Model-Based OCD is the computational engine behind optical scatterometry — using electromagnetic simulation (RCWA, FEM, or FDTD) to compute the expected optical response for a parameterized geometric model, then fitting the model parameters to match the measured spectrum.

Model-Based OCD Workflow

• Geometric Model: Define a parameterized profile (trapezoid, multi-layer stack) with parameters: CD, height, sidewall angle, corner rounding.
• Simulation: Use RCWA (Rigorous Coupled-Wave Analysis) to compute the theoretical spectrum for each parameter combination.
• Library: Build a library of pre-computed spectra spanning the parameter space — or use real-time regression.
• Fitting: Match measured spectrum to library using least-squares or machine learning — extract best-fit parameters.

Why It Matters

• Accuracy: Model accuracy directly determines measurement accuracy — the model must faithfully represent the physical structure.
• Correlations: Parameter correlations limit the number of independently extractable parameters — model complexity must be balanced.
• Floating Parameters: Only a few parameters can "float" (be extracted) — others must be fixed or constrained.

Model-Based OCD is solving the inverse problem — computing what the structure looks like by matching measured optical signatures to electromagnetic simulations.