RAMM: A new random-model-based method for solving ab initio crystal structure using the EXPO package

摘要

The new method RAMM (random-model-based method) has been developed and implemented in the EXPO computing program for improving the ab initio crystal structure solution process. When the available information consists of only the experimental powder diffraction pattern and the chemical formula of the compound under study, the classical structure solution approach follows two main steps: (1) phasing by direct methods (or by Patterson methods) in order to obtain a structure model (this last is usually incomplete and/or approximate); (2) improving the model by structure optimization techniques. This article proposes the alternative procedure RAMM, which skips step (1) and supplies a fully random model to step (2). This model is then submitted to effective structure optimization tools present in EXPO - wLSQ (weighted least squares), RBM (resolution bias minimization) and COVMAP (a procedure of electron density modification based on the concept of covariance between points of the map) - which are able to lead to the correct structure. RAMM is based on a cyclic process, generating several random models which are then optimized. The process stops automatically when it recognizes the correct structure.