The nature of high‐redshift galaxies

Using semi‐analytic models of galaxy formation set within the cold dark matter (CDM) merging hierarchy, we investigate several scenarios for the nature of the high‐redshift ) Lyman‐break galaxies (LBGs). We consider a ‘collisional starburst’ model in which bursts of star formation are triggered by galaxy–galaxy mergers, and find that a significant fraction of LBGs are predicted to be starbursts. This model reproduces the observed comoving number density of bright LBGs as a function of redshift and the observed luminosity function at and with a reasonable amount of dust extinction. Model galaxies at have star formation rates, half‐light radii, colours and internal velocity dispersions that are in good agreement with the data. Global quantities such as the star formation rate density and cold gas and metal content of the Universe as a function of redshift also agree well. Two ‘quiescent’ models without starbursts are also investigated. In one, the star formation efficiency in galaxies remains constant with redshift, while in the other, it scales inversely with disc dynamical time, and thus increases rapidly with redshift. The first quiescent model is strongly ruled out, as it does not produce enough high‐redshift galaxies once realistic dust extinction is accounted for. The second quiescent model fits marginally, but underproduces cold gas and very bright galaxies at high redshift. A general conclusion is that star formation at high redshift must be more efficient than locally. The collisional starburst model appears to accomplish this naturally without violating other observational constraints.