a Combinations of these oxides, plus one.
b Oxide only, keep the error within the given ± range.
c For boron oxide, the factors in the second row are square terms; thus, the sum for each T(n.n) has to be expanded by a term 10 000·a(B2O3)·[y(B2O3)/y(SiO2)]2.
Figure 4 Temperature change brought about by a replacement of 1% of SiO2 by another oxide in the base glass composition 74SiO2, 10CaO, 16Na2O; left side: for oxide amounts by wt; right side: for molar amounts.
3.2 Liquidus Temperatures
Most industrial glass‐melting processes run in a continuous way 365 days per year over periods of 2–15 years. This operation time depends on the type of glass and on the corrosion wear of both the refractory lining of the melting tank and channels guiding the melt to the working stations. A maximum temperature of about 1500 °C is needed to achieve homogenous fusion; at the exit of the furnace, the melt still is at about 1350 °C. This temperature makes it necessary to cool steadily the melt down to T(3.0) while keeping a safety margin ∆T above the liquidus temperature Tliq of the particular composition to prevent precipitation of crystals. This yields a constraint of a minimum temperature of T(3.0) + ΔT required to ensure a crystal‐free glass. This constraint is especially critical for continuous glass fiber production (Chapter 1.5), but it applies to any other process as well before the forming step.
Traditionally, liquidus temperatures have been determined experimentally to be represented graphically for simple‐enough systems in the form of phase diagrams (Chapter 5.2, [2]). For complex compositions of industrial interest, they are generally determined with a gradient furnace whereby a series of 5–10 samples are heated for typically 24 hours in a temperature gradient spanning the expected range of Tliq. After quenching, the samples are examined by optical microscopy. The liquidus temperature is then bracketed by the treatment temperatures of the last homogeneous glass and that of the first sample in which crystallites are observed. In a more accurate approach, samples from the gradient furnace containing tiny amounts of crystals are reheated in a heating‐stage microscope at a rate well below 1 K/min, and the temperature at which the last crystal dissolves is adopted as Tliq.
Thanks to the progress made in thermodynamic modeling of melts (Chapter 5.3), an increasingly useful approach is to predict liquidus temperatures with one of the dedicated softwares designed to calculate phase equilibria relevant to glass making (e.g. 20). Empirically, however, simple rules have long been known to predict the effects on a given oxide on liquidus temperatures. Illustrating again the predominantly pairwise nature of atomic interactions, they rely particularly on the topology of binary phase diagrams (Figure 5). The tremendous freezing‐point depressions of SiO2 brought about by addition of alkali oxides as well as by boron and lead oxides are in fact so conspicuous that they have historically been at the basis of the development of glass technology. As illustrated by a comparison made between Li2O and K2O (Figures 4 and 5), of particular interest is the fact that the alkali oxide that most strongly decreases viscosity is the least effective in lowering liquidus. Only boron and lead oxides cause strong decreases of both viscosity and liquidus. This may shed some light on the technological challenge raised by the replacement of lead oxide in the glass formulae of modern tableware and solder glasses.
In ternary systems, the Na2O–CaO–SiO2 and CaO–Al2O3–SiO2 phase diagrams are especially important as they serve as references for soda‐lime and most stonewool and reinforcement‐fiber glasses, respectively (Figure 6). From Figure 6a, it is easy to understand why, in Antiquity (Chapter 10.3), glasses with silica contents of 70–74 wt % and amounts of lime not exceeding 12 wt % were already mass produced: Thanks to typical T(3.0) values of 1200 ± 10 °C, it is comparatively easy to comply for them with the constraint Tmin = T(3.0) + ΔT. The particular composition “16‐10‐74” has been investigated in many scientific studies. It may be considered as a reference and the mother of all mass‐produced glasses, including of course today's float glass. That the Tmin constraint represents in contrast a real challenge for CaO–Al2O3–SiO2‐based glasses is readily apparent in Figure 6b where only a narrow range around the ternary eutectic between tridymite