Follow-up from Impact Ionization
The following discussion from !972 (merged) should be addressed:
-
@pschutze started a discussion: I would like to leave a few considerations of @simonspa and mine to be resolved in an issue:
- When propagating within or through a large gradient of the electric field, using
(E_i + E_i+1)/2
for the determination of the impact ionisation coefficient can lead to larger errors. Imagine e.g. a sharp field drop at 1/8th of a step or a steep linear fall. This could be overcome by detecting large gradients and re-sample and integrate the coefficient over a step, or use nested intervals to find the point where the field crosses e.g. a 50% level and scale the coefficient by distance fractions. - In this implementation, the number of secondaries is calculated for every charge carrier of a group. This adds computing time, but could be reduced.
- We decided to add/deposit charge carriers at the end of a step. One could consider moving this to the center, which might be more precise from a physics point of view, but introduces implementation/computational overhead, as the induced current would require a proper determination to maintain e-h symmetry.
- When propagating within or through a large gradient of the electric field, using