Efficient generation of realistic model systems of amorphous silica

TL;DR

This paper presents a computational approach combining classical molecular dynamics and ab initio methods to efficiently generate and optimize realistic models of amorphous silica, matching experimental properties.

Contribution

It introduces a hybrid simulation protocol that reduces computational cost while maintaining structural accuracy of amorphous silica models.

Findings

Geometry optimization is five times faster than Car-Parrinello annealing.

Final configurations are consistent across different optimization methods.

Structural and electronic properties agree with experimental data.

Abstract

We used classical molecular dynamics and the van Beest Kramer van Santen (BKS) potential to generate small model systems of amorphous silica. We further optimized the classically equilibrated configurations using plane wave based density functional theory and a generalized gradient (GGA) approximation. Within ab initio treatment we showed that both geometry optimization and Car-Parrinello annealing lead to the same final configurations but the CPU time required for the geometry optimization to reach convergence is one fifth of the time needed by a Car-Parrinello annealing. In addition during the optimization or the annealing no substantial change occurs on the topology acquired by the vitreous silica at the end of the classical quenching protocol. Structural and electronic properties have been calculated and compared to experiments.