Flat-spectrum symmetric objects with ~1 kpc sizes I. The candidates

In order to understand the origin and evolution of radio galaxies, searches for the youngest such sources have been conducted. Compact-medium symmetric objects (CSO-MSOs) are thought to be the earliest stages of radio sources, with possible ages of<10^3 yrs for CSOs (<1 kpc in size) and 10^4-10^5 yrs for MSOs (1-15 kpc). From a literature selection in heterogeneous surveys, we have established a sample of 37 confirmed CSOs. In addition, we only found three confirmed flat-spectrum MSOs in the literature. The typical CSO resides on a z<0.5 galaxy, has a flat radio spectrum (a_thin<0.5; S_v proportional to v^-a), is<0.3 kpc in size, has an arm length ratio<2, and well-aligned (theta<20 deg) opposite lobes with a flux density ratio<10. In order to populate the 0.3-1 kpc size range (large CSOs) and also in order to find more flat-spectrum MSOs, we have built a sample of 157 radio sources with a_{1.40}^{4.85}<0.5 that were resolved with the VLA-A 8.4 GHz. As first results, we have 'rediscovered' nine of the known CSO/MSOs while identifying two new ~14 kpc MSOs and two candidate CSO/MSOs (which only lack redshifts for final classification). We were able to reject 61 of the remaining 144 objects from literature information alone. In the series of papers that starts with this one we plan to classify the remaining 83 CSO/MSO candidates (thanks to radio and optical observations) as well as characterize the physical properties of the (likely) many 0.3-15 kpc flat-spectrum CSO/MSOs to be found.


The evolution of extragalactic radio sources
The origin and evolution of extragalactic radio sources is one of the outstanding problems in Astronomy (e.g. de Vries et al. 1998a) and has been a fundamental problem in the study of active galaxies and their nuclei (AGN). These come in a variety of sizes, from compact (< 1 kpc) to very large (> 1 Mpc). This wide range of sizes has been interpreted as evidence for size evolution of the radio structure (e.g. Blandford & Rees 1974;Carvalho 1985;Fanti et al. 1995;Readhead et al. 1996a,b). In the standard model of AGN, a central supermassive black hole, with ∼ 10 6 -10 9 M⊙ feeds on the material of the host galaxy to produce two opposing radio emitting jets, thus creating a symmetric source that might only be disturbed by the environment/speed of the jets, unless its source runs out of fuel. Mature radio galaxies fit into this picture and are mostly split up into Fanaroff & Riley (1974) type I and type II radio galaxies (FRI and FRII). Up to 10 5 times smaller, compact-medium symmetric objects (CSO-MSO) might be their precursors (e.g. Readhead et al. 1996a,b).
Traditionally, CSO/MSOs have always been considered high-power radio sources. However, low-power sources must be considered as well, if we really want to tell a story about the evolution of small (and young) radio sources all the way until becoming large radio galaxies (FRII or FRI) -e.g. Marecki et al. (2003). Fanti et al. (1995) already pointed out the hypothesis of MSOs evolving not into FRIIs but into FRIs and despite their bias towards high-power CSSs they concluded that, really, only the most powerful MSOs could be the precursors of FRIIs. Similar conclusions were reached by Readhead et al. (1996b), while Middelberg et al. (2004) go as far as proposing the radio structure of NGC7674 (a Seyfert galaxy) as the one of a (very weak) CSO. Begelman (1996) considered both hypothesis: lower power CSO/MSOs would evolve into FRIs while the high power ones would become FRIIs. In order to constrain models it is important to extend the radio power range .
'Hot-spots' in CSOs are so close (<1 kpc) to the nucleus that they might help towards the understanding of the central engines in AGN (Readhead et al. 1996a,b). Furthermore, they are unique probes to the physics of the gas clouds of the broad line-emitting region - Readhead et al. (1996a). MSOs, being larger (1-15 kpc), are ideal to probe the ISM further away from the nucleus (including the clouds in the narrow line-emitting region (NLR) and extended NLR). They might also be the middle link for the hypothetical evolution of CSOs into FRIIs or FRIs.
There are still two possibilities for the origin of CSO/MSOs, summarized in what follows.
Youth scenario: The most popular view is that in which CSOs evolve into MSO/Compact Steep Spectrum Sources † (CSSs; α thin > 0.5 with Sν ∝ ν −α ) which, in turn, evolve into FRIIs -e.g. Phillips & Mutel (1982); Carvalho (1985); ⋆ The formal boundary from Fanaroff & Riley (1974) is at 178 MHz: 5.3 × 10 25 W/Hz with our cosmology. A typical radio galaxy (α 1.4 0.178 = 0.8) has L 1.4 = 1 × 10 25 W/Hz while a flat spectrum CSO/MSO (α 1.4 0.178 = 0.4-0.5) has L 1.4 = 2 × 10 25 W/Hz. † The original definition is on Peacock & Wall (1982), with α 5 2.7 ≥ 0.5 (now as far as α 1.4 0.325 -e.g. Tschager et al. 2003), who also define almost half of their sample as "compact" and with "steep" spectra; in an historical perspective, up to this time compact ⇔ flat spectrum and extended ⇔ steep spectrum. Phillips & Mutel (1982) demand an optically thin regime with α ν 2 ν 1 ≥ 0.5, ν 1 , ν 2 > 1 GHz. We define α thin from a full spectrum linear fit to the part that is optically thin for frequencies greater than a given peak; if there is no peak, it is inferred to lie at some still unobserved low frequency and the full spectrum is used. Begelman (1996); Readhead et al. (1996b); Kunert et al. (2002); Perucho & Martí (2002). The intermediate ∼ 1 kpc stage should be a CSS in the case of self-similar expansion (lobes expand with growth) or a flat-spectrum MSO in case the expansion is non-self-similar (hot spots remain compact, if seen at all). Maybe less luminous CSOs evolve into FRIs via a Giga-Hertz Peaked Spectrum Source (GPS) stage (O'Dea 1998;de Vries et al. 1998b). Re-born scenario: From an analytical model of the evolution of double radio sources < 100 kpc, Alexander (2000) extended Kaiser & Alexander (1997) model to ∼ kpc scales: a population of 'short-lived' sources is predicted, where the jets are disrupted before reaching the ∼ 1 kpc core radius (King density profile) of the host galaxy. This could be interpreted in the context of "re-birth". For example, Baum et al. (1990) show the 47 pc CSO B0108+388 to have weak radio emission on tens of kpc scales; this might be an unrelated source or evidence for recurrent activity.

Definitions
Over the last twenty years, a panoply of names have been used to classify < 15 kpc-sized sources which might be the precursors of the much larger FRI/FRII radio galaxies. Usually applied in the 'young radio source' context, we have CSOs, MSOs, CSSs, GPSs, and, the oldest of all, compact doubles (CDs). It is still disputed whether CSOs are included in the GPS class (e.g. Snellen et al. 1999, O'Dea 1998, Marecki et al. 1999vs. Stanghellini et al. 1997a, Stanghellini et al. 1999, Fassnacht & Taylor 2001. In Figure 1 we summarize the current (confused) status and in Table 1 we propose a 'non-grey zone' radio classification for all these sources, which can be used for the time being, at least operationally: since CSO/MSOs are a more homogenous class than GPSs are (Fassnacht & Taylor 2001), we propose to split up the two main sets of 'young sources' into the ones selected by morphology (CSO/MSOs) and the ones selected spectrally (CSS/GPSs). For CSO/MSOs, in particular, it should be made clear that it is not necessarily true that an edge brighened lobe is an hotspot. It might just be a knot in a longer jet. However, the likelihood that we get two of those opposed to each other and they are not hotspots is small. It is on this basis that CSO/MSOs with only two components are confirmed. When we come to triple (and more) component sources the hotspot/edge-brightened lobe definition relaxes: if we identify a central core component then we have a CSO/MSO structure (even if no obvious hotspots or edge-brightening is seen in any terminal lobe).
Historically there has been a bias against steepspectrum CSOs and flat-spectrum MSOs (e.g. the "CSOfinding" α < 0.5 Caltech-Jodrell Bank surveys (e.g. Wilkinson et al. 1994); the "CSS (⊃MSO) finding" α > 0.5 Bologna-Jodrell-Dwingeloo survey -e.g. Fanti et al. 1995 CSO (hitherto overlooked) MSO (α < 0.5) Figure 1. The currently messy (and incomplete) situation in the definition of CSOs, MSOs, CSSs, GPSs and CDs. CDs and CSSs cover all ranges of sizes. The latter includes GPSs and MSOs, the first generally < 1 kpc and the latter showing symmetric structure straddling a (putative or not) central nucleus. CSOs also appear with flat-spectrum (α < 0.5) and, in this sense, they cannot be paralleled to GPSs (even if only the ones identified with galaxies); in any case, there is no evidence for complete overlap of GPSs and CSOs. MSOs with a flat-spectrum (upper right corner) have been overlooked. Table 1. Our proposed dual-view radio classification for CSOs, MSOs, CSSs and GPSs. CDs fall out of this picture since they can be all or none (when one-sided core-jet sources). The scheme below must be viewed as operational only, since in the coming years more knowledge of the sources involved might suggest a different one. Note that we give strict morphological criteria to confirm the classification of CSO/MSOs; also that the latter are similar, apart from size (MSOs can be flat-spectrum radio sources).

CSOs (< 1 kpc)
A CSO Conway et al. 1994) is a compact radio source with two outer edge-brightened lobes/hotspots or twin-jets plus a (possibly putative) central core. Symmetry is essential so, operationally, the arm ratio should be ≤ 10, although the flux density ratio (between lobes) is not constrained (it is frequency-dependent). CSOs have weak polarization and variability (< 10% in a few years): some are so stable that they might be excellent VLBI flux density calibrators - Fassnacht & Taylor (2001).

CSSs (<15 kpc)
This class of source, with a subgalactic size and a steep spectrum (α thin > 0.5), has more pronounced lobe flux density ratios and/or arm ratios than CSOs (Fanti et al. 1990;Fanti et al. 1995;Dallacasa et al. 1995;Sanghera et al. 1995;O'Dea 1998). When with a spectrum peak at ν > 0.5 GHz they are classified as GPSs (Section 1.2.4) while when symmetric (most -e.g. Readhead et al. 1996a;Fanti et al. 1995;Kunert-Bajraszewska et al. 2005) they are called MSOs (Section 1.2.5). They show low radio polarizations and little vari-ability although up to an order of magnitude more variable than the most stable CSOs (Fassnacht & Taylor 2001).

GPSs (< 1 kpc)
In most properties, GPSs are similar to CSSs. The main difference is in the spectral peak (c.f. Table 1; e.g. Tornikoski et al. 2001): the canonical turnover frequency of GPSs is 1 GHz -de Vries et al. (1997). Also, many are highly variable (mostly identified with quasars - Torniainen et al. 2005) jeopardizing their usual classification when based only on sparse spectral data points (both in observing epochs and in frequency) - Stanghellini et al. (1998) and Tornikoski et al. (2001).

MSOs (1-15 kpc)
So far regarded as steep-spectrum sources, we here point out the existence of α < 0.5 flat-spectrum MSOs (c.f. Sections 2.3 and 4) as a hitherto not considered type of source (they fill the "empty corner" in Figure 1). Augusto et al.
(1998) mention many candidates for such sources. Flatspectrum MSOs could be the sources into which CSOs evolve when the expansion is non-self-similar (de Young 1997;Tschager et al. 2000). The statistics of MSOs are relevant in order to inspect which evolutionary scenario (non-selfsimilar vs. self-similar expansion) dominates.

The optical hosts
Not much is known at optical wavelengths about CSOs since only a few cases have been studied . Readhead et al. (1996a) and Bondi et al. (1998) found that the hosts of five CSOs are mostly mV ∼ 20-22 elliptical galaxies (0.3-1 L * ) with strong narrow emission lines; the continuum is characteristic of an old stellar population. Detailed HST views of three nearby (z < ∼ 0.1) CSOs (Perlman et al. 2001) confirm residence in normal ellipticals but with ten times more dust than radio elliptical galaxies. A lot more is known in the optical about CSS/GPSs, which have similar redshift distributions and have as hosts 0.1 < ∼ z < ∼ 1 regular giant elliptical galaxies (many interacting), like FRIIs do, a fact consistent with a GPS → CSS → FRII source evolution (O'Dea 1998;de Vries et al. 2000). GPS galaxies (z ∼ 0.3) show a CSO morphology while the quasars (z ∼ 2) do not (O'Dea et al. 1991;de Vries et al. 1998b;Snellen et al. 1999;Stanghellini et al. 2001).

This paper
The total number of confirmed CSOs is relatively small (37 -Section 2; 25 have linear size information) for a two orders of magnitude size range (0.01-1 kpc). Worse, only three α < 0.5 flat-spectrum MSOs (1-15 kpc) and four 'large CSOs' (0.3-1 kpc) are confirmed, so far. The lack of 'large CSOs' and flat-spectrum MSOs might be explained by a CSO 'preferred' evolution into CSSs (e.g. Section 1.1; Augusto et al. 1998), but we need better statistics.
The aim of the series of papers which starts with this one is to find a fairly large number of α 4.85 1.40 < 0.5 flatspectrum CSO/MSOs with ∼kpc sizes (large CSOs and MSOs). We start by establishing the current sample of confirmed CSOs as well as describing their overall properties (Section 2). In Section 3 we build up a 157-source sample out of which we expect a few tens to be confirmed as CSO/MSOs when our study is complete. We conclude with a brief summary (Section 4).

CONFIRMED SYMMETRIC SOURCES
The literature abounds with examples of confirmed CSOs (summarized in Section 2.1) while MSOs are only abundant as CSSs, i.e., with a steep spectrum. Flat-spectrum MSOs are rare (Section 2.3).

The sample of CSOs
In Table 2 we present all currently known confirmed CSOs, proved as such from maps (or kinematics, in a few cases) in our extensive literature search. We were very rigorous in our classification, using the criteria of Table 1. All relevant maps/information have been compiled and carefully scrutinized before listing a given CSO as "confirmed" beyond any reasonable doubt. Everytime the candidate had three or more components (even when some doubt remains about which of the central components really is the core), we required a confirmed central core usually from, at least, twofrequency data. If having only two components, they must be edge-brightened lobes: assumed is a putative central core (c.f. Taylor et al. 1996a;Bondi et al. 1998;Polatidis et al. 1999). 'Hotspots' are not necessarily required for sources with three (or more) components; all we need is emission on both sides of the core (even if jetlike). This is the usual way CSOs have been identified (see, for example, Readhead et al. 1996a;Stanghellini et al. 1997a;. Since CSOs have sizes < 1 kpc, we rejected all sources with size > 1 kpc and since, by definition, they are symmetric sources (e.g. Readhead et al. 1996a,b), we ruled out any with an arm ratio R > 10. Figure 2 defines and explains the calculation of the radio map parameters in Columns (10) and (13)-(16) of Table 2.
In five CSOs studied, Taylor et al. (1996a) find considerable flux density ratio asymmetries in the two opposing jets, possibly due to differences in density of the surrounding medium (Stanghellini et al. 1997a). Furthermore, flux density ratios depend on frequency. Hence, it seems more dangerous to place a limit on such ratio and we do not do it. We also do not constrain arm angles (column (16) of Table 2) since, for example, we have φ = 148 • (misalignment θ is 32 • ) for a 'classic' CSO (B2021+614) and only three CSOs in the Table are more misaligned (reaching a minimum of φ = 134 • for B1543+005, a  CSO).
Comments on the sources marked with * in Column (1) of Table 2 follow: B0046+316: This is a Sy2 galaxy; it possibly has a core-jet radio structure in a weird geometry (Anton et al. 2002). B0424+414, B0500+019, B0646+600, B0703+468, B0710+439: These sources are also classified as GPSs (e.g. O'Dea et al. 1991;Marecki et al. 1999;Stanghellini et al. 2001). B1934-638: This is the archetype GPS (e.g. Tzioumis et al. 1989).  Table 2, we here present a case with an actually located core (without which, only Columns (10), (13) and (14) might have values). Most maps in the literature are presented, like this one, with the lowest contour at three times the r.m.s. noise on the map (3σ). Then, in order to estimate its angular size (LAS), we measure the largest possible extent on the 6σ contour. Next, we locate the core and each lobe peak (marked with crosses) and immeditately identify the strongest lobe (bl) from the contours alone, deriving a peak flux density estimate. The same is made for the faintest lobe (f l) and the nucleus, whereby calculating the values in Columns (13) and (14). Finally, joining by segments the crosses that correspond to the two lobes and the nucleus, we estimate φ (Column (16)) and R = a 1 a 2 , a 1 > a 2 (Column (15)).

Statistics
Since all confirmed and candidate CSOs of Table 2 have been extracted from different samples in the literature with no other selection criteria except for morphology, the statistical results must be taken with caution since they might not be representative of the CSO class. We list 41 sources in Table 2 out of which four (labelled "CSO?") still might be MSOs if their sizes turn out to be 1-15 kpc: we keep them in the table until we have enough data to finally classify them. This leaves us with 37 certain CSOs which we use in the statistical study that follows.
The optical information on the 27 CSOs that do have it (73% completeness) shows that galaxies are clearly the typical host (23 or 85%) while only four sources (15%), at most, reside in quasars. In Figure 3 we present the redshift distribution of the sample (25 sources; 68% complete). We clearly see a concentration towards low redshifts, with 17 (68%) sources at z < 0.5, implying a nearby galactic host population. In fact, except for one quasar, all CSOs reside  in z < 1 hosts. The 25 CSO median is ‡ z25 = 0.36 +0.16 −0.14 . The quasar statistics (only three: they do not change the median at all) are still too poor to conclude that, like for GPSs, the hosts/redshifts imply two independent populations.
As regards to spectral indices, since we use a flatspectrum sample defined from α 4.85 1.40 < 0.5 (Section 3), we included this quantity for all CSOs in Table 2, in addition to α thin . Usually, to select CSOs or CSSs from weak samples, only a two frequency spectral index is used/available -see e.g. Kunert et al. (2002). Figure 4 shows the α 4.85 1.40 /α thin distribution for 33/32 of the 37 CSOs of Table 2 (89%/86% complete). Although CSOs, by definition, have no spectral restrictions, it turns out that the majority have a flat radio spectrum (23, or 70% with α 4.85 1.40 ≤ 0.5; 18, or 56% with α thin ≤ 0.5). The medians are α 4.85 1.40 (33) = 0.3 +0.2 −0.1 and α thin (32) = 0.5 ± 0.1. There is a tendency for α 4.85 1.40 being flatter than α thin . In fact, defining ∆α = α 4.85 1.40 − α thin , we have ∆α29 = −0.1 ± 0.1 (only 29 CSOs have both α 4.85 1.40 and α thin values available). ‡ The subscripts in the medians show the actual number of sources with values available for each calculation. We give the asymmetric error of the median at the 95% confidence level.  In Figure 5 we show the ratio in flux densities between core and brightest lobe for 29 CSOs (78% complete). The remaining eight did not have a visible, properly located, core but we still show the upper limits (estimated from the maps). It turns out that a bright nucleus (ratio > 1 with respect to the bright lobe) is present in 11 (30%) of the sources (including the eight sources with upper limits in the statistics), in one case about 40 times brighter. At the other end, five sources (with upper limits) have a nucleus more than 100 times weaker than the brightest lobe. The median is Sn/S bl (29) = 0.7 +0.8 −0.6 . We have studied the ratio in flux densities between the two opposed lobes for all the 37 CSOs ( Figure 6): 27 (73%) have ratios < 10 (of which 21 (57% of the total) of 1 to 3) but one CSO has it as large as 113: in total, ten (27%) of the CSOs have > 10 ratios. The median is S bl /S f l (37) = 2.2 +2.1 −0.2 . A test of symmetry, which was adopted as definition for CSO/MSOs, is the arm length ratio (Figure 7). From the 29 CSOs (78% complete) with data (core located, from  where each arm is measured) we find that none has an arm ratio above 4.6 and that two-thirds (20) have it smaller than two. The spread is not large and we have for the median: R29 = 1.6 +0.7 −0.4 . Both in median and in distribution of arm ratios CSOs seem to lie somewhere between large FRII radio galaxies (symmetric) and CSS galaxies (asymmetric) -e.g. Saikia et al. (2003).
Although not formally established, it is generally understood that a CSO/MSO should be fairly well aligned, similarly to FRI/FRIIs. The inter-arm angular (φ) distribution for 29 CSOs (78% complete; they must have the core located in order to measure the angle) is plotted in Figure 8 but in the form of the misalignment angle (θ), obtained by subtracting φ from 180 • . We have the following medians: φ29 = 171 +5 −9 deg or θ29 = 9 +9 −5 deg. We can then see that, as it was expected, CSOs are fairly well aligned sources, with θ ≤20 • for 22 (76%) of them.
For the 25 CSOs (68% complete) with measured redshifts, we plot, in Figure 9, their projected linear size (l) distribution. The median is l25 = 0.14 +0.07 −0.05 kpc. This indi-kiloparsec symmetric objects I. The candidates 7  cates a lack of 0.3-1 kpc sources. In fact, from the Figure, we note that CSOs, in general, abound at < 0.3 kpc (21, 84% of the total) but are scarce over the rest of the way up to 1 kpc. In Figure 10 we show the 1.4 GHz power distribution for the 23 CSOs (62% completeness) that have this information. The median [log L1.4 = 25.9 +0.4 −0.2 (W/Hz)] and the distribution clearly reflect the fact that the selection and classification of CSOs, so far, has implied high-luminosity sources, namely with L1.4 > 10 25 (W/Hz). The bias is so strong that, using the formal definition at 178 MHz of the FRI/FRII border, with the help of the power extrapolation via α thin and assuming a power decrease with the inverse square of size (e.g. Begelman 1996), only one of the 23 CSOs in Table 2 with enough information will be powerful enough to become a 1 Mpc size FRII: 4C+32.44. If we relax their future to 100 kpc FRIIs, only two more will be added to the list: B0428+205 and 4C+62.22. In any case, we have searched for a correlation between size and 1.4 GHz radio power but found none.

CANDIDATE
1.40 information were also kept.
The revised parent sample now contains 1743 sources (Table 4), whose redshift and spectral index distributions are presented in Figures 11 and 12, respectively. We note that these distributions use the full sample rather than just a representative subsample (c.f. Augusto et al. 1998). A full discussion on the implications of the revision of the parent sample is made in Appendix A.

The 157-source sample
Although small-size (VLBI scale: 1-300 pc) CSOs had dedicated searches/surveys in order to find them (e.g. , bringing the current number of confirmed cases to 37 (c.f. Table 2 and Figure 9), the problem is that large-scale CSOs and α < 0.5 flat-spectrum MSOs (0.3-15 kpc size range; see also Table 3) have only seven confirmed cases. It was vital first, no doubt, to establish CSOs as new, worth of studying, objects and the VLBI efforts had the ideal impact, showing (many of) them as young sources. We believe that the time has come to start populating the 0.3-1 kpc size-range with CSOs, if we really want to learn  point out that at frequencies below ∼ 1 GHz, interstellar scintillation might induce extrinsic variability in extragalactic radio sources. Hence, our selection with α 4.85 1.40 should be safe, as compared to other selections made with α high−ν 0.3 , likely more affected by such variability. ¶ Globally. There is a further complication since some sources that are in the revised sample were not in the old one (e.g. B0218+357) and vice-versa.
RA (2000) Dec (2000)  about the full story of the small/young 1-300 pc CSOs evolution all the way into FRIIs or FRIs. In this respect, more flat-spectrum MSOs are also needed. Augusto et al. (1998) presented 23 CSO/MSO candidates included in a 55-source sample selected from the parent sample by showing a greater than 25% decrease in their kiloparsec symmetric objects I. The candidates 9 8.4 GHz VLA-A visibilities (usually corresponding to strong radio features with <7:1 flux density ratio and > ∼ 0. ′′ 1 apart; at z > 0.2, the projected linear size is > ∼ 0.3 kpc). However, the 55-source sample was biased towards finding gravitational lenses: many sources were excluded from their final sample using a further surface brightness criterion in that sources with a bright and compact component plus other fainter and resolved components would be rejected (as well as any sources with dominant components > 0. ′′ 3 apart)the full details are in Augusto (1996).
Since we now do not apply to the parent sample any of those extra criteria and only use the visibility criterion, we end up with a sample that includes 157 objects (including the 55 sources of Augusto et al. 1998) - Table 5, some of which are very extended objects. We do not claim that this sub-sample of 157 CSO/MSO candidates is complete since neither is JVAS for the reasons presented, e.g, in Augusto et al. (1998) and Patnaik et al. (1992a).
Using the same classification criteria as for building Table 2 (mentioned in Table 1), a literature search ruled out 61 sources as CSO/MSO candidates, also "rediscovering" nine of the CSO/MSOs in Table 2/Section 2.3 (including "CSO?" cases). We have discovered two new ∼ 14 kpc MSOs (J0751+826, J1454+299) and two CSO/MSOs (4C+66.09, J2055+287; a redshift is needed for each, for final classification; the latter case can be an MSO only if its redshift is < 0.09).
An obvious test to our criteria would be to check how many of the northern hemisphere confirmed flat-spectrum (α 4.85 1.40 < 0.5) CSOs in Table 2 were not selected by us from the parent sample, and why. To identify them it is easy, since they are the ones that are both in Tables 2 and 4 and not in Table 5: they total to 13, of which 12 have sizes ≤ 40 mas, hence they would never have been selected by our criteria. Recalling that we typically can only identify sources that have, at least, two components > ∼ 0. ′′ 1 apart and with < 7 : 1 flux density ratio at 8.4 GHz, as regards to the only remaining (B1413+135), although with a global size of ∼ 0. ′′ 18, it has a very faint lobe (Sn/S f l ≃ 35) while the bright lobe (Sn/S bl ≃ 1.6) is too close to the nucleus (∼35 mas). So, it could not have been selected.
In what follows we describe in detail the sources that refer to the text in Table 5: J0013+778: This is a bright core large symmetric object (LSO), which has detailed information with 1.6 GHz VLBI ) and 1.4 GHz VLA . We locate its core thanks to the JVAS map, since at 8.4 GHz with the VLA-A the middle VLBI component has the most inverted spectrum of all while the northeast component also shows a more modest inverted spectrum; the one of the southwest component is very steep. The overall size is about 8 ′′ in a north-south direction (thanks to further, weaker VLA 1.4 GHz components) which, at its redshift of 0.326, gives it ∼ 32 kpc. 4C+36.01: This is a radio galaxy with an extended halo, giving it an overall size of ∼ 40 kpc, as can be seen in a VLA 1.4 GHz map . J0123+307: This source is a good example on why JVAS is not a complete sample. As explained (and imaged) in detail in Augusto (1996) this is, in fact, a VLBA point source (500:1 map) that had its position in error by an amount suf-ficient to cause bandwidth smearing and confuse our visibility selection criterion (see also Augusto et al. 1998;Patnaik et al. 1992a). J0259+426: As for the previous source, this was also a JVAS failure and the selection was made erroneously. Although not quite a VLBI point source (it is a triple source ∼ 15 mas in size - Henstock et al. 1995) it should never have been selected. 3C108: This is a triple source with a candidate core at the centre from a MERLIN 1.7 GHz map in Saikia et al. (1990). In JVAS, the 8.4 GHz VLA-A map confirms the central component as a core (α 8.4 1.7 ≃ 0.2) while the source redshift of 1.215 implies that its 5. ′′ 88 angular size means a linear projected size of ∼40 kpc. Hence, this source is rejected. J0654+427: Bondi et al. (2001) give two VLBI maps at different resolutions (and frequencies: 1.6 and 4.9 GHz) that leave no room for doubt that the structure is that of a corejet source rather than a CSO/MSO: the brightest component in both images is the core since it has α 4.9 1.6 = 0.0 (using peak brightnesses), likely becoming inverted if model fitting is applied. J0656+321: Yet another source that is an example of why JVAS is not a complete sample. As explained (and imaged) in detail in Augusto (1996) this is, in fact, a MERLIN point source (670:1 map) that had its position in error by an amount sufficient to cause significant bandwidth smearing (see also Augusto et al. 1998;Patnaik et al. 1992a). J0751+826: Also presenting VLBI compact structure Xu et al. 1995), with an easily identified core, the large scale structure of this source, easily seen in the VLA-A 8.4 GHz map of JVAS, looks like a ∼ 2 ′′ wideangle tail. Its 1.991 redshift implies a global size of ∼ 14 kpc, just at the border of still classifying it as an MSO. J0815+019: This source is in the appendix of Augusto et al. (1998), discarded by them from the 55-source sample (due to an erroneous spectral index evaluation -see Section 3.1). It is now recovered into the 157-source sample. Augusto et al. (1998) presented a MERLIN 5 GHz map of the source. We must locate the core in future multi-frequency follow ups. J0817+324: As explained (and imaged) in detail in Augusto (1996) this is, in fact, a MERLIN point source (500:1 map) that had its position in error by an amount sufficient to cause significant bandwidth smearing (see also Augusto et al. 1998;Patnaik et al. 1992a). J0837+584: All evidence seems to point to a core-jet source. In addition to the JVAS map and visibility, hinting at a strong unresolved component plus a very weak and distant (∼ 0. ′′ 6 away) blob, the two frequency 1.6 and 5 GHz VLBI maps of Polatidis et al. (1995) and Xu et al. (1995) locate the nucleus as the westernmost component, with α 5 1.6 = 0.15 as opposed to the α 5 1.6 = 1.4 value of the other strong component ∼ 8 mas away. J0855+578: This source was observed with the VLBA at 5 GHz by Taylor et al. (2005). It has one of the lobes edgebrightened but the other is not so convincing. There is no core detected. We must find a core with higher frequency observations, or a more convincing structure to pass our strict criterion for a CSO confirmation; Taylor et al. (2005) have not managed to detect this (weak) source at 15 GHz. 4C+66.09: This source is in the appendix of Augusto et al. (1998), discarded by them from the 55-source sample. It is now recovered into the 157-source sample. Augusto et al. (1998) presented a MERLIN 5 GHz map of the source which leaves no room for doubt that this source is either a CSO or an MSO, depending on its unknown redshift: edge-brigthening is seen in both lobes, although no core is detected. VLBA observations have been conducted for this source at 1.7, 4.8 and 15 GHz (Rossetti et al. 2005) confirming the classification and finding hotspots at both ends, although the nucleus still remains undetected. 4C+55.17: The multi-frequency maps of Reid et al. (1995) and the source redshift of 0.909 imply a large size (∼ 53 kpc), although the source is symmetric indeed (an LSO). J1015+674: Augusto (1996) shows it as a MERLIN point source (300:1 map) that had its position in error by an amount sufficient to cause significant bandwidth smearing (see also Augusto et al. 1998;Patnaik et al. 1992a). J1041+525: This is a well-studied large scale quasar (∼ 150 kpc in size), easily seen also in VLBI scales (e.g. Henstock et al. 1995;Taylor et al. 1996b). J1058+198: With a global size of ∼ 420 kpc (62 ′′ at a redshift of 1.11), this is a very large source, possibly a radio galaxy (e.g. Hooimeyer et al. 1992). J1110+440: Augusto (1996) shows it as a core-jet source (a very compact and strong nucleus and an extended, 50 times weaker, jet). J1306+801: This is a very large triple source (∼110 kpc - Taylor et al. 1996b), possibly an LSO, since the core appears to be located in the middle component (from multifrequency data). J1324+477: Augusto (1996) shows it as a point source, in a 200:1 MERLIN 5 GHz map, so this source was erroneously selected due to bandwidth smearing. J1424+229: This is a well known arcsecond-scale gravitationally lensed multiple-image system (e.g. Patnaik et al. 1992b).  Hintzen et al. (1983) at 1.4 GHz) which, given its redshift of 1.358 and from its angular size of ∼ 15. ′′ 5, has a global size of ∼ 110 kpc. Arp220: This is a very well known radio galaxy with ultraluminosity at IR wavelengths, presenting a double radio/IR nucleus (Norris 1988;Graham et al. 1990) and also maser emission. Too many observations at all wavelengths exist for this source to mention here, so just as essential ex-amples we cite Emerson et al. (1984); Soifer et al. (1984); Norris et al. (1985); Shaya et al. (1994); Heckman et al. (1996); Scoville et al. (1998); Clements et al. (2002). It is not a CSO/MSO since it is thought that most of its radio emission comes from strong starburst activity (e.g. Rovilos et al. 2003). 4C+49.26: As already pointed out by Augusto et al. (1998) this source is an LSO with a 6 ′′ size which, at its redshift of 0.7, makes it ∼ 36 kpc in total. J1607+158: This is a core-jet source, from VLBI (Beasley et al. 2002) up to 8.4 GHz VLA-A scales (JVAS). 4C+12.59: From several multi-frequency maps (Saikia et al. 1990;Lonsdale et al. 1998;Dallacasa et al. 1998) it is still not clear whether this source is a core-jet or an LSO. From our point of view this is irrelevant, since its angular size of 3. ′′ 3 and redshift of 1.795 make it ∼ 24 kpc in size. J1715+217: A recent VLBA map on this source (Gurvits et al. 2006, in prep.) shows it as a core with a jet containing a strong feature about ∼ 60 mas from the core. The VLA-A 8.4 GHz visibilities, however, suggest larger scale structure as well. Future MERLIN 5 GHz observations should find it. J1749+431: All extant multi-frequency maps (Henstock et al. 1995;Taylor et al. 1996b;Beasley et al. 2002) strongly suggest that this source has a core-jet structure. J1753+093: All evidence seems to identify this radio source with a galactic star (Thompson et al. 1990). NGC6521: Condon et al. (2002) find it likely that this source has a core plus two lobes on each side, with an overall size of 5 ′ , giving it a size of ∼ 150 kpc at its 0.027462 redshift.  (1998) presented both MERLIN and VLBA 5 GHz maps of this source after which there is no doubt to classify its structure as a core-jet.
As regards as the redshift distributions, the difference is obvious by eye, with the 157-source sample containing more low-z sources than the parent sample. Furthermore, the latter has a smooth redshift distribution, roughly flattening around z ∼ 0.7 and having an average redshift coincident with that of other flat spectrum radio source samples (<z> ≃ 1.2; Munoz et al. 2003). The median values also suggest a selection of the closest radio sources (0.88 vs. 1.12), albeit Table 6. Comparison of the redshift (z) and spectral index distributions (α 4.85 1.40 ) between the parent sample and the 157-source sample. The medians are indicated with their asymmetric errors (95% conf. level). We also give the number of sources for this calculation in each case (#)    with intersecting values, within the 95% conf. level errors. More formally, we have applied a KS-test to compare the two distributions and reject the hypothesis that they are similar at the 95% confidence level. As for the spectral indices distributions we have, again, a smooth distribution for the parent sample (roughly flattening at α 4.85 1.40 ≃ 0.2) while for the 157-source sample the distribution is also smooth but still rising when it reaches the limit of α 4.85 1.40 = 0.5. This time, the distributions are clearly different by eye and medians (whose errors do not overlap). We performed the formal KS-test to compare both distributions and rejected the hypothesis that they are similar at the 99.9% confidence level. This result is similar to the one of Augusto et al. (1998) and also similarly explained by the fact that we are selecting resolved sources from the parent sample and this (normally) implies steeper spectrum sources.

SUMMARY
In what follows we briefly summarize the main conclusions from this paper: • In order to understand the origin and evolution of extragalactic radio sources in the context of the standard model of AGN, several VLBI searches have been conducted trying to identify the youngest such sources ( < ∼ 10 3 yrs), of which compact symmetric objects (CSOs) are the most serious contenders. Ideally, we should also follow the evolutionary track at later stages, by identifying somewhat older sources (∼ 10 4 -10 5 yrs), possibly medium symmetric objects (MSOs). In this paper we summarize all confirmed cases of CSOs that we found from the literature, which total to 37, and three α 4.85 1.40 < 0.5 flat spectrum MSOs. • By studying the sample of the currently confirmed 37 CSOs we conclude the following (the completeness of the statistics is ≥ 62% but beware that the sample might not be representative of the CSO class due to the heterogeneous surveys from where the sources were selected): i) 85% of the optical hosts are galaxies, typically residing at z < ∼ 0.5; the remaining are quasars, with a large spread in redshift range; ii) most CSOs have flat radio spectra (70% with α 4.85 1.40 < 0.5; 56% with α thin < 0.5); iii) most (17, 59%) CSOs follow the "classical" Conway et al. 1994) definition where the brightness of the nucleus is < 10% of the one of the brightest lobe; one-third of the CSOs present nuclear components that are brighter than the brightest of the two opposed lobes -is this evidence for boosting?; iv) all CSO/MSOs were defined to have arm length ratios R ≤ 10 (for symmetry); the maximum value on the present sample is R = 4.6, with 90% (all but three) having R ≤ 3.0; v) 73% of the CSO/MSOs also present symmetry in the flux density ratios between the two lobes (≤ 10); however, this ratio can be as large as 113, among the remaining; vi) 76% of CSO/MSOs have well aligned opposing structures (θ ≤ 20 • ) but values as large as θ = 46 • can be found; vii) CSOs have a median linear projected size of 0.14 +0.07 −0.05 kpc, with 84% smaller than 0.3 kpc.
• The aim of the series of papers starting with this one is to improve, many times, the number of confirmed large CSOs and of flat-spectrum MSOs (0.3-15 kpc), which currently sits at six. In this paper, in particular, we present a sample of 157 sources, drawn from a parent sample of 1743 flat spectrum (α 4.85 1.40 < 0.5) sources by selecting the ones with radio structure on > ∼ 0. ′′ 1 scales. This resulted in the selection of the lowest redshift and steepest spectrum sources including > ∼ 0.3 kpc CSO/MSO candidates. Although we have immediately rejected, based on literature information, 61 of the sources, 83 are still left with data either to be analysed or to be gathered. As for the remaining thirteen sources, nine were already listed as CSOs/flat-spectrum MSOs from the literature and are, thus, a good quality control for our selection. As for the final four, 4C+66.09 is a CSO/MSO (needs a redshift to identify which type exactly); J0751+826 and J1454+299 are ∼ 14 kpc MSOs; J2055+287 might be an MSO too, if at z < 0.09.
As regards the statistical conclusions of Augusto et al. (1998) we must revise them by comparing the "old" and the "new" situations, now that we have both revised the parent sample and the 55-source sample (which, now, has all its sources included in the larger 157-source sample). We focus on the α 4.85 1.40 distribution only. Starting with the parent samples we first note that, while for the new sample we used the maximum number possible of values (1311) for the old one a representative sub-sample of 373 sources was selected and it is from this one that the statistics of Augusto et al. (1998) are worked out. Comparing both through a KS-test we cannot reject the hypothesis that they are similar. Finally, comparing the α 4.85 1.40 distributions for the 157-source (actually 123 values) and 55-source samples through a KStest we cannot reject the hypothesis that they are similar.