This spreadsheet provided by Jeff Hieronymus based on the stability spreadsheet given above. Jeff reports that the numerical solution seems to have an effective diffusion coefficient that is higher than it should be (see graphs). See what you think, and see what you can do about it.