3.1. Submm continuum surveys
The spectral energy distribution (SED) of galaxies over the radio, mm and FIR domains has a characteristic maximum around 60-100 µm due to dust heated by newly born stars, and the interstellar radiation field (see Fig. 3). This maximum depends on the dust temperature, and the curve is that of a grey-body, where the optical thickness can be modelled by a power-law in frequency, , where ~ 1.5 - 2, according to the nature of dust. In the Rayleigh-Jeans domain (frequencies lower than the maximum), the flux increases almost as 4, and this creates what is called a negative K-correction, i.e. it begins to be more easy to detect objects at high redshift than low redshift, at a given frequency, and sky surveys could be dominated by remote objects (see e.g. Blain & Longair 1993, 1996). The source counts could be inverted in favor of high-redshift objects, if they exist in equal numbers (not depopulated by strong evolution effects). The millimeter domain becomes then a privileged tool to tackle galaxy formation.
Figure 3. Spectral flux distribution for a typical ULIG starburst source in the radio and far-infrared, for various redshifts z = 0.1, 1, 3, 6, 10, 30, 60 (H0 = 75km/s/Mpc, q0 = 0.5). At right is a synchrotron spectrum, in a power-law of slope -0.7, and left the emission from dust, modelled by PAHs, very small grains and big grains, as in Désert et al. (1990) to fit the Milky Way data. It has been assumed here that the dust properties are the same as in our Galaxy, and that the power of the starburst is the same at any redshift, i.e. Tdust6 - Tbg6 is conserved.
In Fig. 3, we see that at low frequency (lower than 100 GHz, or 3mm in wavelength), the radio spectrum is due to synchrotron processes. There is a marked flux minimum that can be used as a redshift indicator (e.g. Carilli & Yun 1999; Blain 1999). Indeed, there is a well-known tight correlation between the synchrotron radio power and the far-infrared emission in star-forming galaxies (Condon 1992). This correlation is thought to arise because both FIR and non-thermal radio emission are both directly proportionnal to recent star-formation (the FIR being a good measure of massive stars luminosity, and the radio of the rate of supernovae). However, the z-indicator is somewhat ambiguous, since there is a degeneracy between increasing the redshift or decreasing the dust temperature.
The bulk of the high redshift galaxies presently known have been discovered in optical. But these surveys could have missed dust-enshrouded starbursts, since we now know that dust and high metallicity occur very early in the universe (cf. the previous sections). Submm and FIR deep surveys are then the best strategy to detect starbursting proto-galaxies, and many such works have been undertaken, either with sensitive array-bolometers on single dishes (IRAM-30m, SCUBA on JCMT, ...) mm-interferometers, or with ISOPHOT and ISOCAM on board of ISO satellite.
The first deep search was made with the SCUBA bolometer (Holland et al. 1999) towards a cluster of galaxies, thought to serve as a gravitational lens for high-z galaxies behind (Smail et al. 1997). The amplification is in average a factor 2. This has the combined advantage to increase the sensitivity, and to reduce the source confusion, since there should be little contamination from cluster galaxies (Blain 1997). A large number of sources were found, all at large redshifts (z > 1), extrapolated to 2000 sources per square degree (above 4mJy), revealing a large positive evolution with redshift, i.e. an increase of starbursting galaxies. Searches toward the Hubble Deep Field-North (Hughes et al. 1998), and towards the Lockman hole and SSA13 (Barger et al. 1998), have also found a few sources, allowing to derive a similar density of sources: 800 per square degree, above 3 mJy at 850 µm. This already can account for 50% of the cosmic infra-red background (CIRB), that has been estimated by Puget et al. (1996) and Hauser et al. (1998) from COBE data. The photometric redshifts of these sources range between 1 and 3. Their identification with optical objects might be uncertain (Richards 1999). However, Hughes et al. (1998) claim that the star formation rate derived from the far-infrared might be in some cases 10 times higher than derived from the optical, due to the high extinction.
Eales et al. (1999) surveyed some of the CFRS fields at 850µm with SCUBA and found also that the sources can account for a significant fraction of the CIRB background (~ 30%). Their interpretation in terms of the star formation history is however slightly different, in that they do not exclude that the submm luminosity density could evolve in the same way as the UV one. Deep galaxy surveys at 7 and 15µm with ISOCAM also see an evolution with redshift of star-forming galaxies: heavily extincted starbursts represent less than 1% of all galaxies, but 18% of the star formation rate out to z = 1 (Flores et al. 1999).
Now that a few dozens of submm sources have been catalogued (Barger et al. 1999a, Smail et al. 1999), the count rates are confirmed, i.e. ~ 1000 source per square degree, above 3 mJy, at 850 µm, and even 8000 above 1 mJy, from the gravitationally amplified cluster fields (Blain et al. 1999). The cumulative count rate can be fitted by a power-law, above 2 mJy, with a slope of -2.2. The main difficulty appears to be the identification of the submm sources with optical or radio counterparts: the spatial resolution of the submm surveys are several arcsecs, with sometimes systematic uncertainties, and some of the previous claimed identifications have been reconsidered (e.g. Barger et al. 1999b, Downes et al. 1999b). Follow-up in the radio (CO lines) or near-infrared, or optical to find redshifts, are much slower than the surveys themselves. At least 20% of the sources reveal an AGN activity, and the bulk of the sources are at relatively low redshift 1 < z < 3 (Barger et al. 1999a).