Two gas Mixing

Brian

Hopefully someone can give me some help on what is happening:

I'm modeling gas mixing in 2-D in Star-CCM+. I have a heavy cold fluid in a volume (area) above a lighter hot fluid, they are connected by a small opening. I'm doing a time-dependent calculation to see how the fluid mixes due to the buoyancy difference. I'm using ideal gas, and my gravity is on.

When I run it the concentration contours looks as you'd expect (with good mixing). But the temperatures are strange. At later time steps I get some cells (just a few) with a temperautre much higher than the initial hot temp, and I get some temps lower than the initial low temp.

If I lower my time step, or lower under-relax for energy, or run more inner iterations/ time step it does better. But there are still a few cells with high temperatures.

But to me it seems like this should be a relatively easy problem, so I'm wondering if there are tips for how to model this. I'm using implicit time solving, segregated solvers and laminar viscous modeling.

Also: I have the same problem in Fluent.

Any help would be grealy appreciated! Thank you much