Hybrid metrology for hot spot prediction in lithographic patterning

Hot spots negatively impact device performance and yield. Minimizing hot spots in the patterning process is critical, yet they are often not identified until after pattern transfer. In this work, we show how we trained a low-cost hybrid metrology model, in the form of physics-enabled artificial intelligence (AI), to identify potential hot spots during etch transfer. The model effectively predicts post-etch 3D behavior and identifies pattern areas susceptible to hot spot formation within a die. Also, we demonstrate the model’s scalability to predict hot spots and etch performance across an entire 300mm wafer.