On 31/01/17 12:49, William Booker wrote:
Dear David,
The choice of theta is meant to be arbitrary in the scheme, and we have had thoughts of looking into an optimal theta for implementations.
So I would prefer to keep the asymmetry in the scheme.
So the point is that if theta = 0.5 is not optimal then it will be *mesh dependent* (because if theta != 0.5 then the best choice is not invariant under renumbering of the mesh).
I'll implement the conditional form in my DIV definition, and get back to you about the results.
Just a quick query, would the facet normal n not have to be assigned to - or + in that expression?
Yes, you want to take an (arbitrarily) restricted side of the expression: so use: dot(f, n)('+')
Also, is there a simpler way to implement the projections that I have ? Going forward with a system that effectively doubles the amount of variables I have is a bit cumbersome.
So the physical variables are effectively diagnosed from the variational derivatives? But then the evolution equation says how the physical variable evolves according to the derivatives? Presumably then the elimination of one set, or the other, of these variables requires that you invert some operator. Is this operator cell-local? Cheers, Lawrence