Figure 5 Liquidus lines of binary silicate systems (left: by wt, right: by mol); all systems comprising a divalent oxide, except BaO, show an extended stable miscibility gap; data source [2].
Figure 6 Ternary phase diagrams in versions of technological relevance; shorthand notation: N = Na2O, C = CaO, A = Al2O3, S = SiO2, Q = quartz, TR = tridymite, CR = cristbobalite; left: the basic system of all commercial hollowware and flat glasses; the triangles mark the positions of the compounds, Na2O·2 SiO2: the circle the position of the base glass 74 SiO2, 10 CaO, 16 Na2O; data source [2]. Right: the basic system of reinforcement‐fiber glasses; industrial compositions flock around the eutectic; calculation using FactSage® [23].
Figure 7 Miscibility gaps. (a) Extension of stable gaps in ternary borosilicate systems with different oxides as third component; the area shaded in gray refers to BaO; (b) isotherms of the (metastable) sub‐liquidus immiscibility dome in the system Na2O–B2O3–SiO2.
3.3 Liquid–liquid unmixing
If phase separation within a condensed system most commonly takes place via partial crystallization, it can also occur as liquid–liquid unmixing (Chapter 5.2). Ternary systems containing boron oxide illustrate that the phenomenon should certainly not be overlooked in glass‐forming systems. For ternary borosilicates, the boundaries of the composition domains where such an unmixing takes place stably, i.e. above the liquidus, are indicated in Figure 7a. If temperature is represented in a third dimension, these domains define the base areas of immiscibility domes that eventually terminate at upper critical points at their tops. The isotherms of the immiscibility dome in the system Na2O–B2O3–SiO2, which is the base composition of all borosilicate glasses, are drawn in Figure 7b. Here, in contrast to Figure 7a, the entire dome comprising its upper critical point (755 °C at composition 25‐05‐70 by wt) is located below the liquidus surface. Hence, liquid unmixing cannot take place during the initial melting step, but at lower temperatures during the forming process. This is one of the reasons why, in order to minimize phase‐separation effects, pharmaceutical and low‐expansion borosilicate glasses are designed around a composition of 80 wt % silica. The system Li2O–B2O3–SiO2 (not shown here) displays a similar topology.
It is only with glasses known under the trade name Vycor Glass that liquid unmixing is exploited on purpose. Here, after forming by conventional technology to the desired shape, phase separation develops upon annealing at an appropriate temperature to yield two interconnected phases, namely an Na2O‐ and B2O3‐rich glass along with another one that contains more than 96 wt % SiO2. Then the former is leached out by a hot strong mineral acid, leaving behind a nanoporous skeleton of high‐SiO2 glass. This material may then be used directly as filter, for example, or sintered at temperatures below 1300 °C to fabricate dense and almost pure silica glass articles much more readily than with pure SiO2.
The numerical calculation of liquid–liquid immiscibility ranges in multicomponent systems (e.g. by using the software and databases mentioned in Section 3.2) is even more challenging than the calculation of solid–liquid equilibria. This is because available experimental data hardly reach beyond what has been sketched in the present section, and such a narrow base of information does not allow to fine‐tune the parameters used in the calculations.
4 Glass Composition – its Relevance to Glass Properties
4.1 Property Optimization
Both search and optimization of glass formulae begin with a given profile of target glass properties. In the following, three properties will be addressed as examples typically targeted in glass development, namely the elastic properties, the thermal expansion coefficient, and the chemical durability. From a scientific point of view, such a task should rest on deep insights on the relationships between chemical composition, glass structure, and glass properties. It is only from such a fundamental approach that ground‐breaking developments of novel glasses with outstanding properties may be expected. But this goal is still a matter of fundamental research as expounded in the following chapters where this most challenging issue is pursued.
For the time being, however, only few manageable tools and procedures of this kind are available for the technological community. To optimize properties, technologists thus rely largely on empirical approaches whereby, as applied to glass viscosity in Section 3.1, they use incremental oxide factors derived by statistical means from large numbers of experiments. One has, however, to keep in mind that these approaches represent only interpolations of what is already known. Hence, limited areas in compositional space leading to truly outstanding properties should be easily overlooked so that developments similar to the famous low‐expansion metallic alloy Invar are very unlikely to be found this way.
4.2 Elastic Properties
Incremental oxide factors for the calculation of the elastic properties from the composition compiled in the right‐hand part of Table 3 are taken from a widely accepted earlier publication [21]; for the sake of clarity, they have been adjusted with respect to the units used, i.e. to cm3/mol for volume, and to GPa for modulus increments. Young's modulus E is then calculated with
Figure 8 Change of Young's modulus E in the base glass composition 74 SiO2