9.1. Star formation and ULIRGs at high redshift
The importance of the molecular medium in the history of the Universe is directly related to that of star formation. The gross features of the global star formation history are now reasonably well established. They are inferred not only from the analysis of the age of stellar populations in local galaxies, but, since about a decade, directly and more accurately from the observation and census of high-z star forming galaxies. The global star formation was much more important in the past, e.g. by about a factor of 10 more than presently at z ~ 1 (see e.g. Le Floc'h et al. 2005 and references therein). The majority of the stellar mass of the Universe was thus formed in progenitors of spiral galaxies between z = 1.5 and 0.5. Another important fraction is formed at higher redshift, including most of the stellar mass of elliptical galaxies. The exploration of the properties of the prominent molecular medium at these climax epochs of star formation in the Universe is fundamental to understand in detail the evolution of galaxies. This is mostly out of reach with current millimetre capabilities whose current limit of CO detection in a typical LIRG such as M 82 is z ~ 0.2 (however, see Combes et al. 2007). It is one of the main drivers of the ALMA project which should be able to detect LIRGs at very high redshift (Section 12).
As discussed in Section 5, the far-infrared (FIR) luminosity is both the most sensitive tracer of the star formation rate in major starbursts, and it is closely related to their molecular emission. The detection of the redshifted FIR emission of galaxies in atmospheric windows close to 1 mm, mainly 850 µm and 1.2 mm, presents the remarkable property of being practically independent of redshift in a very broad range, z ~ 0.5-5 (Blain & Longair 1993). For a fixed observing frequency, there is a strong increase of the luminosity with z at the corresponding restframe frequency because of the very steep emission spectrum, rest4.5 in the submm range. This almost exactly compensates for the luminosity distance factor Dl-2. Such an extraordinary advantage of the mm/submm window explains the remarkable success of mm/submm surveys of high-z ULIRGs first with SCUBA/JCMT and then with MAMBO/IRAM-30m (Coppin et al. 2006, Voss et al. 2006, and references therein) which have detected hundreds of high-z galaxies with LFIR sensitivity of a few 1012 L potentially up to z ~ 6.
The comoving density of `submillimetre galaxies' (SMGs) at z ~ 2 is more than two orders of magnitude larger than locally (e.g. Hughes et al. 1998, Greve et al. 2004, Coppin et al. 2006). They contribute significantly to the submillimetre background (see e.g. Smail 2006 and references therein). Their redshift distribution peaks at z ~ 2-3 (Chapman et al. 2003, 2005; Pope et al. 2006). The starburst origin of their far-infrared emission is confirmed from their radio emission, CO emission (next section) and the absence of strong X-ray emission in most of them (Alexander et al. 2005a, 2005c, 2006 and references therein). With typical star formation rates of a few hundred M yr-1, they contribute significantly to the global star formation rate at z 1 (Le Floc'h et al. 2005); and they are among the main contributors to star formation at z ~ 2-3. It is agreed that they are probably the progenitors of massive elliptical galaxies and that their strong starburst is a major episode of the star formation history of these galaxies. From the general correlation between molecular gas and FIR emission, one may infer that they should be the objects with the largest molecular masses in the Universe. This is indeed confirmed by the millimetre detection of CO in a number of the most luminous ones at z ~ 1-3 (next section).
9.2. CO studies
To date, millimetre CO emission has been detected in about 40 sources with redshift > 1, most of them with z > 2. Molecular gas emission at high redshift has been reviewed by Solomon & Vanden Bout (2005) (see also Cox et al. 2005). Therefore, here we only summarise the situation of this major topic for the discussion of molecules in galaxies, refering the reader to this review for details and a complete list of references.
After the early millimetre detection of dust and CO in exceptional high-z objects, mostly strongly lensed, (Brown & Vanden Bout 1991, 1992, Solomon, Radford & Downes 1992, Barvainis et al. 1992, 1994, McMahon et al. 1994), it was realized that CO could be detected with current equipment in the brightest high-z ULIRGs even without the help of gravitational amplification. After about ten years of major efforts with millimetre facilities, there are 36 CO detections with z between 1.06 and 6.42 reported in Table 1 of Solomon & Vanden Bout (2005) (this list is constantly increasing with new detections, e.g. Kneib et al. 2005, Iono et al. 2006b, Willott et al. 2007, Smail et al. in prep.). The detectability of CO at such prodigious distances may be explained in a way similar to the easy detection of dust emission (Section 9.1). At large redshifts, there is practically always at least one CO line in the 3 mm atmospheric window which is the most sensitive for such detections. For large redshifts, this line corresponds to a high value of the rotation number J, with a rotational line luminosity which strongly increases with J and hence with z, and again almost compensates for the effect of the large distance. The gas is dense and warm enough in these starbursts so that the CO energy distribution peaks in line 4-3 or higher.
The list of CO detections quoted in Appendix 1 of Solomon & Vanden Bout (2005) includes 14 SMGs (and one Lyman Break Galaxy) with z = 1.06-3.41 (see e.g. Neri et al. 2003, Greve et al. 2005, Tacconi et al. 2006), 16 QSOs with z = 1.42-6.42 (see e.g. Omont et al. 1996a, Ohta et al. 1996, Walter et al. 2003, Bertoldi et al. 2003) and five radio galaxies with z = 2.39-5.20 (e.g. Papadopoulos et al. 2000, De Breuck et al. 2005, Klamer et al. 2005). Note that all the 11 objects with z > 3.5 are prominent AGN. About half of the sources are strongly amplified by gravitational lensing with magnification in the range ~ 2-20 (it reaches even 45 in a new detection by Kneib et al. 2005 displayed in Fig. 6). Hovever, it is also certain that another large fraction has no significant magnification. The first CO detection in more than 80% of these sources was achieved with the IRAM-PdB Interferometer, and most of the remaining ones with the OVRO interferometer (see Table 1 of Cox et al. 2005).
In practically all cases with z > 2, at least one line was detected in the 3 mm atmospheric window, corresponding to relatively high rotation number, mostly J = 3-2, 4-3 and 5-4 (in a few cases up to J = 7-6). In some cases, more than a single millimetre CO line was detected, with different J values (see e.g. Fig. 5), including sometimes detections in the 2 mm window with the IRAM 30m-telescope (e.g. Weiss et al. in prep.). Low-J lines, J = 1-0 or 2-1, were detected in the cm range in ten cases, mostly with VLA (see e.g. Fig. 5) (see also one direct detection at z = 5.19 with ATCA by Klamer et al. 2005, and detections with the NRAO Green Bank Telescope (GBT) and the MPIfR Effelsberg 100 m telescope by Riechers et al. 2006c). The value of the ratio of the intensities may provide information about the CO rotational temperature, and, hence, the kinetic temperature and the H2 density. However, an accurate modelling of CO line formation remains difficult in the absence of detailed information about the extension and the structure of CO emission, and the clumpiness of the molecular gas (see e.g. Combes, Maoli & Omont 1999, Solomon & Vanden Bout 2005). This limits the corresponding diagnosis that it could provide for the emitting galaxy. However, systematic programmes such as the one recently carried out on SCUBA-MAMBO SMGs at IRAM-interferometer (Neri et al. 2003, Greve et al. 2005, Tacconi et al. 2006), are efficient in providing basic information about the properties of such starburst galaxies at high redshift. The emerging picture is that they share many features with local ULIRGs, and in particular the CO emission is generally concentrated within the central kpc or so. However, starbursts where CO is currently detectable, are more extreme than typical local ULIRGs, with larger FIR and CO luminosities, higher dust temperatures and larger H2 masses. Very recent observations have shown that some SMGs with velocity spread ~ 1000 km/s, are unresolved with the ~ 0.2" resolution now available with the IRAM interferometer (Tacconi et al. in prep.). This indicates that the 1000 km/s spread occurs within a radius less than 1 kpc, and implies masses of 1010.5 M or greater are enclosed in this volume. Many cases of powerful SMGs display evidence of interaction with neighbouring sources. There are a number of cases of resolved, often double, CO sources, among SMGs (see Section 3.4.2 of Solomon & Vanden Bout 2005, Genzel et al. 2003 and Tacconi et al. 2006), radio galaxies (Papadopoulos et al. 2000, De Breuck et al. 2005, Greve et al. 2006a, and references therein) and prominent QSOs, especially the spectacular double CO source at z = 6.42 (Walter et al. 2004) (Section 9.4).
Even detections of a single CO line may provide interesting information about basic properties of these high-z starburst galaxies. The general H2 / ICO relation, properly scaled for such objects (e.g. Solomon & Vanden Bout 2005), provides an estimate of the mass of the H2 gas, which may give an indication about the future duration of the starburst. Assuming some standard spatial extension for such a nuclear starburst, similar to the few cases where the sources have been resolved, one may infer an estimate of its total dynamical mass from the velocity width of the CO line, but it depends on the unknown inclination angle. The dozen of SMGs detected in the large programme with the IRAM interferometer have CO luminosities in excess of 1010 K kms-1 pc2, and thus MH2 of the order of a few 1010 M, and dynamical masses in the range of 1011 M. This implies very massive systems dominated by baryons in their central regions.
Most of the sources where CO has been detected have FIR luminosities LFIR ~ 1013 L. Compared to the correlation between CO and FIR luminosities in local ULIRGs (Section 5), there is a trend of larger values for the ratio of the FIR luminosity to the CO luminosity. It thus appears that the star formation rate per unit mass of molecular gas was higher in these massive high z starbursts than in the less luminous local ULIRGs.
Figure 5. Spectra of redshifted CO and CII lines of the QSO J1148+5251 (zCO/CII = 6.42): CO(6-5) (93.2 GHz) and CO(7-6) (108.7 GHz) from IRAM interferometer (Bertoldi et al. 2003); CO(3-2 (46.6 GHz) from VLA (Walter et al. 2003); and CII (2P3/2 - 2P1/2) (256.2,GHz) from IRAM 30m-telescope (Maiolino et al. 2005). (Note that the spectrum of CO(6-5) is reproduced twice to allow an easier comparison with the other spectra of CO and CII lines).
9.3. Molecules in the host galaxies of high-z AGN
The fact that more than half the high-z galaxies where CO has been detected, are AGN, mainly bright QSOs and powerful radio galaxies, is probably due to the combination of two reasons. First, there is a high probability of finding strong starbursts in the host galaxies of such AGN as proven by the many detections of dust submm emission: MAMBO-IRAM studies have shown that the probability to detect 1.2 mm continuum with a flux density 2 mJy is as high as 25-30% around bright QSOs with z 2 (Omont et al. 1996b, Carilli et al. 2001, Omont et al. 2001, 2003, Beelen 2004). The probability is even higher in prominent high-z radio galaxies observed with SCUBA-JCMT by Archibald et al. (2001). Furthermore, such objects are very good cases for searching CO emission for two main reasons: there is a general good correlation between CO and FIR emission, and the FIR emission is strong in many of these AGN. It is also much easier to have a good determination of the redshift of the molecular gas of AGN, so that it is accurate enough to warrant that the CO line lies within the narrow bandwidth of millimetre detectors used up to now (note that the bandwidth is significantly increased to more than 5000 km/s with the current new generation of receivers, and soon to larger values as already 25000 km/s at Mopra-22m). This explains in particular the absence of non-AGN sources among current CO detections with z > 3.5. Practically no SMGs are known with such redshift, probably partly because they are rarer and mostly because their radio detection needed for redshift determination is too difficult (see e.g. Chapman et al. 2003, 2005).
Compared to CO-detected SMGs without strong AGN (Section 9.2), the properties of the molecular emission of these AGN host galaxies are not fundamentally different. They have similar H2 masses, a few 1010 M, total virial masses, and CO linewidths ~ 200-800 km/s, with QSOs rather at the lower end of this width range and radio galaxies at the upper end (e.g. Carilli & Wang 2006). They also display a high fraction of interacting objects, especially for radio galaxies (Papadopoulos et al. 2000, De Breuck et al. 2005). There is nevertheless some indication that their temperature might be somewhat higher than regular SMGs without strong AGN, as suggested by the dust temperatures induced from 350 µm emission (Bendford et al. 1999, Beelen et al. 2006).
This class of high-z CO AGN includes the brightest high-z AGN known, especially strongly lensed ones, and the QSO with the largest redshift known, discussed in Section 9.4. These AGN also include most of the few cases where other lines than CO (HCN, HCO+, HNC, CN, CI, C+) have been detected up to now.
9.4. Other species and detailed studies through strong gravitational lensing
9.4.1. Detectability of other millimetre lines All other emission lines from the bulk of the interstellar gas of these high-z starbursts are significantly more difficult to detect than CO. There are already six species outside of CO currently detected - HCN, HCO+, HNC and CN and the fine structure lines of CI and CII - all in a very few sources, from one to five (H2O was also searched in several sources, without success up to now, e.g. Riechers et al. 2006b, Wagg et al. 2006). Most of the detections have been achieved thanks to strong gravitational amplification in the few most exceptional lensed sources discussed in Section 9.4.2. Solomon & Vanden Bout (2005) have devoted a detailed discussion to detections of HCN, CI and CII which were known at the time, as well as to very exceptional sources. We summarise this below with updates.
HCN. As discussed in Section 5, HCN emission is a very good tracer of dense gas and star formation, including major starbursts such as those of high-z Emission Line Galaxies. However, because of its large electric dipole, high-J rotational levels are difficult to excite, so that HCN detection is easier in practice in the J = 1-0 line redshifted into the cm range (however see 3 mm detections in Wagg et al. 2005 and Guélin et al. 2007). There is thus no possibility of compensating the effect of high-z large distances by observing stronger higher-J lines as for CO. Waiting for Extended VLA (EVLA) (and SKA in the long term ), the detection of high-z HCN(1-0) is just at the limit of current facilities (VLA, GBT) for the strongest sources. Four detections have been reported (F10214, Cloverleaf, APM 08279, VCV J1409) together with a few upper limits (see Section 9.4.2; Solomon & Vanden Bout 2005, Carilli et al. 2005 and Greve et al. 2006b). In all cases, the ratios of the FIR and HCN luminosities are within the scatter of the relationship between HCN and far-IR emission for low-z star-forming galaxies, although they have a trend to be larger than the average value of this ratio at low redshift (Carilli et al. 2005).
HCO+ emission is a star formation indicator similar to HCN, tracing dense molecular gas (Section 5). It has been recently detected in the Cloverleaf and APM 08279+5255 (Riechers et al. 2006a, García-Burillo et al. 2006b). HCO+ and HCN have similar luminosities, and there is evidence that they come roughly from the same circumnuclear region.
HNC and CN. The J = 5-4 line of HNC has been detected, and the N = 4-3 line of CN tentatively detected in APM 08279+5255 by Guélin et al. (2007, see also Riechers et al. 2006d). Both lines intensities are about half that of HCN J = 5-4, so that the [HNC]/[HCN] abundance ratio seems similar to its value in the cold Galactic clouds and much larger than in the hot molecular gas associated with Galactic HII regions.
[CI]. Atomic carbon may offer an interesting diagnosis of the molecular gas, especially relatively cold, with its two submm ground state lines and their simple excitation pattern (see e.g. Bayet et al. 2006 for a comparison of CI and CO in nearby galaxies). High redshifts often bring these lines in a much easier atmospheric frequency range than at z = 0. The relatively small intensity ratio, ~ 1.5-5, between adjacent CO lines and CI lines, makes the latter detectable in a significant fraction of the high-z sources where CO is detected. CI has been detected up to now in five high-z sources: F10214, Cloverleaf, SMM J14011, PSS J2322 and APM 08279 (including one detection of the higher excitation line 3P2 - 3P1). The luminosity ratio between the CI and CO lines is a potential diagnostic of the gas excitation and carbon chemistry.
[CII]. The 2P3/2 2P1/2 fine-structure line of C+ at 157.74 µm (1900.54 GHz) is known as the most powerful emission line of the interstellar gas of galaxies. It traces in particular photo-dissociation regions associated with star formation, and is thus an important potential tool for studying the corresponding molecular gas. In local galaxies, with far-infrared luminosities LFIR ~ 1010-1011 L, the ratio of the C+ luminosity to the far-infrared luminosity, L[CII] / LFIR, is typically a few 10-3. However, for ULIRGs with LFIR 1012 L, this ratio drops by about an order of magnitude (see e.g. Fig. 2 of Maiolino et al. 2005). The search for C+ emission at high z has for long suffered from various handicaps: the absence of any `inverse K-correction' compensating the distance factor; the lack of sensitivity in the submm range where the line is redshifted at z 6; the lack of known adequate sources at z 6.4 where the line enters the 1.2 mm band of current sensitive equipment. Therefore, the repeated efforts to detect the CII line at high z remained unsuccessful for more than ten years until two sources were detected very recently: J1148+5251 at z = 6.42 with the IRAM 30m-telescope (Maiolino et al. 2005; see Fig. 5) and with more details with the IRAM interferometer (Walter et al. in prep.); BR1202-0725N at z = 4.69 with the SMA (Iono et al. 2006a). The ratio L[CII] / LFIR is ~ 2-4 × 10-4, i.e. comparable to the value observed in the most luminous local ULIRGs such as Arp 220. These results are important because they confirm that, with the gain in sensitivity of at least two orders of magnitudes with ALMA, the CII line will be easily detectable in all high-z ULIRGs and even LIRGs in the redshift ranges corresponding to the atmospheric submm windows.
9.4.2. Strongly lensed and other prominent sources Among the ~ 40 high-z sources where CO detection has been reported, some deserve a special mention either because they have extraordinarily large amplification by gravitational lensing allowing early detection and detailed studies, or they have the largest redshifts or luminosities, or they are representative of various special types. The most prominent are:
IRAS FSC10214+4724 (z = 2.286). It was discovered by Rowan-Robinson et al. (1991) as an extraordinarily bright high-redshift IR source among IRAS data. The detection of CO J = 3-2 emission (Brown & Vanden Bout 1991; Solomon, Radford & Downes 1992) was the first detection of molecular gas at high redshift. It was later found to be gravitationally lensed by a factor ~ 10 (FIR and CO) to ~ 50 (mid-IR). The correction for magnification reduces its properties to those typical of local ULIRGs. It is still the best studied ultraluminous infrared galaxy at high redshift. It contains both a dust-enshrouded quasar obscured by Compton-thick material, responsible for the mid-IR luminosity (Alexander et al. 2005b), and a much larger molecular ring starburst responsible for a substantial fraction of the FIR luminosity. The high magnification has allowed the angular resolution of the CO emission and the detection of HCN and CI.
The Cloverleaf (HH1413+1143, z = 2.558). This broad absorption line QSO is a lensed object with spectacular four bright image components. Barvainis et al. (1992) discovered strong FIR and submillimeter radiation similar to that of IRAS F10214, showing that bright optical high-z quasars may also be extremely FIR (and mid-IR) luminous. The CO lines are stronger than in any other high-z source (Barvainis et al. 1994), owing to both powerful intrinsic line luminosities and magnification. As a result, this source is the best studied in detail both for the number of CO lines detected (Barvainis et al. 1997, Weiss et al. 2003), and for the angular resolution (Yun et al. 1997, Alloin et al. 1997, Kneib et al. 1998, Venturini & Solomon 2003, Weiss et al. 2003 and references therein). It is also one of the few high-z sources where HCN, HCO+ and CI lines have all been detected.
APM 08279+5255 (z = 3.911). This extremely bright broad absorption line quasar has both one of the highest magnifications (µ ~ 100 at optical-IR wavelengths but only ~ 7 for CO emission), and one of the largest intrinsic FIR and CO luminosities. This has again allowed multi-line CO studies, including strong J = 9-8 emission from ~ 200 K hot gas of sub-kiloparsec size (Downes et al. 1999), and even the J = 11-10 line (Weiss et al. in preparation), as well as angular resolution of the central nuclear emission (Lewis et al. 2002, Papadopoulos et al. 2001). Both HCN J = 5-4 and HCO+ J = 5-4 line emission from the dense molecular gas has been reported (Wagg et al. 2005; García-Burillo et al. 2006b), as well as CI (Wagg et al. 2006), HNC and CN (Guélin et al. 2007).
BR 1202-0725 (z = 4.69). This prominent quasar is one of the brightest sources at 1.2 mm (McMahon et al. 1994), and also the third one which was discovered in CO emission (Omont et al. 1996a, Ohta et al. 1996). Both CO and 1.2 mm emission are split in two components, 4" apart, of comparable strengths; each of them being among the strongest high-z sources, with apparent (if non lensed) LFIR 1013 L (but modest combined mid-IR luminosity, Hines et al. 2006). There is an enormous difference in optical/near-IR extinction between the QSO component and the northern source which is not optically detected (even in the near-IR K-band at its centre). Radio emission has also been detected (both components), as well as recently (Iono et al. 2006a) X-ray (mostly QSO component) and CII line emission (northern component). The origin of the double structure, lens or two interacting prominent starburst galaxies, remains a puzzle; see Solomon & Vanden Bout (2005), Carilli et al. (2002), Iono et al. (2006a), Sameshima (2006) for arguments pro and con each possibility. However, two galaxies seem more probable.
SDSS J1148+5251 (z = 6.42). This is the most distant quasar known to date (Fan et al. 2003). Its multi-wavelength detection (see references in Solomon & Vanden Bout 2005, Hines et al. 2006, Charmandaris et al. 2004, Beelen et al. 2006), including CO(3-2) with VLA (Walter et al. 2003) which proves to be resolved (Walter et al. 2004), and CO(6-5) and CO(7-6) with IRAM interferometer (Bertoldi et al 2003) (see Fig. 5), are typical of other high-z strong starburst QSOs. This shows the presence of a giant starburst and large amount of molecular gas, with heavy elements, less than one billion years after the Big Bang. It is also the first high-z source where the CII line emission was detected (Maiolino et al. 2005).
Figure 6 presents the results of CO observations from the IRAM interferometer of the triple image of another spectacular lensing case, the submillimetre galaxy SMM J16359+6612 lying at z = 2.516 behind the core of the massive cluster A 2218, which produces a magnification of ~ 45 of an intrinsically weak SMG (Kneib et al. 2005).
Figure 6. (Reproduced from Figure 1 of Kneib et al. 2005). Observational results at ~ 3 mm from the IRAM interferometer of a faint submillimetre galaxy, SMM J16359+6612 (SMM1), lying at z = 2.516 behind the core of the massive cluster A 2218. The foreground gravitational lens produces three images with a total magnification of 45. (Left) The 12CO(3-2) map of SMM1 superposed on the optical HST image of A 2218. Three millimetre images SMM1-A, B and C are clearly identified, all three well centered on their optical counterparts. (Right) This panel displays (from top to bottom) the velocity profile of the three different submm images SMM1-A, B and C, and of the sum of these three spectra. A double-peak profile is clearly observed for SMM1-B and -C, as well as in the sum of all three components (bottom panel). The two components are separated by ~ 280 km/s and agree well with two features ( and ) also identified in optical spectroscopy.