Thanks for your reply Rob,

Yes it's a real gas model. I specified the backflow properties close to the flow properties at the outlet. These properties are the turbulence intensity and hydraulic diameter and temperature. 

Right now, the temperature does not change in my simulations as all the heat sources are off and flow remains at 25C temperature similar to inlet. 

I was hoping that assigning these real backflow properties may help the convergence and resolve the backflow but it didn't work.

I think that the solver is getting trapped at a wrong answer and needs a little push toward the right answer. One idea that I had was pitching a 0 pressure on the part of the domain between the last layer and the outlet. Then I would first run it under this condition for a few iterations, and then will remove this forced condition and let it naturally converge. But I'm not sure how I can pitch a property to a specific volume in the domain. 

What do you think?