[CIG-CS] Discontinuous viscosity

[Wolfgang Bangerth]

> One interesting detail is that it is necessary to change the
> fundamental variables, because I do not have expressions for the jumps
> in velocity.

That is similar to using mixed finite elements for the Poisson equation 
with variable coefficients:
   - div  k(x)  grad p = f
is rewritten as
   k^{-1} u + grad p = 0
   -div u = f
and not
   u + k grad p = 0
   - div u = f

The former mixed formulation has a reasonable error behavior as k 
becomes more non-uniform, whereas in the latter case (as in the original 
non-mixed formulation) the error grows proportionally to
   max k(x) / min k(x)

I suppose something similar could be shown for your formulation.

Best
  W.

------------------------------------------------------------------------
Wolfgang Bangerth               email:            bangerth at math.tamu.edu
                                 www: http://www.math.tamu.edu/~bangerth/