X-ray scatterometry using deep learning

With the development of semiconductor manufacturing processes, critical dimension small-angle x-ray scattering (CDSAXS) has been identified as a potential solution for measurement. It is worthy of exploring how to achieve fast parameter extraction. In this paper, we propose a XSCNN model based on deep learning to reconstruct the parameters related to structure and measurement conditions. Simulation experiments performed on a trapezoidal grating have demonstrated that XSCNN can produce satisfactory results. It is expected that deep learning will provide a practical solution in CD-SAXS.