Computational Scanning Electron Microscopy

Current methods for reconstructing surface topography from SEM images are either not suitable at nanometer scales, restricted to simple shapes described by a limited number of parameters or involve time consuming steps. We describe a reconstruction algorithm whereby an initial surface is iteratively updated. This algorithm performs a full three dimensional reconstruction at the resolution of the SEM image. It employs accelerated techniques to compute a simulated SEM image at each iteration so that the simulated image is obtained in less than 10% of the time taken by current reported Monte Carlo methods. The simulated SEM images are compared with actual SEM images and this comparison leads to further refinement of the reconstructed surface. Further acceleration can be achieved using approximate models for the generation of SE‐I, SE‐II and backscattered electrons.