In Fig. 4a the histogram of the galaxy velocities is shown. There are at least three peaks (labelled as A, B, C in the figure): the first is at a velocity of km/s, the second and the third at km/s and km/s, respectively. Peak A is at the same velocity as A 548 and presents a clear bimodality.
Although no significant differences (through a K-S test) are found between the overall distributions of galaxies with and without emission lines, a more detailed analysis of the three peaks reveals that the two distributions inside the single peaks are in fact different (see Fig.4b). In particular, for peak A a separation in velocity is evident, as the population of emission line objects is dominant in the clump at lower velocity: the percentage of emission line galaxies with respect to the total in this clump is , while it is in the higher velocity clump. Because galaxies with and without emission lines have different luminosity functions (Zucca et al. 1997), it could be suspected that their different distribution is a consequence of a change in the relative values of their selection functions: however, the width of peak A is relatively narrow (less than 1000 km/s) and therefore this effect is more likely due to a real variation in morphological composition in the two subclumps.
We estimated the dynamical parameters (mean velocity and velocity dispersion) of the three peaks with the biweight estimators of location and scale (Beers et al. 1990). The advantage of these estimators, with respect to the standard mean and dispersion, is that of minimizing biases from interlopers, giving less weight to data with higher distance from the median. The confidence intervals of the two estimators are calculated boostrapping the data with 100 random catalogs. In order to find the velocity range in which the cluster members lie, we have assumed that the velocity distribution of cluster galaxies is Gaussian, as expected when the system has undergone a violent relaxation (see details in Bardelli et al. 1994).
For the case of peak A, in which the presence of a substructure was suspected on the basis of both a visual inspection of the velocity histogram and the shape estimators a, , and I (see Bird & Beers 1993), we checked if the distribution is consistent with a single Gaussian or if it is bimodal applying the KMM test (Ashman et al. 1994), using the program kindly provided by the authors. This test gives the likelihood ratio between the hypothesis that the dataset is better described by the sum of two Gaussians and the null hypothesis that the dataset is better described by a single Gaussian.
In Table 8, we report the dynamical parameters for the velocity excesses found in our sample (see the discussion below). Column (1) refers to the peak identification, column (2) reports the number of velocities used, columns (3) and (4) are the estimated mean velocity and velocity dispersion.
3.1. Peak A
Peak A, formed by 64 galaxies, is at the same velocity as A 548 and has km/s and km/s. The shape estimators a and revealed a significant deviation from the null hypothesis of Gaussianity (at more than significance level). The KMM test revealed that the distribution is significantly better described by two Gaussians, both in the homoscedastic and in the heteroscedastic case. Assigning the objects to the two groups on the basis of the a posteriori probability given by the KMM algorithm (see Ashman et al. 1994) and estimating the dynamical parameters with the biweight estimators, we found km/s and km/s (based on 28 velocities) and km/s and km/s (based on 36 objects). Fig. 5 shows a close up of the velocity distribution of galaxies in peak A, with superimposed the two Gaussians of parameters , and , . No significant differences are found, inside each clump of peak A, between the dynamical parameters of galaxies with and without emission lines.
These mean velocities can be compared with those of the subclumps found by Davis et al. (1995) in A 548. Our value of is well consistent with their Clump a (see their Table 4b). Our is 2.6 different from the mean velocity of their Clump b: however assuming that the error on the Davis et al. determination (not reported in their paper) could be similar to ours, the discrepancy may be reduced to less than 1 . Moreover, our values are well in agreement with those reported by Escalera et al. (1994).
More noticeable is the difference in the velocity dispersions. Clumps a and b have values of 638 km/s and 553 km/s respectively, while we estimated km/s for both groups. This fact could be an indication of a dependence of the velocity dispersion on the distance from the group centers.
The dynamical situation of peak A is therefore very similar to that of A 548 and this peak has to be considered an extension of the nearby cluster rather than a separated entity. In particular, the subcondensation of peak A at lower velocity is part of Clump a, while the higher velocity group of peak A is associable to Clump b.
In order to see the relative distance of the clumps detected in A 548 from our galaxies, we plotted on the isodensity contours the positions of their centers and the distribution on the sky of our sample. In Fig. 6 stars represent galaxies with km/s and triangles refer to objects with km/s. The big crosses are the reported centers of the extended emissions found in the ROSAT observations labelled as S1 and S2 in Tableda 2 of Davis et al. and coincident with their optical Clumps a and b. Their source S3 falls outside the figure. We reported also the position of their Clump c as an asterisk. Note the good coincidence between these positions and peaks in the density field.
The extension in A 3367 of Clump a of Davis et al. seems to have two condensations (see stars in Fig. 6) at , and , , at the distances of and arcmin from the corresponding substructure in A 548, respectively.
The extension in A 3367 of Clump b (see triangles in Fig. 6) is more smooth and spans all the distances between and from its center, corresponding to 2-3 h-1 Mpc.
Finally we note that the mean velocity reported by Postman et al. (1992) for A 3367 is consistent with the redshift of peak A. The velocity of the brightest cluster member (Postman & Lauer 1995) reveals that this galaxy is part of the higher velocity clump of peak A: therefore this galaxy is probably associated to A 548 rather than to A 3367.
3.2. Peaks B and C
The estimated dynamical parameters of the second peak seen in the redshift histogram are km/s and km/s, based on 40 velocities. The percentage of emission line galaxies is . In Fig. 6, the objects belonging to this structure are plotted as black dots and seem to be concentrated around a density peak at and . The shape estimators indicate that the distribution is consistent with a Gaussian. Given the fact that this density excess is the largest one inside one Abell radius from the nominal center of A 3367 and considering that its velocity dispersion is typical of a cluster, we suggest that the name A 3367 should be attributed to peak B.
Finally, the estimated mean velocity and velocity dispersion for peak C are km/s and km/s, determined with 17 objects. The Gaussianity of the distribution cannot be excluded.
© European Southern Observatory (ESO) 1998
Online publication: August 6, 1998