2 Optical observations, data reduction and analysis

2.1 Observations and reduction

Multi-Object Slit (MOS) spectra for both clusters were acquired using the ESO 3.6-m telescope at La Silla in two runs, one in 1993 and one in 1999 (Table 2). The first run was equipped with EFOSC1 (ESO Faint Object Spectrograph and Camera), Tektronix $512\times512$ CCD (pixel size 27 $\mu$m and $0.61\hbox{$^{\prime\prime}$ }$/pixel) and Grism B300, covering the range 3740-6950 Å, with central wavelength 5250 Å and dispersion of 6.2 Å/pixel. The observations in 1999 were performed with EFOSC2, CCD #40 $2048\times2048$ (pixel size 15 $\mu$m, $0.157\hbox{$^{\prime\prime}$ }$/pixel) and Grism O300 which covers 3860-8070 Å  with central wavelength 5000 Å and dispersion 2.06 Å/pixel. However, the final wavelength range depends on the position of each slit on the mask. The seeing was 1.0-1.2 $^{\prime\prime}$for both 1993 and 1999 runs. The corresponding airmasses and the exposure times for each mask/slit configuration are given in Table 2.

Figure 1: Abell 1451: V-band image from the Danish 1.5-m telescope showing the objects included in the spectroscopic study. Numbers correspond to the object identification in Table 3; those denoted as diamonds are either stars or non-cluster members. The dashed box is the central part, shown zoomed in Fig. 2.

All spectra were processed within the ESO-MIDAS or IRAF environment to produce the final wavelength calibrated, sky subtracted spectra. A HeAr lamp was used for wavelength calibration, and a second-order cubic spline or polynomial fit to 5-12 comparison lines gave rms residuals <1 Å.

We have determined redshifts using the Tonry & Davis (1979) cross-correlation technique as implemented in the RVSAO package (Kurtz & Mink 1998). Before processing with RVSAO we have masked out spectral ranges where significant residuals from sky subtraction could occur, and also any possible emission lines. Generally we have cross-correlated our observed spectra with 3 to 6 galaxy and star templates with good signal-to-noise, and also with the composite absorption-line template (fabtemp97) distributed with RVSAO. For each target object we adopted the redshift value from the best (highest correlation coefficient) template. The redshifts were then checked by identifying the prominent absorption features (CaII H & K, G band, MgI triplet) in each spectrum. If emission lines were present in the spectra, redshifts were determined from them.

The errors in redshift were computed from $\Delta z = k/(1+r)$, where r is the cross-correlation coefficient and k was determined empirically by adding noise to a high signal-to-noise spectrum and correlating it with template spectra with known velocities. For our observational configurations we found k=0.003, which gives a velocity error of about 200 km s^-1 for a redshift estimate with r=3. We have checked our redshift measurements and uncertainties using a few galaxy spectra repeated in different runs or masks and they are all in excellent agreement, consistent with the derived errors.

We have not converted our measurements to the heliocentric system because the correction is about 1 km s^-1 for the 1993 run and -8.8 km s^-1 for the 1999 run, well below the uncertainties.

The final redshift catalogue of objects for Abell 1451 is shown in Table 3 and for RXJ1314-25 in Table 4, with corresponding finding charts in Figs. 1 and 3.

2.2 Optical data analysis

2.2.1 Abell 1451

Cluster membership was determined by using an interactive version of the Beers et al. (1990) ROSTAT package (Valtchanov 1999) and confirmed by the 3-$\sigma$ clipping method of Yahil & Vidal (1977). The cluster redshift distribution is shown in Fig. 4 and some relevant characteristics, taking redshift errors into account, are given in Table 5. The distribution is very close to Gaussian, which has been quantified by various statistical tests for normality (D'Agostino & Stephens 1986) - the Wilk-Shapiro test accepts normality at the 99% level and the Anderson-Darling test at the 95% level.

There is no single dominant cluster galaxy. The brightest cluster members (BCMs, #34, #39, #36 and #40) are linearly aligned in projection (see Fig. 2) and their spectra are typical elliptical galaxy absorption spectra with no emission lines. The projected galaxy density distribution from SuperCOSMOS (Hambly et al. 2001) data shows an elongation in the same direction. The velocity dispersion among the BCMs is $\sim$800 km s^-1, somewhat less than the overall cluster velocity dispersion.

Figure 2: Abell 1451: zoomed image of the central part of Fig. 1. The plus sign marks the X-ray emission centre (see Sect. 3) and the contours are ATCA 13 cm radio observation with levels = 0.3, 0.5, 1, 2, 5 and 10 mJy/beam, rms noise level is $60~\mu$Jy/beam (see Sect. 4). The object in brackets is a background galaxy.

   
2.2.2 RXJ1314-25

Statistics of the redshift distribution are shown in Table 6 and the redshift histogram in Fig. 5. The distribution clearly shows two peaks and the test for the hypothesis of a unimodal distribution - the dip test (Hartigan 1985) - gives an insignificant probability of a single mode. Going further, we have applied the KMM method for detecting bimodality (Ashman et al. 1994), which confirmed the dip test negative result for unimodality and also gave us the probable group membership, assuming a bi-modal distribution. In Table 6 we show the same statistics for both groups. Although the number of cluster members is relatively small for giving high weight to the KMM result, the segregation of the galaxies on the sky supports that conclusion (see Fig. 6).

The optical image of the cluster (Fig. 3) shows two dominant galaxies (#48 and #19, marked also in Table 4) as the brightest cluster galaxies (BCG). Both have typical giant elliptical galaxy absorption line spectra, without any apparent emission lines. The second-ranked BCG (#19, z=0.2503) is at rest with respect to the galaxy members of the western group, while the first-ranked galaxy (#48, z=0.2463) differs by $\sim$800 km s^-1 from the mean redshift of the eastern group, but lies closer to the overall cluster mean. The separation between the two groups in velocity space is $\sim$1700 km s^-1, substantially greater than their combined velocity dispersions. With optical data alone we can not determine whether the two groups are gravitationally connected.