In the framework of the evolutionary model for radio sources mentioned above and described in detail in Chapter 2, the well defined anti-correlation between the radio turnover frequency, Vp, and the projected angular size, O (O'Dea & Baum 1997; Fanti et al. 2002) means that the youngest objects have the highest turnover frequencies and that the
peak frequency is expected to move towards lower frequencies as the source expands and the energy density decreases. In principle, the highest the turnover frequency, the youngest the radio source is.
The aim of this thesis is to search and study \extreme" GPS sources, with spectral very close to their birth. These objects, however, are likely to be very rare because their expansion rate, and therefore the rate of decrease of their peak frequency, is thought to be very rapid. The shape of the radio spectrum and the turnover frequency have been used as selection tools for this class of objects. A complete sample of bright candidate GPS sources peaking above ' 5 GHz has been obtained by Dallacasa et al. (2000), who termed such sources \High Frequency Peakers" (HFPs). Only a fraction of their objects are expected to be truly newborn sources since the sample is contaminated by beamed,
at-spectrum quasars (see also Snellen et al. 1999, Stanghellini et al. 2003) caught during a flaring phase of a strongly self-absorbed component that dominates the radio spectrum. Since the process that characterizes both
flaring blazars and young radio sources is the expansion of an emitting blob in the ISM, i.e. the lobes in the case of young sources, a knot of the jet that dominates the radio spectrum for blazars, it is quite diffcult to distinguish the two phenomena with a single observation. Howeever, their time evolution has quite different timescales: GPS/CSS sources are thought to expand at mildly relativistic velocities with marginal or no influence on the radio spectrum while large Doppler factors in blazars boost the variability amplitudes in the spectrum and decrease the corrisponding observed timescales. In order to disentangle the two classes in our sample we have carried out various observational projects (described in Chapter 3) to:
- study the spectral shape evolution and the variability properties of the sample;
- determine the polarisation properties;
- look for extended emission that may indicate that we are really dealing with an evolved-old source but can also be the relic of previous large scale radio activity in a re-born radio source;
- study the milliarcsecond morphology of the candidates.
The information on variability of HFP candidates, on their morphology at different resolutions and the polarimetric study can help us to define criteria to discriminate between the two classes of objects. In Chapter 4 we discuss various criteria and the consistency of their indications on the source nature.
Since it has been proved that the incidence of blazar objects in a spectral selected sample increases with increasing selection frequency and, as a consequence, with increasing turnover frequency (Stanghellini 2003), this kind of analysis is crucial to select truly young radio sources. Clearly, a serious blazar contamination of GPS samples may lead astray analyses of their statistical properties and implies that evolutionary models based on them need to be reconsidered. To effectively explore the luminosity and peak frequency evolution of GPS sources, we need samples that provide a wide coverage of the turnover frequency and flux density (Vp-Sp) plane. In Chapter 5 we give an account of the data sets we used. We also analyze the effect of the different selection criteria adopted to define the samples. We have finally studied an evolutionary model in the framework of the self-similar evolution scenario by Fanti et al. (1995) and Begelman (1996) comparing the redshift and peak frequency distribution yielded by the model with the observed ones. A summary of the main results are given in Chapter 6.