The liquidus temperatures and viscosities of the primary melts forming during the early stages of batch melting are plotted in Figure 3 in the cases of binary salt‐like melts formed from soda ash and another compound (Figure 3a) and of the systems Na2O–SiO2, Na2B4O7–SiO2, and B2O3–SiO2 (Figure 3b).
Owing to their extremely low viscosities, the salt‐like primary melts play an important role in bringing about high turnover rates during batch melting. When they are lacking, as in alkali and boron‐free continuous fiber glasses, the products in contrast remain in a granular state until they reach their lowest eutectic temperature (compare with Figure 6b in Chapter 6.1).
These different stages of early batch melting are sketched in Figure 4 for a soda lime silicate glass batch. Before a soda‐ash melt forms, the batch remains in a granular state (Figure 4a). Then, a primary salt‐like, low‐viscosity melt rapidly spreads throughout the batch, thereby wetting the solid grains (Figure 4b). This stage is characterized by a large ratio between the liquid interface and the melt volume. It is predominantly at this time that diffusion paths for oxygen exchange are short and that the surrounding atmosphere effectively interacts with the melt to determine its final redox state.
Upon further melting (Figure 4c, d), the melt becomes increasingly viscous and the ratio of liquid interface to melt volume decreases. In a real batch heap, stages (a) to (d) proceed longitudinally along the L axis (see Figure 2) and vertically from the outside to the inside of the batch. The batch melts from both its top and bottom side, typically in an almost symmetrical way as the heat fluxes from above and below are of the same order of magnitude. The release of gases is in contrast asymmetric since those from the upper parts readily escape whereas those coming from the lower parts remain trapped below the batch. In a successful primary melting process, the majority of solids are digested, only a minor part being released to the rough melt. This requires good batch mixing and a well‐balanced granulometry of the raw materials. The issue can be tested at the lab scale by so‐called batch‐free time crucible tests. In these simple tests, batch samples of 50–100 g are exposed to a laboratory furnace at 1400 °C and the progress of melting is inspected visually after a given time.
Figure 4 Early stages of batch melting, manually sketched after the scanning electron microscopy micrograph. (a) Open‐pore stage with granular solids and gas, the gas composition being dominated by the equilibrium between CO2 and O2 from trapped air and the furnace atmosphere. (b) Closed‐pore stage with the development of a widespread primary liquid, a large ratio s of effective liquid interface (solid/liquid and solid/gas) and liquid volume, and a gas composition dominated by CO2, redox active materials, and polyvalent ions in the primary melt. (c) Reaction‐foam stage characterized by large volumes of granular solids, bubbles, and melt, and by progressive melting of solids and decreasing s ratios. (d) Rough‐melt stage, the melt being the predominant phase coexisting with considerable amounts of bubbles and undissolved grains and showing on top a seam of the primary foam formed.
4.3 Sand Dissolution
All solids surviving primary batch melting have to dissolve in the viscous rough melt by slow diffusion processes under comparatively low driving chemical forces. This is one of the reasons why, even today, long dwell times are required for the fusion process. By mass, the sand represents the major part of solids that have to dissolve in this way. The process suffers from an especially unfavorable feature (Figure 5): the decrease of the silica concentration from the sand grain to the melt phase represents a strong chemical gradient that causes the grain to be surrounded by a seam of melt with a high viscosity and a low basicity. This gradient affects not only mass transport but also the solubility of gases, which generally decreases with decreasing basicity (cf. Chapter 5.7). Thus, gases dissolved in the rough melt tend to form bubbles around a dissolving sand grain.
Figure 5 Schematic view of a dissolving sand grain; the grain is surrounded by a solid reaction layer (e.g. tridymite) followed by a liquid high‐viscosity diffusion seam with decreasing SiO2 concentration, hence decreasing acidity, from inside to outside; gas bubbles – mostly O2 – precipitate at the interface solid/liquid; upon complete dissolution of the sand grain, a bubble cluster remains in the melt.
In addition, temperature‐induced reduction of ferric iron takes place as described by the reaction
(2)
describing how firm [Fe3+O4] oxygen complexes give rise to the weak [Fe2+O6] complexes formed by ferrous iron. The equilibrium constant of the reaction is given by
(3)
so that, at constant redox state, tiny oxygen bubbles emerge at the boundary of the dissolving grain. Any dissolving sand grain leaves behind it a cluster of small bubbles, removal of these bubbles makes sense only if their generation is over. This is one of the reasons why sand dissolution and the fining process need to take place in separate parts of the furnace.
In summary, successful sand dissolution is a prerequisite for successful fining. Even apparently small differences in the grain‐size distributions of sands have a big impact in this respect. This statement will be demonstrated for two different sands. Let us assume that a spherical sand grain with radius r dissolves according to Jander's kinetics:
(4)