$\begin{figure} \includegraphics [width=8cm,clip]{ds1753f2.ps}\end{figure}$

Figure 2: Histogram of low ( $\alpha_{\rm l}$ )- and high-frequency ( $\alpha_{\rm h}$ )spectral indices. The blank area corresponds to $\alpha_{\rm l}$ , the hashed one to $\alpha_{\rm h}$

$\begin{figure} \includegraphics [width=8cm,clip]{ds1753f3.ps}\end{figure}$

Figure 3: Colour-colour diagram of B3VLA sources. Galaxies are symbolized by crosses, quasars by dots

Table 2 presents the whole database. Column 1 lists the B3-VLA name, Cols. 2 and 3 the radio centroid (equinox J2000.0) from Vigotti et al. (1989; computed as the geometric mean of the source components). The following 12 columns list the flux densities and errors at 151 MHz, 327 MHz, 408 MHz, 1.4 GHz, 4.85 GHz and 10.6 GHz, respectively (all in mJy). The last column contains the sources' optical identifications, abbreviated as follows: g: radio galaxy identified on the POSS-I, most of which are $z\le 0.5$ ; G: far radio galaxy with measured redshift ( $0.5\le z\le 3.5$ ); Q: spectroscopically confirmed quasar; b: blue objects (i.e. non-confirmed quasars); BL: BL Lac; F: featureless spectrum; a blank means "empty field'', i.e. it lacks any optical counterpart down to the POSS-I limit (more that 90% are distant radio galaxies, the remaining ones being quasars with magnitudes fainter than the POSS-I).

For 19 sources the 408 MHz data are not reported. In 15 cases the flux density is affected by a nearby strong source. In four cases the B3-VLA sources were not resolved by the 408 MHz beam.

In order to complete the spectral database we observed 164 sources at 4.85 GHz whose flux densities were not available in the catalogues listed in Table 1; 314 sources with detected polarization at 10.6 GHz were re-observed at 4.85 GHz for future polarization studies. An analysis of the polarization data will be published in a forthcoming paper. In Fig. 1 we show the plot of our measurements versus the GB6 flux densities. Intrinsic source variability is likely to increase the scatter of the plot.

For each source we computed two spectral indices: a low-frequency index $\alpha_{\rm l}$ (0.3-1.4 GHz) and a high-frequency one, $\alpha_{\rm h}$ (4.8-10.6 GHz). Figure 2 shows the histograms of $\alpha_{\rm l}$ and $\alpha_{\rm h}$ (shaded) of 1034 sources, for which four flux densities are available. The resulting median values for the two distributions are $\langle\alpha_{\rm l}\rangle = -0.853$ and $\langle\alpha_{\rm h}\rangle = -1.053$ ( $S\propto\nu^{\alpha}$ ).

In Fig. 3 we show a radio colour-colour diagram illustrating the different population areas of radio galaxies and quasars. As already evident in Fig. 2, $\alpha_{\rm h}$ covers a wider range of values (dispersion 0.40) than $\alpha_{\rm l}$ (dispersion 0.23). This is to be expected if spectral steepening due to synchrotron and inverse Compton energy losses is important: it changes $\alpha_{\rm h}$ first, before the sources have aged sufficiently such as to affect $\alpha_{\rm l}$ as well. The corresponding evolutionary track in the $\alpha_{\rm l}-\alpha_{\rm h}$ diagramme is that populated by the radio galaxies in Fig. 3: if these sources commence their lives with flat (injection) spectra, they should gradually move downward at a faster rate than leftward. Also evident in Fig. 3 is that radio galaxies (crosses) have on average steeper high-frequency spectra than quasars; in particular, the radio galaxies dominate the lowest part of the diagramme.

Some sources (essentially quasars) exhibit extreme values, especially those with flat $\alpha_{\rm l}$ and/or flat $\alpha_{\rm h}$ (populating the upper and right-hand portion of the plot). These may possess self-absorbed components that become visible in different frequency regimes, depending on their optical thickness. A thorough analysis and interpretation of our results will be presented in a forthcoming paper.

We thank Helge Rottmann for his help during the Effelsberg observations. We are grateful to Dr. Heinz Andernach whose comments on the manuscript helped to improve the paper significantly. Part of this work was supported by the Deutsche Forschungsgemeinschaft, grant KL533/4-2, and by the European Commission, TMR Programme, Research Network Contract ERBFMRXCT97-0034 "CERES''. We thank the Italian Ministry for University and Scientic Research (MURST) for partial financial support (grant Cofin98-02-32). We acknowledge the use NASA's SkyView facility (http://skyview.gsfc.nasa.gov) located at NASA Goddard Space Flight Center.

4 Discussion