## 4. Results and discussionFor each source we have compiled flux densities at different frequencies from the literature, mostly from Kühr et al. (1979) and from the CATS database (Verkhodanov et al. 1997). Our new measures at 230 GHz have been added to the compilation. All flux densities have been brought to the BGPW scale (Baars et al. 1977). Examples of source spectra are shown in Fig. 1. For the fit algorithm described in Sect. 4.2 we have assumed flux densities less than 3 to be upper limits only.
Most of the sources show significant departure from the classical power law which describes a zero age transparent synchrotron spectrum from a relativistic electron population with power law energy distribution. The deviations from the power law are of the following type: a) a low-frequency turnover (the most conspicuous deviation); b) a steepening at high frequencies. High-frequency flattening, if any, is quite rare. The above deviations are interpreted as due to synchrotron self-absorption and to particle energy losses, respectively. In order to describe them, we have fitted the compiled flux densities with a synchrotron aged spectrum (described in the next section), modified by low-frequency absorption, as follows (Pacholczyk 1970): where is the frequency at which the optical depth is equal to 1. In case of an homogeneous synchrotron self-absorbed source , while is the not aged spectral index in the transparent frequency range. ## 4.1. The synchrotron aged spectrum modelWe assume that the radio source evolution is described by a
The whole spectral shape cannot be described by an analytic equation, the two behaviours described being only the asymptotical ones, and has to be computed numerically. This model is referred as continuous injection (CI). Fitting the spectral data to the numerically computed CI spectrum, one obtains the break frequency , from which the source age is obtained if the magnetic field is known. This simple model does not consider expansion effects, which may be important if the source is young. So, the simplest variant of the original model is the one in which the radiating particles loose energy through expansion and the magnetic field changes according to flux conservation (Kardashev 1962, CIE). An alternative possibility to be considered, in an expanding source, is that the magnetic field changes less rapidly than in the flux conserving assumption because of a continuous magnetic flux input associated to the fresh particles injection. We set , where for m = 2 we get the flux conserving expansion. This is referred as CIm model. In our special case we assume m = 1, consistent with the models applied by Baldwin (1982) or Begelman (1996). Although the theoretical background to these models is in the Kardashev paper, we have decided to present in the Appendix the detailed development. The spectral shapes of the CIE and CIm models have been computed again numerically. The break frequency for the CIE model is sixteen times higher than in the CI case, for equal elapsed time and final magnetic field intensity (Kardashev 1962). In the CIm model, instead, one finds that the break frequency is larger than in the CI model. The asymptotic behaviours at frequencies lower and larger than the break frequency are the same as in the CI model. However the steepening occurs over a broader frequency interval. In order to better emphasize the differences between these three models it is useful to consider, together with the usual flux-frequency representation, the point-to-point spectral index defined as Both representations are shown in Fig. 2. While in the flux-frequency plane the differences are hardly visible, they can be much better traced in the point-to-point spectral index behaviour. Note that the displayed figures only show the theoretical differences. In practice the observed spectra permit only fits in the flux-frequency plane.
## 4.2. The spectral fitsSpectral fits to the spectra have been made with the CI, CIE, and CIm models. They allow us to determine the non-aged spectral index and the break frequency , that, together with the normalization, are the three free parameters characterizing all the models. The best fits are, surprisingly, obtained with the CI model. The other models have steepenings which are too gradual for the majority of the spectra. An example is given in Fig. 3. The reduced of the models including adiabatic expansion and a variable magnetic field is always greater (typically twice) than that of the CI model (Fig. 4). The CI fits appear quite good, even in cases of high values, which, at visual inspection, appear more due to an under-estimation of the flux density errors than to a poor fit of the spectral model on the data.
The majority of the spectra show a clear break frequency, with a change of slope . We stress that there is no evidence for spectral steepening with significantly larger than 0.5. Only a few sources are fitted by simple power laws. In these cases could be either very high ( GHz) or very low (). In these sources we have preferred the low frequency choice, since for the high frequency one would have implied abnormally high values for as compared to the other sources. We also note that only a very small fraction of the CSS sources, if any, shows some evidence of flux density excess at high frequency, as it would be caused by a flat spectrum core or by thermal dust emission. Perhaps the only case is 3C138, where the core, known from VLBI observations, shows up in the integrated spectrum at 230 GHz only. It may appear surprising that the fits with the assumed spectral model are so good. In fact the sources of the sample consist of several components, as lobes and jets and hot spots, where physical conditions can differ from one another and therefore also break frequencies may be different. It is likely that the spectrum is dominated by the brighter component(s). In addition, one could imagine that some confusion might have occurred between genuine spectral steepening due to energy losses and the low-frequency turnover due to absorption processes. We feel that this is a minor problem, but it is difficult to quantify it (see, however, next section). The break frequencies range from a few hundred MHz to tens of GHz. At low frequencies, as said above, the limit is set by confusion with the effects of absorption processes. The injection spectral index ranges from 0.35 to 0.8, with . The typical errors of the break frequencies and the injection spectral indices as given by the fit algorithm are up to 40% and 0.05, respectively. The results of the CI fits are compiled in Table 2.
## 4.3. Radiative ages and the nature of CSS sourcesIn order to determine radiative ages, from Eq. (2) and variants of
the other models, the magnetic field Using the value of , we obtain,
from Eq. (2), radiative ages ranging
from to
years. Since the intrinsic magnetic
fields of the CSS sources in our sample strongly overweight the
magnetic field equivalent to the cosmic microwave background
(), the latter can be neglected in
Eq. (2). Therefore, if the source magnetic field deviates by a factor
The existing structure information on our sample, mostly from MERLIN and VLBI observations, allows us to split the sources in two groups: those in which the overall spectrum is dominated by lobes (classified "type a" in Table 2); those in which the spectrum is dominated by a bright jet or hot spot (classified "type b" in Table 2). For the B3 VLA sample the available information does not yet allow such a morphological sub-division. The sources of class b have radiative ages systematically lower than those of class a, as we expected. Furthermore, while the synchrotron age is well correlated with the source size for class a, it seems that there is no correlation at all between the linear size and the radiative age for class b sources (see Fig. 5). We have further computed for each source the expansion velocity , where LS is the source largest dimension. The two classes show very different distributions: Provided the assumed magnetic field is reasonably correct, the
above values for the radiative ages indicate that
In order to maintain the One could think that the correlation between linear sizes and the radiative ages shown in Fig. 5 could be a partial consequence of the equipartition assumption. In fact, implies that . This is not the case for the following reasons: 1) no correlation between linear sizes and radiative ages is obvious for type b sources. 2) In particular, the break frequencies seem to be correlated with the linear sizes for type a sources (Fig. 6).
These means, at least for class a sources, that the break frequency is an effective clock indicating the source age. The correlation seen in Fig. 5 is not an artifact completely introduced by the assumption of equipartition in Eq. (2). © European Southern Observatory (ESO) 1999 Online publication: April 28, 1999 |