Faint galaxies in semi-analytic models: how robust are the predictions?

In spite of their overall success, semi-analytic models of galaxy formation and evolution predict slopes of luminosity functions which are steeper tha n the observed ones. This discrepancy has generally been explained by subtl e surface brightness effects acting on the observational samples. In this p aper, we explicitly implement the computation of surface brightness in a si mple semi-analytic model (with standard CDM),and we estimate the effect of observational surface brightness thresholds on the predicted luminosity fun ctions. The crucial free parameter in this computation is the efficiency ep silon of supernova feedback which is responsible for the triggering of gala ctic winds. With the classical formalism for this process, it is difficult to reproduce simultaneously the Tully-Fisher relation and the Bat slope of the observational luminosity function with the same value of epsilon. This suggests that the triggering of galactic winds is a complex phenomenon. The highly uncertain formalism for supernova feedback that is used by semi-ana lytic models produces large uncertainties in the results. However, once a v alue of epsilon has been chosen, the various luminosity functions observed in different wavebands (B, r, K) and at different surface brightness thresh olds, are consistently reproduced with the surface brightness thresholds qu oted by the observers. This seems to show that these observations do see su bsamples of the same underlying populations of ''sub-L*" and dwarf galaxies . The conclusion of this heuristic paper is that a more realistic descripti on of SN feedback is needed, and that surface brightness effects should not be neglected in the modelling of galaxy formation.