Locally, only dwarf galaxies have sufficiently high comoving density to be the counterparts of the faint blue galaxy population at intermediate redshifts; this fact led some to suggest that excess faint galaxies might be dwarf galaxies brightened by bursts of star formation [\protect\astronciteBabul and Rees1992][\protect\astronciteBroadhurst et al.1988]. The amount of star formation required by this scenario has important implications for metal abundance and gas loss in Irr galaxies [\protect\astronciteLilly1993].
The Broadhurst et al. broad88 ``starbursting dwarfs'' model is based on the fact that gas content M
of a galaxy is related to its luminosity, L
, via a simple relation of the form L
, and on the reasonable assumption that star-formation rates are proportional to the gas content during a burst. This model could be used to determine a relationship between star formation and galaxy luminosity. For the case 
= 1 (late-type systems have > 1), the burst strength is independent of absolute magnitude whereas a large value of (> 4) produces a strong magnitude-dependence for the burst 
i.e. evolution is much more important for lower luminosity galaxies. Dwarf galaxies are therefore selectively brightened up to L
whereas galaxies now at L see their luminosity virtually unchanged. This luminosity-dependent luminosity evolution would explain the absence of a high redshift tail in the redshift distribution of faint blue galaxies and the low median of the redshift distribution. A 0.1 Gyr burst converting 5 per cent of the galaxy mass into stars is sufficient to increase the B-band luminosity of the galaxy by 2.2 mag.
The amount of star formation produced by this model is limited by metal production considerations. A rough estimate of the far-ultraviolet (2500Å) luminosity of the faint blue galaxies population suggests that, over a timescale of 6 h
Gyr (0.2 < z < 1), a global metal density of approximately 10
M
Mpc
(independent of H
and q) will be produced. This is an order of magnitude larger than the metal density (
10
h
M Mpc) seen in the local Irr galaxies with comparable comoving density [\protect\astronciteLilly1993]. To avoid this metal enrichment problem, star-bursting galaxies would have had to remove 90
of their metal enriched gas. This is not ruled out by current gas loss models [\protect\astronciteDekel and Silk1986], but the intergalactic medium would be considerably enriched.
Babul and Rees babul92 proposed a model in which faint blue galaxies were an entirely new population with a comoving density higher than any local population. In their model, faint blue galaxies were low-mass galaxies experiencing their starburst at z 1. To explain the formation of faint blue galaxies, they used a generic hierarchical model in which mini-haloes (M10
M) condensed from the expanding background and virialized at a redshift of about 3. The onset of star formation was delayed by the UV intergalactic background radiation produced by quasars and/or by young galaxies. At z=2, this UV background was sufficiently intense to keep the gas trapped in the potential wells of the mini-haloes in a photoionized state. Thus, the gas could not concentrate towards the center to become gravitationally unstable and form stars. From z=2 and the present epoch, the UV background strength fell sharply by a factor of about 100. It ceased to be able to photoionize a protogalaxy around z = 1, and star formation began at that epoch. A few million years thereafter, the first generation of supernovae began to explode. These explosions pumped enough energy in the interstellar medium to trigger a gas outflow. In low density regions where the pressure of the intergalactic medium (IGM) was low, all the gas was ejected from the galaxy. The loss of gas quenched star formation, and the galaxy faded away. In high density regions, the IGM stopped the escaping wind, and the gas then fell back on to the core where it was reprocessed in repeated bursts. These galaxies then evolved to become the present-day dwarf ellipticals seen within regions of high IGM density i.e. around giant galaxies. It is interesting to note that the result of such an episodic star formation scenario may have been observed in the Carina dwarf spheroidal [\protect\astronciteSmecker-Hane et al.1994].