5. Spectral indices
Now that both WENSS and NVSS are nearing completion it is fairly straightforward to derive spectral indices beween 325 and 1400 MHz. It should be noted however that the NVSS gives components , obtained by doing Gaussian fits on the NVSS images. This means that different "sources" may refer to one and the same radio source, because they represent for example the core and the lobes separately. However, WENSS has already grouped such components into a radio source. Since the large quantities of data force us to use automatic procedures, we made a cross correlation with a deliberately large search radius of four arcmin, between the WENSS minisurvey sources and the NVSS. As a next step we then applied a source-dependent search radius, requiring that NVSS components should be considered as the 1400 MHz counterpart of the WENSS source only if i) the NVSS component was not more distant from the WENSS position than the WENSS beam size (taken as 54 arcsec), in the case of unresolved WENSS sources, ii) the NVSS component was closer than 1.5 times the largest angular size, in the case of resolved WENSS sources. This procedure reduces the contamination of the 1400 MHz flux by unrelated background sources to a minimum, while at the same time guaranteeing the inclusion of most components (except perhaps for the very biggest sources).
There is one point which needs some more attention: in trying to derive radio spectral indices (using the NVSS) we should be careful that there are no, or only insignificant, hidden biases in the flux densities at 325 MHz and 1400 MHz. This is all the more important, as the enormous quantity of data forces one to follow automatic procedures. The routines used to derive integrated flux densities are slightly different in WENSS and NVSS: whereas NVSS makes gaussian fits of components, which are forced to give a source size that is at least as big as the beam, this is not so in WENSS, where the source size is directly related to the fit parameters (and these are allowed to be smaller than the beam). For strong sources the bias in spectral index should be so small as to be insignificant, but unfortunately for weaker sources ( mJy) this is not true, as is shown in Fig. 4.
WENSS tends to underestimate the integrated fluxes of the weak sources and we estimate that WENSS fluxes are on average too small by a factor of the order 10 % for sources with peak flux densities in the range mJy (see Fig. 4). An empirical correction factor was determined, by which the integrated fluxes of Table 1 should be divided: , where is the peak flux at 325 MHz in mJy. This correction is small above 100 mJy; it shifts the distribution of the ratio integrated/peak flux densities back to being symmetrical (in the logarithm) around unity, as we would expect if the scatter is purely due to the map noise. Of course, flux densities below 100 mJy should anyway be treated with caution.
For our galaxy identifications we find that the average spectral index between 325 and 1400 MHz is (standard deviation of the mean). The spectral index distribution is shown in Fig. 5.
In particular the faintest sources (below 100 mJy) tend to have flat spectra, but we should keep in mind that we had to apply an empirical correction to the integral flux densities of WENSS, which may not have always succeeded in recovering the flux completely.
For WENSS sources with mJy we have: .
This spectral index is significantly flatter than what is usually found for samples of radio sources. It is known, however, that the spectral index depends on radio power (Laing & Peacock 1980). Their correlation is mainly based on FR II sources, but it is most likely that the correlation extends down to the fainter end of the FR I sources (see Laing & Peacock 1980). In fact, if we take all minisurvey sources with mJy (roughly 3100) and repeat the cross-correlation we get: . This is expected, as the bulk of the radio sources will be associated with quasars and galaxies of higher radio power, at larger distances than the sources of our nearby galaxy sample. We also checked that the spectral index of the galaxy sample was not significantly biased by resolution effects: since the sources are nearby they may have large angular diameters and consequently some flux may have been lost at 325 MHz. We therefore took only unresolved WENSS sources, for which we get a median , or 0.67 if we exclude sources associated with spiral galaxies. This shows that any resolution effect is so small as to be insignificant. Using the relation between spectral index and power given by Laing & Peacock (1980), and assuming a median of 0.67, our sample should consist of FR I radio galaxies peaked around a median power (at 325 MHz) of . The sample of elliptical galaxies given in Table 1 has a median and is therefore entirely compatible with the Laing & Peacock relation.
© European Southern Observatory (ESO) 1998
Online publication: September 30, 1998