We have summarized in previous paragraphs the main properties of local galaxies when observed at long wavelengths, and emphasized the unique capability of these observations to unveil classes of sources, unnoticeable at other wavelengths, but extremely luminous in the IR. It was clear from this that the most luminous objects in the universe and the most violent starbursters can be reliably studied only at these wavelengths.
Our previous discussion has also illustrated the complexity and difficulty of modelling the long-wavelength spectra of galaxies, heavily dependent on the relative geometries of stars and dust.
Now, assumed we have a decent understanding of the local universe and its IR galaxy populations, we dedicate the next Sections to illustrate and discuss new emerging facts about their distant counterparts, which entail important discoveries for cosmology.
The IRAS survey in 1983, allowing the first sensitive all-sky view of the universe at long wavelengths, is considered as the birth date of IR astronomy. Most of our knowledge about local IR galaxies, as previously discussed, comes from the IRAS database. The fair sensitivity of the IRAS surveys, coupled with the prominent emission of IR galaxies at 60-100 µm, have also allowed to sample and study galaxies at cosmological distances and to derive first tentative indications for evolution.
Counts of IRAS galaxies (mostly at 60 µm, where the S/N was optimum, S including the source signal and N the instrumental and sky [cirrus] noise) have been obtained by Hacking & Houck (1987), Rowan-Robinson et al. (1991), Gregorich et al. (1995), Bertin, Dennefeld, Moshir (1997). Samples at 60 µm with optical identifications and radial velocities have been published by Saunders et al. (1990, 1997), Lonsdale et al. (1990), and Oliver et al. (1996).
Early evidence in favour of evolution for IRAS-selected galaxies have been
discussed by
Hacking et al. (1987),
Franceschini et
al. (1988) and
Lonsdale et al. (1990),
among others.
In the models by Franceschini et al. and
Pearson & Rowan-Robinson
(1996),
a sub-population of starburst galaxies including a substantial fraction
(30%) of all galaxies in the local universe evolves as
L(z) = L(0)(1 + z)3.1
(Pearson & Rowan-Robinson) or
L(z) = L(0)e 4.3
(z)
(Franceschini et al.), roughly
reproducing counts and redshift distributions.
However, the IRAS sensitivity was not enough to detect galaxies
at substantial redshifts, apart from a handful of exceptions
(essentially due to gravitational lensing amplifying the flux): the most
distant were found at
z 
0.2 - 0.3. Any conclusions based on IRAS data are to
be considered as
preliminary, large-scale inhomogeneities badly affecting these shallow
samples.
Another problem for the IRAS surveys was the uncertain identification
with faint optical counterparts, because of the large
[
1 arcmin2] IRAS error-box: this
implied a systematic bias towards
associating IRAS sources with the brightest galaxy falling inside it, which
may systematically miss the fainter higher-redshift correct identification.