Plasma Process Classification Using Causal Discovery Technique

Abstract

The plasma etching process for semiconductor fabrication is too complex to specify the causal structure of the mechanism especially of process variation. Therefore, prediction of etching performance is affected by correlation but not actual causal relationship to process variation. Such correlation is called pseudo correlation. In this research, we introduced the causal discovery technique to clarify the causality of the parameters in process. This method has been applied for experimental process data with consumed parts. The causal structure has been estimated reasonable and a model based on the structure have been achieved better prediction precision for process performance and parts consumption.