From a random model to the correct structure: the VLD algorithm
Carmelo Giacovazzo, Maria Cristina Burla, Giampiero Polidori · Acta Crystallographica Section A Foundations of Crystallography · 2010
Recently there has been a growing interest in iterative algorithms with regards to structure determination of small molecule X-ray crystallographic data.The roots of iterative algorithms originate in electron microscopy community with the Gerchberg-Saxton algorithm [1] , which was later refined by Fienup into the Hybrid input-output (Hio) algorithm [2] , and Elser produced the difference map algorithm [3] .The reason for particular interest in iterative algorithms is their distinct advantages over the classical direct methods, namely not requiring any knowledge of the crystal symmetry, or composition in order to solve a structure.In classical direct methods the measured intensities are used directly to estimate the phases, this is not the case for the iterative algorithms making them more robust to experimental errors in the measured intensities, which is a distinct advantage for powder and electron crystallography datasets.Electron crystallography data is particularly problematic for direct methods because the underlying assumption of kinematic data does not always hold, due to the much stronger interaction of electrons with matter when compared to x-rays.An introduction to iterative algorithms with a historical perspective will be presented, with a particular emphasis being placed on the application of charge flipping algorithm [4] to the structure solution of electron crystallographic data sets.