SURFACE VISCOSITY AND THE MELTING OF GLASS

Daniel R. Swiler

Owens Illinois

OH Perrysburg

ABSTRACT

The extreme corrosion at the melt line of glass tanks has been attributed to the Marangoni effect, with the driving force being glass currents driven by surface tension. It is challenging to use this model to completely describe upward drilling by gas bubbles, downward drilling by metals, and rapid dissolution of porous refractories.

In 1975, Woldemar Weyl discussed how many properties of glass could be explained if the surface of glass had a viscosity lower than the bulk. This concept of surface viscosity is explored to describe refractory corrosion. A viscosity model is proposed where parameters can be modified for the surface as well as the bulk. The model uses the equation: ?(T) = 1/(M(T-T0)exp(-E/(T-T0))), where the E term is linearly correlated to the coordination number of network bonds in the glass network, M is the ability for the structural units to rotate, and T0 is related to bonding in the glass not due to the network bonds.

Accelerated corrosion of glass contact refractories in melting furnaces at the following locations:

• Melt line corrosion
• Corrosion at block joints and cracks
• Corrosion due to porosity in blocks
• Downward drilling due to metals
• Upward drilling due to bubbles
• High temperature regions
• High flow regions

• It will be discussed how the first 5 of these processes are highly affected by surface viscosity, and the challenges to reduce this accelerated corrosion.

GLASS FURNACE CORROSION

The corrosion in glass furnaces contributes significantly to the cost of glass production. In container glasses the life of a furnace ranges between 10 and 20 years. Rebuilding incurs significant capital costs as well as business lost from supply interruptions. The ability to both predict and mitigate the most severe corrosion mechanisms in a furnace is critical for profitable melting furnace operation.

A look inside an end-of-life glass melting furnace leads to the conclusion that bulk corrosion is not the factor that limits its lifetime. Over half the volume of fused cast AZS refractory remains when glass is at risk of penetrating through high corrosion rate regions of the furnace.

The corrosion resistance of AZS refractory in contact with glass is very good due to the high concentration of Zr and Al in the oxide structure, and the presence of a Si oxide glassy phase that minimizes porosity and thermomechanical stresses. A simplified model is that the zirconia is sparingly soluble in glass providing structural rigidity and the alumina increases the viscosity near the glass / refractory interface, slowing the corrosion process.

There are areas in the melter where the refractory rich viscous layer is significantly thinner due to high shear rates at the refractory glass interface. These regions are typically near the doghouse, where batch being pushed into the melter scrapes across the surface of the refractory, around the throat, and near bubblers and electrodes where the glass moves more rapidly.

Other areas that show abnormally high corrosion that are not due to bulk glass flow. The most visible region is along the surface of the melt, or metal line. Less visible corrosion regions are due to block porosity, upward drilling, downward drilling and at refractory joints. These locations of high corrosion are always where three phases are present, the solid refractory, the glass melt, and a phase that is not bonded to either the glass melt or the refractory. Typically, this third phase is a gas, however in downward drilling the phase is a liquid metal.

For more than 50 years the consensus was that this rapid corrosion was due solely to surface tension induced flow, or the Marangoni effect.1 While this mechanism clearly occurs, there are challenges to explain the corrosion pattern and rate based only on this mechanism.

OBSERVED END OF LIFE FURNACE JOINT CORROSION

During end-of-life furnace inspections O-I has identified corrosion at joints and cracks as one of the most critical for furnace life. Earlier work had identified porous blocks additional areas of concern, which has been significantly mitigated by a ground penetrating radar inspection technique that detects and allows us to reject blocks with high levels of porosity.2 Flux line corrosion can be detected by thermal monitoring and can be mitigated by cooling wind and overcoating. This leaves joint and crack corrosion as the major concern in furnace sidewalls.

Figure 1 shows a wall section of an end-of-life furnace. This wall section was scanned using laser point mapping to measure the depth of corrosion. The images in Figure 2 are the result of this mapping, and show the deep penetration of the corrosion front into the joints. The right side of the image shows the joint corrosion visualized from the back side of the blocks. In the inspected furnace joint corrosion in several of these locations are deeper than the block thickness.

Figure 1 Photo of AZS sidewall at furnace end of life

Figure 2 Image of furnace surface created from laser dot mapping. Left is front of block, right is visualization of the surface from the back side.

The shape of the corrosion into the joints is shown in Figure 3. It can be seen that rapid corrosion occurs deep into the joint. This rapid corrosion occurs at the glass-refractory-air interface. Because it is well below the glass surface it poses a higher risk for a catastrophic leak than flux line corrosion. It was our goal to develop a model that would explain this corrosion mechanism.

Figure 3 Corrosion at the joints as seen from the back and bottom of the furnace sidewall

CORROSION AT THREE PHASE INTERFACES

Melt line or metal line corrosion is the most well-known and studied mechanism. This is possibly due to its visibility on the surface of the melt. Cable3 showed a figure with the expected convective currents at the glass line, similar to Figure 4. Cable also noted that the maximum corrosion occurs above the glass level and the region of these convective currents.

Figure 4 Expected convective currents

Pötschke and Brüggmann4 were able to describe the shape of the corrosion curve based upon the assumption that the wetting angle and increase in surface tension at the glass refractory interface defines the shape. This leads to the observed shape as shown in Figure 5.

Figure 5 Observed corrosion shape at metal line.

Their observations were that the corrosion occurred where the glass thickness was expected to be thinnest. The difference between Figure 4 and Figure 5 shows an unresolved problem with this model. How can convective currents produce severe corrosion at very thin layers? This problem becomes even more challenging when applied to upward, downward, and crack and joint corrosion where the difference in composition across the bubble diameter is small, yet corrosion is rapid.

Figure 6 Observed corrosion patterns for upward drilling, downward drilling, and sidewall joint (as seen from above) corrosion.

In the case of upward drilling, downward drilling and corrosion at the end of joints, there is less compositional gradient to drive convective flow, yet this corrosion at the glass surface is much more rapid than at the glass refractory interface.

SURFACE VISCOSITY

The concept of a much more reactive and lower bonding glass structure at the surface was postulated by W.A. Weyl in papers written in 19495 and 1975. 6 Compared to the bulk of the glass, the ions on the surface have incomplete coordination. He expressed this defect state as being related to the polarizability of bonds at the surface, and he made the observation, "No matter how we look upon the subsurface layer, its rheology and chemical reactivity behave as if it had a higher temperature." Swiler7 and Sreeram8 described the change in glass transition temperature as a function of crosslinking of glass forming networks. The relationship between average coordination number and glass transition temperature in Ge-Sb-Se non oxide glass is shown in Figure 7.

Figure 7 Glass transition temperature as a function of crosslinking in Ge-Sb-Se glasses

Use of this concept can be expanded into a model for glass viscosity of silicate glasses. This viscosity model, based on structure, will be applied to understand the high corrosion rate where thin layers of glass are present.

GLASS VISCOSITY

In 1925, Fulcher published his paper proposing a formula for glass viscosity as a function of temperature.9 This formula is currently referred to as the VFT or Vogel Fulcher Tammann formula, and is commonly used by glass technologists.

This equation has the form:

While accurately fitting the form of glass viscosity curves over a wide range of temperatures, it is generally considered to lack theoretical significance.

The MYEGA formula is also a three parameter fit equation, and is based upon a thermodynamic approach to viscosity based upon the number of accessible structures.10 The...