Is your rigorous simulator accurate enough for advanced DUV and EUV lithography?
As the feature size gets smaller, the accuracy requirement for rigorous simulators becomes more stringent. If your simulator cannot consistently predict the diffraction spectra of DUV and EUV mask patterns accurately to at least three decimal places, it may not be accurate enough for advanced lithography.
At Fastlitho, we take great care to make sure that our rigorous simulator gives accurate results. To verify the accuracy, we compare our results with those of the rigorous coupled wave analysis (RCWA) method, which is widely regarded as the most accurate rigorous method. Furthermore, to ensure complete objectivity in the comparison, we use a third-party RCWA tool for reference.
Until recently, a third-party RCWA tool may cost hundreds, if not thousands of dollars, and so such a tool was not readily available for use as a reference. However, recently, a very popular RCWA tool, GSolver, was released to the public free of charge. This means that users can now check the accuracy of their own rigorous simulators by comparing with this free, third-party RCWA tool.
However, since RCWA is very time consuming for large problems, it can be used to generate reference solutions accurate to four decimal places only for small, one-dimensional grating problems. So, we propose that any rigorous simulator which claims to be accurate must, at the minimum, agree with RCWA to at least three decimal places for one-dimensional grating problems. Otherwise, the simulator may not be accurate enough for advanced lithography.
To see how accurate Fastlitho's rigorous simulator is compared to RCWA, click the links below for DUV and EUV examples:
Hint: If you do not wish to run GSolver yourself, you can simply run the two grating examples with your own rigorous simulator and compare your results with the GSolver results given in the links. This way, you can at least find out whether or not your simulator passes these two basic tests. (Note: Some simulator's results may need to be adjusted for normalization before comparison with GSolver.)