Minimizing defects is an integral part of producing high-quality glasses.
Figure 1. Group of 11 stone defect samples (red circles) embedded in plate glass pieces.
Figure 2. The defects exhibited a white, cloudy morphology with a “wispy” pattern.
Glass defects are typically unavoidable, but their occurrence must be minimized during the production of high-quality glasses. Common glass defects include bubbles, stones, knots, cords and cat scratches. One unusual stone defect in a pattern glass was recently discovered, and a scrupulous methodology was necessary to investigate its source.
Numerous plate glass samples with white-colored stone defects were received for analyses (see Figure 1). These defects were generated in an end-fired furnace that pulls approximately 240 metric tons of soda-lime pattern glass per day. The relatively new furnace has been in operation for only 18 months. The hot spot temperature was reported in the range of 1540 to 1570°C, while the refiner temperature was in the 1240 to 1250°C range. The stone defects were unusually large-greater than 2 mm-and showed a “wispy” milky-white pattern (see Figure 2).
Figure 3. SEM backscatter electron image of white defects. The beads’ interiors are an alumina/silica/zirconia composition with no sodium (1), while their outer zone contains some sodium diffused from tank glass (2). The zirconia (ZrO2) snowflake crystals are remnants of beads after alumina and silica dissolved into the tank glass.
Polished cross-sections of several representative samples of each stone type were prepared for analyses by the reflected light microscopy, scanning electron microscopy (SEM) and energy dispersive X-ray (EDS) methods. Figure 3 shows a backscatter SEM image of the milky-white defects embedded in the host glass. This image reveals a collection of mostly rounded or spherical features within the stone body.
The defect’s chemistry was measured across its width (i.e., from the defect core to its edge). The defect core consisted of alumina (Al2O3), silica (SiO2) and zirconia (ZrO2). The immediate outer zone also contained alumina, silica and zirconia, but it contained a small amount of sodium oxide (Na2O) as well (see Figure 3). Finally, the outermost zone mostly consisted of fine snowflake-type zirconia crystals that were carried away by the molten glass circulation; this resulted in the formation of the wispy tail structure from the stone. Table 1 shows the chemistry of each phase measured with EDS.
Figure 4. High-temperature insulation fiberglass blanket roll.
The remaining white stones that were not prepared for SEM/EDS analyses were cut out from their plate glass pieces, crushed and analyzed by X-ray diffraction for phase identification. The results showed several weak diffraction lines for tetragonal zirconia superimposed on a broad diffraction band, which was indicative of an amorphous (glassy) structure. Since stone defects are typically crystalline, the amorphous nature of these stones added to the mystery of their source.
Table 1. Defect chemistry as measured by SEM/EDS.
The defects were unusual in that each milky-white stone did not contain just one large solid inclusion. Instead, the bulk of the defect consisted of several small stones that were essentially round and-in some cases-like spherical beads. These small stones, ranging from about 100 to 200 microns, displayed an amorphous core consisting of Al2O3, SiO2 and ZrO2(34%, 51% and 14%, respectively), as well as a low level of Na2O in the outer zone. The extreme outer zone consisted of fine snowflake-type zirconia crystals.
With these results in hand, the first thoughts about the likely source of the defects focused on fusion-cast alumina-zirconia-silica (AZS) refractory. However, this possibility was quickly dismissed due to the absence of Na2O and much higher level of Al2O3 in the defect core; these measurements are not consistent with what would typically be found in defects sourced from AZS. The low level of Na2O in the outer zone of the defect was attributed to diffusion from the host glass during the residence time of the defect in the furnace. Bonded zircon-mullite refractory was also considered as a possibility, but it was ruled out due to the absence of zircon pseudomorph structure around the zirconia grains.
The unusual structure and chemistry of this defect was puzzling, as no refractory or raw material source made sense. The focus of the examination then turned to other ceramic materials used in furnace construction, such as insulation fiber blankets. Ceramic fiber blankets are commonly used for the insulation of glass furnaces to minimize heat losses (see Figure 4). Most high-temperature insulating fiber blankets have high alumina content or some ratio of alumina and silica. However, an AZS-based fiber blanket is also used as insulation in glass tank furnaces.
Figure 5. Stereomicroscope image of fiber blanket sample with noticeable shot pieces, many of which are beads.
In order to confirm AZS fibers as the likely source of the defects, samples of fibers were first analyzed in their original condition using SEM/EDS. Fiber blankets are generally spun into fibers of less than four microns in diameter from a glass melt of specific alumina-silica based chemistry. Typically, a fiber blanket also contains rounded, non-fiber pieces called “shot.” They are made of the same composition as the fibers and can be up to 100 microns in diameter. These undesirable shot pieces reflect portions of the fiberglass melt that failed to spin into a fiber.
Figures 5 and 6 are stereomicroscope and SEM secondary electron images of an AZS fiber blanket that clearly exhibit shot particles. The chemical similarities between the AZS fiber and the defects, as well as the dimensional similarity between the fiber shot and defect beads, strongly suggested an AZS high-temperature insulating fiber as the source of the defects.
Figure 6. Higher-magnification image showing individual shot. The diameter of the spherical shot particle (~100 µm) is similar to the defect beads.
Next, a commercial sample of high-temperature insulating blanket (Al2O3, 30%; SiO2, 53%; and ZrO2, 16%) was used in the lab to attempt to recreate the defects. In this test, pieces of fiber were mixed with defect-free parts of the glass samples received from the customer and melted in a fusion-cast AZS crucible at 1300oC. After the test, the samples were prepared for microscopy analyses.
Figure 7 shows a backscatter image (SEM) of a sample that was exposed for 19 hours. The results showed features identical to those seen in the defects with the wispy tails and fine-grained zirconia crystals. The level of corrosion was more advanced than that of the actual defect sample; however, pseudomorphs of the original shot features with a similar EDS spectrum to that of the defect chemistry were evident. Analyses of the shot by SEM/EDS methods in the as-received fiber sample and lab-tested sample matched that of the beads in the defects.
Figure 7. Microstructure and chemistry of AZS fiber sample after crucible test in plate glass after 19 hours at 1300°C. Fibers and shot fragments are breaking down to zirconia snowflake crystals, while shot sphere remnants are also evident.
Several unusually large stone defects in a pattern glass were analyzed. The white stones all contained alumina/silica/zirconia amorphous and rounded beads with fine, tetragonal zirconia crystals that formed wispy tails from the outer edges.
The common sources of a zirconia defect, such as fusion-cast AZS or bonded AZS refractory lining or raw batch, were ruled out. The stones were found to have a near-perfect chemistry match to a commercially available AZS fiber blanket sample. The shot size in the as-received fiber also matched the rounded bead size in the defect.
The lab experiment duplicated the breakdown structure of the fiber with the wispy tails in the glass. Consequently, evaluations and comparisons of the chemical and microscopic analyses of the defect samples, as-received AZS high-temperature insulation, and lab-tested high-temperature insulation fiber, confirmed-with a good degree of confidence-that the defect source was alumina/silica/zirconia high-temperature insulation fiber exposed to the tank glass melt.
It was later confirmed that the pattern glass manufacturing company did indeed use AZS insulation fiber over open gaps of the feeder channel crown. For more information regarding glass defect analysis, contact RHI Monofrax Ltd. at 1870 New York Ave., Falconer, NY 14733; (716) 483-7238; fax (716) 655-2478; e-mail firstname.lastname@example.org; or visit www.rhi-monofrax.com.