[sampaio]

Dear All,
I am trying to simulate a "pathologic" (cause it has no physical meaning, and its purpose is just to test numerical squemes) scalarTransportFoam case, in which a unitary square box, 20 divisions in each edge, has prescribed (fixed) temperature, in all 4 edges: 0 in the bottom and left edge; -1 in the top edge; 1 in the right edge.

The velocity is uniform (1, 1, 0).

The initial temperature is also 0 everywhere.

termal diffusivity DT is set to 10^-10.

I am solving it as a transient problem with timestep = 0.005, with backward differences squeme.

One should expect that the downwind BC would not influence the interior domain, at least when Upwind is used. When Central Differences is used, we get the expected "checkered board" problem.

However, with QUICk and even with Upwind, I get an unbounded solution that grows indefinitely. In the upwind solution, the first interior cell-center, just upwindward of the downwind BC (top and right edges) grows beyond the bounding limits (-1, 1) as the time is advanced.

Is this something expected, that I am missing because of my ignorance?

I can post the case here if anyone wants to look at it.

Cheers,
luiz