[Snowball]

I am trying to simulate a density-driven flow which is induced by a cooling source inside the computational domain. The flow will drop from a certain height (underneath a inlet boundary) due to temperature and negative buoyancy, and then impinge on a wall. I used pressure inlet and outlet boundaries but results (temperature) got divergent. I used k-omega SST model for turbulence and ideal gas law for fluid property.

I did this successfully in a 2-D axis-symmetric domain but got divergent in a 3-D simulation.

Does anyone know what boundary conditions should I use in this case? What caused the divergence?