Application of design of experiments methodology to optimization of classical molecular dynamics generation of amorphous SiO2 structure

Abstract

We introduce a design of experiments (DOE) methodology to the optimization of a-SiO2 structures using classical molecular dynamics (MD) simulations. This approach produces defect-free systems in good agreement with experimental results of radial distribution functions and angular distribution functions. Application of DOE techniques allows systematic development of optimal MD simulation sequences of amorphous structures. A DOE approach allows control over the number of defects in the sample, as well as other properties, which may not be possible simply by maximizing annealing time and minimizing the cooling rate. In addition, DOE statistical analysis can give detailed insight into the dependences between preparation parameters and the properties of the samples so generated. For some response values, analytical forms have been fit, while for some other responses, accurate mapping on a space of affecting parameters has been performed. Although our investigation is restricted to the generation of a-SiO2 structures, the approach is very general and can be effectively used for development and optimization of arbitrary amorphous systems.