Intermediate Variable Extraction for Digital Twin Modeling in the Semiconductor Manufacturing Process

Constructing digital twin models for multiple processes presents challenges owing to the dependence of subsequent processes on the previous results. In this study, we examined the issue of reduced model reusability caused by the dependency of process models on current process configurations. Using a case study involving the post-annealing and etching of an oxide film formed by atomic layer deposition, we developed a model that divided processes based on intermediate variables extracted through physical analysis. The results demonstrated that these intermediate variables sufficiently represented the impact of previous processes, with a high coefficient of determination indicating model effectiveness. Our analysis of the class structure emphasized the importance of confining variables within device classes for effective process abstraction. Applying this model design approach enhances process model abstraction and supports the model’s suitability for representing the entire process.