3. GAS AND DUST

"Sometimes you can't stick your head in the engine, so you have to examine the exhaust" Donald E. Osterbrock

Gas and dust, in various states, are important constituents of HzRGs and must play a fundamental role in their evolution. Gas has been observed over a wide range of temperatures and having a variety of forms.

3.1. Hot ionised gas - Radio depolarization

X-ray observations show that hot ionized gas is widespread at the centres of clusters and around radio galaxies at low redshifts. Typical derived parameters are temperatures of ~ 10^7.5 K, densities of 0.05 cm^-2, and masses of a few × 10¹² cm-3, with pressures of ~ 10^-9 dyne cm^-2. Such a gas would be sufficient to confine both the radio-emitting plasma and the Ly halos (see Section 3.2) of HzRGs. For low and intermediate redshift clusters, bubble-like structures in X-rays have been attributed to energy injection into such a hot intracluster medium by radio jets (e.g. McNamara et al. 2000).

The sensitivity of the present generation of X-ray telescopes is only marginally sufficient for detecting similar hot gas at z ~ 2. X-ray emission has been observed from the Spiderweb Galaxy at z = 2.2 (Figure 4) (Carilli et al. 2002). The X-ray emission is extended along the radio source axis and attributed to thermal emission produced by shocks as the synchrotron-emitting relativistic jet that propagates outwards.

Figure 4. X-ray emission from the Spiderweb Galaxy, PKS 1138-262 at z = 2.2, observed with the Chandra X-ray Telescope [From Carilli et al. (2002)]. X-ray contours are superimposed on a VLA gray scale representation of the 5 GHz radio continuum emission at 0.5" resolution. The cross marks the position of the radio galaxy nucleus. Note that the X-ray and radio emission are aligned with each other. See also Figure 16.

An alternative technique for detecting and studying this hot gas is through its effect on the polarisation of background radio emission (e.g. Miley 1980). As it propagates through a magnetoionic medium the polarisation of linearly polarised radiation is rotated through an angle that is proportional to the square of the wavelength, (cm).

= 5.73 × 10^-3 RM ², where R is the observed rotation measure, given by RM is RM = 812 n_t(s)B_||(s)ds ~ 8.1 × 10⁸ n_t B_||s rad m^-2.

Here s(kpc) is the path length through the medium, n_t (cm^-3) is the density of thermal electrons and B_|| is the component of magnetic field along the line of sight. The intrinsic rotation measure is RM = RM_obs × (1 + z)²

Observations of radio polarisation data for > 40 HzRGs (Carilli et al. 1994, 1997, Athreya et al. 1998, Pentericci et al. 2000, Broderick et al. 2007) revealed that several sources have extremely large rotation measures (RMs), up to more than 1000 rad m^-2 in the rest frame (Carilli et al. 1997, Athreya et al. 1998) (e.g. Figure 5), with considerable variation across the sources. An obvious interpretation of the large observed rotation measures is that HzRGs are embedded in dense gas. At low-redshifts the largest rotation measures are observed for radio sources that are located in X-ray emitting clusters that are inferred to be "cooling flows" (Ge and Owen 1994). Hence the large rotation measures of HzRGs are strong circumstantial evidence that these objects are also located in cluster environments at high redshifts (Section 7).

Figure 5. The observed polarisation position angle as a function of the square of the wavelength2 for a radio knot of 4C 41.17 at z = 3.8. [From Carilli et al. (1994)]. The knot is the brightest of the 4 radio knots shown in Fig. 2. Note the excellent fit to a 2 dependance and the large resultant rotation measure for the knot.

3.2. Warm ionized gas

The bright emission lines emitted by luminous radio galaxies allow their redshifts to be determined and have been an important reason why radio galaxies have been such important cosmological probes. The emission lines are emitted by warm (10^4.5 K) gas and provide a powerful tool for studying physical conditions within the HzRGs.

Emission lines observed from active galactic nuclei are both "permitted" recombination lines and collisionally-excited "forbidden" lines. When a free electron is captured by an atomic nucleus it cascades to the ground state emitting a series of recombination lines. The most prominent recombination lines are due to the most common elements, hydrogen and helium. However, some of the strongest lines in the spectra of emission nebulae correspond to so-called "forbidden" transitions that have miniscule probability of occurring relative to the permitted transitions, but whose energies lie within a few kT of the ground levels and can therefore be easily populated by collisions. Each element has a critical density below which most de-excitations are radiative and a line is produced. Forbidden lines are only produced by gas with densities in the range ~ 10 - 10⁵ cm^-3 and the line ratios contain important information about the density, temperature, ionisation and abundances of the emitting gas (Osterbrock and Ferland 2006, Peterson 1997).

A composite emission-line spectrum for HzRGs, and a list of emission line strengths were given by McCarthy (1993). During the early nineties it was found that, for high-redshift radio galaxies, nuclear emission lines are generally accompanied by an additional component that is highly extended spatially. The presence of giant luminous ionized gas nebulae are amongst the most remarkable features of HzRGs (e.g. Reuland et al. 2003). These "halos" have sizes of up to ~ 200 kpc (e.g Fig 16) and their study provides a wealth of information about kinematics and physical conditions within and surrounding the HzRGs and on the origin of the gas.

During the last 15 years considerable progress has been made in studying the properties of Ly halos, using large ground-based optical/IR telescopes. Observations of their emission line spectra have been extended into the infrared (optical rest-frame), where additional useful physical diagnostics can be obtained. Also, the kinematics of several HzRGs have been mapped in detail. An excellent recent review of the emission-line properties of HzRGs has been given by Villar-Martín (2007). For detailed information the reader is referred to that review and the references therein.

The ionised gas halos have Ly luminosities of typically ~ 10^43.5 erg s^-1 and line widths varying from a few hundred km/s in their outer parts to > 1000 km s^-1, near the galaxy nuclei. The usual emission-line diagnostics (Osterbrock and Ferland 2006) show that the gas has temperatures of T_e ~ 10⁴ K - 10⁵ K, densities of n_e ~ 10^0.5 - 10^1.5 cm^-3 and masses of ~ 10⁹ - 10¹⁰ M. The warm gas occupies a relatively small fraction of the total volume of the HzRGs, with filling factors estimated to be ~ 10^-5 compared with unity for the hot gas. Although the topology of the gas is not well determined, the properties of the emission lines and the covering factor deduced from statistics of absorption lines (Section 3.3.2) clouds led to a model (van Ojik et al. 1997) in which the halos are composed of ~ 10¹² clouds, each having a size of about 40 light days, i.e. comparable with that of the solar system. van Ojik et al. (1997) speculate that the clouds might be associated with the early formation stages of individual stars.

The morphologies of the Ly halos are clumpy and irregular. Their overall structures are often aligned with the radio axes and sometimes extend beyond the extremities of the radio source. There appear to be two distinct regimes in the halos that often blend together. The inner regions close to the radio jets are clumpy, with velocity spreads of > 1000 km s^-1. They appear to have been perturbed by the jets. The outer regions are more quiescent, with velocity ranges of a few hundred km s^-1). They appear to be more relaxed than the inner regions (e.g. Fig. 7).

Figure 7. Long slit spectroscopy of the ionized gas halo in MRC 0943-242 at z = 2.9. [From Villar-Martín (2007), Humphrey et al. (2006)]. VLA radio map overlaid with the WFPC2 HST image (left), spatially aligned with the 2-dimensional Keck spectra of the main UV rest frame emission lines. The impact of jet-gas interactions on the observed properties of the giant nebula (Size ~ 70 kpc) can be seen in the much broader and brighter emission lines within the radio structures compared with the faint emission detected beyond.

3.2.1 Ionisation

Various mechanisms have been proposed for exciting the gas. These include (i) photoionisation from an AGN, (ii) photoionisation from stars, (iii) photoionisation by ionizing X-rays emitted by shocked hot gas and (iv) collisional ionization from shocks. Plots of optical-line ratios have been used extensively to study the ionisation of gas in nearby active galaxies, where evidence for both jet- and accretion-powered shocks and for photoionization by the central AGN has been found (e.g. Villar-Martin and Binette 1997, Villar-Martin et al. 1997, Bicknell et al. 2000, Groves et al. 2004, Groves et al. 2004).

Although these relationships have been most accurately calibrated in the rest-frame optical region of the spectrum, line diagnostic diagrams have also been developed for use in the ultraviolet (Allen et al. 1998, Groves et al. 2004), the spectral region of HzRGs sampled by optical observations. Interpretation of emission line ratios are complicated by effects of dust and viewing angles (Villar-Martin and Binette 1996). Also most HzRG spectra are spatially integrated over regions of 10 or more kiloparsec, where conditions and sources of excitation may change. In a recent comprehensive study using as many as 35 emission lines throughout the rest-frame UV and optical spectra (Humphrey et al. 2006) concluded that photoionization is the dominant source of excitation in the quiescent gas. A harder source of photoionisation than stars is needed, consistent with photons from an AGN.

If a quasar is exciting the warm gas in the halo, why don't we see it? The usual explanation is that the quasar emits radiation anisotropically. It is highly absorbed in the direction in which we view it, but not along the radio axis. Support for this idea is provided by the large optical polarisations measured for some HzRGs. (Section 3.5.2). However, if highly anisotropic flux from a hidden quasar is indeed the dominant source of excitation, it is difficult to understand why many of the Ly halos are approximately symmetric in shape. An alternative explanation is that the quasar activity is isotropic and highly variable, with short sharp periods of intense activity and longer periods of relative passivity, when the quasar is dormant (see Section 5.1).

Although photoionisation by a quasar is presently the "best bet" for the dominant source of excitation, it is not likely to be the only culprit. Considerable variation is observed in the relative emission line strengths from object to object and within individual HzRGs. There is strong evidence that there is also collisional excitation from shocks, particularly close to the radio jets - see (e.g. Bicknell et al. 2000, Best et al. 2000) and Section 4.3.

3.2.2 Abundances and star formation

Although Ly is by far the brightest emission line emitted by the warm gas halos, other lines are also detected. In 4C 41.17 at z = 3.8, kinematical structures in the Ly line are closely followed in the carbon CIV and oxygen [OII] and [OIII] lines, with the [OIII] 5007 emission extending by as much as 60 kpc from the nucleus (Reuland et al. 2007). See also Fig. 7

The chemical abundance of the halo gas is close to solar (Vernet et al. 2001, Humphrey et al. 2006), consistent with the HzRG having undergone prodigious star formation at earlier epochs. Further indications that the star formation rate in HzRGs was higher in the past come from measurements of the Ly luminosities and Ly/HeII ratios that are both systematically larger for HzRGs with z > 3, than for those with 2 < z < 3 (Villar-Martín et al. 2007, Villar-Martín et al. 2007).

The relative intensity of the NV 1240Å line varies from being an order of magnitude fainter than the carbon and helium lines (De Breuck et al. 2000) to being as luminous as Ly (e.g. van Ojik et al. 1994). This has been interpreted as indicating that there are large variations in metallicity between HzRGs.

3.2.3 Inner Kinematics: Outflows, jet interactions and superwinds

In considering the kinematics of the ionized gas halos, we shall separately consider the nuclear turbulent regions, where there is evidence for outflows and the outer passive regions where the most likely dominant systematic motions are inwards. The inner regions show shocked gas closely associated with the radio lobes. These display disturbed kinematics and have expansion velocities and/or velocity dispersions > 1000 km s^-1.

Besides synchrotron jets there is evidence that starburst "superwinds" (Armus et al. 1990, Zirm et al. 2005) are also present in the inner regions of the Ly halos. Both the jets and the superwinds will exert sufficient pressure on the warm gas to drive it outwards. Interactions between radio jets and the ambient gas is important in some low-redshift radio galaxies (e.g. Heckman et al. 1982, Heckman et al. 1984, van Breugel et al. 1984, van Breugel et al. 1985, van Breugel et al. 1985). The radio sources are observed to excite, disturb and entrain the gas. Likewise, the gas can bend and decollimate radio jets and enhance the intensity of their radio emission through shock-driven particle acceleration. In general small radio sources show more jet-gas interactions than large ones (Best et al. 2000). At z > 2, HzRG exhibit signatures of even more vigorous jet-gas interactions (e.g. Villar-Martin et al. 1998, Villar-Martín et al. 1999, Villar-Martín et al. 2003, Humphrey et al. 2006). The kinematics is more turbulent and the ionisation is higher in the region of the jet than in the quiescent outer halo (Humphrey et al. 2006).

Nesvadba et al. (2006) recently carried out an important study of the Spiderweb Galaxy with an integral field spectrograph in the optical rest frame. They make a convincing case that there is an accelerated outflow of warm gas in this object. The only plausible source of energy for powering this outflow (few × 10⁶⁰ erg) is the radio jet and, even then, the coupling between the jet and the ISM must be very efficient to account for the observed kinematics. The pressure in the radio jets can drive gas outwards from the nuclei for tens of kiloparsecs and play an important role, together with starburst-driven superwinds, in "polluting" the intergalactic medium with metals. There is spectroscopic evidence for the ejection of enriched material in 4C 41.17 at z = 3.8 up to a distance of 60 kpc along the radio axis (Reuland et al. 2007). See also Fig. 7.

In Section 4.3 we shall discuss strong evidence that the jet-gas interactions can also trigger star formation in HzRGs.

Figure 6. Velocities and velocity dispersions across MRC 2104-242 at z = 2.5 measured with VIMOS on the VLT. [From Villar-Martín (2007)]. This HzRG is surrounded by a giant Lya nebula that extends by ~ 120 kpc The position of the radio core is indicated with a cross. The velocity field appears symmetric and ordered implying either rotation or radial motions. See also van Ojik et al. (1996).

3.2.4 Outer kinematics: Infall of the quiescent halos

An important diagnostic in tracing the origin of the warm gas is the kinematics in the outer region of the giant halos, where they are apparently unperturbed by the radio jet. The outer halos displays systematic velocity variations of a few hundred km/s (e.g. see Figure 7). Are these systematic velocity variations the result of rotation (van Ojik et al. 1996, Villar-Martín et al. 2003, Villar-Martín et al. 2006), outflows (Zirm et al. 2005, Nesvadba et al. 2006) or infalling motions?

It is difficult to discriminate between the various kinematic scenarios from velocity data alone, but comparison of spectroscopic and radio data provides additional relevant information. In a study of 11 HzRGs with redshifts 2.3 < z < 3.6, several correlated asymmetries were found between the halo kinematics and assymetries in the radio structures (Humphrey et al. 2007, Villar-Martín et al. 2007). On the side of the brightest radio jet and hot spot, the quiescent nebula appears systematically redshifted (receding) compared with the other side. On the assumption that the bright radio jet and hot spot are moving towards us and brightened by Doppler boosting (Rees 1967, Kellermann 2003), the quiescent gas must be moving inwards. The brightest radio hotspot is also the least depolarised one, as expected if it is on the closest side of the HzRG and consistent with the infall scenario.

Could the inflowing motion of the gas be a result of cooling flows, that have long been studied in clusters of galaxies at low redshifts, (e.g. Fabian 1994, Kaufmann et al. 2006)? Recent XMM and Chandra observations have shown that the cooling rates are reduced by an order of magnitude below the simple cooling flow models at temperatures < × 10⁷ K, probably through interaction of the gas with radio sources associated with galaxies in the cluster. An alternative diagnostic of cooling flows that are more feasible for HzRGs than X-rays is the measurement of roto-vibrational lines from H₂ molecules at ~ 2000 K (Jaffe and Bremer 1997, Jaffe et al. 2005). At high redshifts these lines are shifted from the near to mid-IR bands, accessible with the Spitzer Telescope. The H₂ lines are a promising tool for studying both the cooling in the gas around HzRGs and the excitation mechanisms.

To summarise, the kinematics of the warm gas is complex. There is evidence that gas in the outer regions of the halo is flowing inwards, providing a source of material for feeding the active nucleus (Section 5) and that gas in the inner region is being driven outwards by pressure from jets and starbursts. The various motions in the ionized gas halos are likely to contribute to feedback processes between the AGNs and the galaxies invoked in current models of massive galaxy evolution (Section 6).

We end our discussion of the halo kinematics on a cautionary note. Most kinematic studies of HzRGs until now have been based on Ly, because of its relatively large equivalent width and its accessibility with optical telescopes. However, Ly is a resonant line and subject to strong scattering and optical depth effects. Hence, the resultant kinematics may not be completely representative of the gas as a whole, particularly in the inner regions.

3.2.5 Relation to non-radio Ly nebulae

There may well be a connection between Ly halos in HzRGs and the disembodied extended Ly nebulae that have been discovered in recent years (e.g. Fynbo et al. 1999, Steidel et al. 2000, Francis et al. 2001, Matsuda et al. 2004, Dey et al. 2005, Colbert et al. 2006, Nilsson et al. 2006). These nebulae also have physical extents 100 kpc and Ly line fluxes of ~ 10^-15 ergs s^-1 cm^-2. Although in many respects, these Ly blobs resemble the giant ionised gas halos around HzRGs, they have < 1% of the associated radio continuum flux and no obvious source of UV photons bright enough to excite the nebular emission. However, millimeter emission has been detected from several of these nebulae (Chapman et al. 2001, Smail et al. 2003, Geach et al. 2005) and Matsuda et al. (2004) suggest that the extended Ly nebulae are also associated with dense environments in the early Universe. Just like the HzRG halos, these nebulae can be excited by quasars that are heavily obscured along the line of sight (Haiman and Rees 2001, Weidinger et al. 2004, 2005), or quasars that undergo recurrant flares (Section 5.1). Alternatively, they may be associated with cooling-flow-like phenomena (Haiman et al. 2000, Dijkstra et al. 2006, Dijkstra et al. 2006).

3.2.6 Nebular continuum

Dickson et al. (1995) pointed out that the nebular continuum emission due to the line-emitting warm gas of HzRGs is significant and must be taken into account when computing the various contributions to the UV continuum. See also (Aller 1984, Vernet et al. 2001, Humphrey et al. 2007). However, in general the nebular component contributes < 25% of the continuum emission at 1500 Å, and is much less important than starlight or emission from a hidden or dormant quasar. The contibution from nebular continuum can be quite accurately predicted by means of the strength of the HeII 4686Å line, or when not available the HeII 1640Å line (Aller 1984).

3.3. Neutral Gas

There are two techniques for studying cool HI gas in HzRGs. One method is to measure redshifted absorption of the 21 cm hydrogen line against the bright radio continuum. The second method is to observe deep narrow absorption troughs that are often present in the Ly profiles. In principle, both these techniques can be used to constrain properties of the neutral hydrogen such as spatial scales, mass, filling factor, spin temperature and kinematics (Röttgering et al. 1999).

3.3.1 HI Absorption

Neutral hydrogen (HI) atoms are abundant and ubiquitous in low-density regions of the ISM. They are detectable by means of the hyperfine transition, emitting at 1420.405751 MHz (~ 21 cm). Observations of atomic neutral hydrogen by means of this line has been one of the most powerful tools of radio astronomy since its inception. Studying the 21cm HI line in absorption has been an important probe of HI around low-redshift radio galaxies see the recent review of Morganti (2006). Limitations of such searches for HI in high-redshift radio galaxies include the availability of low-noise receivers that cover the observing frequencies dictated by the target redshifts and problems due to radio frequency interference. In 1991 HI absorption was detected in the HzRG, 0902+34 at z = 3.4, by Uson et al. (1991). See Figure 8. Since then progress in this field has been disappointing. For a review see Röttgering et al. (1999). Besides some follow-up work on 0902+34 (Briggs et al. 1993, De Bruyn 1996, Cody and Braun 2003, Chandra et al. 2004), there have been only two tentative but unconfirmed additional detections of HI absorption in 0731+438 at z = 2.4 and 1019+053 at z = 2.8.

Figure 8. Absorbing neutral hydrogen. Top: The redshifted HI absorption profile of 0902+34 at z = 3.4. [From Cody and Braun (2003)]. Bottom: The redshifted Ly absorption profile of 0943-242 at z = 2.9, with a model overlaid. [From Jarvis et al. (2003)].

The HI absorption line provides a measure of the average column density of the absorbing material weighted by the flux density of the background source. The total column density is ~ 4.4 × 10¹⁸ T_S. Assuming a spin temperature of T_S ~ 10³ K, (De Bruyn 1996) derives an HI column density of ~ 2 × 10²¹ cm^-2 for 0902+34 and a corresponding mass of neutral hydrogen of M_HI ~ 3 × 10⁷ M.

Why has HI absorption not been detected in a larger number of HzRGs, despite extensive searches? A possible explanation is that the HI absorption is caused by small ~ 100 pc-sized disks or torus-like structures, aligned perpendicular to the radio source (Röttgering et al. 1999). The radio emission of 0902+34 is more centrally concentrated than that for most HzRGs, consistent with the hypothesis that the absorption is produced by such a small disk. For the more extended radio sources associated with most HzRGs, the disk covering fraction would be very small and the disk would not produce significant absorption.

3.3.2 Ly absorption.

More than a decade ago van Ojik et al. (1997) discovered that strong absorption features are common in the Ly profiles of HzRGs. Such features were present in the majority of the 18 HzRGs that they studied with sufficient spectral resolution. Derived column densities were in the range 10¹⁸ - 10^19.5 cm^-2.

The absorption, usually interpreted as being due to HI surrounding the HzRG, provides an interesting diagnostic tool for studying and spatially resolving neutral gas surrounding HzRGs. Because the spatial extension of the absorbing region can be constrained, the Ly absorption lines provide information about properties of the absorbing gas (e.g. dynamics and morphologies) that cannot be studied using quasar absorption lines. Since in most cases the Ly emission is absorbed over the entire spatial extent (up to 50 kpc), the absorbers must have a covering fraction close to unity. From the column densities and spatial scales of the absorbing clouds, the derived masses of neutral hydrogen are typically ~ 10⁸ M.

Additional information about the properties of the HI absorbers was obtained by Wilman et al. (2004). In a study of 7 HzRGs with 2.5 < z < 4.1, they identified two distinct groups of H I absorbers: strong absorbers with column densities of N_HI ~ 10¹⁸ to 10²⁰ cm^-2 and weaker systems with N_HI ~ 10¹³ - 10¹⁵ cm^-2. They suggest that the strong absorbers may be due to material cooling behind the expanding bow shock of the radio jet and that the weak absorbers form part of the multiphase proto-intracluster medium responsible for the Ly forest. Furhermore, Krause (2005) carried out hydrodynamic simulations of a HzRG jet inside a galactic wind shell and showed that strong HI absorption could be produced.

Absorption is occasionally observed in the profiles of other emission lines than Ly, such as CIV - (Röttgering et al. 1995). Jarvis et al. (2003) studied the profiles of two of the most prominent absorbing HzRGs 0943-242 at z = 2.9 and 0200+015 at z = 2.2 with high spectral resolution. The data are consistent with a picture in which the absorbing gas has low density and low metallicity and is distributed in a smooth absorbing shell located beyond the emission-line gas. However, the metallicity, inferred from the C IV absorption, is considerably lower in 0943-242 than in the slightly larger source 0200+015. This difference in metallicity is explained as due to chemical enrichment via a starburst-driven superwind (Section 3.2.3). Further observations and modeling of the spectrum of 0943-242 by Binette et al. (2000) indicate that in this object the absorbing gas may actually be ionised - see also (Binette et al. 2006). However, 0943-242 may be a special case. It has a relatively small radio size and one of the deepest Ly absorption trough of all known absorbers.

3.4. Molecular Gas

Star formation is generally observed to occur in molecular clouds - cold dark condensations of molecular gas and dust, that are observable in the millimeter and near-IR. In these clouds atomic hydrogen associates into molecular hydrogen, H₂, a species that unfortunately does not emit easily observable spectral lines. The next most abundant molecule is carbon monoxide. Because the rotational transitions of the dipolar molecule ¹²CO are caused primarily by collisions with H₂, CO is an excellent tracer of molecular hydrogen. The most important redshifted CO transitions for the study of high-redshift objects are J=(1-0), (2-1), (3-2), (4-3) and (5-4) at 115.2712, 230.5380, 345.7960, 461.0407 and 576.2677 GHz respectively. These lines are an important diagnostic for probing the reservoir of cold gas available for star formation.

Intensive searches for CO emission from HzRGs during the early 1990s were unsuccessful (Evans et al. 1996, van Ojik et al. 1997). Since then, the sensitivity of (sub)millimetre receivers has been improved and the high-redshift Universe has been opened to molecular line studies (see reviews by Solomon & vanden Bout (2005) and Omont (2007)).

To convert the CO to H₂ mass, one often assumes a standard conversion factor. For high redshift CO studies, this factor is calibrated based on observations of nearby ultra-luminous infrared galaxies (ULIRGS) (Downes et al. 1993). With the assumption that this value is also applicable to high redshift objects, one can use the strength of the (1-0) CO transition to derive the mass of the molecular gas. Because current centimetre telescopes do not allow observations of the (1-0) transition at z < 3.6, one needs to observe higher order transitions shifted to the atmospheric windows at 3, 2 and 1.3 mm.

The inferred masses of H₂ in the CO-detected galaxies are between 10¹⁰ and 10¹¹ M, indicating that there is a large mass of molecular gas in these objects and a substantial reservoir of material available for future star formation. However, the calculated masses should be treated with caution, because their derivation is based on a large number of assumptions (e.g. Downes et al. 1993). Observations of higher CO transitions are biased to the detection of denser gas than ground-state observations and can result in an underestimate of the total molecular gas mass.

Observations of multiple rotational transitions allow the temperature and density of the molecular gas to be constrained using large velocity gradient models. The results indicate that the CO properties are heterogeneous. TN J0924-2201 is only detected in the ground-state (1-0) transition, but 4C 41.17 is only detected in CO(4-3), despite sensitive searches for the (1-0) transition (Ivison et al. 1996, Papadopoulos et al. 2005). This high excitation level in 4C 41.17 implies large gas densities n(H₂) > 1000 cm^-3, consistent with gas fueling a nuclear starburst.

In a few cases the CO appears spatially resolved (Papadopoulos et al. 2000, Greve et al. 2004, De Breuck et al. 2005), extending over 10 - 20 kpc (e.g. see Figure 9), providing kinematic information about the molecular gas. Klamer et al. (2004) have claimed that there are alignments between the molecular gas and radio morphologies in some of the detected CO HzRGs, as might be expected from jet-induced star formation (Section 4.3). However, higher resolution, larger signal to noise and more statistics are needed before any conclusions about possible alignments can be drawn.

Figure 9. CO in 4C 41.17 at z = 3.8 with the Plateau de Bure interferometer. [From De Breuck et al. (2005)]. Shown is the velocity plotted against position offset of CO(4-3), extracted along a PA of 51° (see radio morphology in Figure 2).The central frequency at 96.093 GHz (z = 3.79786) is based on the wavelength of the optical HeII 1640Å emission line. The position-velocity diagram shows the two components of the CO emission. One of these is coincident with the radio core, while a fainter component is spatially offset from it towards the southwestern lobe A.

Offsets between the velocities of the molecular gas (CO) and those of the warm gas (e.g. HeII 1640) of up to 500 km/s have been measured (De Breuck et al. 2003, 2005). Because the H₂ masses exceed those of the warm ionized gas Ly halos by an order of magnitude, the CO lines provide a better measure of the systemic redshift of the HzRGs than the UV and optical emission lines.

The CO detections listed in Table 2 are not representative of the distribution of CO in HzRGs. There are several observational biases inherent in current CO studies:

• Many searches for CO at high redshifts pre-select targets on the basis of prior dust detections, i.e continuum millimetre flux (see Section 3.5.1). However, 850 µm continuum emission was not detected for two of the CO-emitting HzRGs - 53W002 at z = 2.390 (Scoville et al. 1997, Alloin et al. 2000) and TN J0924-2201 at z = 5.2 (Klamer et al. 2004). This indicates that the gas/dust ratios in high redshift CO samples could be be systematically underestimated due to bias in favour of dusty targets.

• Positive detections are more likely to be published than non-detections. Furthermore, the derived limits are often dubious in view of relatively large uncertainties in the targeted redshifts. These uncertainties are due to large offsets that are often present between the CO and optical redshifts, the unknown widths of the CO emission lines, and the limited bandwidth of correlators used for the detections.

• For higher redshift objects progressively higher CO transitions need to be observed consistent with the atmospheric transmission bands acessible to millimetre arrays. Hence, measurements of CO in will tend to trace denser gas in HzRGs than at lower redshifts. However for z > 3.6, the ground-state transition becomes observable with centimetre facilities, such as the VLA and ATCA, resulting in a discontinuity in the studies of multiple CO transitions needed to constrain the densities and temperatures of the gas.

Table 2: HzRGs detected in CO

Name	z	Transition	VCO	SCO V	M(H2)	Ref. 1
			km s-1	Jy km s-1	1010 M
53W002	2.390	(3-2)	420	1.2 ± 0.2	1.2	1,2
B3 J2330+3927	3.086	(4-3)	500	1.3 ± 0.3	7	3
TN J0121+1320	3.517	(4-3)	700	1.2 ± 0.4	3	4
6C 1909+72	3.537	(4-3)	530	1.6 ± 0.3	4.5	5
4C 60.06	3.791	(1-0)	...	0.15 ± 0.03a	13	6
		(4-3)	>1000	2.5 ± 0.4	8	5
4C 41.17	3.792	(1-0)	...	<0.1	...	7
		(4-3)	1000	1.8 ± 0.2	5.4	8
TN J0924-2201	5.197	(1-0)	300	0.09 ± 0.02	~10	9
		(5-4)	250	1.2 ± 0.3		9
a Only broad component; narrow component has SCO V = 0.09 ± 0.01 Jy km s-1.
1 References: 1 = Scoville et al. (1997), 2 = Alloin et al. (2000), 3 = De Breuck et al. (2003), 4 = De Breuck et al. (2003), 5 = Papadopoulos et al. (2000), 6 = Greve et al. (2004), 7 = Papadopoulos et al. (2005), 8 = De Breuck et al. (2005), 9 = Klamer et al. (2005).

The increased sensitivity of ALMA will also open the study of fainter molecular and atomic lines in HzRGs. Lines such as HCN or HCO⁺ probe at least an order of magnitude denser molecular gas than CO, and are therefore better tracers of the dense molecular cores in star-forming regions (Papadopoulos 2007).

3.5. Dust

Dust is an important constituent of HzRGs and an additional diagnostic of star formation. It is both a major constituent of the molecular clouds from which stars generally form and an indicator that substantial star formation has already occurred. The presence of dust means that chemically enriched material is present. Thermal re-radiation from dust often dominates the spectral energy distribution of HzRGs at millimetre and sub-millimetre wavelengths (Fig. 1) and dust is observable as a polarizing and absorbing medium in the optical and ultraviolet.

3.5.1 Millimeter and sub-millimeter emission

Since the detection of 4C 41.17 at z = 3.8 (Dunlop et al. 1994, Chini and Kruegel 1994, Ivison 1995) a large number of HzRGs have been observed at millimeter wavelengths (Archibald et al. 2001, Reuland et al. 2004). Reuland et al. (2004) analysed a sample of 69 radio galaxies with 1 < z < 5, detected at 850 and/or 450 µm. Isothermal fits to the submillimetre spectra give dust masses of a few × 10⁸ M at temperatures of ~ 50 deg (Archibald et al. 2001).

Possible sources of heating for the dust are X-rays from an AGN (Section 5) and UV photons from young stars. The typical submillimetre luminosity (and hence dust mass) of HzRGs strongly increases with redshift, with a (1 + z)³ dependence out to z ~ 4. This is evidence that star formation rates were higher and/or the quasars brighter in HzRGs with z > 3.

Disentangling whether the dust is heated by an AGN or young stars is difficult, because the two processes produce dust at partially overlapping temperatures. Evidence for heating by young stars comes from the slope of the Rayleigh-Jeans part of the thermal dust emission - observed at millimetre and submillimetre wavelengths. This indicates that the dust temperatures are relatively cool, ~ 50K, consistent with starburst-heated emission (e.g. Archibald et al. 2001, Stevens et al. 2003, Reuland et al. 2004). However, there is also evidence for a warmer(~ 300 K dust) component in radio galaxies, consistent with AGN heating (Rocca-Volmerange and Remazeilles 2005). Furthermore, recent Spitzer observations at 5 < _obs < 70 µm showed that the Wien tail part of the dust SED is inconsistent with low dust temperatures. Dust with temperatures of ~ 300 K or higher is required, consistent with AGN heating. Although the submillimetre luminosity of the HzRGs is uncorrelated with radio luminosity, suggesting no strong dependence on the strength of the quasar/AGN emission. (Reuland et al. 2004), the rest-frame 5 µm luminosity does appear to be correlated with radio luminosity (Seymour et al. 2007).

It is likely that young stars and AGN both play a role in heating the dust. Clarification of their relative importance await accurate measurement of the SED near the peak of the thermal dust emission by the Herschel Telescope and high spatial resolution observations of the radial profiles of the thermal dust emission with ALMA. Once properly isolated, starburst-heated dust emission will be a powerful tool for measuring the star-formation rates in radio galaxies. On the assumption that most of the submillimetre emission is heated by stars, derived star formation rates are up to a few thousand M yr^-1, consistent with those obtained from UV absorption line measurements (Section 4.2).

3.5.2 UV continuum polarisation

The rest-frame UV polarisation is an important probe of dust in the inner regions of HzRGs (di Serego Alighieri et al. 1989, di Serego Alighieri 1997, Cimatti et al. 1993, Cimatti et al. 1998, Vernet et al. 2001). The fractional polarisation provides a unique tool for determining the contribution and nature of of scattered light, while the polarisation angle allows the location of the scattering medium to be pinpointed. However, polarisation measurements of HzRGs require both high sensitivity and high precision.

Eight of 10 HzRGs studied by Vernet et al. (2001), show high continuum polarization just redward of Ly, with fractions, f_p ranging from 6% to 20% (e.g. Fig. 10). The shape of the polarised flux (= percentage polarisation × total intensity) is very similar to that of an unobscured quasar (Type 1 AGN) in both the slope of the continuum and the presence of broad emission lines (Vernet et al. 2001). This is strong evidence for the presence of an obscured quasar in the nuclei of HzRGs. The spectral energy distribution of a typical quasar is bluer than most other components that contribute to the SED of HzRGs between the UV and near-IR. The dilution by the other components is smaller in the rest-frame UV. At z > 2, this waveband is redshifted into the optical, where the most sensitive (spectro-)polarimeters exist on large telescopes, allowing detailed studies despite the cosmological brightness dimming.

Figure 10. Spectral and polarization properties of 0211-122 at z = 2.3 taken from observations with the Keck Telescope. [From Vernet et al. (2001)]. Plotted from top to bottom are: the observed total flux spectrum in units of 10-17 erg s-1 cm-2 Å-1 at two different scales, the first to show strong emission lines and the second to show the continuum, the crosses respectively indicate continuum and narrow emission lines (including their underlying continuum). Shaded areas indicate regions of bright sky emission.

The polarisation angle is in most cases closely perpendicular to the radio structure, indicating that the scattering occurs within the cones traced out by the radio jets (Vernet et al. 2001). The spatially resolved narrow emission lines (see Section 3.2) are generally not polarised (e.g. Fig. 10), indicating that the scattering medium must be located between the broad and narrow line regions.

The wavelength dependence of the polarised emission is consistent with both grey dust or electron scattering. However, electron scattering can be excluded as it is much less efficient and would imply masses of the scattering medium which are close to the total galaxy mass (Manzini and di Serego Alighieri 1996). A clumpy scattering dust medium can also produce the observed continuum slope. Using spectropolarimetric observations near the rest frame 2200 Å dust feature, Solórzano-Iñarrea et al. (2004) argue that the composition of the dust is similar to that of Galactic dust.

Cimatti et al. (1993) presented a dust-scattering model that explains the structure, the polarization properties and the spectral energy distribution of the ultraviolet aligned light with optically thin Mie scattering of quasar radiation emitted in a cone of ~ 90° opening angle. The required amount of spherically distributed dust is (1-3) × 10⁸ M, consistent with the estimate for the dust mass from the sub-millimeter data (Section 3.5.1). However, more recent predictions Vernet et al. (2001) indicate that the dust responsible for scattering the AGN emission should have a sub-millimetre emission that is an order of magnitude fainter than observed.

Not all HzRGs have significant polarisation. Two of eight objects studied by Vernet et al. (2001) had fractional polarisation f_p < 2.4%. There is no evidence that the nuclei of these HzRGs contain obscured quasars (Section 5.1). We discuss information derived from the UV polarisation observations about the obscured or dormant quasar in Section 5.1.