Amorphous materials are ubiquitous — from the high-tech amorphous silicon in e-book readers to the more mundane glass in windows. Despite their familiarity, there are still many open questions about the properties of even the simplest amorphous materials that require a robust understanding of their atomic-level structures. The key problem is that the lack of long range order (a defining property of amorphous materials) prevents the application of well-established concepts from crystallography — our conventional means of probing atomic-scale structure. Indeed their structural disorder also means diffraction data are not enough on their own to allow unique structure solution. Consequently, one of the key challenges in the field is the development of robust algorithms for solving the structures of disordered materials,
In a recent paper, we have suggested two new general principles that can be used to assess the realism of models of amorphous materials. Crucially, our approach does not assume any prior knowledge regarding the material structure — so it can be applied without bias to disordered materials for which little is currently known. The first principle concerns the extent to which chemically-equivalent atoms are found in similar spatial environments (a similar concept was explored here). The second principle concerns the ‘simplicity’ of the configuration — the extent to which atomic environments are of the highest possible local symmetry permitted by the data. Our two new metrics aim to compensate for the inherent bias in structure refinement methods such as the ‘Reverse Monte Carlo’ (RMC) technique, which tend to favour complex — and hence unrealistic — structural models of disordered materials.