Plasma process optimization using machine learning

Machine learning (ML) techniques were applied to optimize the plasma-enhanced atomic layer deposition (PEALD) process conditions and control the plasma distribution. The latter is crucial for achieving uniformity in film thickness and etch rate on a wafer, which are important factors in semiconductor manufacturing. In this study, the ML approach involved using an algorithm to build regression models with tuning parameters as explanatory variables, and the uniformity of film thickness and etch rate as response variables. To optimize the uniformity of film thickness within a wafer, an engineer and the ML approach separately conducted experiments. The engineer found it challenging to achieve consistent uniformity after five trials, while the ML approach effectively settled the variation. These results indicate that the ML approach can quickly find the optimal conditions and reduce variation. Once the ML suggests key parameters, human judgment can be used to achieve the desired target in the extrapolation area. Essentially, the knowledge gained from ML can be transferred when the process conditions are similar. Therefore, building a database using easily obtainable data is an efficient approach.