Implementation of dispersion
A dispersion interface should be implemented that hopefully works for all discretizations.
Old flyspray description
Metadata
| Property | Value |
|---|---|
| Project | dumux |
| Category | General |
| Reported by | Alexander Kissinger (alexander.kissinger@iws.uni-stuttgart.de) |
| Reported at | Mar 30, 2016 15:18 |
| Type | Bug Report |
| Version | 2.9 |
Description
I noticed that the implementation of dispersion in the 1p2c model may not be correct for applications using the cell centered discretization. The dispersion tensor needs to be evaluated at the interface between two dofs. It requires the velocity at that location. For the Box method this can easily be achieved using the potential gradient evaluated from multiple dofs in the element. For the CC method however the potential gradient calculated currently at the interface only depends on the two opposing dofs (i and j). This will generally not result in the correct velocity. A correct implementation would require an interpolation of the two cell velocities i and j at the interface.
Example: Flow parallel to the interface between i and j would result in a zero velocity at the interface as there is no potential gradient between i and j and the velocity component parallel to the interface is not considered. As a result also the transverse dispersion would be zero.