Dependence of a cooling rate on structural and vibrational properties of amorphous silicon: A neural network potential-based molecular dynamics study

Amorphous materials have variable structural order, which has a significant influence on their electronic, transport, and thermal properties. However, this difference in structure has rarely been investigated by atomistic modeling. In this study, a high-quality machine-learning-based interatomic potential was used to generate a series of atomic structures of amorphous silicon with different degrees of disorder by simulated cooling from the melt with different cooling rates (10(11)-10(15) K/s). We found that the short- and intermediate-range orders are enhanced with decreasing cooling rate, and the influence of the structural order change is in excellent agreement with the experimental annealing process in terms of the structural, energetic, and vibrational properties. In addition, by comparing the excess energies, structure factors, radial distribution functions, phonon densities of states, and Raman spectra, it is possible to determine the corresponding theoretical model for experimental samples prepared with a certain method and thermal history.