When gas is continuously fed through a sparger into a downflowing liquid in a pipe a ventilated cavity is often formed. The cavity remains attached to the sparger even in the presence of high liquid flow rates that would wash away a free slug bubble. Small bubbles are shed from the base of this cavity by the falling liquid film at the wall of the pipe and these bubbles are swept downwards forming a bubbly flow that is highly effective for mass transfer. The ventilated cavity is undesirable since it reduces the driving force for liquid circulation when the pipe is the downcomer of an external air loop fermenter or analogous gas/liquid reactors. The cavity also reduces the available interfacial area for mass transfer. It has been shown [Thorpe et al., 1997. Proceedings of the Fourth International Conference on Bioreactor and Bioprocess Fluid Dynamics; Lee, 1998. Ph.D Thesis, University of Cambridge, UK], that the length of the cavity can be reduced by replacing the common industrial design of a horizontal sparger (HS) with two novel spargers; a peripheral sparger (PS) and a plunging jet sparger (PJS) (Fig. 3). In this paper we investigate the effect of PS and PIS on mass transfer and the resulting bubble size. Experiments were carried out with air and water in a large circulating rig with a 0.105 m diameter test section. The local average bubble size in the bulk two-phase flow region below the ventilated cavity was determined using photography for three combinations of liquid and gas volumetric flow rates. The average bubble size was essentially the same (differences within 10%) for the PS, central spranger (CS) and HS. The PS created the largest bubble in all cases examined. The PIS created smaller bubbles than all the other spargers and did not allow the formation of cavities, which suggests that it has the superior performance. The estimated increase in k(L)a due to the smaller bubble size for the PJS was by a factor of 1.3. In order to check this result, the effects of sparger type on the volumetric mass transfer coefficient (k(L)a) were also measured. The k(L)a was determined with a dynamic method, by using unsteady state absorption of oxygen. The results confirmed the apparent superiority of PIS over the other spargers. An average increase of 19% in the k(L)a was observed when the PJS was used instead of the industrial design (HS). The CS and PS showed similar k(L)a values again within 10% of the HS. However the power consumption is larger when the PIS is used instead of the industrial design HS. Hence an attempt was made to adjust the bubble size and mass transfer coefficients of the PJS to account for the differences in energy consumption. When this is done the PIS and HS produce roughly the same bubble size and have the same mass transfer performance. Still the PIS had the important operational advantages of producing shorter cavities and having the greater resistance to stall at low liquid flow rates. (c) 2005 Elsevier Ltd. All rights reserved.