2. Comparison of the properties of emission and non-emission line galaxies
In a recent paper based on ENACS data, Biviano et al. (1997, the third paper of the series) gave a good review of the distribution and kinematics of emission line (ELGs) versus non-emission line (NoELGs) galaxies. Their approach is a statistical one, leading to a general picture which fits well in a current scenario of cluster formation. On the other hand, the present study is devoted to a specific cluster, ABCG 85, with already well known general properties; the spirit of this work is therefore closer to that of Mohr et al. (1996).
We will first compare the distributions of ELGs versus No ELGs as a function of various parameters. As pointed out by Mohr et al. (1996), this roughly corresponds to a separation into gas-rich and gas-poor galaxies, with some contamination of the gas-poor sample expected. It can also be considered as a separation into spiral and non-spiral galaxies, with an underestimation of spirals, since not all spirals are ELGs.
The numbers of ELGs and NoELGs in our sample are 102 and 449 respectively. These numbers are reduced to 33 and 272 respectively in the cluster velocity range, defined as the 13000-20000 km s-1 interval by Durret et al. (1998a). This corresponds to an ELG fraction of 0.11 in the cluster. Such a fraction is in the range derived by Biviano et al. (1997) for the ENACS sample (0.08-0.12), but is notably smaller than the value of 0.34 estimated by Mohr et al. (1996) for ABCG 576.
A classification of the ELGs belonging or not to ABCG 85 based on the equivalent widths of the main emission lines will be performed in a forthcoming paper. The possible existence and influence of environmental effects on the presence and level of activity in galaxies will be discussed in that paper.
2.1. Magnitude distribution
The histograms of the distributions of ELGs, NoELGs and all galaxies as a function of R magnitude are displayed in Fig. 1 for all the galaxies in our redshift catalogue. A similar histogram is drawn in Fig. 2 for galaxies belonging to the 13000-20000 km s-1 velocity range. NoELGs are seen to be brighter than ELGs in both samples, as in ABCG 576 (Mohr et al. 1996).
The fraction of emission line galaxies (defined in each magnitude bin as the number of emission line galaxies divided by the total number of galaxies) as a function of magnitude in the R band is displayed in Fig. 3 for cluster members. We note an increase of this fraction with magnitude, as expected since for a given exposure time redshifts are easier to obtain for emission line galaxies, which can therefore be measured for fainter objects than for non-emission line galaxies. Note however that this increase becomes steep for R17, i.e. even at magnitudes for which there is usually no problem to measure absorption line redshifts, possibly suggesting that ELGs are intrinsically fainter, though it is difficult to ascertain this result since our redshift catalogue is by no means complete in the entire region (its completeness is 85% in a circular 2000 arcsec radius region for R18, then drops at larger radii, see Table 2 in Durret et al. 1998a). In order to quantify this effect, we calculated the average luminosities for ELGs and NoELGs in the range 14.5R17.9. We find average luminosities corresponding to magnitudes of 16.83 and 16.41 for ELGs and NoELGs respectively; the standard errors on the luminosities give corresponding magnitude ranges of [16.65, 17.04] and [16.35, 16.47], which do not overlap. The Kolmogorov-Smirnov and Student tests indeed give probabilities that these two samples originate from the same parent population of 0.0014 and 0.03 respectively, confirming that ELGs are indeed intrinsically fainter that NoELGs. This result is in agreement with that found on a much larger sample by Zucca et al. (1997) of field galaxies.
2.2. Spatial distribution
The spatial distributions of ELGs and NoELGs are different for galaxies in the velocity interval 13000-20000 km s-1. It is generally believed that ELGs are more frequent in the outer regions of the cluster than in the inner zones (Biviano et al. 1997, Fisher et al. 1998). This result is indeed confirmed here in ABCG 85. Fig. 4 shows the fraction of emission line galaxies (estimated as the ratio of the number of emission line galaxies to the total number of galaxies in concentric rings 500 arcsec wide around the cluster center) as a function of projected distance to the cluster center for the galaxies in the 13000-20000 km s-1 velocity range. This fraction increases towards the outer regions of the cluster, implying a difference in the spatial distributions of ELGs and NoELGs.
In order to visualize better the spatial distributions of both kinds of galaxies, we have drawn adaptive kernel maps (e.g. Pisani 1993) of the spatial density distributions of ELGs and NoELGs; these are shown in Figs. 5 and 6 for all galaxies in the spectroscopic sample, disregarding their velocities. The distribution of NoELGs (Fig. 5) is comparable to that derived from our much larger photometric catalogue for all galaxies, independently of spectral features and cluster membership (see Fig. 1 in Slezak et al. 1998): it is elongated along and shows a strong concentration around ABCG 85, a secondary peak towards the south east coinciding with ABCG 87 (see for example Table 2 in Durret et al. 1998b) and an enhancement roughly at the position of ABCG 89 to the north west (as discussed by Durret et al. 1998b). The galaxy distribution of ELGs (Fig. 6) is quite different: its peak does not coincide with ABCG 85, but is close to the position of ABCG 87; it shows a weak secondary maximum in the north-northeast direction and elongations along several PAs, all quite different from .
2.3. Velocity distribution
The mean and median velocities, as well as the velocity dispersions are quite different for ELGs and NoELGs; the mean velocities are 15968 and 16627 km s-1, the median velocities are 15701 and 16732 km s-1, and the velocity dispersions are 1606 km s-1 and 1109 km s-1 for ELGs and NoELGs respectively, suggesting that the morphology-density relation is coupled with kinematic differences. Biviano et al. (1997) found differences in the average velocities of ELGs and NoELGs at a level larger than 2 only for 12 clusters out of their sample of 57; their interpretation was that in these 12 clusters ELGs are a non-virialized population falling onto the main cluster. In a much smaller sample of 6 clusters, Zabludoff & Franx (1993) also found a difference in mean velocity between spirals and early type galaxies in 3 clusters; on the other hand, Mohr et al. (1996) found that in ABCG 576 ELGs and NoELGs had the same average velocity, but with ELGs having a larger velocity dispersion, as in ABCG 85.
The velocity distributions displayed in Figs. 7 and 8 were obtained simultaneously using profile reconstructions based on a wavelet technique and classical histograms. We remind the reader that the features obtained with the wavelet method are significant at a 3 level above the noise (estimated at the smallest scale, see Fadda et al. 1998). While the velocity distribution of NoELGs shows only one peak around 16800 km s-1 close to the mean or median previously given, that of ELGs shows a peak at about 15300 km s-1 and a much smaller one around 19250 km s-1. Therefore, calculations of a mean value and of a velocity dispersion for the total sample of ELGs in ABCG 85 do not characterize the ELG velocity distribution properly. The mean (median) velocity for galaxies belonging to the main ELG component (corresponding to v18000 km s-1) is 15394 km s-1 (15411 km s-1) and the velocity dispersion is 900190 km s-1, much smaller than the value of 1606 km s-1 previously calculated. The second small peak includes five galaxies, and has a mean (median) value of 19170 km s-1 (19320 km s-1) and a velocity dispersion of 320 km s-1. Three of these five galaxies are very close both spatially and in velocity space, suggesting that they are part of a physical group. Notice that the velocity distribution of NoELGs is not gaussian. A more detailed velocity analysis of the total sample of galaxies can be found in Durret et al. (1998b).
Furthermore, it is interesting to note that the minimum of ELG velocity density is obtained for km s-1 close to the maximum for NoELG velocity density ( km s-1); the two samples are therefore different in velocity space, as confirmed by statistical tests (see Table 1 and Sect. 2.4).
Table 1. Statistical tests on the properties of ELGs and NoELGs
The fraction of ELGs seems to increase with velocity. It is about 0.12 within the cluster velocity range, and increases to 0.25-0.33 for background objects (those with velocities larger than 20000 km s-1), with an extreme value of 0.45 in the last velocity bin (velocities larger than 60000 km s-1). Such variations are obviously at least partly due to a selection effect (the redshifts of faint background galaxies are easier to measure for emission line than for absorption line spectra), but are difficult to interpret because of the incompleteness of our velocity catalogue at large radial distances.
2.4. Are the ELG and NoELG properties significantly different?
As shown above, the spatial, magnitude and velocity distributions of ELGs and NoELGs are different. In order to quantify statistically these results, we have tested the null hypothesis which would assume both samples to be issued from the same parent population, both in velocity space (v) and spatially (right ascension , declination and projected distance to the cluster center D).
The statistical tests used are the unpaired comparison t-test and the Kolmogorov-Smirnov (K.-S.) test. The former is based on the comparison of the means of both samples (ELGs and NoELGs). Since we have seen that the velocity distribution of ELGs is bimodal, we have applied this test in two velocity ranges: (A) [13000-20000 km s-1], corresponding to the entire cluster velocity range, and (B) [13000-18000 km s-1], where the second small peak in velocity distribution of the ELGs is eliminated, while the NoELG distribution is not strongly affected.
The results are shown in Table 1, indicating that the null hypothesis is rejected with a high degree of confidence. This result is confirmed by the non-parametric K.-S. test, which was applied to the four characteristic properties of both samples. Except for the variable, all the other quantities give weak probabilities for the null hypothesis.
2.5. Physical interpretation
We have previously shown (Durret et al. 1998b) that ABCG 87 is made of several small groups falling onto the main body of ABCG 85. This picture also allows us to give a general explanation of the ELG properties described above. The density-morphology relation (e.g. Adami et al. 1998a and references therein) shows that the less dense a region, the larger the rate of late type galaxies. Spirals, and consequently ELGs (which are mainly spirals) would therefore tend to avoid the central region of ABCG 85. Since groups of galaxies are less dense, spirals tend to be more numerous in the ABCG 87 region (both in space and in velocity space, see Sects. 2.2 and 2.3). Moreover, the arrival of these groups onto ABCG 85 probably creates shocks, which leads the temperature of the X-ray emitting gas to increase, as indeed observed in the ASCA temperature map obtained by Markevitch et al. (1998). The shocks induced by the merging of groups into the main cluster may well trigger star formation in gas-rich spiral galaxies and account for the increase in the number of ELGs in the ABCG 87 region (assuming that the general identification of spirals with ELGs is valid). Such a picture is consistent with numerical simulations (e.g. Bekki 1999), in which merging phenomena in clusters trigger star formation, and therefore enhance the numbers of ELGs in merging regions.
The velocity dispersion in the main ELG density peak is high (900 km s-1). In the general picture described above, the large velocity dispersion found for ELGs can be explained as resulting from the convolution of the velocity dispersion in each blob (typically 300 km s-1) with the velocity of each blob. However, due to the small number of ELGs, it is difficult to show this directly from the ELG data.
In their interesting statistical analysis of the properties of ELGs in nearby clusters based on the ENACS data, Biviano et al. (1997) emphasize some results, in particular the fact that ELGs appear to avoid the central regions of clusters. They propose a schematic model with two types of components, one with a velocity offset relative to the average cluster velocity and a fairly small velocity dispersion, and the other with no velocity offset and a large velocity dispersion.
These properties, combined with others, suggest that ELGs are falling into the central region without having been previously in it. Such a result is also found by Carlberg et al. (1996) for a sample of about 15 clusters with redshifts between 0.17 and 0.55. Notice that the larger velocity dispersion of ELGs compared to NoELGs can at least partly be due to the difficulty of separating various velocity subsamples; it is only in the case of large amounts of data and when the distribution is clearly asymmetric that it is possible to improve the analysis, as in our case. When a particular cluster is studied in detail, one of the two types of components prevails. Infall is generally not spherically symmetric, because it occurs preferably along filaments (van Haarlem & van de Weygaert 1993, West 1994); this is the case in ABCG 85.
© European Southern Observatory (ESO) 1999
Online publication: March 1, 1999