Optical scatterometry with relaxed model constraints

Optical scatterometry extracts structural and material properties of nanostructures from optical signal by solving the optical inverse problems (OIPs). It typically requires prior information for the establishment of structural models. Existing modeling methods are almost based on strict model constraints, which may oversimplify structural details. Here, we propose a strategy with relaxed model constraints. To deal with the expanded solution space caused by the relaxed constraints, the inverse design inspired optimization for optical scatterometry(IDO-OS) is developed by integrating automatic differentiation. We evaluated our method under both numerical simulation and experimental conditions. The results of numerical simulation demonstrate the performance of our method with an optimized spectral goodness-of-fit of 0.99 and a high degree of consistency with target structures. In the experiment, we integrate IDO-OS with hybrid metrology and thin plate splines algorithm to generate relaxed model constraints, and the reconstructed structures show excellent agreement with the measurements of transmission electron microscopy. This work establishes what we believe to be a new strategy for solving inverse problems in optical scatterometry, offering enhanced structural reconstruction capabilities while maintaining computational efficiency.