Abstract
The corrosion of float glass is a big problem in the glass industry. It manifests itself during exposure of glass to humid conditions or when in contact with water. The corroded surface impairs the adhesion of coatings and causes optical defects at the glass surface. Glass corrosion is a complex phenomenon involving diffusion and ion exchange accompanied by silica network dissolution and the formation of corrosion products. Treatment of the glass surface with zinc ions has been proposed to inhibit the corrosion process, however little rigorous study on the inhibition of glass corrosion using zinc has been published. In this current work the corrosion of float glass in contact with humid air has been studied by a gravimetric technique. This has also been investigated by surface analysis techniques (SIMS, ion-beam analysis, XPS and AFM). A numerical simulation of water uptake due to glass corrosion has also been developed, based on a Cellular Automaton approach. The gravimetric analysis was performed on crushed float glass, enabling direct quantification of corrosion in real time by running experiments that last for several days. The primary assumption for the measurement of corrosion is that the mass uptake by the sample is purely a result of water adsorption to the surface, where the water adsorbed is replacing the water that has diffused into the bulk glass. Results of the gravimetric studies showed increased corrosion for higher temperatures and higher relative humidity values. Corrosion flux values were obtained by evaluating the specific surface area of the samples from BET analysis of water sorption isotherms. A striking effect of zinc treatment was observed. The corrosion flux value for an untreated sample was 11.8 times higher that of a zinc treated sample exposed to 15C, 9.25 times higher at 20C and 2.85 times higher at 30C. The effectiveness of zinc treatment therefore appears to decrease with increasing temperature. From the three temperatures studied, it appears that the corrosion flux follows an Arrhenius relationship with temperature, consistent with the published behaviour of water diffusion through glass. Values of the activation energy (Ea) were obtained. For untreated glass the value of Ea recorded in this present work was 50.9 kJ.mol-1, comparing favourably with previously published values for float glass. The value of Ea obtained for zinc-treated glass was 122.3 kJ.mol-1. This increase in activation energy gives further quantification of the effectiveness of zinc treatment. The beneficial effects of zinc treatment were also apparent from the significant reduction in water mass uptake for zinc-treated samples when measuring water adsorption isotherms. Furthermore, SIMS surface analysis of an untreated glass surface showed much more sodium depletion after exposure to saturated air, compared to zinc-treated glass. Previous work has attributed the effectiveness of zinc treatment to zinc ions blocking the silica interstices, thus inhibiting the ion-diffusion process. However the current work reveals that zinc treatment effectively reduces the amount of water adsorbed. It is therefore proposed that zinc ions combine with the surface hydroxyl groups thus rendering the surface less hydrophilic. Gravimetric analysis suggests that zinc treated glass surfaces lose their corrosion protection with time. This ageing effect is supported by data from surface analysis studies, indicating that zinc ions diffuse into the glass over long periods of time. Further work is recommended to characterise this effect. During routine water adsorption isotherm measurement for zinc-treated glass, an anomalous decrease in mass was observed at a relative humidity of between 30 and 40%. No explanation can currently be offered to explain this phenomenon and further work is therefore required. A Cellular Automaton numerical simulation technique has been applied to model float glass corrosion. This technique was found to give simulation results that compared favourably to experimental data for mass uptake of water. The CA technique can readily be adapted to include other mechanisms and therefore it is recommended for further use in the field of glass corrosion. A sensitivity analysis revealed that the CA model was mostly influenced by parameters relating to the gas phase and surface reaction. The model was insensitive to parameters relating to the bulk glass phase, including diffusivity. Further development of the model is recommended.