ENACS galaxies is shown (dashed histogram)

For the 75 clusters used in the comparison, the overall fraction of ENACS galaxies that have no COSMOS counterpart is 9%. For individual clusters the fraction varies between tex2html_wrap_inline1417 4% and tex2html_wrap_inline1417 30%. The clusters with smaller number of galaxies show somewhat larger fractions of "missing" COSMOS galaxies, probably mostly as a result of discretization effects due to small numbers.

3.2. Positions

  The sample of 3896 galaxies (in 75 clusters) for which the position difference between COSMOS and ENACS is less than 7 arcsec, has been used to investigate the relation between the positional systems in the two sets of data. There could be differences on scales of a few arcsec, because the astrometry for ENACS was done on fairly small sections of the same Schmidt plates for which, in the COSMOS catalogue, one overall solution was made. However, it appears that such differences are small. The average offsets per cluster are between -2 and +2 arcsec, in both coordinates, and the rms position offset is slightly less than 1 arcsec. In the clusters themselves, the position differences for individual galaxies are of the same order. The overall distribution of position differences, made with all galaxies in common between ENACS and COSMOS (taking out the average, small offset for each cluster), is well described with a Rayleigh distribution with a dispersion of 0.9 arcsec.

For 3 clusters, viz. A3128, A3354 and A3744, all of which were observed with more than one Optopus plate, there is some evidence for an offset of the positions in one Optopus area wrt those in the other Optopus areas. The offsets are probably due to the fact that the astrometry for the different Optopus areas within a cluster was not always based on the same set of standard stars. However, the offsets are small, viz. at most 2.5 arcsec, and often it is not possible to be sure which positions are correct. Although the offsets are probably significant, we have not attempted to correct them; fortunately they are of the same order as the offsets between different clusters, as well as the random position errors.

3.3. Magnitudes

 

3.3.1. Summary of the ENACS photometry

For the 3896 galaxies that we used in Sect. 3.2 (click here), we have also analyzed the relation between the R25 magnitudes in the ENACS catalogue and the bj magnitudes of the COSMOS catalogue. Before we can discuss the results, we must briefly summarize how the ENACS R25 magnitudes were derived.

When we produced the galaxy catalogues for the Optopus spectroscopy, by scanning the copies of survey plates with the Astroscan measuring machine, we also obtained accurate photographic photometry. The survey plates that we used, and on which the photographic photometry was done (by measuring the sum of photographic densities, i.e. approximately the amount of silver in the galaxy image), were of two kinds. First, and for most clusters, we used film copies of the SERC survey (with green-sensitive IIIa - J emulsion) and secondly, for the other clusters, we used glass copies of the red POSS-I plates (with red-sensitive 103a-E emulsion).

This photographic photometry was calibrated with CCD-imaging. Because of the limited amount of time available, we did most of our CCD-imaging in R-band and only a small fraction in (the more time-consuming) B-band. Even so, we only managed to calibrate the photometry of about 40 clusters. For those, we determined and applied individual zero-points, while for the other clusters we applied the average calibration curve for the clusters with CCD-calibration (see Paper I).

In the case of the IIIa-J plates, we actually measured a photographic bj magnitude, which we transformed into a calibrated R25 magnitude, by effectively subtracting the average apparent bj-R25 colour of those galaxies that were used for the calibration. In other words: for the IIIa-J plates, the R25 magnitudes are, in effect, bj magnitudes on a pseudo R25-scale, so that differences in the ENACS R25-values are in reality differences in bj. On the other hand: for the red POSS-I plates, we really calibrated photographic R-magnitudes with R-magnitudes derived from the CCD-imaging, and differences in ENACS R25 are differences in real R25.

In Table 4 (click here) we indicate, for each of the 107 clusters in the ENACS, on which type of optical material the magnitudes are based and how these were calibrated. In this table, we indicate if the R25-magnitudes are pseudo R25-values (indicated by G, corresponding to IIIa-J) or real R25-values (indicated by R, corresponding to 103-E), and whether the zero-point that we applied was the average value (a), or individually determined from the CCD-calibration (i).

 

A0013 G a A0087 R a A0118 G a
A0119 R a A0151 R i A0168 R i
A0229 R a A0295 R i A0303 R a
A0367 G i A0380 G a A0420 R a
A0514 G i A0524 G a A0543 G a
A0548 G a A0754 R i A0957 R i
A0978 R i A1069 R i A1809 R i
A2040 R i A2048 R i A2052 R a
A2353 R a A2354 R a A2361 R a
A2362 R a A2383 G i A2401 G a
A2426 R i A2436 R a A2480 G i
A2500 G a A2502 R a A2569 R a
A2644 R a A2715 R a A2717 G i
A2734 G i A2755 G a A2764 G i
A2765 G a A2778 G a A2799 G a
A2800 G a A2819 G i A2854 G a
A2871 G a A2911 G a A2915 G i
A2923 G a A2933 G a A2954 G a
A3009 G i A3093 G a A3094 G i
A3108 G i A3111 G a A3112 G i
A3122 G i A3128 G i A3141 G i
A3142 G i A3144 G a A3151 G a
A3158 G i A3194 G a A3202 G a
A3223 G i A3264 G i A3301 G a
A3341 G a A3354 G a A3365 G a
A3528 G i A3558 G a A3559 G a
A3562 G a A3651 G i A3667 G i
A3677 G a A3682 G a A3691 G i
A3693 G a A3695 G a A3696 G a
A3703 G a A3705 G a A3733 G a
A3744 G a A3764 G a A3781 G a
A3795 G i A3799 G a A3806 G a
A3809 G i A3822 G i A3825 G a
A3827 G a A3864 G i A3879 G a
A3897 G a A3921 G a A4008 G a
A4010 G a A4053 G a
Table 4: The magnitude types and offsets

 

3.3.2. Comparison of ENACS and COSMOS photometry

As explained above, the comparison between the COSMOS bj and the ENACS R25 magnitudes for galaxies for which we did the photographic photometry on the red POSS-I plates, is a comparison between magnitudes in different spectral bands. Such a comparison therefore involves the individual colours of all galaxies, as well as an offset (i.e. the average colour of the calibrator galaxies). In Fig. 2 (click here) we show the relation between bj and R25 for the galaxies for which bj-R25 measures a real colour. In this figure we have corrected the bj magnitudes by 1.5 mag, which approximately takes into account the average colour of the galaxies. The result is quite reassuring: there do not appear to be serious problems with either of the magnitude scales, and the fairly wide colour distribution of the galaxies is clearly visible.

  figure362
Figure 2: The relation between bj-1.5 (i.e. the COSMOS bj magnitude corrected for the approximate average colour) and R25, for the galaxies in the clusters for which R25 was based on 103a-E plates

On the other hand, the comparison between bj and R25 magnitudes for galaxies measured on IIIa-J plates, is a comparison between two measures of the same thing, because R25 is actually bj on a pseudo R25 scale. In this case, the offset between bj and R25 is equal to the average colour of the galaxies that were used for calibration. In addition, there is some noise due to different sampling of the brightness distributions, small differences in the definition of the aperture over which the brightness was integrated, and possibly some noise generated by the two measuring machines.

The difference between the two cases (bj vs. real R25 and bj vs. pseudo R25) is clearly visible in Fig. 3 (click here), where we show two distributions of relative colour, viz. bj-R25 of an individual galaxy referred to the average colour tex2html_wrap_inline1855 of its cluster. The upper histogram refers to 59 "clusters" scanned on IIIa-J plates (one cluster, A3264, was not included in the upper histogram because it has only 5 galaxies in common between ENACS and COSMOS, so the average colour is not very well defined); the lower histogram refers to the 17 clusters with photographic photometry on 103a-E plates.

It is clear that the lower histogram is significantly wider than the upper one, as a result of the appreciable range of galaxy colours. This is indicated not only by the dispersions in tex2html_wrap_inline1857, which are 0.23 and 0.11, respectively but also by the long tails in the lower distribution. The dispersion for the IIIa-J plates, of 0.11 mag, is quite satisfactory in view of the estimated random errors in the individual magnitude estimates of about 0.15 mag.

We have checked if there are differences between the two subsets of the COSMOS catalogue, i.e. EDSGC and non-EDSGC and, as expected, we indeed find that the EDSGC subset has a better magnitude-calibration than the non-EDSGC subset. This is apparent from the following numbers: if one makes separate versions of the upper histogram in Fig. 3 (click here), for EDSGC and non-EDSGC we find dispersions of 0.097 and 0.114 respectively. Even stronger evidence is provided by the dispersions in the individual values of tex2html_wrap_inline1855 which are 0.25 and 0.37 for EDSGC (14 clusters) and non-EDSGC (44 clusters) respectively. However, within the errors the average colours are the same, viz. 1.54 tex2html_wrap_inline1861 0.06 for the EDSGC and 1.44 tex2html_wrap_inline1861 0.06 for the non-EDSGC part of the COSMOS catalogue.

  figure386
Figure 3: The distribution of the colour difference (bj-R25) for galaxies that are common to COSMOS (the source of bj) and the ENACS (the source of R25). Note that all colours are referred to the average colour tex2html_wrap_inline1855 of the "cluster" to which the galaxy belongs. The upper histogram is for "clusters" for which R25 was estimated from IIIa-J plates, the lower histogram for clusters for which R25 was measured on 103a-E plates

Note that in the previous paragraph we have tacitly assumed the ENACS magnitudes to provide a reference system for the COSMOS magnitudes. However, the distribution of the average colours tex2html_wrap_inline1877, in principle also contains information on the quality of the ENACS magnitude calibration. As with the magnitudes, the meaning of these average colours depends on the type of photometric photometry that was calibrated with the R-band CCD-imaging

For clusters with Astroscan data from 103a-E plates, tex2html_wrap_inline1877 is a real colour, viz. the average colour of all galaxies in the cluster for which we have R25 as well as a bj available (i.e. not just those used in the calibration). Differences in average colour between clusters can thus be due to significantly different total galaxy populations in different clusters. In addition, the zero-point of the calibration for a given cluster is not known with infinite precision; however, zero-point errors are measurement- and limited statistics- errors only, and are not dependent on the colours of the calibrating galaxies.

There are 17 clusters for which we used 103-E plates for the photographic photometry; 6 of these were calibrated individually with CCD-imaging, while for the other 11 clusters we applied the average relation derived for those 6. For the 6 clusters we find average cluster colours and dispersions of 1.69 and 0.24, with the offsets applied. If we do not apply the individual offsets, we find 1.78 and 0.27. Clearly, the statistics is not overwhelming, and the assumption of a universal (bj-R25)-distribution may not be a very good one for such a limited number of clusters. Yet, there is some evidence that the application of the individual zero-points for the 6 clusters makes sense as the dispersion around the average value of tex2html_wrap_inline1877 increases from 0.24 to 0.27 (and from 0.22 to 0.32 for the 4 clusters with at least 5 galaxies with CCD-imaging), if one does not apply the 6 individual zero-points. However, we note that the dispersion of the mean colours of the 11 clusters for which we applied the average calibration, is only 0.21. This must mean that differences in the average real colours of the calibrator galaxies in the individually calibrated clusters do indeed play a rôle. On the other hand, the latter also indicates that the average calibration is quite good.

On the other hand, for clusters with Astroscan data from IIIa-J plates, we are not dealing with real average galaxy colours because the measured bj and the pseudo R25 magnitudes are based on the same images on the same IIIa-J plates. Therefore the average colour tex2html_wrap_inline1877 in this case does not reflect the average colour of the total galaxy population, but only the real average colour of the calibrating galaxies. Especially when the calibration is based on a fairly small number of galaxies, differences in the real average colour of the calibrating galaxies may be as important as errors in the determination of the zero-point.

In the COSMOS-ENACS comparison 58 clusters have photographic photometry from IIIa-J plates; for 23 of those an individual CCD-calibration was available. From the average colours of the latter, it is immediately clear that there is a serious problem with the calibration for A3559 (and therefore also A3558 and A3562, which have identical calibration). The apparent value of tex2html_wrap_inline1877 for A3559 is 0.1 rather than about 1.5, as found for the other clusters. This means that the large zero-point correction of 1.7 that we found must indeed have been incorrect (as we already suspected in Paper I, but could not "prove" without the COSMOS magnitudes). Therefore, in the ENACS catalogue we have, for A3558, A3559 and A3562, not applied the zero-point derived in Paper I, but the average zero-point.

The remaining 22 clusters with individual calibration show a clear relation between the apparent value of tex2html_wrap_inline1877 and the number of galaxies with CCD-imaging, on which the calibration is based. The observed average values are 1.62 for clusters with N > 6 and 1.24 for clusters with tex2html_wrap_inline1903 6. When the number of calibrating galaxies is low one is more likely to have a difference in average colour between the calibrating galaxies and the (much more) numerous galaxies used in the COSMOS-ENACS comparison. That this should produce a colour bias is not immediately evident, but not difficult to explain either. When the distribution of galaxy colours is skewed (see Fig. 2 (click here)), or if a magnitude limit in one of the colours induces a colour selection, a bias could easily result. For the 11 clusters with tex2html_wrap_inline1903 6, the average value of tex2html_wrap_inline1877 differs so systematically and considerably from the average value for the other clusters that we have decided, in those cases, not to apply the individual zero-points derived in Paper I. The clusters in question are: A2480, A2717, A2734, A2915, A3009, A3094, A3108, A3141, A3809, A3822 and A3864.

For the 11 individually calibrated clusters with N > 6, the dispersion in average colour is 0.17, which must be compared with the corresponding value of 0.25 for the 36 clusters without individual calibration. So, indeed there is some evidence that the individual zero-points are worth applying, even though they differ only by a few tenths from the average zero-point. However, if one applies the average zero-point for the 11 calibrated clusters, the dispersion in the average colour does not increase noticeably. This is consistent with the fact that the dispersion is dominated by the 36 clusters without individual calibration.

In summary, we conclude from the comparison of the COSMOS and ENACS magnitudes that:
- the average calibration applied to the majority of the ENACS clusters is well supported by the magnitudes in the COSMOS catalogue
- the zero-points obtained for individual clusters are somewhat, but not very much, better than the average zero-point derived from all clusters with photometric calibration
- we found good reasons for not applying the individual zero-points derived in Paper I of 14 clusters: A3558, A3559, A3562 and the 11 clusters listed above.


next previous
Up: The ESO Nearby

Copyright by the European Southern Observatory (ESO)