Abstract
Advances in both reactor and feature scale modeling have been achieved in plasma processing. Models are being used to address several practical issues in industry. However, accurate predictions of physical phenomena that occur during the process need an integrated multiscale approach that couples reactor scale to feature scale. The implementation of multiscale codes is still a challenge where robustness plays a critical role. This paper addresses the weakest points of such codes especially at the feature scale and provides proper choices for successful implementations. It focuses on the level set methods and fast marching algorithms implemented on quadtree (2D) and octree (3D) data structures and possible extensions of semi-lagrangian methods.