Optimizing High-Product-Mix Manufacturing with Advanced Process Control through Machine Learning-Based Virtual Metrology

The semiconductor foundry encounters challenges in addressing multifaceted customer requirements and complicated manufacturing procedures. Variations within the chemical vapor deposition (CVD) process directly impact transistor characteristics and overall yield. Advanced machine learning (ML) software gives rise to an extended virtual metrology (VM) model enhanced by design feature and real-time fault detection and classification (FDC) data. Through the integration of an advanced process control (APC) system, the manufacturing process effectively mitigates film thickness variations within high-product-mix foundry fabs. This remarkable outcome is validated through control simulations.