"We have traced the broad outline of the development in regions of exceptional fruitfulness of the first settled village communities into more populous towns and cities" "The city community arose around the altar of the seed-time blood sacrifice." The Outline of History, H. G. Wells, 1920
Within standard Cold Dark Matter (CDM) scenarios the first stars and stellar systems should form through gravitational infall of primordial gas in large CDM halos (e.g. White and Rees 1978). Simulations suggest that these halos merge and form web-like networks traced by young galaxies and re-ionized gas (e.g. Baugh et al. 1998, De Lucia and Blaizot 2007). The most massive galaxies, and the richest clusters emerge from regions with the largest overdensities.
7.1. Finding protoclusters using HzRGs
To investigate the emergence of large-scale structure in the Universe, it is important to find and study the most distant rich clusters of galaxies. Conventional methods for finding distant galaxy clusters usually rely on the detection of hot cluster gas using X-ray techniques. These are limited by the sensitivity of X-ray telescopes to 1.3, i.e. much smaller than the redshifts of HzRGs.
Since HzRGs are amongst the most massive galaxies in the early Universe, they are likely to inhabit regions that are conducive to the formation of rich galaxy clusters. During the mid-nineties observational evidence emerged that HzRGs are located in dense cluster-type environments. First, the measurement of large radio rotation measures (RM) around some HzRGs, indicated that the host galaxies are surrounded by a hot magnetized cluster gas (Carilli et al. 1997, Athreya et al. 1998) (Section 3.1). Secondly, an excess of "companion galaxies" was found near HzRGs (Röttgering et al. 1996, Pascarelle et al. 1996, Le Fevre et al. 1996, Knopp and Chambers 1997). These results together with the indications that HzRGs hosts are forming dominant cluster galaxies (Section 6) prompted the initiation of direct searches for galaxy clusters in the vicinity of HzRGs.
Conventionally the term "galaxy cluster" refers to a bound structure of several hundred galaxies. At z ~ 3, the Universe is only about 2 Gyr old, i.e. insufficient time for a galaxy with a velocity of a few hundred km/s to have crossed cluster-scale structures. Hence any overdense structure of galaxies observed at high redshifts must still be forming and cannot be virialised and bound. Following (Overzier 2006), we shall use the term "protocluster" for an overdense structure in the early Universe (z > 2), whose properties are consistent with it being the ancestor of a local bound galaxy cluster.
During the last few years there have been several successful direct searches for protoclusters around HzRGs, with 8 to 10m-class optical/ infrared telescopes. One search technique is to use narrow-band imaging to detect emission line objects at redshifted wavelengths corresponding to those of the target HzRGs. A second technique is to carry out broad-band imaging with colours chosen to detect "dropout objects" due to redshifted features in the continuum spectral energy distributions. The protocluster candidate galaxies detected in these imaging experiments were generally followed up by multi-object spectroscopy to confirm their redshifts and their membership of the protoclusters. The most important emission line for establishing the redshifts of the protocluster galaxies is Ly (Pentericci et al. 1997, Kurk et al. 2000, Venemans et al. 2002, 2004, Croft et al. 2005, Venemans et al. 2005, 2007). Other relevant emission lines for such searches for z > 3 are H (Kurk et al. 2004, Kurk et al. 2004) or [OIII] 5007. There are two relevant features in the continuum spectra of galaxies that are exploited for such protocluster searches - the Lyman break around the 912Å Lyman continuum discontinuity (Miley et al. 2004, Overzier et al. 2006, Overzier et al. 2006, Intema et al. 2006) and the 4000Å break or Balmer break close to 3648Å (Kurk et al. 2004, Kajisawa et al. 2006, Kodama et al. 2007). The various galaxy detection techniques are complementary, because they tend to select stellar populations with different ages. For example, Ly excess galaxies and Lyman break galaxies have young on-going star-forming populations, while the Balmer technique is sensitive to stellar populations that are older than a few × 108 y.
The measured overdensities of Ly emitters in the radio-selected protoclusters are factors of 3 - 5 larger than the field density of Ly emitters at similar redshifts (Venemans et al. 2005, 2007). A recent study with the HST of 4000Å break objects in MRC 1138-262 at z = 2.2, the protocluster that contains the Spiderweb Galaxy (Zirm et al. 2007, submitted to Ap.J) shows that the overdensity of red galaxies is 6.2, compared with non-protocluster fields. The photometric redshifts of galaxies in this field shows a significant "redshift spike" for 2 < z < 2.5. Although such observations at z > 5 are sensitivity limited, a significant overdensity of both Ly emitters (Venemans et al. 2004) and Lyman break galaxies (Overzier et al. 2006) has been established around TN J0924-2201 at z = 5.2, the HzRG with the highest redshift known to date.
Possible protoclusters have also been detected in non-targeted optical surveys, with several overdense regions found at large redshifts in the form of "redshift spikes" or "filaments" (e.g. Steidel et al. 1998, Keel et al. 1999, Steidel et al. 2000, Möller and Fynbo 2001, Shimasaku et al. 2003, Hayashino et al. 2004, Matsuda et al. 2005, Steidel et al. 2005, Ouchi et al. 2005). Although caution must be exercised in deriving overdensities from the redshift distributions alone, because the effects of peculiar galaxy velocities can influence the apparent clumpiness (Monaco et al. 2005). However, the observed structures around the HzRGs have two additional ingredient expected of ancestors of rich clusters, namely (i) the presence of the HzRG hosts with properties expected of progenitors of dominant cluster galaxies and (ii) sharp peaks in the redshift distributions of Ly emitters that lie close to the predetermined redshift of the radio galaxy. Taken together, the measured galaxy overdensities combined with the presence of HzRGs is strong evidence that the galaxy structures around HzRGs are indeed the ancestors of rich local clusters.
For some of the HzRG-selected protoclusters there is additional evidence for a dense environment from excess counts at millimeter (Ivison et al. 2000, Smail et al. 2003, Stevens et al. 2003, De Breuck et al. 2004, Greve et al. 2007) and X-ray (Pentericci et al. 2002, Overzier et al. 2005) wavelengths. Also, the X-ray counts indicate an enhanced AGN fraction in protoclusters compared to the field.
Figure 17. HzRG-selected protocluster around TN J1338-1942 at z = 4.1. [From Venemans et al. (2002, 2007)]. Shown are the spatial distribution of the spectroscopically confirmed z = 4.1 Ly emitters (circles and diamonds), additional Ly excess candidates from narrow-band imaging (stars) and the radio galaxy (square). The circles represent emitters with redshift smaller than the median redshift and the diamonds with those larger than the median. The size of the circles is scaled according to the relative velocity of the object. Larger circles and diamonds represent a larger relative velocity compared with the median velocity. The structure appears to be bound in the northwest of the image and unbound in the south.
Figure 18. HzRG-selected protocluster around TN J1338-1942 at z = 4.1 as in Figure 17. [From Venemans et al. (2002, 2007)]. Velocity distribution of the confirmed Ly emitters. The peak velocity of the radio galaxy is indicated by an arrow. The solid line represents the selection function of the narrow-band filter normalized to the total number of confirmed emitters. Note that the velocity distribution of the detected emitters is substantially narrower than the filter width and centered within 200 km s-1 of the redshift of the radio galaxy. Similar data are available for 6 protoclusters.
Figure 19. Morphologies of galaxies in the protocluster TN J1338-1942 at z = 4.1 with the ACS on the Hubble Space Telescope [From Overzier et al. (2006)]. The filter is i775 and each image measures 3'' × 3'', corresponding to ~ 20 kpc × 20 kpc at z ~ 4. Left are images for a selection of the overdense g475-dropout objects that are candidate Lyman break galaxies (LBGs) in the protocluster. Right are images of 12 spectroscopically confirmed Ly excesses (LAE) protocluster members. There are a wide range of morphologies present, with some galaxies showing clear evidence for small-scale interactions.
7.2. Properties of radio-selected protoclusters
The largest and most comprehensive HzRG protocluster search project was based on a study of 8 radio galaxies, with redshifts ranging from 2.2 to 5.2, using the narrow-band Ly search technique with the VLT (Pentericci et al. 2000, Kurk et al. 2000, Venemans et al. 2002, 2004, Croft et al. 2005, Venemans et al. 2005, 2007). Followup observations have been made of candidates in several of the targets using a variety of ground-based telescopes as well as the Hubble and Spitzer Telescope (Kurk et al. 2004, Miley et al. 2004, Overzier et al. 2005, 2006, Intema et al. 2006, Kodama et al. 2007). Typically, ~ 3 Mpc-scale regions around the HzRGs were covered in the Ly searches. Ly redshifts were determined for 168 objects. Almost all these objects are star-forming galaxies at similar redshift to the HzRG, with typical star formation rates of a few M per year, derived from the UV continuum and the Ly luminosities.
Six of the 8 fields were found to be overdense in Lya emitters by a factor 3 - 5 as compared to the field density of Ly emitters at similar redshifts. The 6 included all of the most radio-luminous objects (i.e. all HRzGs with radio luminosities L2.7 GHz > 6 × 1033 erg s-1 Hz-1 sr-1) (Venemans et al. 2007). Because these targets were chosen arbitrarily, (redshifts that place Ly in an available narrow-band filter), it is possible that all HzRGs with such large radio luminosities are embedded in such overdensities.
Measured velocity dispersions are in the range ~ 300 - 1000 km s-1, centered within a few hundred km/s of the mean velocity of the radio galaxies. Taking account of the relatively narrow peaks in the velocity distributions of the Ly emitters compared with the widths of the imaging filters, the measured galaxy overdensities range from ~ 5 - 15. Estimates for the sizes of the protocluster structures are limited by the typical size of the imaging fields used in the searches to date (~ 8' equivalent to ~ 3 Mpc). Nevertheless, the protocluster sizes are estimated to be in the range ~ 2 to 5 Mpc (Intema et al. 2006, Venemans et al. 2007).
Because cluster-size overdense structures at high-redshifts cannot be old enough to have become bound, the usual method of calculating cluster masses using the virial theorem cannot be applied. However, estimates for the masses of the protoclusters can be obtained from the volume occupied by the overdensity, the mean density of the Universe at the redshift of the protocluster, the measured galaxy overdensity, and the bias parameter (e.g. Steidel et al. 1998, Venemans et al. 2007). The masses obtained (a few times 1014 M - 1015 M, Kurk et al. 2004, Venemans et al. 2005) are comparable to the masses of local rich clusters.
It is interesting to inquire what is the relation of the protocluster structures to the general large-scale structure of the Universe. Is the topology of the protocluster filamentary, or do HzRG-selected protoclusters illuminate the densest most tangled regions of the cosmic web. For TN J1338-1942 at redshift z = 4.1, there are indications that the Mpc-sized protocluster of Ly excess galaxies may be part of a larger-scale structure. A 25' × 25' survey for Lyman break galaxy candidates (B-band dropouts) showed several significant density enhancements amidst large voids (Intema et al. 2006) and (Overzier et al. 2006) showed that this large scale structure ties in closely with significant sub-clustering across the 3' × 3' ACS field near the radio galaxy.
Future wide-field narrow-band imaging and spectroscopy around HzRG-selected clusters should allow the topology of the Universe in the region of protoclusters to be mapped in detail, providing a glimpse of large-scale structure emerging in the early Universe for comparison with computer simulations.
7.3. Are radio-selected protoclusters typical?
We have seen that protocluster-like structures have been found around almost all of the most luminous radio galaxies with z > 2 that have been targeted. However only ~ 50% of powerful radio sources at z ~ 0.5 are located in rich clusters and radio sources appear to avoid clusters at low redshift (Hill and Lilly 1991). What is known about the environment of radio galaxies having redshifts 1 < z < 2? This range is difficult to study spectroscopically because of the redshift desert (Section 1.3). However, there have been several reports of detections of clusters and/ or excesses of red galaxies around radio AGN in this redshift range (Hall et al. 1998, Kodama and Bower 2003, Nakata et al. 2002, Best et al. 2003). We note that there is a strong correlation between the radio luminosities and redshift of objects in such studies due to Malmquist bias. Taking all the data together, there appears to be a substantial increase in the density of the environment around radio galaxies as a function of redshift and/or luminosity.
The statistics of the luminosity function of radio galaxies are consistent with every brightest cluster galaxy (BCG) having gone through a luminous radio phase during its evolution (Venemans et al. 2002, 2007). This statement is based on the facts (i) that the space density of luminous steep-spectrum radio sources decreases by ~ 100 between 2.5 z > 0 (Section 1.4) and (ii) that the radio synchrotron lifetimes (few × 107 yr) are ~ 100 smaller than the cosmological time interval corresponding to the observed redshift range (few × 109 yr). Distant radio galaxies may therefore be typical progenitors of galaxies that dominate the cores of local clusters. Likewise, the ancestor of every rich cluster in the local Universe may have gone through a phase in which it hosted a HzRG. This would imply that HzRG-selected protoclusters are typical ancestors of local galaxy clusters.
7.4. Protocluster evolution
Radio-selected protoclusters are powerful laboratories for tracing the emergence of large scale structure and for studying the evolution of galaxies in dense cluster environments. An interesting parameter that can be measured for line-emitting galaxies is the velocity dispersion of the protoclusters. Although more statistics are needed, the velocity dispersion appears to decrease with increasing redshift (Venemans et al. 2007), consistent with the predictions from simulations of forming massive clusters (e.g. Eke et al. 1996).
A population study of the protocluster around the Spiderweb Galaxy PKS 1138-262 at z = 2.2 was carried out by Kurk et al. (2004), using deep optical and infrared observations with the VLT. Besides Ly emitters, the study included objects having apparent H excesses and extremely red objects, with colours characteristic of old galaxies with 4000 Å breaks at the redshift of the protocluster. An intriguing result of this study is that candidate H emitters and 4000 Å break objects appear more concentrated towards the centre of the protocluster than the Ly emitters. This indicates that the galaxies that are dominated by old stellar populations are more settled into the gravitational potential well of the protocluster, consistent with them being older.
An important diagnostic of galaxy and cluster evolution is the cluster colour-magnitude diagram (CMD). The presence of a narrow "red sequence" in the cluster CMD is well established out to a redshift of 1.4 (Stanford et al. 2005). The red colours imply that the galaxies are dominated by older stars, whose SED peaks redwards of the 4000 Å break. The presence of this red sequence can be used to set a lower limit to the redshift at which the stellar populations formed. It is extremely difficult to measure the CMD at z > 1.4, even with the largest ground-based telescopes.
However, indications of an emergent red sequence has been found in the CMD of the protocluster MRC 1138-262 at z = 2.2, both with a wide-field near-IR imager on the Subaru Telescope (Kodama et al. 2007); Fig. 20 and with the much narrow field but more sensitive NICMOS on the HST (Zirm et al. 2007). The galaxy colours indicate that while some relatively quiescent galaxies exist in the protocluster, most of the galaxies are still undergoing star formation Furthermore, Kodama et al. (2007) studied 3 other radio-selected protoclusters and found that the fraction of red galaxies in the 4 protoclusters increases between z ~ 3 and z ~ 2. To summarise, there is strong evidence that most protocluster galaxies undergo substantial star formation between z ~ 3 and z ~ 2 and that the bright end of the red sequence is still being formed during this epoch.
Figure 20. Color-magnitude diagram of J - KS plotted against KS for the MRC 1138-262 protocluster at z = 2.2. [From Kodama et al. (2007)]. Filled circles indicate protocluster member candidates selected using colour selection criteria. Large stars mark the targeted radio galaxies. Dotted error bars show 1s photometric errors. Solid, dashed and dotted lines show the expected location of the CMD at the relevant redshift for passive evolution. The iso-stellar mass lines of 1011 M and 1010 M are shown by thick dotted lines. The dot-dashed lines indicate 5 (KS) and 2 (J) detection limits.
Information about the early evolution of stellar populations in clusters has been obtained at even higher redshifts. In the protocluster surrounding the HzRG TN J1338-1942 at z = 4.1, the distribution of Ly emitters (LAEs) is highly filamentary and appears to avoid the locations of Lyman break galaxy candidates (Overzier et al. 2006). A similar spatial segregation between LAEs and LBGs was observed in a structure around QSO SDSS J0211-0009 at z = 4.87 (Kashikawa et al. 2007). This indicates that an age- or mass-density relation was emerging little more than 1 Gyr after the Big Bang, when the Universe was only 10% of its present age.
We note that derived masses of z > 2 LBGs and LAEs are 10 smaller than the masses of early-type galaxies in local clusters, indicating that a large fraction of the stellar mass still has to accumulate through merging (Overzier et al. 2006). Detailed observations of protocluster regions on much larger scales (~ 50 co-moving Mpc) are needed to test if the number density of LBGs is indeed consistent with forming the cluster red sequence population through merging. Simulations show that clusters with masses of > 1014 M can be traced back to regions at z = 4 - 5 of 20 - 40 Mpc in size, and that these regions are associated with overdensities of typical dark matter halos hosting LAEs and LBGs of g ~ 3 and mass overdensities m in the range 0.2 - 0.6 (Suwa et al. 2006). Also, recent numerical simulations of CDM growth predict that quasars at z ~ 6 may lie in the center of very massive dark matter halos of ~ 4 × 1012 M (Springel et al. 2005, Li et al. 2007). They are surrounded by many fainter galaxies, that will evolve into massive clusters of ~ 4 × 1015 M at z = 0.
The addition of the new IR - oprical camera WFC3 to the HST and the advent of several new wide-field imagers and multi-object spectrographs on ground-based telescopes will allow the detailed evolution of protoclusters to be studied in great detail during the next few years.