Modeling of continuous random networks: a case study for vitreous GeO2. I.Model generation

Here a molecular modeling program called TUMBLEWEED which generates and ana lyzes spherical models (clusters) for covalent amorphous solids of the cont inuous random network (CRN) type is introduced. The algorithm sequentially adds atoms to a core in accordance with topological and geometrical constra ints. These constraints comprise the type and number of atomic neighbors, i ntervals for bond lengths and angles, and minimum distances for non-bonded atoms. The program includes a new mechanism to control the model density vi a selection from a pool of models, and a procedure to avoid infinite loops if atoms are repeatedly added to and removed from the model surface. The me thod is demonstrated for glassy GeO2, for which CRN models are presented fo r the first time. Various distributions are derived from the models to asse ss their quality. It is found that the models are homogeneous and have a de nsity of (3.60 +/- 0.03) g/cm(3) [experimental data: (3.65 +/- 0.01) g/cm(3 )]. The bond angle distributions are similar to those derived from synchrot ron experiments. Neutron scattering data are simulated for the model struct ures and compared to experimental data. The normalized root mean square dev iation, R-x, between the simulated and the measured total correlation funct ions was found to be as small as 0.027 +/- 0.003 over the interval 0-1 nm i ndicating that CRN models are well suited to describe the structure of vitr eous germania. (C) 1999 Elsevier Science B.V. All rights reserved.