Answers are grounded in this paper and cite their sources.
BackgroundAs semiconductor device scaling progresses into the sub-10 nm critical-dimension regime, the adoption of extreme-ultraviolet (EUV) lithography for high-volume manufacturing (HVM) of advanced nodes has brought stochastic variability and its resulting failures to the forefront of yield-loss concerns. EUV photons possess high energy, complicating absorption by photoresists and exacerbating random defectivity as feature sizes shrink. Although continuous improvements have been made in photoresist performance, mask technology, and post-lithography processes, stochastic defects remain a major source of yield loss in HVM. Experimental approaches for identifying, analyzing, and correcting stochastic defectivity are costly, requiring large numbers of wafers, extensive metrology resources, and considerable time.AimWe discuss a cost-effective predictive failure probability model (FPM) and demonstrate that it is an essential tool for studying and reducing stochastic defectivity in EUVL.ApproachBy leveraging rigorous probabilistic computational lithography methods, we construct accurate EUVL failure probability models. We use actual defect inspection tools to show how many more defects can be detected using the FPM’s predicted hotspots or weak points (WPs) compared with a conventional WP detection method.ResultsOur results confirm that the defectivity predictions generated by the FPM are in good agreement with e-beam inspection measurements. Larger-area inspection of space bridge defects was performed with optical defect inspection tools. In this case, defect detection was enhanced by use of FPM predictive simulation: defect detection using FPM output was 1.48× higher versus the standard methods.ConclusionBy leveraging FPMs grounded in rigorous probabilistic computational lithography, we are able to make robust predictions of stochastic defectivity over large layout areas, thereby establishing an optimal patterning process.
No takes yet. Share an insight, caveat, or question.
Choi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: