Foamed food has become very popular since the 1990's. In food foaming process the key point is to disperse a high quantity of gas with small bubbles diameter (lower than 50 mu m) within a continuous phase. At industrial scale, this operation is generally carried out in mixing units such as rotor/stator or scraped surface heat exchangers. The experimental set-up in our laboratory is a simplified version of the latter, consisting in a "Narrow Annular Gap Unit" (NAGU), equipped with flatbladed impellers. In part one of this work, it has been shown experimentally that the performance of foaming depends on the relative position of the impellers (angle and clearance between two successive impellers). In this part, Computational Fluid Dynamics (CFD) is used to perform flow pattern: velocity, pressure, shear and elongation rate and dissipation rate for the same configuration as the ones used in experimental investigation. The angle between two impellers induces the multiplication of Taylor vortices, making the flow more chaotic and enhancing the mixing efficiency. The CFD analysis shows also an equal magnitude of the shear rate, both at the wall and in the bulk for all the geometries. The dispersive mixing seems to be controlled by the blade/wall gap where the strain rate magnitude is much higher than in the bulk where the shear rate has a minor role in the mixing. The CFD analysis also highlights that there is no direct link between the power consumption and the yield of dispersion. Hence, the explanation of the better performance of the compact/shifted configuration can be interpreted by the distributive mixing performances. The negative effect of the clearance between two impellers can only be explained by the stress relaxation zones in the flow between two impeller arrays, which overrate the coalescence. (c) 2012 Elsevier Ltd. All rights reserved.