Running the 2D-production test (#473) before and after the nonorthogonal operator implementation into neutral_mixed (#285) for a longer time leads to a reduction of Nesep by at least 50%. The 2D-production test has a grid with nonorthogonality at the core and at the PFR boundary, but orthogonal targets.
Repeating the test on an orthogonal grid results in no change. A bisection of the changes in the PR found that if we revert to the old operators but keep all other changes from the PR, the issue goes away. Changing the first operator to nonorthogonal already causes an issue. I have not tested the other operators individually.
Plotting a total particle integral reveals a drop. This means that Nesep is being reduced through a new domain particle sink. I have plotted some domain particle sources and found that there is a very large particle sink of neutrals in the PFR:
Now, the boundary condition for neutral density, pressure and temperature is set to Neumann everywhere but the core. This means there shouldn't be any radial flow into the boundary, as there is no radial pressure gradient.
Picking a radial slice near a target reveals that the radial pressure gradient is indeed zero at the PFR boundary, but there is a non-zero flow. The same appears to be the case on the SOL boundary:
We should make the nonorthogonal operators optional while we work on resolving this.
Running the 2D-production test (#473) before and after the nonorthogonal operator implementation into neutral_mixed (#285) for a longer time leads to a reduction of Nesep by at least 50%. The 2D-production test has a grid with nonorthogonality at the core and at the PFR boundary, but orthogonal targets.
Repeating the test on an orthogonal grid results in no change. A bisection of the changes in the PR found that if we revert to the old operators but keep all other changes from the PR, the issue goes away. Changing the first operator to nonorthogonal already causes an issue. I have not tested the other operators individually.
Plotting a total particle integral reveals a drop. This means that Nesep is being reduced through a new domain particle sink. I have plotted some domain particle sources and found that there is a very large particle sink of neutrals in the PFR:
Now, the boundary condition for neutral density, pressure and temperature is set to Neumann everywhere but the core. This means there shouldn't be any radial flow into the boundary, as there is no radial pressure gradient.
Picking a radial slice near a target reveals that the radial pressure gradient is indeed zero at the PFR boundary, but there is a non-zero flow. The same appears to be the case on the SOL boundary:
We should make the nonorthogonal operators optional while we work on resolving this.