Top-profile-prior-constrained inversion for bottom-parameter reconstruction of extreme ultraviolet mask phase defects

Area of Science:

• Semiconductor Manufacturing
• Lithography Technology
• Metrology and Imaging

Background:

• Accurate characterization of buried multilayer phase defects in extreme ultraviolet (EUV) lithography is crucial for defect repair and compensation.
• Current methods struggle to recover the 3D morphology of buried defects, particularly bottom-profile parameters, from single aerial-image observations.

Purpose of the Study:

• To address the information bottleneck in reconstructing buried defect morphology from aerial images.
• To develop a novel computational approach for accurate recovery of bottom-profile parameters of EUV mask defects.

Main Methods:

• Reformulated defect reconstruction as a conditional inverse problem, incorporating the measurable top profile as an explicit prior.
• Developed a dual-stream neural network, CoPriNet, utilizing latent-conditioned feature-wise linear modulation.
• Enforced physical constraints, specifically the multilayer deposition smoothing inequality, within the network architecture.

Main Results:

• CoPriNet significantly reduced the mean relative error for bottom-profile parameters of simulated EUV mask defects.
• Achieved 0.91% error for bump defects and 0.79% for pit defects.
• Demonstrated robustness to noise in the reconstructed complex field and moderate errors in the top-profile prior.

Conclusions:

• The CoPriNet approach effectively overcomes the limitations of single aerial-image observations for buried defect metrology.
• Incorporating physically coupled priors, like the top profile, is a viable strategy for improving reconstruction accuracy in lithography defect analysis.
• The developed method offers a path towards more reliable defect repair in advanced semiconductor manufacturing processes.