dense medium cyclone simlation-help
 

I am doing unsteady multiphase simulation of a 3D model of dense medium cyclone. For modelling the turbulence we use RSM,The boundary conditions are velocity inlet at inlet and pressure outlet at both overflow and underflow with gauge pressure set to 0 at both outlets. I observe that invariably the air core starts developing only from the overflow. It then extends till the centre and reaches the spigot. This appears to be in contradiction to what actually happens in the hydrocyclone, air getting sucked in from the spigot and developing from the underflow. Many CFD literatures also claim that the air core develops from the underflow. Why is it not happening i our case?