Next Contents


For several decades, extragalactic radio surveys remained the most powerful tool to probe the distant universe. Even `shallow' radio surveys, those of limited radio sensitivity, reach sources with redshifts predominantly above 0.5. Since the 1960s, the most effective method for finding high-z galaxies has been the optical identification of radio sources, a situation persisting until the mid-1990's, when the arrival of the new generation of 8-10 m class optical/infrared telescopes, the refurbishment of the Hubble Space Telescope, the Lyman-break technique (Steidel et al. 1996) and the Sloan Digital Sky Survey (York et al. 2000) produced an explosion of data on high-redshift galaxies.

This is not a historical account (see Sullivan III 2009); but listing the revolutions in astrophysics and cosmology wrought by radio surveys serves to set out concepts and terminology. On the astrophysics side we note the following:

  1. Active Galactic Nuclei (AGNs) The discovery of radio galaxies (Bolton et al. 1949, Ryle et al. 1950) whose apparently prodigious energy release (Burbidge 1959) suggested Compton catastrophe, calling the cosmological interpretation of redshifts into question.

  2. Synchrotron emission The identification of synchrotron emission (1951 Ginz-burg, 1952 Shklovskii) as the dominant continuum process producing the apparent power-law spectra of radio sources.

  3. Quasars The discovery of quasars, starting with 3C 273 (Hazard et al. 1963, Schmidt 1963), leading to the picture of the collapsed supermassive nucleus (Hoyle & Fowler 1963), and hence to the now-accepted view of the powerful Active Galactic Nucleus (AGN) - massive black-hole - accretion disk systems (Lynden-Bell 1969) powering double-lobed (Jennison & Das Gupta 1953) radio sources via `twin-exhaust' relativistic beams (Blandford & Rees 1974, Scheuer 1974).

  4. Relativistic beaming The discovery of superluminal motions of quasar radio components (Cohen et al. 1971), this non-anisotropic emission (anticipated by Rees 1967) resolving the Compton non-catastrophe (Woltjer 1966) and leading to the development of unified models of radio sources: quasars and radio galaxies are one and the same, with orientation of the axis to the viewer's line of sight determining classification via observational appearance (Antonucci & Miller 1985, Barthel 1989, Urry & Padovani 1995).

On the cosmology side we note the following:

  1. Scale of the observable Universe An irrefutable argument by Ryle & Scheuer (1955) placed the bulk of `radio stars' beyond 50 Mpc, and it was quickly realized when arcmin positional accuracy became available (Smith 1952) that the majority of the host galaxies were beyond the reach of the optical telescopes of the epoch. Minkowski (1960) measured a redshift of 0.46 for 3C 295, the redshift record for a galaxy for 10 years. Astronomers had discovered a set of objects substantially `beyond' the recognized Universe. By 1965 the redshift record was 2.0 for the quasar 3C 9 (Schmidt 1965). Only after the turn of the century did the redshift record become routinely set by objects discovered in surveys other than at radio wavelengths (e.g. Stern 2000).

  2. History of the Universe Early radio surveys generated a passionate and personal debate, the Steady-State vs Big-Bang controversy. It was rooted in the simplest statistics to be derived from any survey: the integral source counts, the number of objects per unit sky area above given intensities or flux densities. As discussed by Ryle & Scheuer (1955), the source count from the 2C radio survey (Shakeshaft et al. 1955) showed a cumulative (integral) slope of ~ -3, far steeper than that expected from the Steady-State prediction, any reasonable Friedman model, or from a static Euclidean universe. For each of these, the initial slope at the highest flux densities is -3/2. (Euclidean case: the number of sources, N, is proportional to the volume, i.e. to r3 for a sphere; the flux density is propto r-2, so that Npropto S-3/2.) Bondi & Gold (1948) together with Hoyle (1948) were uncompromising proponents of the new Steady-State theory. Ryle et al. interpreted the 2C apparent excess of faint sources in terms of the radio sources having far greater space density at earlier epochs of the Universe. Confusion, the blending of weak sources to produce a continuum of strong sources, was then shown to have disastrous effects on the early Cambridge source counts. From an independent survey in the South, Mills et al. (1958) found an initial slope of -1.65 after corrections for instrumental effects, significantly lower than that found for 2C. Scheuer (1957) developed the P(D) technique, circumventing confusion and showing that the interferometer results of 2C were consistent with the findings of Mills et al. But the damage had been done: cosmologists, led by Hoyle, believed that radio astronomers did not know how to interpret their data.

In 1965 the `source-count controversy' became irrelevant in one sense. Penzias & Wilson (1965) found what was immediately interpreted (Dicke et al. 1965) as the relic radiation from a hot dense phase of the Universe. The Big Bang was confirmed.

Ryle was right all the time. Integral source-count slopes of -1.8 or even as shallow as -1.5 were nowhere near what the known redshifts plus Steady State cosmology - or even any standard Friedman cosmology - predicted. These all come out at -1.2 or -1.3, shallower than the asymptotic -1.5 as sources of infinite flux density are not observed, and nobody has ever claimed the initial source count slope at any frequency to be as flat as this. The discovery of the fossil radiation (see Peebles et al. 2009) may indeed have shown that a Big Bang took place; but the source counts demonstrated further that objects in the Universe evolve either individually or as a population - a concept not fully accepted by the astronomy community until both galaxy sizes and star-formation rates were shown to change with epoch.

Source counts from radio and mm surveys - with errors and biases now understood - are currently recognized as essential data in delineating the different radio-source populations and in defining the cosmology of AGNs. These counts are dominated down to milli-Jansky (mJy) levels by the canonical radio sources, believed to be powered by supermassive black-holes (e.g. Begelman et al. 1984) in AGNs. At fainter flux-density levels, a flattening of slope in the Euclidean normalized differential counts (i.e. counts of sources with flux density S, within dS, multiplied by S2.5, see Section 2) was found (Windhorst et al. 1984, Fomalont et al. 1984, Condon & Mitchell 1984), interpreted at the time as the appearance of a new population whose radio emission is, to some still-debated extent, associated with star-forming galaxies.

Radio-source spectra are usually described as power laws (Snu propto nu-alpha) 1; the early low-frequency meter-wavelength (e.g. 178 MHz) surveys found radio sources with spectra almost exclusively of steep power-law form, with alpha ~ 0.8. Later surveys at cm-wavelengths (higher frequencies, e.g. 5 GHz) found objects of diverse spectral types, some with spectra rising to the high frequencies, some with steep low-frequency portions flattening and rising to the high frequencies, and yet others with a hump in the radio regime, or indeed two or more humps. In general, anything which was not `steep-spectrum' in form was called `flat-spectrum', an inaccurate nomenclature: very few truly flat-spectrum sources have been found and even then the flatness persists over only a limited frequency range. Nevertheless AGN-powered radio sources are traditionally classified in two main categories: steep- (alpha > 0.5) and flat-spectrum (alpha < 0.5). Broadly speaking, to radio telescopes the steep-spectrum objects showed extended double-lobed structures, while the flat-spectrum objects were point sources, unresolved until the Very-Long-Baseline Interferometry (VLBI) technique provided sub-arc-second mapping. The compact nature of flat-spectrum sources led to the conventional interpretation of synchrotron self-absorption at frequencies below the bump(s), implying brightness temperatures of ~ 1011 K for the estimated magnetic field strengths.

From a physical point of view, it is appropriate to consider the integrated spectra as composites, built of the combination of different components of radio sources. Unified models provide a framework for such a discussion.

In the widely accepted `unification' scheme (Scheuer & Readhead 1979, Orr & Browne 1982, Scheuer 1987, Barthel 1989) the appearance of sources, including this steep-spectrum/flat-spectrum dichotomy, depends primarily on their their axis orientation relative to the observer. This paradigm stems from the discovery of relativistic jets (Cohen et al. 1971, Moffet et al. 1972) giving rise to strongly anisotropic emission. In the radio regime (Fig. 1), a line-of-sight close to the source jet-axis offers a view of the compact, Doppler-boosted, flat-spectrum base of the approaching jet. Doppler-boosted low-radio-power (Fanaroff & Riley 1974, type I (FRI; edge-dimmed)) sources are associated with BL Lac objects, characterized by optically-featureless continua, while the powerful type II (FRII; edge-brightened) sources are seen as flat-spectrum radio quasars (FSRQs). The view down the axis offers unobstructed sight of the black-hole - accretion disk nucleus at wavelengths from soft X-rays to UV to IR, and this accretion-disk radiation may outshine the starlight of the galaxy by 5 magnitudes. The source appears stellar, either as a FSRQ or a BL Lac object. FSRQs and BL Lacs are collectively referred to as blazars. In the case of a side-on view, the observed low-frequency emission is dominated by the extended, optically-thin, steep-spectrum components, the radio lobes; and the optical counterpart generally appears as an elliptical galaxy. A dusty torus (Antonucci & Miller 1985) hides the black-hole - accretion-disk system from our sight (Fig. 1). At intermediate angles between the line-of-sight and the jet axis, angles at which we can see into the torus but the alignment is not good enough to see the Doppler-boosted jet bases, the object appears as a `steep-spectrum quasar'.

Figure 1

Figure 1. Unified scheme for high radio-power Fanaroff-Riley (1974; FRII) sources (following Jackson & Wall 1999).

In general, then, each source has both a compact, flat-spectrum core and extended steep-spectrum lobes (Fig. 2). This already implies that a simple power-law representation of the integrated radio spectrum can only apply to a limited frequency range. The reality is even more complex (Wall 1994). External absorption or, more frequently, self-absorption (synchrotron and free-free) can produce spectra rising with frequency at the low-frequency optically-thick regime, while at high frequencies the synchrotron emission becomes optically thin, power law, and energy losses of relativistic electrons ("electron ageing", Kellermann 1966) translate into a spectral steepening.

Figure 2

Figure 2. Spectral behaviour in the millimeter band of the radio galaxy NGC6251 (left panel) and (right panel) 11-GHz isophotes overlaid on the 0.3 GHz map (Mack et al. 1997). The low-frequency spectrum is due to the steep-spectrum outer lobes while at higher frequencies the flatter-spectrum core-jet system dominates.

Two classes of ultra-steep-spectrum (alpha > 1.3) sources have been discovered. One is associated with galaxy clusters; the objects are of relatively low luminosity and generally are not associated with any host galaxy. They are diffuse and of several types, including cluster `radio halos', `radio relics' and `mini-halos', and each type appears to involve reactivation of the hot intra-cluster medium by shocks or cooling flows, the observed form depending on the cluster evolutionary state (Feretti 2008). These `radio ghosts' will not be discussed further here. The second class of ultra-steep-spectrum source is very radio-luminous and these are mostly identified with very-high-redshift radio galaxies. The high redshifts tempt the suggestion that the steep spectral index is due to the effect of redshift moving the steepest part of the spectrum (where electron ageing effects are strong) into the observed frequency range. However, Klamer et al. (2006) demonstrated that this is not the dominant mechanism, and that high-redshift radio galaxies, discovered by the steep-spectrum technique, have intrinsically power-law spectra. The selection of ultra-steep-spectrum sources is a very effective, but not the only (Jarvis et al. 2009), way to find high redshift radio galaxies (see Miley & De Breuck 2008 for a comprehensive review), including the one holding the current record, TN J0924-2201 at z = 5.19 (alpha0.3651.4 appeq 1.6; van Breugel et al. 1999). The highest-redshift radio-loud quasar known to date, the z = 6.12 QSO J1427+3312 (McGreer et al. 2006), also has a steep radio spectrum (alpha1.48.4 = 1.1) although it was not discovered through this characteristic.

In very compact regions, synchrotron self-absorption can occur up to very high radio frequencies, giving rise to sources with spectral peaks in the GHz range. At high radio luminosities this category comprises the GHz Peaked Spectrum (GPS) sources (O'Dea 1998) some of which peak at tens of GHz (High Frequency Peakers; Edge et al. 1998, Dallacasa et al. 2000, Dallacasa et al. 2002, Tinti et al. 2005). At low luminosities, high-frequency spectral peaks, again due to strong synchrotron self-absorption, may be indicative of radiatively inefficient accretion, thought to correspond to late phases of the AGN evolution, with luminosities below a few percent of the Eddington limit (advection-dominated accretion flows (ADAF, (Quataert & Narayan 1999) or adiabatic inflow-outflow scenarios (ADIOS, Blandford & Begelman 1999, Blandford & Begelman 2004).

As the `flat' spectra are actually the superposition of emitting regions peaking over a broad frequency range (Kellermann & Pauliny-Toth 1969, Cotton et al. 1980), whose emission is strongly amplified and blue-shifted by relativistic beaming effects, a power-law description is a particularly bad approximation. The spectral shapes are found to be complicated, and generally show single or multiple humps. Many of these show flux-density variations, attributed to the birth and expansion of new components and shocks forming in relativistic flows in parsec-scale regions. The variations may be on times scales from hours to months or even years, and substantial resources have been devoted to monitoring these variable sources, led by groups at Michigan (USA) and Metsahövi (Finland) (e.g. Aller et al. 2003, Valtonen et al. 2008). The latter reference shows how global (multi-wavelength and multi-telescope) these monitoring programmes have become; moreover the quasi-periodicity for the object in question, OJ 287, indicates that it is probably a binary black-hole system. With regard to flux variations, we also note the `Intra-Day Variables' (IDVs), blazars whose flux densities vary wildly on time scales from minutes to days: these are flat-spectrum objects with extremely small components that show inter-stellar scintillation (ISS) via the turbulent, ionized inter-stellar medium (ISM) of our Galaxy (e.g. Lovell et al. 2007). Detailed discussion of all these variable objects is beyond the scope of this review.

The discovery of Compact Steep Spectrum sources (CSS; Kapahi 1981, Peacock & Wall 1982, O'Dea 1998) originally appeared to be an exception to the conventional wisdom that steep and flat spectra are associated with extended and compact sources respectively. CSS sources are unresolved or barely resolved by standard interferometric observations (arcsec resolution), and the integrated spectra show peaks at < 0.5 GHz, above which the spectral indices (on average, alpha ~ 0.75) are typical of extended radio sources. There is compelling evidence that these objects, as well as GPS and associated types of object (HFPs and CSOs - Compact Symmetric Objects) are young radio galaxies, as summarized concisely by Snellen (2008).

It follows from the above that the conventional two-population approach (flat- and steep-spectrum) assuming power-law spectra is particularly defective at high radio frequencies, where several different factors (emergence of compact cores of powerful extended sources, steepening by electron energy losses, transition from the optically-thick to the optically-thin synchrotron regime of very compact emitting regions, etc.) combine to produce complex spectra (see Fig. 3). Nevertheless, for many practical applications the conventional approach remains useful in describing the bulk population properties of AGN-powered radio sources.

Figure 3

Figure 3. Examples of radio-source spectra at mm wavelengths: a flat-spectrum source (top left panel); a steep-spectrum source (bottom left panel); a source whose spectrum flattens at nu ~ 10 GHz (top right panel); a High Frequency Peaker (HFP) source (bottom right panel). Data from the NEWPS Catalogue' (López-Caniego et al. 2007, [asterisks]) and from the AT20G Survey (Massardi et al. 2008a, [diamonds]).

The radio emission of star-forming galaxies is mostly optically-thin synchrotron from relativistic electrons interacting with the galactic magnetic field, but with significant free-free contributions from the ionized interstellar medium (Condon 1992, Bressan et al. 2002, Clemens et al. 2008). At mm wavelengths, however, the radio emission is swamped by (thermal) dust emission, whose spectrum rises steeply with increasing frequency. The well-known tight correlation between radio and far-IR emission of star-forming galaxies (Helou et al. 1985, Gavazzi et al. 1986, Condon et al. 1991) vastly increases the body of data relevant to characterize, or at least constrain the evolutionary properties of this population. However, to date few attempts have been made to build comprehensive models encompassing both radio and far-IR/sub-mm data (but see Gruppioni et al. 2003).

In this paper we first review the observed radio to mm-wave source counts (Section 2), the data on the local luminosity function of different radio source populations (Section 3), and the source spectral properties (Section 4). Next (Section 5) we look at evolutionary models for the classical radio sources as well as for individual populations, such as GPS sources, ADAF/ADIOS sources, and (Section 6) star-forming galaxies and gamma-ray afterglows at radio wavelengths. We deal briefly with the Radio Background (Section 7) and the Sunyaev-Zeldovich effect on cluster and galaxy scales (Section 8). Section 9 contains a summary of the information on large scale structure stemming from large-area radio surveys. Finally, in Section 10 we summarize perspectives for the future, and Section 11 contains some conclusions.

Unless otherwise noted, we adopt a flat LambdaCDM cosmology with OmegaLambda = 0.7 and H0 = 70 km s-1 Mpc-1.

Figure 4

Figure 4. Differential source counts at 150, 325, 408, 610 MHz normalized to c Snu-2.5, with c = 1000, 100, 10, 1 respectively. Reference codes are spelt out in the notes to Tables 1, 3, and 4. The lines are fits yielded by an updated evolution model (Massardi et al., in preparation).

1 We note that this negative sign convention for alpha is not universal; however the convention has been adopted for the K-corrections of optical quasars and for the extrapolation from optical to X-rays (`alphaox'). Back.

Next Contents