To study the status and the evolution of clusters of galaxies at intermediate redshifts we built a sample of candidate clusters using radiogalaxies in the NRAO VLA Sky Survey as tracers of dense environments.

From the NVSS maps we extracted a catalogue of radio sources over an area of $\approx$ 550 square degrees, and made optical identifications with galaxies brighter than $b_{\rm J}~=~20.0$ in the EDSGC Catalogue, resulting in a sample of 1288 radiogalaxies (Zanichelli et al. 2001, Paper I).

In this paper we have presented the detection technique we applied to select candidate groups and clusters associated to NVSS radio sources. The method is based on the search of excesses in optical surface galaxy density nearby NVSS radiogalaxies. To keep low the probability of spurious radio-optical identifications, as well as to preferentially select clusters at redshifts $z \mathrel{\rlap{\lower 3pt\hbox{$\sim$ }} \raise 2.0pt\hbox{$>$ }}0.1$ , we restricted the cluster search to the 661 radiogalaxies having radio-optical distance $\le$ $7\hbox{$^{\prime\prime}$ }$ and magnitude $b_{\rm J}\ge 17.5$ .

The search of regions having high optical galaxy density has been made using the EDSGC galaxy catalogue, building matrices of galaxy counts down to magnitude $b_{\rm J} =20.5$ . This choice allows to find density excesses surrounding the faintest radiogalaxies (identified down to $b_{\rm J} = 20$ ) without introducing significant incompleteness effects in the optical data. Smoothing of galaxy counts has been done using a Gaussian filter with ${\it FWHM} = 2\hbox{$^\prime$ }$ . The mode and standard deviation of smoothed galaxy counts have been used to define a detection threshold for the surface density excesses: we selected as cluster candidates those density excesses whose centroid is within $4\hbox{$^\prime$ }$ from a radiogalaxy. This search radius for candidate clusters associated to NVSS radio sources corresponds to an Abell radius of a cluster at $z \sim 0.45$ .

By applying this cluster detection strategy to 661 radiogalaxies over $\approx$ 550 sq degrees at the South Galactic Pole, we obtained a sample of 171 cluster candidates. The estimated contamination level is about $28\%$ .

$\begin{figure} \par\includegraphics[width=8.8cm,clip]{ms10560f5.ps}\end{figure}$

Figure 5: Velocity dispersion versus redshift for the 9 spectroscopically confirmed clusters. Error bars are $1 \sigma$ . The lack of correlation between redshift and velocity dispersion suggests that structures with different richness are well represented at any distance in the redshift range covered by our candidate cluster sample.

Out of these 171 candidates, 76 correspond to already-known clusters, while 95 cluster candidates in our list do not have any known counterpart in the literature and have been the subject of subsequent spectroscopic follow-up. The full sample of radio-optically selected cluster candidates will be presented in a following paper. Multi Object Spectroscopy aimed to confirm the detection of clusters has been successfully acquired at the 3.6 m ESO telescope for a subset of 12 candidates. In 2 cases the radiogalaxy does not lie at the same redshift as any other observed target, while 9 candidates have been confirmed as clusters of galaxies in the redshift range $0.13 \le z \le 0.3$ , thus confirming that this joint radio-optical cluster selection technique can be used as a powerful tool for the detection of cluster candidates at intermediate redshifts. For one additional candidate, the very low number of measured redshifts does not allow any conclusion on the presence of a cluster surrounding the radiogalaxy, and further observations are needed. Velocity dispersions of the 9 spectroscopically confirmed clusters vary from values typical of moderately rich clusters to those typical of groups or poor clusters, thus strengthening the assumption that this technique is equally efficient in selecting structures over a wide range of richness at different redshifts. If confirmed by future spectroscopic follow up, this last result could be of great interest as this technique would offer the possibility to study the properties of different environments, such as groups or rich clusters, in a homogeneously selected cluster sample.

7 Conclusions