2 The radio sample and the optical identification procedure

The parent radio sample used in this study is based on the Parkes-MIT-NRAO radio survey (PMN, Griffith et al. 1995 and references therin), selected at 5 GHz. The data cover the southern sky with $\delta < 10^\circ$ to a flux density limit of 20-72 mJy depending on declination. No additional flux density limit was applied when carrying out the survey, but note that when considering the statistics of the final quasar sample, we consider sources with $S\ge 72$ mJy, since the PMN completeness at lower flux density levels is patchy (see Fig. 1).

To provide accurate positions and spectral index information, the PMN catalogue was matched to the 1.4 GHz NRAO-VLA Sky Survey (NVSS, Condon et al. 1998), which covers the declination range $\delta \ge -40^\circ$. All NVSS sources matching within a 2 arcmin radius of a PMN position were kept.

Since the beam sizes of the PMN and NVSS surveys are different (4.2'for PMN and 45'' for NVSS), the 1.4 flux density used to determine spectral indices were calculated by summing the total flux density from all NVSS sources whose positions were within a 2 arcmin radius of the PMN position. Flat spectrum objects with $S_{\rm PMN} \ge 50$ mJy were then selected using the criterion $\alpha^{5}_{1.4} \ge -0.5$ where $S\propto\nu^{\alpha}$. This gave 5976 PMN radio sources.

The NVSS positions were matched to optical catalogs which were derived from scans of UK Schmidt Telescope (UKST) plates. The scans were produced using the Automatic Plate Measuring (APM) facility (McMahon & Irwin 1992). The optical identification procedure was similar to that used in the past to make POSS-based identifications of radio samples, described in Hook et al. (1995, 1996, 1998). However the current survey also makes use of I-band plates for the first time.

The plate data used is from the UKST survey of the southern sky in the B, R and I bands, reaching limiting magnitudes of approximately 22.6, 21.0 and 19.5 mag respectively. The blue plates are in the $B_{\rm J}$passband, $\rm 3950{-}5400$ Å, and have 606 centres spaced at intervals of 5 degrees. In most cases the matching R plate used was from the OR survey with a passband of $\rm 5950{-}6900~\AA$, although in some fields not yet covered by the OR survey, the old R plate was used ( $\rm 6300{-}6900~\AA$). The I plates used IV-N emulsion + RG715 filter giving a passband of $\rm 7150{-}8900~\AA$.

The maximum redshift of quasars that our survey can detect in principle is defined by the redshift at which the $\rm Ly-\alpha$ line is redshifted out of the redder passbands. For the R-plates this occurs at redshifts above z=4.7 and for the I plates at redshifts above z=6.3.

2.1 Area covered by the survey

Figure 1: Area covered by the radio data at the time of the survey. light dots show PMN sources with $S\ge 72\rm ~mJy$ which had a counterpart in the NVSS catalogue, and dark dots show those PMN sources which did not have an NVSS counterpart. Usually this was because the NVSS survey had not been completed in those areas at the time (March 1998).

Figure 2: Area covered by the optical data. Crosses show UKST fields whose centres satisfy $\vert b\vert>30\deg$, $-40\deg<\delta<0\deg$, open circles show fields that have B and R plates, and filled circles show fields that have B, R and I plates. A total of 313 fields have at least B and R plates available, corresponding to an area of 7525 sq deg. The shaded area represents the declination range of the survey, defined by the overlap of the NVSS, PMN and UKST data.

Figure 1 shows the spatial distribution of the radio sources which form the basis of the survey. These lie in the region of overlap of the PMN and NVSS surveys. The effective area of our quasar survey is defined by the subset of this region that is covered by APM scans of UKST plates.

All the available B, R and I plates have now been scanned in the region of overlap with our radio sample ( $-40^{\circ} < \delta < +2.5^{\circ}$, avoiding the galactic plane, $\vert b\vert > 30^{\circ}$). At the time of the spectroscopic observations reported here, an area of 7525 sq deg was covered by plates in the B and R bands (4637.5 sq deg in the South Galactic Cap region and 2887.5 sq deg in the North Galactic Cap region). A total of 3887.5 sq deg also had I-band plates. The spatial distribution of the plates is shown in Fig. 2. Of this, two regions around $\rm RA=13$ h, $\rm DEC=0^{\circ}$ have no radio data, as shown in Fig. 1. The total area of overlap with the B, R data is 7265.5 sq deg of which 4637.5 sq deg is in the South Galactic Cap (SGC) region and 2628.0 sq deg is in the North Galactic Cap (NGC). Of this a total of 3637.5 sq deg also had I data.

In addition there are small areas within the region of radio/optical overlap that were not yet covered by the NVSS survey at the time our QSO survey was carried out. This resulted in some PMN sources not being matched with an NVSS counterpart in certain regions, as can be seen in Fig. 2. When considering the statistics of our final quasar sample, we take this incompleteness into account in a statistical way by calculating the fraction of PMN sources with $S_{\rm PMN}>72$ mJy that do not have NVSS counterparts. In the SGC the matched fraction is 0.955 and in the NGC it is 0.871.

Figure 3: Distribution of differences between the optical (APM) and radio (NVSS) positions. The solid line is the histogram for all sources, the dashed line is for objects classified as stars and the dotted line is for objects classified as galaxies. A matching radius of 3 arcsec was used in this survey.

2.2 Selection of the spectroscopic sample

Optical identifications were made based on positional coincidence of the NVSS position with an optical counterpart on the UKST plates. A matching radius of 3.0'' between the NVSS and optical positions was used (see Fig. 3 for the distribution of positional differences).

From these identifications, red, stellar objects were selected for spectroscopic follow up. Figure 4 shows colour-magnitude and colour-colour diagrams for optical identifications within 3'' of the NVSS position. The first spectroscopic sample contained any stellar object with $B-R\ge 1.5$. A second sample, the "I-band sample'', was then defined, which contained stellar objects with $R-I \ge 1.0$, or $B-I \ge 2.0$ that were not already included in the first sample. The criterion for being considered a stellar object was $\sigma_{\rm class} \le 3.0$ where $\sigma_{\rm class}$ is the APM classification parameter measured from the R plate. If the object was not detected in R then the I plate classification was used. These selection criteria resulted in a sample of 228 sources. Of these, 33 had known redshifts from the literature prior to the start of this project.