The Use of Stochastic Algorithms for Phase Retrieval in High Resolution Transmission Electron Microscopy

The usefulness of stochastic algorithms to retrieve the exit-plane wave function from periodic high-resolution electron microscopic images is investigated. In contrast to “classical” focal series reconstruction methods which need approximately twenty input images, fully non-linear reconstructions of periodic wave functions are possible in most cases from only two images using stochastic algorithms. The efficiency and accuracy of two different algorithms, a Simulated Annealing algorithm and a genetic algorithm, are compared with each other. Simulated high-resolution images of the intermetallic alloy Ni 4 Mo and experimental images of the high-T c superconductor YBa 2 Cu 3 O 7 are used as a test input.