Studies of the structure of spray pyrolysed bioactive glasses using electron diffraction and DFT simulations

Abstract

Bioactive glasses have received considerable attention during the past few decades. Recently, Shih et al. demonstrated that Si‐Ca‐P based glasses prepared by spray pyrolysis can have better bioactivity than glasses prepared by other methods [1]. To understand the reason behind these improved properties we have studied the structure of several Si‐Ca‐P glasses with different compositions. In this study structural information from bioactive glass samples was obtained using electron diffraction. Compared to X‐ray and neutron diffraction electrons can be easily focused on specific nano volumes and used to probe nanoscale variations in structure. Using reduced density function (RDF) analysis local structural parameters of the materials can be extracted from experimental diffraction patterns with high precision [2]. However, experimental diffraction data alone is not sufficient to build reliable atomic models. Therefore, we used DFT molecular dynamics simulations of liquid‐quench to obtain atomic models of the materials, which serve as initial models for further structure refinements. All experimental data presented here was collected using a JEOL JEM2100 transmission electron microscope (TEM) operating at 200kV. Electron diffraction patterns were recorded on a Gatan Orius CCD that eliminates charge overflow to the neighbouring pixels, even when saturated. Using a camera length of ~180mm and the central beam positioned at the edge of the detector, the usable range of scattering vectors, q extends to 18Å ‐1 , comparable with X‐ray experiments. Figure 1 (a‐d) shows typical TEM images and diffraction patterns of two specimens, with the corresponding RDF curves shown in Figure 1 (e). DFT simulations were carried out on ARC supercomputer facilities at the University of Oxford. The CASTEP software [3] was used to perform molecular dynamics simulations of liquid quench from 3000K to 300K, with cooling rate of 2·10 14 K/s. Energy optimizations with a 300eV pseudo potential cut‐off were performed after quenching. Figure 2 (a) shows a model of 60%SiO 2 ‐35%CaO‐5%P 2 O 5 glass simulated by Reverse Monte Carlo using only experimental diffraction data and the model after DFT simulation is presented in Figure 2 (b), indicating considerable structural changes. This model already demonstrates a good agreement with the experimental RDF curve (Figure 2 (c) and has been used for further RMC refinements. We will discuss correlations between bioactivity and structural data in these materials. Financial support from the European Union under the Seventh Framework Program under a contract for an Integrated Infrastructure Initiative (Ref 312483‐ESTEEM2) is gratefully acknowledged.