Predicting Film Thickness and Properties in CESL Process Based on CGAN

Contact etch stop layer (CESL) process plays a pivotal role in semiconductor manufacturing, where accurate predicting the THK and RI of silicon nitride film in CESL process is crucial. This paper introduces a Conditional Generative Adversarial Network (CGAN)-based prediction model, which simultaneously predicts film thickness and refractive index by CESL process recipes. By incorporating adversarial training, the model achieves superior accuracy compared to traditional methods such as Support Vector Regression (SVR) and XGBoost. This advancement allows for improved process optimization, reduced dependence on physical experiments, and better production quality control.