Atomic-scale description of interfaces in rutile/sodium silicate glass–crystal composites

Abstract

In this work interfaces between (Na2O)x(SiO2)1-x glasses (for x = 0.0, 0.1 and 0.2) and TiO2 crystals are simulated using molecular dynamics and empirical potentials. Interfaces are presented for the distinct terminating surfaces of TiO2 with Miller indices ≤2, the properties of which have been investigated using atomistic models. Simulations showed that partially ordered layers had been induced in the glass close to the interfaces, with successive oxygen-rich and cation-rich planes being noted. The first silicate layer in contact with the crystal tended to be highly-structured, with Si ions occupying well-defined positions that depend on the orientation of the crystal at the interface, and showing 2-dimensional ordering depending on glass composition. Finally, interface energies were calculated. These indicated that the interface formation may stabilise a crystal surface in comparison to maintaining a free surface. Results are presented suggesting that the structural flexibility of the glass network allows it to conform to the crystal, thereby providing charge compensation and avoiding large relaxation of the crystal structure close to the interfaces. Such interfacial properties could be crucial to improving phenomenological models of glass-crystal composite properties.

Atomistic modeling approach to the thermodynamics of sodium silicate glasses

“…The BO spectrum is similar to the O spectrum from SiO 2 . The NBO modes, however, do not contribute to the peaks (2) and 3, and have a high-frequency peak at a different energy than BOs. This results in a broadened high-energy band in the total spectrum.…”

“…Some of the information calculated using LD can also be retrieved from the analysis of ions trajectories resulting from MD simulations. The normal mode DoS, which can be obtained from the ionic velocities This can be used to determine constant-volume heat capacity and vibrational entropy using equations similar to that of LD (2) and 3, respectively and The main limitation compared with LD is that the modes cannot be separated: the mode frequencies and coordinates are unaccessible. However, projected densities of states can still be obtained by calculating g(ω) separately for different groups of atoms.…”

“…Typically, this problem is approached from a numerical perspective with either Density Functional Theory (DFT) models, or classical empirical force fields. Both classes of models are commonly used to their respective strengths, the former yielding more accurate results, whereas the latter enable simulation of phenomena with larger characteristic time ( eg, diffusion 1 ) or length scales (eg, interfaces 2 ). Because of the size of the structure required to represent accurately a random glass network, most of the studies of vibration properties are based on empirical potentials, 3‐10 although some DFT calculations have also been carried out with smaller supercells 11‐14 .…”

“…have the right energies and the right relative amplitudes, apart from peak (2). Its other limitation is its inability to reproduce the split of peak (4).…”

“…Its other limitation is its inability to reproduce the split of peak (4). In contrast, BKS offers a better description of the high-energy B 4 band, at the expense of a worse behavior at low energies and an almost inexistent peak (2). The spectra from the other potentials are all significantly worse.…”