As seen in #295, the NTV torque model can have sharp peaks in torque density at the far edge. If calculated self consistently, the large kinetic terms damp out perturbations here and there should not be an issue. But when calculating in a faster workflow from ideal perturbed equilibrium, this can cause unphysically large total torque values. It also makes the total torque very sensitive to the edge truncation.
In reality, the NTV theory if all for thin banana widths. We should not trust it in regions where banana widths (or even ion orbit widths) are crossing the separatrix. Some simple improvements would be to,
- Report the standard banana width approximation for thermal ions as a radial profile output like omega_E, nu, etc.
- Truncate the kineticForces integration at
max(psilim[1], 1-0.5*rho_banana(1)) where rho_banana(1) could be approximated by the psihigh value or extrapolated from a profile spline.
As seen in #295, the NTV torque model can have sharp peaks in torque density at the far edge. If calculated self consistently, the large kinetic terms damp out perturbations here and there should not be an issue. But when calculating in a faster workflow from ideal perturbed equilibrium, this can cause unphysically large total torque values. It also makes the total torque very sensitive to the edge truncation.
In reality, the NTV theory if all for thin banana widths. We should not trust it in regions where banana widths (or even ion orbit widths) are crossing the separatrix. Some simple improvements would be to,
max(psilim[1], 1-0.5*rho_banana(1))where rho_banana(1) could be approximated by the psihigh value or extrapolated from a profile spline.