I have a model of a vertical section through a probe design. I have run a solution in CFX-5 (heat transfer only, no flow) and I seem to be having problems at some of the interfaces – some cells at the interfaces between the different regions show spurious temperatures lower than any of the temperatures set in the boundary conditions. If you go into Pre and the domain interface (which is created automatically when you have a 1-1 interface) and change the Connection Type to GGi, this will eliminate the spurious temperatures. For a solid-solid interface where the conductivity is very different for the two solids, this is done automatically. Your case is a solid and a fluid, but the fluid has no advection, so it is essentially acting as a solid with a low conductivity and is therefore causing the problem. In summary, it should be fine to have automatic as the interface except in cases such as yours.