Rigorous calibration and validation of high sigma defect prediction models

Accurate prediction of patterns susceptible to rare stochastic failure has been a key goal for EUV lithography modeling since the initial use of EUV lithography in production processes. This goal has been difficult to achieve for several reasons, including quantification of stochastic effects in photoresist chemistry interactions, mask variation contributions in all three-dimensions, and measurement uncertainty. Another difficulty has been the extremely large number of experiments needed to calibrate and validate stochastic models for high sigma (very rare) pattern defectivity rate, both for wafer experiments and for rigorous Monte Carlo statistical simulations. In this paper, we investigate and show results from algorithmic enhancements to generate significantly faster rigorous statistical simulations which can enable accurate rigorous high sigma pattern defect simulation in a reasonable time. We then use these enhancements to build and validate accurate full-chip capable through-sigma defect printability compact models.