Simulation of FIB-SEM images for segmentation of porous microstructures

FIB tomography yields high quality 3D images materials microstructures at the nanometer scale combining serial sectioning using a focused ion beam with scanning electron microscopy (SEM). However, SEM images represent the projection of a slice of unknown thickness. In FIB tomography of highly porous media this leads to shine-through-artifacts preventing automatic segmentation of the solid component. To overcome these difficulties, we simulate the SEM process. Monte-Carlo techniques yield accurate results, but are too slow for FIB-SEM requiring hundreds of SEM images for one dataset. Nevertheless, a quasi analytic description of the specimen and acceleration techniques cut down the computing time by orders of magnitude, allowing the simulation of FIB-SEM data. Based on simulated FIB-SEM image data, segmentation methods for the 3D microstructure of highly porous media from the FIB-SEM data can be developed and evaluated. Finally successful segementation enables quantitati ve analysis and numerical simulations of macroscopic properties.