Finding crystal structures from few diffraction data by a combination of a random search with genetic algorithms

Abstract

A new procedure for performing structural analysis of crystalline materials from diffraction data, using internal coordinates, is described. For starting information only unit-cell content, space group, chemical formula, molecular connectivity and a limited amount of diffraction data are required. After first selecting a number of solutions using a Monte Carlo approach with severe filters, which reject the most unrealistic solutions, genetic algorithms (crossover and mutations) are applied. In fact, the initial selection step alone is, frequently, a powerful tool for discovering structures, without recourse to the genetic algorithms. The procedure, while suffering from the limitation that connectivity must be known, is effective in cases where direct methods are not applicable because the diffraction data are scarce, are limited to low diffraction angles or are missing in specific portions of the reciprocal space. The main features of the algorithm are described and examples of validation given. The routines are now available as part of the freely distributed general-purpose program TRY. The program is available on the Web at http://www.theochem.unisa.it/try.html. © Attilio Immirzi et al. 2008.