Читать книгу Encyclopedia of Glass Science, Technology, History, and Culture - Группа авторов - Страница 330

Ab initio simulations of glasses have now become standard to determine observables of high practical relevance, but usually inaccessible to simulations with effective potentials. Despite their computational cost, they constitute a most useful tool to gain insights into microscopic properties of glasses, especially when their composition is complex, and even to develop novel types of glasses with exotic compositions. Their accuracy makes them most useful when a multitude of possible local atomic environments coexist either in the bulk or at glass surface or when important structural heterogeneities are found at a microscopic scale. This applies, for instance, to the properties of the charged and neutral defects in amorphous silica, as their presence can strongly affect the performances of electronic and optical devices based on SiO₂ [30]. The technologically important chalcogenide glasses (Chapter 6.5) can only be simulated with force‐fields accounting correctly for homopolar bonds, i.e. ab initio simulations. Because chalcogenides do not contain oxygen, an element that is relatively costly to deal with, they can be investigated with smaller efforts per particle and, thus, quite large systems sizes and accessible timescales [31–33].

Other important topics are the effects of water on bulk and surface properties because the high reactivity of hydrogen induces local modification of the structure which can change the properties of the glass even at very low concentrations. The difficulty is then that fairly large systems must be investigated, which is the reason why relatively few simulations have been done on complex water‐bearing glasses [9, 15]. At present, the main problem faced indeed remains the large computational effort required and hence more efficient methods are being devised to reduce the computing time. Examples are the so‐called “order‐N” algorithms, with which the computational cost increases only linearly with the system size [34], or the second‐generation Car–Parrinello approach [35]. Although these advances will make it possible to deal with much larger systems, including, for instance, 10⁴ oxygen atoms, it is unlikely that they will allow to gain one more order of magnitude in system size. To study problems like the mechanical behavior of glasses on the mesoscale, simulations will thus have to be done with effective potentials. But even in these cases ab initio simulations will be very useful by providing more accurate potential parameters [36–38].