CHALLENGES AND PROGRESS IN UNDERSTANDING GLASS MELTING

Mathieu Hubert and Irene Peterson

Corning Research and Development Corporation

Painted Post, NY, USA

ABSTRACT

Glass is one of the oldest materials manufactured by mankind. However, many aspects of the reactions that convert batch materials to the melt are still only partially understood. Improvements to fundamental understanding on how different batch material characteristics, tank process variables and atmosphere control the melt evolution are needed to allow industry to produce high quality glass in the most efficient way, while reducing environmental impact. Recent improvements in measurement techniques have driven progress in understanding of the reaction pathway and kinetics of the batch-to-melt conversion on the laboratory scale. The development of advanced computing tools has increased the ability to visualize heat and mass flow inside the production tanks on the macro-scale. However, significant gaps remain in the ability to scale up experimental results from small-scale testing to production, availability of robust in-situ measurements for production tanks, and in incorporation of experimental data on the batch-to-melt conversion into mathematical models of glass production.

INTRODUCTION

For millennia, highly skilled craftsman improved glass quality through their choices of batch materials, mixing methods and thermal treatments. Advances in technology were closely guarded secrets. More recently, this responsibility has shifted to scientists and engineers. Sharing of information between academia, industry and national laboratories disseminates improvements throughout the world. Advances in techniques for measurements and calculations have led to tremendous improvements in glass quality, expanded the breadth of chemistries produced, and improved the efficiency of energy utilization. Quality, as measured by visible bubbles and seeds, inhomogeneity, surface staining and durability have improved significantly, along with product yields - while energy requirements and furnace emissions have decreased [1].

A large and growing number of experimental techniques are in use to follow, quantify and understand the effect of batch material chemistry and particle geometry, heating and cooling profiles, oxidation state and furnace atmosphere on glass formation. Because of the multi-scale, interactive and complex nature of the batch-to-glass conversion process in a production tank, a variety of different techniques are used together to illuminate different parts of the process. This paper, while not meant to be a comprehensive review, will discuss a few of the methods in most common use, and present some recently developed techniques.

Because of the multiple variables at different scales that control the melting process, modeling is a critical part of building a useful and practical understanding of melting behavior. The ultimate goal is to use data gathered in the lab to build new math models, which would be able to combine the effects of the different variables from the micro-scale to the tank scale to find the best melting solutions.

A MULTISCALE CHALLENGE

The conversion of a mixture of raw materials into a homogeneous melt involves a large number of different and interacting variables which act on scales from the microscopic to the tank. The same final carbonates or minerals can be used as an alkali source. Each batch material powder has a size distribution and a particle shape. The primary particles of a batch material can be agglomerated by spray drying, or particles of different materials can be mixed and compacted into pellets or briquettes.

A typical glass melting batch will include major components that make up a relatively large proportion of the batch, minor components to adjust the properties of the glass, and additives in very small amounts to assist fining and control the oxidation state. For example, a soda lime silicate batch can contain sand as 50-75 weight % of the total, minor additions of soda ash and limestone, and additives such as sodium sulfate for fining (typically <1 wt%), nitrate as an oxidizing agent, or coke as a reducing agent (usually <0.5 wt%). Despite their small percentages, additives that control the oxidation state of the melt have a considerable impact on fining behavior [2] and the color of the glass. The redox state of the melt changes its heat transfer behavior and thermal profile in the melting tank, and thus the energy required to melt a given glass; more details are given in [3]. As shown by this example, changes to any of the batch materials can have unforeseen and far-reaching consequences for the rest of the process.

The batch materials undergo a complex sequence of reactions on the pathway to forming a liquid. An example of this sequence was given by Hrma, Kruger, and Pokorny [4]:

• Water evaporation
• Gas evolution
• Melting of salts
• Borate melt formation
• Reaction of borate melt with molten salts and amorphous solids
• Precipitation of intermediate crystalline phases
• Formation of continuous glass-forming melt
• Growth and collapse of primary foam
• Dissolution of residual solids
• Formation of secondary foam

A variety of experimental methods are used to study specific stages and aspects of the batch-to-melt conversion at different scales from microscopic to macroscopic. Visual observation, either in-situ using a side view furnace, or by melt-and-quench methods are a useful first step. Gas formation and release can be studied using Thermo-Gravimetric Analysis (TGA), Differential Scanning Calorimetry (DSC) and Evolved Gas Analysis, with FTIR or Mass Spectroscopy attachments. Phase evolution can be measured using x-ray (XRD) and neutron diffraction techniques. The evolution of the microstructure of the reacting batch can be observed using x-ray tomography. In this paper, these methods will be briefly discussed, along with their advantages and limitations.

However, the conditions under which the batch reactions occur in a tank are dramatically different than in a crucible. A schematic of the furnace interior is shown in Fig. 1. The batch is introduced into the furnace on top of the molten glass, and is heated from above by heat transfer from combustion flames, and from below by heat transfer from the glass melt. As the batch heats up, it undergoes a complex series of reactions which are controlled by kinetic and thermodynamic factors, and heavily influenced by both tank design and process setup. Because of all the interacting variables, computer modeling is critical to understanding tank behavior. Currently, modeling of the batch-to-melt conversion is still in an early stage of development, and the measurements from the tank that are needed as input are difficult to obtain.

Figure 1 Schematic of heat and mass behavior inside a furnace

VISUAL OBSERVATIONS

Visual observation is a useful first step towards understanding batch reactions. The batch free time (BFT) experiment is a simple and commonly used approach (see [5]). In BFT experiments, a standard quantity of batch can be heated either isothermally or following a desired time-temperature profile. Samples are removed from the furnace after different amounts of time, quenched and analyzed using an optical microscope. The amount of time required for all the solid material to dissolve under a particular set of conditions is called the "batch-free-time". This approach is very useful for studying the effects of different types of sand, or different amounts of cullet. For example, the effect of adding cullet as particles or as briquettes on melting behavior of a soda-lime-silicate batch was studied by Deng et al [6]. In Batch A, the cullet was added as powder (as 84.5 wt% in the batch), while in Batch B, some of the cullet was added as a briquette (15 wt% added as briquettes made from fine cullet particles). Batches were heated up at 4°C/min then held at the indicated temperature for 4 hours before quenching. The samples were annealed and then cross-sectioned, and are shown in Fig. 2. At temperatures above 1200°C, the two batches showed similar melting behavior.

Figure 2 Effect of adding cullet as either fine powder or as briquettes on melting behavior. Reproduced from [6] with permission of Wiley.

The effects of different types and amounts of fining agents can be studied using the same technique. Samples containing differing kinds of fining agents are heated under the same conditions, then quenched. The minimum amount of time needed under a given set of conditions to remove all the blisters, called the "bubble-free time" can be used to compare the behavior of different fining agents. The size distribution and number of blisters, and the gases present inside them can also provide useful guidance about fining behavior.

These melt-and-quench experiments are very flexible. They can be performed using a variety of different batch sizes, from grams to kilograms, and are inexpensive and simple to perform and analyze. Different laboratories use different variations of this test: some melt from batch starting at room temperature and heat the batch at a constant rate, others put batch on top of a glass melt at high...