packages that have been developed over the years. Various groups use different approaches to maintain and develop these packages, the most popular ones being CPMD, VASP, Quantum Espresso, CP2K, Siesta, CASTEP, etc. (see on http://www.psi‐k.org/codes.shtml for a more extended list). Each of them has advantages and disadvantages regarding the scaling of computational effort with system size, accuracy, ensembles that can be simulated (microcanonical, canonical, constant pressure, etc.), quantities that can be calculated, etc., and therefore the best choice will depend on the application.
To give an idea of what can be done at present with ab initio simulations, we compare in Table 1 ab initio and classical simulations. Since the computational load for the former does depend on the system considered (owing to the different electronic structure for the atoms), we will consider the example of an oxide glass. We note that, for a given computing time, there is trade‐off between the size of the system and the time span covered by the calculation. For a classical system the relevant number is basically the product of the two quantities so that doubling the system size increases the computer time by a factor of two. For ab initio simulations the situation is not that favorable since doubling of the system size usually leads to an increase of the computational load by a factor of 2α , where α is about 3. As a consequence, system sizes are smaller and accessible timescales are shorter in ab initio than in classical simulations. Therefore, the quench rates used to cool a system from its liquid to its glass state are usually on the order of 1013–1015 K/s. Such rates are thus significantly larger than those of classical simulations (1010–1014 K/s) and of course ways higher than experimental values (10−2 − 106 K/s). Despite these huge differences, the resulting glasses are surprisingly similar since many of their properties depend only in a logarithmic way on these rates. Therefore, it does make sense to use ab initio simulations to investigate the properties of glasses on the microscopic scale.
Table 1 Comparison of various features of large‐scale computer simulations carried out with classical and ab initio methods.
| Classical MD | Ab Initio MD | |
|---|---|---|
| Size | 1 000–500 000 atoms | 100–500 atoms |
| Box size | ∼100 Å | ∼15–20 Å |
| Trajectory length | ∼1 ns–10 μs | ∼20–100 ps |
| Transferability | No | Yes |
3 Structural Properties
In this section we will present some examples to illustrate how ab initio simulations can help to gain a better understanding of the local structure of complex glasses. The very first ab initio MD simulations for a glass‐former were carried out by Sarnthein et al. in 1995 on silica [7]. Using the Car–Parrinello approach, they generated a glass model by equilibrating for 10 ps a liquid sample with 72 atoms (!) at 3500 K and subsequently quenching it to 300 K. Despite the smallness of the system and the high quench rate (1015 K/s!), the resulting glass structure was surprisingly similar to that of the real material with a network of SiO4 corner‐sharing tetrahedra and a neutron structure factor compatible with the that from scattering experiments. These authors also found that their electronic density of states matched well the X‐ray photoelectron spectroscopy data although the predicted band gap of 5.6 eV underestimated the experimental value of about 9 eV, a flaw that often occurs in DFT calculations [2].
This pioneering paper was followed by further investigations in which more complex glass‐formers were studied, such as alkali silicate glasses, calcium‐alumino‐silicates [8–11], and other glass‐formers [12, 13]. These investigations allowed to obtain detailed insight into the local arrangement of the atoms, how network modifiers like Na or Ca change these arrangements, and the connections between local structure in real space with structural features as determined in neutron or X‐ray scattering experiments.
As an example we will briefly discuss some results obtained for sodium borosilicate of composition 30% Na2O–10% B2O3–60% SiO2 (in mol %), an important glass‐former in which boron atoms bond to either three or four oxygen neighbors, which makes its structure rather complex, Chapter 7.6, [6]. One important quantity to characterize the structure is the partial radial distribution functions gαβ(R) which is directly proportional to the probability that two atoms of type α and β are found at a distance r from each other (Chapter 2.2). Thus, this function is defined as [14]
(8)
where 〈.〉 represents the thermal average, V is the volume of the simulation box, Nα is the number of particles of species α, and δαβ is the Kronecker delta.
For the boron‐oxygen pair (Figure 1a), the first peak of the radial distribution function is relatively large and slightly asymmetric. By considering three‐ and fourfold coordinated boron atoms, [3]B and [4]B, respectively, we can decompose this peak into two contributions and see that the broadening of the total peak is a consequence of the presence of these two populations, the [3]B─O and [4]B─O bonds giving rise to the smaller and larger distances, respectively. We can decompose these distributions further by considering the nature of the second nearest neighbor of the central boron atoms, i.e. whether the nearest oxygen is bridging connected to a Si or B atom, or whether it is a non‐bridging oxygen. We see that for the case of the [3]B atoms, the nature of this second nearest neighbor species influences the B─O distance since the position of the corresponding peak depends on the species, whereas this is not the case for the length of the [4]B─O bonds. The presence of [3]B and [4]B units is also reflected in the OBO triplet angle distribution (shown in Figure 1b) by the presence of two peaks.
Figure 1 Contributions of the various structural units of a sodium borosilicate glass to structural features [6]. (a) Distribution of B─O distances around a given B atom as given by the first peak of the partial radial‐distribution function. (b) Distribution of the angle formed by two oxygens connected to the same boron.
Although this detailed information is valuable to understand better the local structure of the glass, it is also important to connect the results from the simulation with experimental data. Even if radial distribution functions cannot be directly accessed experimentally, it is possible to measure the static structure factor, which
