Structural evolution of silica: molecular dynamics comparison of monomer-derived and melt-quenched networks with experimental support from rice husk

Abstract

Molecular dynamics (MD) simulations of silica (SiO2) crystallization remain challenging due to high energy barriers that frequently trap systems in amorphous states. Although experimental synthesis commonly relies on pH-controlled sol-gel routes, computational studies predominantly employ extreme melt-quench protocols. Here, we systematically compare two computational pathways for silica network generation: a Monomer-Derived Bottom-Up Route modeled using the Takada potential and Well-Tempered Metadynamics, and a top-down melt-quench approach utilizing Vashishta and BKS potentials. Structural analysis shows that the biased monomer-derived route yields an amorphous network dominated by cristobalite-like 6-membered rings (77.0%), with a Si–O bond distance of 1.621 ± 0.006 ˚A upon cooling to 300 K. Further analysis of the underlying free-energy landscape demonstrates why the applied bias failed to induce long-range order: deposition did not converge within 5 ns, and the biased Steinhardt coordinates showed degeneracy by reaching crystalline values within a network that remained macroscopically vitreous. These computational findings are supported by experimental sol-gel synthesis using rice husk-derived silica, which confirms that neutral conditions (pH 7) optimally promote cristobalite phase formation after calcination. By aligning simulated structural evolution with empirical observations, this study provides actionable guidance for selecting appropriate interatomic potentials and sampling strategies in future silica crystallization modeling.