Collisions can violate parallel boundary condition?
Issue #192
new
In the parallel direction the boundary condition is zero incoming particles, i.e. h = 0 at -ntgrid for particles going from left to right and h = 0 at ntgrid for particles going from right to left (where gnew ~ h + a * phi + b * bpar) . The particles travelling in the other direction are unconstrained. This has the potential to produce a discontinuity in the boundary distribution function at the v||=0 point (could this be related to the issues with the WFB we sometimes see?).
Collisions can scatter the distribution function in v|| and one would expect collisions to act to smooth out discontinuities such as those that might arise at the boundaries. The result is that after collisions are applied there are now incoming particles at the boundary and the solution no longer satisfied the parallel boundary condition we wish to impose.
It’s not clear that this is a major problem, but issue #160 describes one situation in which not satisfying the boundary conditions can result in a numerical oscillation. Might the boundary ghost point suggested in issue #187 possibly also help here?
Comments (5)
-
reporter
-
reporter
Here’s an example showing h as a function of energy for the incoming particles at the incoming boundary for simulations with and without collisions (one time step). Note that the collisionless case retains hr=0 at both boundaries whilst this is not true for the collisional simulation.
-
reporter
Note the WFB doesn’t satisfy the zero incoming boundary conditions with or without collisions by default. Choosing wfbbc_option = ‘passing’ causes it to display the same behaviour as the passing pitch angles.
-
reporter
One may also note that the collisional result is not symmetric in vpar/boundary even though the system should be up-down symmetric.
-
reporter
You might be interested in this @Colin Malcolm Roach
- Log in to comment
One solution could be to change the boundary conditions of the collision operators only at the boundaries.