Physics-Informed Neural Networks (PINN) for OPC

Physics-Informed Neural Networks (PINNs) integrate lithography physics directly into neural network training by embedding physical laws into the loss function.

Why PINN for OPC?

• Pure ML lacks physical guarantees
• Lithography is governed by known physics (Hopkins, resist models)
• PINNs reduce data dependency and improve generalization

Concept

Physics loss enforces consistency with:

• Optical imaging equations
• Resist diffusion constraints
• Process window behavior

Benefits

• Better extrapolation to unseen patterns
• Reduced retraining for process drift
• Higher physical reliability

Key Takeaway

PINNs bridge the gap between physics-based OPC and data-driven ML.

↩ Go to OPC / RET Technical Notes Index
↩ Go to Home – Tech Notes Park