Silica glass structure generation for ab initio calculations using small samples of amorphous silica

Multiple small samples of amorphous silica have been generated and optimized using classical dynamics and the van Beest-Rramer-van Santen (BKS) empirical potential function. The samples were subsequently optimized and annealed using density functional theory (DFT) with both the local density and the generalized gradient approximations. A thorough analysis of the local and medium-range structure of the optimized samples obtained from the different methods was carried out. The structural characteristics obtained for the average of small systems each containing ca. 100 ions are compared for each of the different methods, and to the BKS simulation of a larger system. The differences found between the DFT and BKS simulations and the effects of volume relaxation on the structures are discussed. Fixed-volume samples are compared to neutron scattering data, with good agreement to 5 Å, the length limit of the sample sizes used here. It is shown that by creating multiple small samples, it is possible to achieve a good statistical sampling of structural features consistent with larger simulated glass systems. This study also shows that multiple small samples are necessary to capture the structural distribution of silica glass, and therefore to study more complex processes in glass, such as reactions.