Modeling of disordered materials: radial distribution function vs. vibrational spectra as a protocol for validation

Abstract

As molecular dynamics is increasingly used to characterize non-crystalline materials, it is crucial to verify that the numerical model is accurate enough, consistent with experimental data and can be used to extract various characteristics of disordered systems. In most cases the only derived property used to test the "realism" of the models has been the radial distribution function. We report extensive ab-initio simulation of hydrogenated amorphous silicon that demonstrates that although agreement with the RDF is a necessary requirement, this protocol is insufficient for the validation of a model. We prove that the derivation of vibrational spectra is a more efficient and valid protocol to ensure the reproducibility of macroscopic experimental features.