1 Introduction

There are several reasons to increase the still rather meagre data on very high-z, powerful radio galaxies (e.g. McCarthy 1993; Pariskij et al. 1997). High-z radio galaxies are unique laboratories for investigating the early stages of galaxy and AGN evolution at look-back times corresponding to more than 90% of the age of the universe derived from Friedmann models. They may be used as tracers of the first generation of galaxy clusters (Peacock & Nicholson 1991; Peacock 1997) and of the physical state of the intergalactic space (Parijskij et al. 1996a). By high-resolution one may study important morphological features related to e.g. merging activity and "star burst'' regions.

At redshift z>2, 120 radio galaxies are known at present (de Brueck et al. 1997), in comparison with about 250 radio loud quasars ($S_{5\,{\rm GHz}}\gt$0.03 Jy in Veron-Cetty & Veron 1996), though the former are intrinsically more abundant. According to the popular unified scheme, both classes are the same thing. One can study the host galaxies and close environments of radio galaxies, but this is difficult for QSOs at a similar redshift.

One aspect, where even a single galaxy may be decisive, is the question of how close in time to the cosmological singularity it is possible to find galaxies, with normal stellar population and supermassive compact objects in their nuclei. Though the use of high-z objects in classical cosmological tests is hampered by severe problems, development of such tests is still one aim of observational cosmology. To identify selection effects and evolution, large samples are required. One must increase identifications of very remote galaxies, also in view of the new generation ground and space telescopes, which will allow their study at high resolution.

1.1 Extension of identified USS sources to fainter fluxes

It has been known since the late 70's (Tielens et al. 1979; Blumenthal & Miley 1979) that radio sources with steep spectra are optically fainter (and hence probably more distant) than sources with flatter spectra. Later it was established that observing faint radio sources with ultra-steep spectra (USS) is an efficient way to detect radio galaxies at high redshifts (see e.g. McCarthy 1993). As the USS Fanaroff-Riley type II (FRII-type; Fanaroff & Riley 1974) radio galaxies are not good "standard radio candles'' and as the reason for the success of the spectrum criterium is not known (see e.g. Röttgering et al. 1994), it is not clear what the outcome will be when USS samples are extended to a progressively fainter flux limit. Fainter flux may imply 1) larger redshifts, 2) similar redshifts, though weaker radio luminosity, or 3) smaller redshifts and still weaker luminosities. The first alternative is most interesting, though cases 2 and 3 are also important: extension of the luminosity range will help one to uncover the influence of radio luminosity on the classical cosmological tests (angular size-redshift; Nilsson et al. 1993 and Hubble diagram; Eales et al. 1997) and to decide whether alignment effect depends primarily on redshift or luminosity.

The flux range where differential normalized source counts show steepening is generally regarded as the most promising hunting place for high redshift objects. Parijskij et al. (1991) pointed out that the bulk of the RATAN-600 sample (see below) has fluxes in the range of 10-50 mJy at 3.9 GHz where the normalized counts show a maximum steepening, usually interpreted as a cosmological effect. A similar steepening in the counts is seen separately for steep spectrum sources (Fig. 6 in Kellermann & Wall 1987). It has been suggested (e.g. Röttgering et al. 1994) that the most effective way to find distant galaxies would be a USS sample with $S_{408}\sim 0.2-1$ Jy. Indeed, this has proven to be so since about 50% of the Röttgering et al. (1994) USS objects have z>2 (van Ojik et al. 1997). The bright end of the USS sources is well studied (e.g. 4C/USS, B2/1 Jy, MRC/1 Jy McCarthy (1993) and references therein) and recently fainter flux limits have been reached (e.g. B3/VLA S₄₀₈>0.8 Jy Thompson et al. 1994; ESO/Key-Project S₃₆₅>0.3 Jy Röttgering et al. 1994). However, in the Röttgering et al. (1994) sample 365 MHz flux density distribution peaks at about 1 Jy.

1.2 RATAN-600 (RC) and UTRAO catalogues

This paper is part of a programme initiated at the Special Astrophysical Observatory (Russia) with the aim of searching distant radio galaxies and investigating the early evolutionary stages of the universe (Goss et al. 1992). We wish to extend the steep-spectrum criteria to fainter fluxes than previously. This is accomplished by RC and UTRAO catalogues (see Fig. 1).

  

Figure 1: Frequency - flux limit diagram with the positions of the RC sample and some other major radio catalogues. UTRAO ($-36\hbox{$^\circ$}< \delta <$ 72$\hbox{$^\circ$}$) is the optimum low frequency catalogue presently available, which can be used for calculating the spectral index for a large part of the RC sample ($\delta \sim 5\hbox{$^\circ$}$). Note that the 6C sample has $\delta \gt$ 20$\hbox{$^\circ$}$.The lines correspond to a source with $\alpha=1$

  

Figure 2: Hubble diagram in R-band for various radio galaxies from the literature. The triangles are from the complete Molonglo sample (McCarthy et al. 1996), the open boxes are from Allington-Smith et al. (1988), Maxfield et al. (1995), McCarthy et al. (1987), McCarthy et al. (1990), McCarthy et al. (1991), Thompson et al. (1994), Windhorst et al. (1991). Filled dots are from Carilli et al. (1997), Chambers et al. (1988), Djorgovski et al. (1988), Dunlop & Peacock (1993), Eales et al. (1993), Hammer & LeFevre (1990), Kristian et al. (1978), Lacy et al. (1994), LeFevre et al. (1988), LeFevre & Hammer (1988), Lilly (1988), Lilly (1989), Owen & Keel (1995), Miley et al. (1992), Spinrad et al. (1995) and filled stars are from the ESO/Key-Project (Röttgering et al. 1995; Röttgering et al. 1996; van Ojik et al. 1994). Open symbols are r-magnitudes, which are transformed as R=r-0.4. The histogram of RC/USS sources R-magnitudes is shown above. The magnitudes are from K95b. Present NOT-observations are concerned with

Our high frequency catalogue is based on a sample of faint radio sources originally discovered using the RATAN-600 radio telescope in the "Kholod" ("Cold") experiment in 1980-81 (Parijskij et al. 1991; Parijskij et al. 1992; Parijskij & Korolkov 1986). In the experiment, performed at 7.6 cm (3.9 GHz), the strip around the sky at $\delta=5\hbox{$^\circ$}\pm20\hbox{$^\prime$}$was surveyed with a limiting flux of about 4 mJy. The RC catalogue resulted in containing 1145 objects. Within the inner strip of $\pm~5\hbox{$^\prime$}$the completeness of the catalogue reaches 80% at the flux limit S_3.9 =7.5 mJy and is almost 100% at 15 mJy (Parijskij et al. 1991). Such flux limits are really quite faint and allow one to identify a large number of steep spectrum sources, if a low frequency catalogue with sufficiently faint flux limit is available. The UTRAO (Douglas et al. 1996) is such a catalogue with a flux limit of $\sim$100 mJy at 365 MHz (see Fig.  1). The RATAN-600 catalogue (RC) provided the first sample which allowed one to calculate the spectral index for practically all UTRAO sources within the region covered by the "Kholod" experiment (Soboleva et al. 1994). Of the original sample of 840 sources (Parijskij et al. 1991), 491 sources matched those of the UTRAO catalogue. Soboleva et al. (1994) could identify optically from POSS (Palomar Optical Sky Survey) 240 sources at galactic latitude $\gt 20 \hbox{$^\circ$}$.

 

Table 1: RC/USS source parameters. The IAU name is in the first column followed by the equatorial, then galactic coordinates and galactic extinction in R-band. The radio spectral index is in the seventh column, followed by 3.9 GHz flux density and the LAS of the radio source. The results of optical identification are in the last column. The data have been taken from Kopylov et al. (1995a,b) and Parijskij et al. (1996a)

1.3 Construction and properties of the RC/USS sample

The present study is concerned with sources in the range 4$^{\rm h}<{\rm RA}<22^{\rm h}$ (Parijskij et al. 1991). The first RC/USS sample consists of 40 steep spectrum ($\alpha \gt$ 0.9, $f_{\nu}\propto\nu^{-\alpha}$), double or triple FRII sources, and optically fainter than the POSS limit. The radio morphology comes from observations with the VLA (Kopylov et al. 1995a). The largest angular size of the radio source (LAS) was not used as a criterion, because only eight sources had LAS larger than 30$\hbox{$^{\prime\prime}$}$.The median LAS of the sample is 7$\hbox{$^{\prime\prime}$}$. The median 365 MHz flux density is 0.5 Jy (average 0.7 Jy) ranging from 0.2 Jy to 3 Jy.

Optical identifications were made from deep observations at the 6 m telescope, down to about m_R=24. These results and the optical fields around the sources have been reported by Kopylov et al. (1995b, here after K95b). Table 1 contains information on the basic RC/USS sample: source name, equatorial and galactic coordinates, spectral index, flux, LAS and m_R. From this list we selected objects which are not unreasonably faint (m_R < 24 mag) for a medium sized telescope.

Figure 2 gives a representative m_R-z Hubble diagram for radio galaxies collected from the literature, together with the magnitude distribution of the RC/USS objects. The Hubble diagram allows one to estimate a lower limit to redshift, because of the rather sharp lower envelope, especially above m_R=21. Where the bulk of the RC/USS galaxies are situated, redshift is expected to be 0.7 as shown in Fig. 2. Soboleva et al. (1994) estimated the maximum photometric redshifts for the RC/USS objects from the requirement that radio luminosity is not higher than optical luminosity: when radio flux is known, the minimum optical magnitude may be calculated, hence the rough maximum $z_{\rm ph}$, which is usually large, >1.

It should be mentioned that one optically bright (m_R=19) object RC 2036+0451 was measured at the 6 m telescope to have z=2.95 (Pariskij et al. 1996b). Though for a quasar, this large z also supports the view that present selection criteria lead to high average redshift.

The aim of the NOT imaging was to study the morphology of the RC/USS sources with high resolution and confirm the optical identifications. This paper is organised as follows. In Sect. 2 we describe our observations and data reduction. Morphology of individual galaxies is discussed in Sect. 3. The results are summarised in Sect. 4.