4 The new sample of GPS spectrum sources

Finally we merged Subsample One (24 sources) with those 98 sources from Subsample Two whose flux densities we had successfully measured with MERLIN at 408 MHz. Since the selection process for this project was carried out, the 5 GHz Green Bank survey of WB92 has been superseded by the GB6 survey (Gregory et al. 1996) and at 1.4 GHz we can now use the the NRAO VLA Sky Survey (NVSS) (Condon et al. 1998) which has a resolution of $45\hbox{$^{\prime\prime}$}$ and an rms sensitivity of approximately 1.5 mJy.

Unfortunately, at the time of writing, NVSS -- although almost complete -- did not cover all areas of the sky within its declination limits. Many NVSS maps ($4\hbox{$^\circ$}\times 4\hbox{$^\circ$}$ each) appear to be "patchy" and the "holes" can sometimes be quite large. Our survey suffered considerably from this shortcoming of the current edition of NVSS -- 9 sources out of those 122 sources we wanted to study were simply not present in the NVSS catalogue. (One source out of these nine was also unavailable in GB6.) Additionally we decided to remove 2 other sources from the further processing: one of these is blended with a nearby source and the second one has an extended structure which should be studied in more detail.

At 1.4 GHz we also tried to use the Faint Images of Radio Sky at Twenty (FIRST) catalogue (White et al. 1997) -- 24 our sources could be found there. For 20 objects out of these we noted a very good compatibility between FIRST and NVSS based fluxes; the 4 objects which showed discrepancy are indicated in Table 6.

The selection process described above gave us finally 111 objects for which we arrayed the flux density values at each frequency. Then we attempted to fit model spectra to the available data using a broken power-law with the following formula (Moffet 1975):

$$S(\nu)={S_0\over1-{\rm e}^{-1}} \cdot (\nu/\nu_0)^k \cdot (1-{\rm e}^{-(\nu/\nu_0)^{l-k}}).$$ (2)

Here k and l are the spectral indices of the rising and declining parts of the spectrum as often used in radio astronomy, while S₀ and $\nu_0$ are just fitting parameters which are not equal to the maximum flux density ($S_{\max}$) and the peak frequency ($\nu_{\max}$) of the fitted spectrum. Even though the broken power-law seems to be the physically more sensible choice for a model spectrum compared to a simple second-order polynomial, it has the disadvantage that it is unconstrained if the peak of the spectrum falls beyond the 2nd highest or below the 2nd lowest available frequency. In these cases we fixed the peak of the model spectrum (i.e. $S_{\max}$ and $\nu_{\max}$) at the peak of the measured data -- this was usually the measurement at 4.85 GHz -- and marked the fit as unconstrained in Table 4. This means that the values for k or l have to be considered as a lower or upper limit respectively (i.e. in reality the spectrum will be more inverted at low frequencies or steeper at high frequencies).

The spectra of 35 sources could not be fitted with such a convex-shaped curve (Fig. 1) and we claim that those sources cannot be termed "GPS sources" at all and most likely are just variable flat-spectrum sources. The 76 spectra that could be fitted with our algorithm are presented in Fig. 2 and the fitting parameters are given in Table 4. As can be seen from Fig. 2 and Table 4, some of the sources with unconstrained model spectra, fit the data relatively poorly at low frequencies or have relatively flat spectral indices (e.g. 0307+380 and 0610+510) and thus are less probable GPS candidates.

 

Figure 1: Spectra of non-GPS sources. Abscissae: frequency in GHz, ordinates: flux densities in mJy

 

Figure 1: continued

 

Figure 2: Spectra of GPS sources. Abscissae: frequency in GHz, ordinates: flux densities in mJy

 

Figure 2: continued

 

Figure 2: continued

 

Figure 2: continued

 

Table 4: GPS sources' spectra fitting parameters

Table 4. continued

 

Table 5: Some other parameters of JVAS GPS sources

Table 5. continued

References of the catalogues:

S4 -- Pauliny-Toth et al., 1978.
S5 -- Kühr et al., 1981b.
FIRST -- White et al., 1997.
B3 -- Ficarra et al., 1985.
WENSS -- Rengelink et al., 1997.
CJ2 -- Taylor et al., 1994.

 

Table 6: Candidate GPS sources

In Table 5 we specified some parameters of our "new" GPS sources gathered from the literature: the names of other catalogues a particular source is a member, the optical identification according to the NASA/IPAC Extragalactic Database (NED) and the redshift. At the time of writing 21 objects from our collection have been identified (3 galaxies, 18 QSOs) and their redshifts are known. Those 3 galaxies have low redshifts (); on the other hand -- as expected -- the majority of quasars have large redshifts: for 6 QSOs , for 7 other QSOs . One QSO, 1338+381, is extremely redshifted: z=3.103.

Most of the GPS sources studied so far hardly show any variability, therefore we checked our sources against possible flux variations. Firstly, because a significant variability of the flux density would mean that the source in question is likely not to be a GPS and secondly -- since our data are not simultaneous -- any variability makes derivation of spectra questionable. The part of sources' spectra around 1.4 GHz is obviously the most "sensitive" with regard to the GPS phenomenon so we compared fluxes at this frequency given in WB92 to those from NVSS. We applied corrections for the different beam sizes of these two measurements. If a particular source had changed its flux between epochs of the GB surveys and NVSS/FIRST more than 25% or the 1.4 GHz GB flux was missing in WB92 we treated such a source as potentially variable, unless we could find a second epoch flux density measurement elsewhere. We assigned a "candidate'' status for such objects and listed them in Table 6. Among these there are 4 sources (0412+447, 1125+366, 1357+404, 2005+642) with inverted spectra only, i.e. apparently having turnovers in their spectra at frequencies larger than 8.4 GHz. This feature was yet another reason to assign them a candidate status.