Automatic differentiation for x-ray grazing-incidence surface imaging using ptychography

In recent years, the drive toward smaller and more complex microelectronics has rendered traditional inspection methods less effective, particularly in wafer bonding processes where interface morphology and surface defects critically affect device performance. X-ray grazing-incidence geometry offers deep penetration depth and short wavelengths for high-resolution imaging. However, dynamical effects impose significant challenges on performing ptychographic imaging of the interfacial structures, particularly near total external reflection. Here, we introduce an automatic differentiation-based ptychographic reconstruction method that incorporates accurate reflection geometry for grazing-incidence experiments. By accounting for surface-normal components in our forward model, we could capture partial dynamic scattering phenomena while still extending beyond the conventional approximation. We validate the proposed method by comparing its reconstruction results with those from a conventional approach, demonstrating improved applicability and performance.