In this article, the conditions for aerobic biofilm formation on suspended particles, the dynamics of biofilm formation, and the biomass production during the start-up of a Biofilm Airlift Suspension reactor (BAS reactor) have been studied. The dynamics of biofilm formation during start up in the biofilm airlift suspension reactor follows three consecutive stages : bare carrier, microcolonies or patchy biofilms on the carrier, and biofilms completely covering the carrier. The effect of hydraulic retention time and of substrate loading rate on the formation of biofilms were investigated. To obtain in a BAS reactor a high biomass concentration and predominantly continuous biofilms, which completely surround the carrier, the hydraulic retention time must be shorter than the inverse of the maximum growth rate of the suspended bacteria. At longer hydraulic retention times, a low amount of attached biomass can be present on the carrier material as patchy biofilms. During the start-up at short hydraulic retention times the bare carrier concentration decreases, the amount of biomass per biofilm particle remains constant, and biomass increase in the reactor is due to increasing numbers of biofilm particles. The substrate surface loading rate has effect only on the amount of biomass on the biofilm particle. A higher surface load leads to a thicker biofilm. A strong nonlinear increase of the concentration of attached biomass in time was observed. This can be explained by a decreased abrasion of the biofilm particles due to the decreasing concentration of bare carriers. The detachment rate per biofilm area during the start-up is independent of the substrate loading rate, but depends strongly upon the bare carrier concentration. The Pirt-maintenance concept is applicable to BAS reactors. Surplus biomass production is diminished at high biomass concentrations. The average maximal yield of biomass on substrate during the experiments presented in this article was 0.44 +/- 0.08 C-mol/C-mol, the maintenance value 0.019 +/- 0.012 C-mol/(C-mol h). The lowest actual biomass yield measured in this study was 0.15 C-mol/C-mol.