Origins of aspect ratio dependent etching in plasma materials processing

Summary form only given. During reactive ion plasma etching (RIE) of semiconductor materials, a strong correlation between etch rate and feature size, or more specifically feature aspect ratio (AR), is often observed. There have been several analytical models proposed to explain this aspect ratio dependent etch-rate (ARDE), also known as “RIE lag.”[1,2] While most analytical models only address individual causes of ARDE, the complexity of most etching reactions implies that several ARDE mechanisms may be at play. This complexity motivates the use of numerical models which are sensitive to all of the probable causes of ARDE to determine the dominant factor in any given etching reaction. To investigate processes leading to ARDE, a 3-dimensional voxel based Monte-Carlo etch profile simulator was developed. In this model, the etch feature is represented by mesh of cubic computational voxels. Energy and angular resolved fluxes of neutral radicals and ions are produced by a companion plasma equipment model. Pseudoparticles are randomly chosen from these distributions and directed to the surface. Each collision of a gas phase pseudoparticle with the profile then adds, removes or changes the identity of material voxels based on probabilities defined in a reaction mechanism. We investigated some of the complex interactions that may lead to ARDE for Cl₂ plasma etching of Si and fluorocarbon plasma etching of SiO₂. By tuning parameters that are uncontrolled in experiments, such as the sticking coefficients of etch products to sidewalls of the feature, we were able to isolate the causes of ARDE in several etch regimes. Even within a single reaction mechanism, the dominant cause of ARDE can change with etch geometry, processing conditions or even with etch depth. For instance, as the profile of a trench or via transitions from a flat etch front to a tapered profile, a transition which occurs at some AR for almost all RIE processes, the dominant cause of ARDE can abruptly shift from radical starvation to lack of ion energy flux. The dominant cause(s) of ARDE will be discussed for a range of operating configurations (experimentally accessible) and model parameters (only simulation accessible). For conditions where a single mechanism dominates we present scaling relations and test the applicability of analytical models.