Virtual Metrology models for predicting physical measurement in semiconductor manufacturing

The semiconductor manufacturing industry has a large-volume multistage manufacturing system. To insure the high stability and the production yield on-line a reliable wafer monitoring is required. The approach, called Virtual Metrology (VM) is defined as the prediction of metrology variables (either measurable or non measurable) using process and wafer state information. It consists in the definition and the application of some predictive and corrective models for metrology outputs (physical measurements) in function of the previous metrology outputs and of the equipment parameters of current and previous steps of fabrication. The goals of this paper are to present a methodology for VM module for individual process applications in semiconductor manufacturing and to present a case study based on industrial data.