Читать книгу Encyclopedia of Glass Science, Technology, History, and Culture - Группа авторов - Страница 313

The main advantage of MD simulation is to provide important structural information that complements conclusions drawn from experimental studies. One such important result was the demonstration that the structure of sodium silicate glasses fits the modified random network model because the spatial distribution of sodium ions showed a clustering tendency [10]. The result was especially significant as the extremely high fictive temperatures of glasses quenched in MD simulations strongly favor a much more random distribution.

For such reasons, MD simulations have become the most popular method to study theoretically glass and liquid structures (e.g. [11, 12]). Their main advantage is to yield from the three‐dimensional coordinates calculated for all atoms a variety of structural information that can often be checked against experimental data. In addition, they also provide information that escape experimental determinations and may thus point to the existence of unknown structural features.

Since they deal with the instantaneous state of a system, MD simulations rest on the Lagrangian function L(r, ) of coordinates r and their time derivatives as defined in terms of kinetic (K) and potential (U_p) energies

(12)

and on the Lagrangian equations of motion

(13)

This leads to

(14)

where m(i) is the mass of atom i and

(15)

where f(i) is the force exerted on atom i.

If the coordinates of all atoms are known at time t₀, all the forces exerted and the resulting velocities are calculated with Eq. (15) and then Eq. (14). All the velocities and coordinates after a time step of Δt are updated by the time integration of the equations of motion (14).

When MD methods are applied to glass or disordered systems, several important points should be noted:

1 As the proper choice of the interaction potential model is extremely important, a model with complex interaction potentials may be required if any dynamic structure or property is to be calculated after along with the static structure.

2 The particular starting configuration is not important as long as equilibration time steps are numerous enough at high temperature. Almost the same structural information should be obtained from different initial configurations. If not, the calculated results are unreliable.

Advanced techniques may be used to calculate structure and properties more efficiently. To omit unimportant contributions to the dynamics, one can, for example, keep constant bond lengths such as O─H during the MD calculation. As an alternative to this dynamic constraint method, one can use nonequilibrium MD, which is especially efficient to calculate transport properties such as viscosity. Unlike with conventional MD, a continual friction force can be imposed on the system and its response be monitored.