1.1 Research Overview
Carbon and, in particular, silicon carbide-based refractories have been used as sidewall materials for the Hall-Heroult cell. The performance of both materials relies on the existence of a frozen electrolyte layer (ledge) that is formed by forcing high heat flux through the sidewalls which leads to very high energy requirement. Once new sidewall materials that have high resistance to the corrosive environment within the cell be developed, the frozen ledge will no longer be required and the energy consumption of the process can be reduced. Recent developments in inert anodes (IA) and drained cathodes cell show very promising prospects. Carbon footprint could be significantly reduced by replacing conventional carbon anodes by IA. However, the application of IA will increase the energy requirement due to the absence of carbon-oxygen reaction that provides some energy for the alumina electrolysis process. Although the use of IA requires the theoretical minimum reaction voltage 1V higher than carbon anodes, it can offer an overall improved operational energy performance than that of carbon anodes due to: (1) elimination of carbon anode manufacturing; (2) reduction of anode polarisation overvoltage; and (3) reduction in anode-cathode distance (ACD) resulting from a stable anode surface. It was calculated that in spite of high decomposition voltage for IA, the electrical energy consumption could still be reduced by 23% (including 17% saving from the elimination of carbon anode manufacturing) by lowering ACD for about 60%. The electrical energy saving energy could be increased to 32% if IA are combined with cathodes technology such as inert/wettable cathode or drained cathode cell. When the electrical energy consumption is reduced, i.e. by lowering ACD, the amount of heat that is generated within the cell would also be reduced. Therefore, it is going to be necessary to incorporate high thermal insulation to the cell in order to reduce the heat loss and bring the cell into the heat balance. This, however, would reduce the stability of the frozen ledge. In some cases, the frozen ledge will melt away, which is considered as disaster for existing cells. New cell designs and constructions and new sidewall materials that can perform well without the protective frozen layer would be needed.
Due to the complexity of the properties requirement, multi layered-structure (combination of materials) for the sidewall is proposed. Research on joining the
Materials and interface characteristic would be conducted in my PhD project.
1.2 Research Objective
1.3 Dissertation Outline
2 Background and Theory
2.1 Production of Aluminium
Aluminium is the most abundant metallic element in the earth's crust, constituting roughly about 8 wt % of the Earth's solid surface. However, due to its very strong chemical reactivity, aluminium is not available in nature in its elemental state. It is always found in its oxidised form, such as aluminates and silicates.
Recent-day production of aluminium from the naturally occurring aluminium oxides ore involves two independent industrial processes. The first process is the production of low impurity aluminium oxide from the ore, continued by the production of aluminium metal from the oxide.
A relatively pure alumina (Al2O3) produced from bauxite ore through three stages of Bayer process which are extraction, precipitation and calcinations (Grjotheim and Kvande 1993). During the extraction process, bauxite is digested in caustic soda at the operating temperature and pressure to dissolve aluminium bearing compound – Gibbsite, Böhmite and Diaspore - and separate it from the insoluble residues called red mud. At the decomposition stage, alumina trihydrate is formed by adjusting the conditions so that the reverse of extraction reaction occurs as:
Al(OH)4- + Na+ ---> Al(OH)3 + Na+ + OH- (2.1)
The alumina trihydrate is then converted to alumina by removing water through calcinations at 1050 °C as:
2Al(OH) ---> Al2O3 + 3H2O (2.2)
Alumina produced by Bayer process can be converted to metallic aluminium through several processes, included:
- The Alcoa Smelting Process / AlCl3 reduction process (Russell 1981; Grjotheim and Welch 1988)
- The Carbothermic Process (Stroup 1964; Choate and Green 2006)
- The Hall-Héroult Process (Grjotheim and Welch 1988; Thonstad, Fellner et al. 2001)
Among the above processes, Hall-Héroult process is the only viable method by which aluminium is produced commercially today. The Alcoa smelting process is not economically competitive due to its high cost of fuel oils, maintenance and waste treatments. It also face problems associated with the presence of harmful polychlorinated hydrocarbons and difficulties in producing and handling the pure - water free – aluminium chloride required for electrolysis (Grjotheim and Kvande 1993). In the case of carbothermal process, the main disadvantages are very high operating temperature of above 2000 oC and large emission of CO and CO2 up to a factor of ten, compared with Hall-Héroult process (Thonstad, Fellner et al. 2001).
Hall Heroult Process is a high energy intensive process. In 2008, about 13 kWh electric energy is needed to produce 1 kg of aluminium (Schwarz 2008). This is being the reason why the amount of aluminium produced per year is much less when it compared to the production of other metals, i.e. steel.
2.2 Hall-Héroult Process
The Hall-Héroult process was firstly introduced in 1886 by Charles Hall and Paul Heroult independently. The process is based on the electrolysis process by which high purity alumina (Al2O3) is reduced to aluminium. Electrolyte that mainly consists of molten cryolite (Na3AlF6) is used to dissolve alumina and passing electricity from the positively charged anode to the negatively charged cathode. The feature of the conventional Hall-Héroult process is the use of anode made of carbon based materials. When the electric current passes through the electrolyte, in which alumina is regularly fed into it, the carbon of the anode combines with the oxygen in the alumina. The chemical reaction produces metallic aluminium and carbon dioxide, as follows:
2Al2O3 (dis) + 3C (s) = 4Al (l) + 3CO2 (g) (2.3)
The electrolysis process take place at about 965 oC inside a large reduction pot made of lined refractory steel. Since the electrolyte has a lower density, the molten aluminium goes down to the bottom of the pot and is siphoned periodically while the carbon dioxide escapes as gas.
There are two basic types of Hall-Héroult cell, namely Prebaked anode cell and Soderberg cell as illustrated in 1 and 2, respectively.
The main difference between prebaked cell and the soderberg cell is in anode design, which is discussed briefly in Chapter 2.2.3.
2.2.1 The Electrolyte
The electrolyte used in the Hall-Héroult cell are mainly consist of cryolite (Na3AlF6) formed from a mixture of sodium fluoride (NaF) and aluminium fluoride (AlF3). Cryolite melts at 1010 oC as shown by the NaF – AlF3 phase diagram in 3. It can also be seen from 3 that the addition of aluminium fluoride to the system lowers the liquidus temperature of cryolite.
The relative amount of sodium fluoride to aluminium fluoride is commonly expressed as Cryolite Ratio (CR), refers to the molar ratio of NaF to AlF3, or Bath Ratio (BR), refer to the mass ratio of NaF to AlF3. Pure cryolite has a CR of 3.00 and decreased with increasing of aluminium fluoride addition. Since the molar mass of aluminium fluoride is almost exactly twice that of sodium fluoride, the bath ratio is half the value of the cryolite ratio.
The relative amount of sodium fluoride to aluminium fluoride is sometimes also described as Excess AlF3, which refers to the mass% AlF3 in excess of the Na3AlF6 composition. The corresponding value of cryolite ratio, bath ratio and excess AlF3 is given in Table 1.
Table 1. Relationship between cryolite ratio, bath ratio and excess AlF3 (reference)
Cryolite Ratio (CR)
Bath Ratio (BR)
Excess AlF3 (wt %)
The electrolyte is not consumed during the electrolysis process, but some losses occur through vaporisation and hydrolysis, and by penetration into the cathode lining.
Additives are used in order to improve the physicochemical of the electrolyte such as lower the density, reduce metal solubility, lower vapour pressure, increase alumina solubility and increase electrical conductivity (Burkin 1987). The most common additives are AlF3, CaF2, LiF, and MgF2. All of the additives reduce the melting point of the electrolyte and thereby the cell operating temperature, but unfortunately they also reduce the alumina solubility. However, the reduced solubility is not as critical in the past since it is now easier to control the alumina content of the electrolyte using modern alumina feeding techniques (Thonstad, Fellner et al. 2001).
In addition to being the solvent for alumina and being the medium for electricity to pass from anode to cathode, the electrolyte also act as a resistor that generate heat and it also prevent the back reaction by physically separate the produced aluminium with the gas produced at the anode.
The main function of alumina in the Hall-Heroult cell is as the raw material to produce aluminium. According to the stoichiometri of reaction (2.3), 1.89 kg of alumina is consumed to produce 1 kg of aluminium. The alumina concentration within the bath is maintained between 2 to 4 % by weight (Grjotheim and Kvande 1993). In modern Hall-Heroult cell, small amount of alumina is periodically added to the bath by special automatic point feeder. This allows the alumina to dissolve and rapidly mix in the bath and avoid the formation of sludge beneath the metal pad which is difficult to remove. When the alumina content is lower than the range, it will lead to anode effect resulting very high cell voltage up to 30 to 50 V which may interrupt the electrolytic process.
Another function of alumina is to adsorb the fluoride gas emission and to preserve heat inside the cell by forming the alumina crust on the top of frozen ledge.
2.2.3 The Anodes
There are two basic anode designs, prebake anodes and the single, continous, selfbaking soderber anode. The original patents of hall heroult process contained prebaked anodes. These are made from a mixture of petroleum coke aggregate and coal tar pitch binder, moulded into blocks and baked in separate anode baking furnaces at about 1100 C. an aluminium copper rod with an iron yoke stubs is cast or rammed into one, two, three, four or six recesses in the top of the anode block to support the anode and conduct the electric current to the carbon surface.
Prebaked anode must be replaced at regular intervals, typically when they have reacted down to about one fourth of their original size, usually after 22 to 30 days. The spent anodes are called butts, and they are cleaned outside the cell in a separate butts-cleaning station. The cleaned butts are crushed and are reused as raw material in the manufacture of new anodes. The adhering electrolyte materials that have been removed from the cell are also recycled.
The soderberg anode is also made from a similar mix of petroleum coke and coal tar pitch, but it typically contains between 25 and 28 wt % pitch, whereas prebaked anodes usually contain between 13 and 16 wt% pitch. Briquettes of soderberg anode paste are added on top of the soderberganode. While the paste passes slowly downwards through a rectangular steel casing, it is baked into an electrically conducting solid composite bz pyrolysis of the pitch by the waste heat generated in the electrolyte and the anode itself. Electric current usually enters the soderberg anode through horizontal studs is still being used in some older and smaller soderberg cells.
2.2.4 The Cathodes
It has become common in the aluminium industrz to describe the whole container of liquid alunminium and electrolyte as the cathode. Thus, the cathode consist of prebaked carbonaceous blocks made of anthracite pluc pitch, anthracite-graphite blends, or more commonly now, of semi-graphitic or semi-graphitized carbon material, joined by a carbonaceous seam mix. The steel current collector bars are inserted into grooves at the bottom of the blocks and fixed to them by pored cast iron or by rammed carbon paste to enable good electrical contact. Beneath these rodded blocks there are layers of refractory and insulation materials. Carbon side wall materials are jointed to the cathode blocks, all of which are contained in a rectangular outer steel shell, typically 9 to 14 mm long, 3 to 5 m wide and 1 to 1,5 m deep. The operating cavity depth is only 0.4 to 0.5 however.
Because no material is known that can resist the combined corrosive actions of the liquid cryolite-containing electrolyte and liquid aluminium, the thermal insulation of the cell is designed carefully to provide sufficient heat loss to freeze a protective layer of solid electrolyte, called side ledge, on the inner side walls. However, this shoul not cover the horizontal bottom surface, which must remain bare for good electrical contact with the liquid aluminium cathode.
Although the aluminium pad protects the cathode bottom during cell operation, the bottom blocks still swell due to electrolyte and sodium penetration. Ultimately, the cathode fails and the cell then has to be shut down for relining. For large, modern cell, relining cost usually exceeds 100 000 USD, so the economic consequences of short cell lives are significant. Typical average cathode life times for most modern cell lines are now between 1800 and 2800 days, while some individual cell can be in operation for more than 4000 days. Still there are considerable variations among cells, even with the same design and construction, indicating the cell operation may have a significant impact on cathode lives.
2.2.5 Sidewalls Materials
The Sidewall for aluminium smelter is usually made from prebaked carbon, manufactured from the same materials as the carbon cathode blocks and therefore having similar properties (Sorlie and Oye, 1998). Hence, chemical reaction between carbon cathode blocks with the bath and molten aluminium could also prevail for the sidewall. During the lifetime of the cell, following reactions could proceed:
1. Aluminium reaction with carbon.
Aluminium can react with carbon according to the following reaction:
4Al(l) + 3C(s) = Al4C3(s) (1)
In the presence of electrolyte, Reaction (1) is thermodynamically favored at all temperatures (Worrell, 1965). The solid layer of Al4C3(s) acts as a barrier for further reaction and virtually stops the diffusion controlled reaction (Grjotheim, 1978). However, Al4C3(s) may be dissolved in the electrolyte and be reoxidized by carbondioxide produced at anode as:
Al4C3(s) + 9CO2(g) = 2Al2O3(sol) + 12CO(g)
making the further carbon consumption be possible (Sorlie and Oye, 1998). The aluminium carbide layer removal is also enhanced by the circulation of electrolyte and aluminium in the bath (Yurkov, 2006).
2. Sodium intercalation followed by bath impregnation
Sodium can be formed at the interface of liquid aluminium and bath according to the following reaction (Thornstadt et al, 2001),
Al + 3NaF = 3Na + AlF3 (2)
Na established at the interface could then impregnate carbon and form sodium intercalation. Due to capillary action, bath could also impregnate through carbon via the porous network formed by the volatilization and shrinkage of the binder during baking. The impregnated bath then reacts with the intercalated sodium and carbon as:
4Na3AlF6(l) + 12Na(C) + C(s) = Al4C3(s) + 24NaF(l)
Compounds that also usually found in the discarded carbon potlining are Al2O3, Na2CO3, Na2.11Al2O3 (ß-alumina), NaCN, AlN and Al-Fe alloys (Sorlie and Oye, 1998).
3. Cathodic dissolution
Carbon (graphite) have high electric conductivity hence sidewall made from carbon based material could somehow act as cathode which tend to undergoes cathodic dissolution. The cathodic dissolution of carbon into cryolite melt takes place according to the following reaction (Skybakmoen, 1999):
C(s) + 3 AlF3(l) + 4 Na+ + 4 e- = Na3Al3CF8(l) + NaF
Sidewalls made from carbon block are also susceptible to degradation due to oxidation. Carbon is unstable towards air and will react according to following reaction:
C(s) +O2(g) = CO2(g)
The rate of oxidation is increase with temperature increase, making thermal conductance be an important design parameters. Thermal conductance becomes more important due to the fact that carbon sidewall material is protected from electrolyte attack by the solid electrolyte ledge. The formation and stability of the frozen ledge is depended on the rate of convective heat transfer from the electrolyte through the sidewall
2.3 Energy Aspect of the Hall-Héroult Process
Hall-Héroult process is energy intensive process. The total specific energy requirement, Wel, of the Hall-Heroult process can be determined through the following relationship by estimating the cell voltage Thornstad et al., 2007 (Thornstad et al., 2007),
where U is the cell voltage, F is Faraday constant, MAl is the molar weight of Al, xAl is the corresponding current efficiency fraction, while │Ve│ and VAl are the stoichiometric numbers of electrons and of product of cathodic reaction, respectively.
From the above equation, it is clear that the energy consumption of the Hall-Héroult process can be minimized by minimizing cell voltage and maximizing current efficiency as close as 100%.
2.3.1 Cell Voltage
The total voltage of the Hall-Héroult composed of three elements which are the decomposition voltage, electrode overvoltages and voltage drops due to the ohmic resistance of various parts of the overall cell (grjotheim and Welch).
The decomposition voltage provides the Gibbs energy for the overall reaction to prevail. It consists of standard electrode potential of the reaction and the activity of the solute species and can be given by Nernst equation:
where R is universal gas constant (8.314 J / K mol), T is absolute temperature (K), z is number of electrons transferred in the reaction, F is Faraday constant (96500 J / V mol) and K is equilibrium constant.
For aluminium electrolysis, the above equation can be written as (buku kuning):
where ΔG° is standard Gibbs energy of reaction, cox is alumina concentration (mass %) and cox(sat) is saturated alumina concentration (mass %).
The overall reaction of the aluminium electrolysis is depended on the anode material being used. In the case of consumable carbon anodes being used, Reaction (2.3) proceeds and the decomposition voltage for the reaction at 1273 K is -1.1697 V . If the cell uses inert anodes, there would be no reaction between the anodes and the oxygen, and the overall cell reaction would be:
2Al2O3(diss) = 4Al(l) + 3O2(g) (2)
with a decomposition voltage of -2.196 V at 1273 K .
Huglen et al (1993 hal 144) buku kuning distinguish the electrode overvoltage into two kinds of overvoltage, which are charge transfer overvoltage (reaction overvoltage) and concentration overvoltage (polarization). Charge transfer overvoltage is an excess applied voltage that is required to make the reaction proceed at an appropriate rate due to the slow rate of reaction (electron transfer) at the electrodes. Concentration overvoltage is due to by slow mass transport through a diffusion layer in the electrode surrounding that is cause the changes in the concentration of the electroactive species at the electrode surface.
Thornstadt (buku kuning hal 144) formulated equations to calculate cathode concentration overvoltage (ηCC), anode concentration overvoltage (ηAC) and anode reaction overvoltage (ηAA), as:
where ic is cathodic current density, ia is anodic current density, icc is critical current density and io is limiting current density; all in (A / cm2).
For a laboratory scale cell, the limiting current density and the critical current density can be calculated by (buku kuning hal 145):
where Aa is anode surface area (cm2)
Part of the cell that contribute to the ohmic resistance voltage drops are bath (electrolyte), bubble, cathode, anode and other external voltage drops. Hyde and Welch (1997) dari tesis auklan formulated the equation to calculate the bath voltage drop (UB) by taking into account the voltage drop due to the gas bubbles (UBU) as:
where hE is interelectrode distance, χ is electrical conductivity of the electrolyte, δ is bubble layer thickness, ta is adhering bubble layer thickness, ε is gas fraction in the bath and fc is the fraction of the anode covered with gas. χ may be calculated by using the following formula (Choudary buku kuning hal 145):
where cF is mass % of CaF2.
As a summary, the total cell voltage (U) of the Hall-Héroult cell can be calculated by:
A typical cell voltage of the Hall-Héroult cell that use carbon anode at the cell temperature of 960 °C, 44mm interelectrode distance, 95 % current efficiency, 10 % a-alumina and 33 % alumina saturation is represented schematically in 4 (kvande and haupin).
In the case of carbon anode is replaced by inert anode, the typical cell voltage for a similar operational condition as above, except that the interelectrode distance is reduced to 22.4mm
2.3.2 Current Efficiency
Current efficiency defined as the ratio of the actual weight of aluminium produced electrolytically per unit of time to the theoretical amount predicted from Faraday laws. It can be expressed as:
where p is measured production rate (g/s), po is theoretical production rate (g/s), P is amount produced at electrode (g), M is Molecular mass (g), Z is number of electron involved in the electrode reaction, F is Faraday constant, I is current (A) and t is time (s).
There are number of variables that influence the current efficiency that are temperature, current density, interelectrode distance, composition of electrolyte and cell design. The primary mechanism for the lowering of the current efficiency is the back reaction between aluminium metal and carbon dioxide. The back reaction is initiated by chemical dissolution of the metal at the metal/electrolyte interface into electrolyte followed by mass transport of the metal through electrolyte to the electrolyte/gas interface. The oxidation of the dissolved aluminium by the anode gas then occurred according to the equation (aluminium smelter technology grjotheim and Welch):
2Al (dis)+ 3CO2 ---> Al2O3 (dis) + 3CO(g)
The carbon dioxide of the above reaction can be either in the gaseous phase or dissolved in the electrolyte.
Further loss of current efficiency may take place due to the following less major mechanism (K Grjotheim, aluminium electrolysis p 230):
- Back reaction of dissolved aluminium with CO gas to form solid carbon:
2 Al (diss) + 3 CO (g) ---> Al2O3 (diss) + 3 C (s)
- The formation of aluminium carbide according to the reaction:
4 Al (l) + 3 C (s) ---> Al4C3 (s)
The above reaction may occur at the interface between the aluminium and the carbon cathode or carbon sidewall in the absence of side ledge. The aluminium carbide can be dissolved in the electrolyte and subsequently oxidized at the anode according to the reaction (Odegard, dari buku elektrolisis):
Al4C3 (diss) + 6 CO2 (g) ---> 2 Al2O3 (diss) + 3 C (s) + 6 CO (g)
- The oxidation of dispersed aluminium droplets or particles formed by turbulence:
Al (diss) ---> Al3+ (diss) + 3 e-
where Al3+ (diss) is actually in the form of Al-F complex anionic species
- Reduction of metals impurities in the electrolyte which consume electric current without producing aluminium.
- Electrical shorting due to direct contact between the anodes to the aluminium pad that may happened during anode changing or tapping.
- Loss of aluminium into cell lining due to penetration into cracks and by dissolution of sodium in carbon.
- Metal losses through spillage during tapping.
2.3.3 Cell Heat Balance
Change in energy consumption (input) have direct influence to the heat balance of the cell
Hi = Hp – Hc + Hl
Heat input, Heat produced by reactions, Heat consumed by reactions, Heat Loss
Heat Loss distribution:
Heat loss through the bottom of the cell, constrains: bath penetration
Heat loss through the top of the cell, constrains: crust thickness
Heat loss through the sidewall, constrains: ledge formation
Heat Loss from sidewalls :
hb = heat transfer coefficient of the boundary layer
Ab = cross-sectional surface area of the boundary layer
Tb = bulk of the bath
Tl = temperature at the bath-ledge boundary
2.3.4 Energy Consumption
Kvande and Haupin  calculated and compared the energy consumption of the carbon anodes cell with the three different scenarios of inert anodes cell as tabulated in Table 1. The energy consumption of 15 kWh/kg Al for the carbon anodes cell was chosen since it close to the average energy consumption for all existing cells in the world. For the calculation of the inert anode cells, it is assumed that the cell amperage and the current efficiency remain the same and no major changes in the cell construction except that the carbon anodes replaced with the inert anodes.
Table I. Cell Voltage and Energy Consumption of Carbon Anodes Cell and Inert Anodes Cells .
Voltage Type (V)
Total Cell Voltage (V)
Energy Consumption (kWh/kg Al)
Enthalpy EAl (V)
Heat Loss (V)
From the above data, it can be seen that for inert anode to be implemented with the same heat loss as carbon anode, in order to maintain the frozen ledge, higher energy consumption is needed due to increase in total cell voltage (scenario 1). By decreasing the electrolyte voltage drop i.e. reducing the electrode gap as in scenario 2 and 3, the total cell voltage can be reduced. However, good thermal insulation is needed in order to reduce heat loss and maintain the heat balance of the cell. In the scenario 2, total cell voltage is reduced and the heat loss manipulated within the region that the frozen ledge is formed. Still, the energy consumption increase by 1.43 kWh/kg Al.
For inert anode to be implemented with the same energy consumption as carbon anode cell (scenario 3) or even with lower energy consumption, less heat needs to be dissipated through the sidewalls. In this case, the protective frozen ledge is impossible to be maintained and the sidewall materials will be exposed directly to the very aggressive environments of cryolite bath causing significant reduction on its life. This conditions call for new sidewall materials that have a high stability towards the corrosive environment that exist within the cell.
U = operational cell voltage (V)
x = current efficiency given as a fraction (CE/100%)
6. Inert anode applications:
The absence of carbon – oxygen reaction will increase the voltage required for electrolysis by about 1 V. Carbon – oxygen reaction provide some energy.
7. Should be compensated by increasing cell voltage (increasing energy consumption) to maintain same heat loss as with carbon anode (preserve heat balance) or reduce the heat loss to maintain or even lowering the energy consumption.
3 Improvement in The Hall-Héroult Cell Materials
Silicon carbide and silicon carbide containing materials have been increasingly used over the last twenty to thirty years (Brooks, 2007). Silicon carbide has higher thermal conductivity than carbon, providing more stable frozen ledge. This condition allows reduction of the ledge thickness which in turn opens the use of larger anodes for higher cell productivity (Skybakmoen, 1999). In contrast with carbon based material that evolve only carbon dioxide gas during oxidation process, the oxidation of silicon carbide material produce passive layer of solid silicon oxide that contribute to the prevention of further oxidation. Silicon carbide materials also have higher electrical resistance and hence do not suffer electrochemical erosion to the same extent of carbon. The major factor that hinders its application is the relatively high cost compare to conventional carbon based material.
The degree of corrosion resistance of silicon carbide material by molten electrolyte is found to be increased by the increase of oxidation resistance (Yurkov, 2006).
The oxidation resistance of the silicone carbide material generally can be increased by controlling the amount and type of impurities, the amount and type of sinter additives, the pore size and distribution (Skybakmoen, 1999)
It should be noted that even tough SiC-based materials showed improved resistance to degradation compare to carbon materials, it will still remain essential to maintain a stable layer of frozen bath during the electrolysis.
3.4 Future Sidewall Materials
3.4.1 Proposed Strategy
3.4.2 Challenge in Joining Ceramics
Joining of ceramics to themselves and to other materials (especially metals) is vitally important for many advanced, high-performance applications. There are three general reasons for joining. First, joining is needed to overcome processing size limitations. Both physical (i.e., material and facility) and economic constraints in producing ceramics limit the size of components that can be made. Inherent brittleness in ceramics leads to the formation of microflaws that degrade tensile properties and toughness in service and can cause gross failure even during processing. This occurs in ceramics at all temperatures. Severe thermal gradients and fast cooling rates in these inherently low thermal conductivity materials, either during processing or later on in service, can lead to gross failure through brittle fracture. Thus, fabricating large objects from ceramics requires the joining of smaller, easier-to-fabricate components.
Second, joining is needed to overcome processing shape limitations. Many ceramic parts, especially those having the most attractive engineering properties, can be made only in relatively simple shapes. This again relates to the inherent susceptibility of these brittle materials to process-induced flaws (e.g., microcracks from shrinkage from various sources, from differential thermal expansion or, especially, contraction, or from thermal shock). Problems are most severe and prevalent at points where section thickness changes occur. Thus, the fabrication of more complex shapes requires machining, which is difficult or impossible because of the inherent high hardness and poor tolerance of point loads and associated stress concentrations.
Obviously, joining of simple details into complex units is an attractive alternative.
Third, joining of ceramics and glasses enables material optimization. Some applications require more than one type of ceramic or glass to be combined in the design to obtain the optimum properties desired in the assembly. It is often preferable to have these properties in an integral or unitized component or structure. Thus, joining of dissimilar ceramics or glasses becomes important. Joining of ceramics or glasses to other materials or to ceramics or glasses of different compositions is usually motivated by technological needs. On the other hand, joining of these materials to make larger or more complex-shaped structures or objects is usually forced by economic considerations in addition to or instead of technological needs.
3.5 Basic Joining Technique for Ceramic
There are essentially four basic joining techniques used with ceramics and glasses, one exclusively for use during the initial production of the ceramic or glass article and three for use in secondary processing or assembly. The first technique can only be used for joining a ceramic to another ceramic of the same or similar microstructure (even if it is of different composition), while the other three techniques can be used to join ceramics to ceramics, glasses to glasses, or ceramics or glasses to one another or to other materials.
The first technique is sinter bonding, a process that is almost exclusively restricted to the joining of ceramics. During the initial production of ceramic articles it is possible, and is often the practice, to join smaller simpler shapes together to form larger and/or more complex shapes by co-firing them. The joint is created by diffusion, often (but not necessarily) during partial melting. Primary chemical bonds of the same type as found in the parent ceramics are formed. The growth of grains (i.e., individual, uniquely oriented crystals in the aggregate) across the initial interface between the abutting pieces obliterates the interface if the process is done properly. This growth is the result of ordinary sintering, in which a small neck forms at a point of localized, intimate contact between particles and grows by diffusion (usually, but not only, in the solid state) to reduce the total surface area and energy. This so-called sinter bonding may require the use of an intermediate material such as a glassy frit or a slurry of the powdered crystalline ceramic or mixed ceramics, or it can occur directly, without the aid of any intermediate material.9
The second technique is mechanical joining. It is possible to join ceramics to other ceramics or even to other materials through the use of mechanical interlocking (i.e., designed-in or processed-in physical features) or, to a lesser extent, mechanical fasteners. The third technique is called ‘‘direct joining.'' It is possible to join ceramics to other ceramics (or glasses to other glasses) by welding, employing either fusion or non-fusion processes. The process is completely analogous to that of metals. In such direct joining no intermediate material is required. This is the most common technique for joining glasses to other glasses of similar or different composition, but it is actually uncommon in the joining of ceramics, where intermediate materials are often required.
The fourth technique involves ‘‘indirect joining.'' This is the most common technique for achieving high-integrity joints between ceramics, between ceramics and glasses, or between ceramics or glasses and other materials. An intermediate bonding material is absolutely required and can be an organic adhesive (in adhesive bonding), glass or glass–ceramic combination (in frits), anhydrous oxide mixture (in cementing or mortaring), a relatively lower melting ceramic (in ceramic brazing), or metal (in metal brazing or soldering, or even in solid state diffusion welding).
Elevated temperature serviceability is often a major factor driving the selection and application of a ceramic. In general, service temperature capability increases for ceramics joined by these processes in the following ascending order: organic adhesives, solders, mortars, metal brazes, inorganic cements, ceramic brazes, mechanical fasteners, mechanical interlocks, welding, and sinter bonding.
BRAZING AND SOLDERING OF CERAMICS
Challenges Posed by Ceramics to Brazing and Soldering
American Welding Society definition: brazing and soldering are joining processes that create atomic-level bonds and adhesion through the capillary flow of molten fillers between properly set up and gapped solid substrates.
Brazing refers to those processes that are carried out with fillers that melt and flow above 4508C (8408F), while soldering refers to those processes that are carried out with fillers that melt and flow below 4508C (8408F). Based on this definition, brazing and soldering represent an entirely different approach to the indirect joining of ceramics using metal intermediary materials in most cases (e.g., metal brazing or soldering) or ceramic, glassy-ceramic, or glass intermediary materials in other cases (e.g., ceramic brazing or soldering with ‘‘solder glasses'').
The major challenge posed by ceramics (as well as glasses) to brazing and soldering relates to the chemical stability of these two closely related materials. By being chemically stable, inert, or non-reactive, ceramics and glasses tend to resist being wetted by other materials, especially molten metals. Without wetting of the substrate by molten metal filler, it is impossible to form strong ‘‘metallurgical-quality'' joints.
A sound brazed or soldered joint can be produced only if some type of filler can be found that wets a ceramic or a glass, or some method is developed to cause traditional (e.g., metal) fillers to wet the ceramic or glass. Even when wetting and subsequent bonding are achieved, there are problems with the inherent incompatibilities in certain physical properties between metal fillers and ceramics (or glasses) that must be dealt with, the most notable being mismatched CTEs. If CTE differences are too great (generally greater than 10% to 15%), induced stresses can become intolerable, leading to failure along or immediately adjacent to the joint in a weak boundary layer.
As a result of these characteristics of ceramics and glasses, a number of brazing and soldering processes for ceramics apply a metal layer to the ceramic, in what is often referred to as ‘‘metallizing.'' Subsequent brazing or soldering is used to join one metallized ceramic or glass to another metallized ceramic or glass.
Characteristics of Brazing Methods for Ceramics
There are fundamentally two general methods for brazing ceramics (virtually identically analogous to soldering): (1) those that use metallic materials as the intermediary material, called ‘‘metal brazing,'' and (2) those that use ceramic materials as the intermediary, called ‘‘ceramic brazing.'' The soldering analogues are simply metal soldering of metallized ceramics (or glasses11) and soldering of glasses (and possibly some glass-containing ceramics), using what is known as a ‘‘solder glass,'' which is just a low-melting glass intermediary. For metal brazing, there are three specific methods that are not necessarily fundamentally different in terms of the mechanism of bonding or adhesion but are differentiated by practice. These are (1) noble metal brazing, (2) active metal brazing, and (3) refractory metal brazing. Metal brazing of ceramics is one of the most common methods of joining ceramics (excluding cements and concretes, where it is never used!), especially for high-performance applications. The basic advantage of metal brazing is that a variety of materials can be joined together by generally simple procedures, producing vacuum-tight joints that have modest to high strength.
For ceramic brazing, the ceramic filler actually fuses or melts and distributes in the joint by capillary action. This distinguishes it from ceramic cementing with fired cements, where no such overt melting and certainly no capillary flow occur. Rather, the cement to be fired is preplaced in the joint to be bonded by firing. An advantage of ceramic brazing over metal brazing is the closer match of properties between the filler and the substrates, especially physical properties like CTE.
Metal Brazing of Ceramics
The obvious problem in trying to braze or solder ceramic substrates is achieving wetting by the molten filler. When metal filler is used, the problem is overcome either by altering the surface of the ceramic to make it like (or make it act like) a metal, or by getting the filler to act like a ceramic. The aforementioned three major methods of brazing ceramics using a metal filler or intermediate layer are described in more detail in this section. It can be seen that in noble metal brazing and in one form of active metal brazing, the filler metal is made to act like a ceramic by having one or more of its components oxidize. In another form of active metal brazing and in refractory metal brazing, the surface of the ceramic is chemically altered by the metallic braze filler to act like a metal. By metallizing the surface of the ceramic (at least in the area to be joined) by depositing or embedding metal by electroplating, sputtering, ion-implanting, or some other means, brazing with a metal filler can be accomplished as is normally done with metal substrates, that is, by simply selecting a filler that is compatible with the metallized surface material.
Noble Metal Brazing
Noble metal brazes are most commonly based with silver or platinum and their alloys, somewhat less often based with copper or nickel, and occasionally based with other noble metals (e.g., palladium and gold). Such brazing is normally done in air or even an oxygenrich atmosphere, with evidence that noble metal oxides form and bond with the ceramic substrates, particularly oxide ceramics. Typical ceramics that have been brazed with Pt, Pd, Au, or Ag, with little or no pressure beyond that needed to hold joint elements in contact, include MgO, Al2O3, ZrO2, UO2 , BeO, ferrite, SiO2, glasses, and graphite.12
Noble metal brazed joints have strengths that are approximately 50–100% of the strength of epoxy bonded joints, their strengths can be much greater for some fillers. Strengths range from 23MPa (3,400 psi) for lead as the filler to 252MPa (36,000 psi) for Pt in Al2O3.
Active Metal Brazing
Active metal brazing is most commonly based on the use of Ti but can also be based on Zr, Nb, Cr, or Y, particularly for Al2O3. Iron, cobalt, and nickel have also been used with certain oxide ceramics. Two procedures have been used with traditional Ti-based active metal brazing. In the first, the Ti (or other active metal) is incorporated into the brazing alloy to aid wetting by forming an oxide and another compound (e.g., carbide, boride) through reaction of the active metal with the ceramic. In the second, the surface of the ceramic to be brazed is coated with the active metal or a compound that decomposes to that metal (e.g., a metal hydride, such as TiH2). In this second approach, done in vacuum, the metal layer is referred to as ‘‘metallization.'' This layer is then wet by an appropriate metal braze alloy for use with the active metallization.
Another method of applying active metals (per Pattee et al., 1968) is to use molten alkali halide salts (e.g., alkali or alkaline earth halides for Ti) to prevent oxidation while the active metal is deposited on the ceramic substrate. In a diffusionbrazing approach, brazing is accomplished below the liquidus temperature of the braze filler alloy, through the formation of a transient liquid phase.
For graphite, a particularly difficult material to join by any means, Ag–Cr, Ag–Ti, Ag–Zr, Au–Zr, and Cu–Cr filler alloys work well by forming carbides with the active metal component. Two commercially available braze fillers usable with graphite are 68.8 wt.% Ag/26.7 wt.% Cu/4.5 wt.% Ti (with a melting range of 830–8508C or 1,525–1,5608F) and 70 wt.% Ti/15 wt.% Cu/15 wt.% Ni (with a melting range of 910–9608C or 1,670–1,7608F). Because graphite reacts readily with oxygen, oxygen must be excluded, usually by brazing under a vacuum of approximately 10_4 Torr. The reaction of an active component and subsequent diffusion are common means for achieving wetting and bonding when brazing substrates that are difficult to wet.
Refractory Metal Brazing
Although some brazing is actually done using refractory metals and their alloys (e.g., Al2O3 or Si3N4 to themselves using Nb or Zr, respectively), these materials are usually used to coat or ‘‘metallize'' a ceramic, with subsequent brazing to the metallized layer (often with a second metal applied by plating). Tungsten and molybdenum are normally used as the refractory metal metallization. The process requires high temperatures (e.g., 1,400–1,6008C or 2,600–2,9008F) and a hydrogen atmosphere. The refractoriness of the resultant braze, along with high strength and reliability, make this an attractive option, especially for Al2O3 and BeO. Strengths are typically 105–150MPa (15,000–30,000 psi) for Mo brazing of Al2O3 at 1,500–1,6008C (2,750–2,9008F).
The operative refractory metal needed to metallize the ceramic's surface can be embedded as a powder during initial processing, mixed with the parent ceramic and fired to develop a coating, chemically or vapor deposited, or sputtered on. Another common method of metallizing with a refractory is called the ‘‘molybdenum–manganese (Mo–Mn) process.'' Here, a paint of molybdenum and manganese metal powder or their oxides is applied to the ceramic as a slurry. The assembly is fired in hydrogen with a controlled dew point so that the manganese is converted to its oxide (MnO) while the molybdenum remains a metal. The MnO then reacts with the ceramic grains and any glass phase to form a controlled amount of glassy phase containing the MnO. The Mo sinters to form a porous coating into which the glassy phase penetrates and interlocks mechanically. In addition, the glass at the interface reacts with the Mo to form MoO, thus forming a chemical bond to the ceramic grains since they are compatible. An electrodeposited coating (e.g., Ni) is often plated over this metallized layer to further facilitate brazing. Alumina and beryllia substrates have been metallized this way using either Mo with Mn or W with Mn.
Ceramic Brazing of Ceramics
Ceramics can be joined to themselves (or, as will be seen in Chapter 15, to metals) by brazing with ceramic fillers or intermediate materials instead of metals. Glasses are common ceramic braze filler materials, but mixtures of glasses and crystalline phases, or of all crystalline phases (often as eutectics) can be used. Ceramic brazes can be applied and processed in the same fashion as metal brazes (i.e., in the solid state by diffusion brazing, or in the fluid or liquid state by conventional brazing). While ceramic brazes tend to be preplaced before brazing rather than applied during brazing, they do distribute uniformly within the joint by capillary action, as all brazes must by definition. Ceramic brazes provide good environmental compatibility (e.g., service temperature and corrosion and oxidation resistance), often better than most metals.
Unfortunately, ceramic brazes tend to be less tolerant of thermal expansion mismatch than most metal brazes, so care must be exercised in their selection and use and in the joint's design. One common ceramic braze filler is Pb–Zn borosilicate glass. Some refractory ceramic brazes are manganese pyrophosphate and MnO eutectic (30:70), Al2O3_MnO_SiO2 for Al2O3, and Al2O3_CaO_MgO_SiO2 for Al2O3.
Normally, ceramic brazing is accomplished using the furnace brazing method with appropriate atmosphere control, which may include a vacuum. Soldering of ceramics to one another or to metals typically employs metallized layers applied by powder processing (during ceramic production), by chemical or physical deposition, by electro- or electroless plating, by sputtering or ion-implantation, or by other means. Soldering is then a matter of finding a compatible solder for the metallized layer and for the intended functional requirements and service conditions. The other alternative is to use In-based solders (see Chapter 8, Subsection 8.5.10).
WELDING OF CERAMICS
Challenges Posed to Welding by Ceramics
Welding is really the oldest technique for joining a major group of ceramics, namely silicate glasses. The joining of two or more glass shapes to one another in the practice of glass working, although not usually referred to as such, is actually welding. Welding of ceramics, in general, and very refractory, single-phase crystalline ceramic materials, in particular, is relatively new. The basic requirements for welding ceramics are twofold. First, the ceramics involved in the joint must be chemically compatible with each other and with the environment in which they will be joined. Second, the ceramics being joined must be mechanically compatible (i.e., have reasonably comparable strengths) and physically compatible (i.e., have coefficients of thermal expansion that differ by no more than approximately 1_1:5 _ 10_6 per 8C). The two general methods for welding ceramics are (1) in the solid state using non-fusion processes, and (2) using melting with fusion processes.
Solid-Phase (Non-Fusion) Welding of Ceramics
Solid-state, solid-phase, or non-fusion welding is accomplished by heating the components to be joined while they are held in intimate contact, usually under substantial pressure. For metals, which are often inherently ductile, this pressure causes plastic deformation of microscopic surface asperities or localized points of contact, bringing more points into contact. With more points of intimate contact, there are more paths for solid-state diffusion. Filler is rarely needed or used. For ceramics, on the other hand, such plastic deformation is difficult or impossible because of the inherent hardness and brittleness of most ceramics. Only limited elastic deformation is able to contribute to increasing the number of points of contact for subsequent diffusion, and these are usually not sufficient to allow enough diffusion in reasonable lengths of time. Thus, in the solid-state welding of ceramics, an intervening layer of ceramic in powdered form is often sandwiched between the joint components. This powder can be of the same composition as the similar substrates being joined, or of one or the other or a mixture of both if the two materials being joined are of different compositions. In extreme cases, where the two materials being joined are chemically or physically incompatible, a series of thin layers of powder, grading from one base material's composition to the other's composition, can be used. 12.8 schematically illustrates such a functionally gradient material (or FGM) joint.
Bonding during solid-phase welding and/or by sintering occurs by the creation and growth of a reaction zone between the two materials to be joined, both of which rely on diffusion. When dissimilar materials are being joined, bonding will usually depend on the reaction and/or interdiffusion of both joint component materials. With similar materials, it depends primarily on the sintering ability of the materials (i.e., the rates of diffusion as well as pressure-induced recrystallization and grain growth). The principal process used for accomplishing solid-phase welding of ceramics to ceramics is diffusion welding, in the form of hot pressing and isostatic pressing. The process of diffusion welding for ceramics is essentially identical to that for metals (Chapter 6, Subsection 6). However, it is usually much slower because of the inherent difficulties posed to diffusion in ionic compounds (where both small cations and an appropriate number of large anions must both diffuse to maintain a balanced electrical state).
Various friction-welding processes have also been successfully employed to join ceramics, but not without great challenges to equipment and to process control. For both diffusion and friction-welding, ductile metal interlayers (‘‘intermediates'') can be used to advantage, provided any inherent chemical incompatibilities can be overcome by either metallizing the ceramic or oxidizing the intermediate layer.
A wide variety of oxide and non-oxide ceramics have been solid-phase welded to themselves, to one another, and to metals.
Fusion Welding of Ceramics
As with metals, fusion welding of ceramics is achieved by filling the joint between parts to be joined with molten material obtained by melting the edges of the parts making up the joint while they are in contact (i.e., in autogenous welding), or with additional molten material from a filler of a similar or compatible material. In fusion welding of dissimilar materials, their melts must also be compatible with one another. Besides chemical compatibility between substrates and any filler, the ceramics being fusionwelded must be compatible with the welding environment and must be physically compatible with one another.
While compatibility with the welding environment depends somewhat on the particular fusion technique being used, an overall requirement is that the ceramics melt properly and then solidify properly. There are several problems associated with attempting to fusion-weld ceramics. First, most ceramics have very high melting temperatures, so getting enough energy into them to cause them to heat enough to melt is not a trivial matter. One method is to employ processes with high energy densities. Second, some ceramics (e.g., BN, Si3N4, and SiC) vaporize without melting (i.e., they sublime), so at normal pressures they cannot be fusion welded. This problem can only be overcome by employing a non-fusion technique. Third, some ceramics (e.g., MgO) have very high vapor pressures at the melting points, so they vaporize after they melt, making fusion welding difficult, if not impossible. This problem can also be overcome by employing a non-fusion welding process. Fourth, the very high temperatures involved in the fusion of many ceramics can cause problems with phase transformations in surrounding heat-affected areas, leading to severe property degradation or fracture. Again, non-fusion welding, with as little elevation of temperature as possible, is the only answer. Fifth, thermal stress fractures, resulting from severe temperature gradients around the fusion zone or from thermal shock on heating or cooling, have long been considered a major obstacle to the fusion welding of ceramics. Supplemental heating around the intended weld, to reduce thermal gradients and stresses, helps greatly, although non-fusion welding is certainly a viable alternative. Other problems arise from the fact that most ceramics are electrically non-conductive, so they cannot be made part of the circuit for arc welding or resistance welding. The principal processes for fusion welding ceramics, in descending order of popularity, are (1) laser beam welding (LBW) (primarily CO2 but also Nd:YAG), (2) electron beam welding (EBW), and (3) arc welding (especially GTAW and PAW), provided the ceramic will support the establishment of an arc. LBW and EBW can be employed with any ceramic that will melt without subliming or thermally decomposing. PAW can be employed on even insulating ceramics by using a non-transferred arc technique.
Sinter Bonding of Ceramics
Most materials that are produced by powder techniques, such as most refractory materials (whether metals, ceramics, or intermetallics), require more than simple pressure compacting of the powder particles to result in strong particle-to-particle bonding as well as high density (i.e., low porosity). They usually also require sintering. Sintering is the process of causing particles of a material (or more than one material) to join together by interdiffusion. The driving goal of sintering to join two particles is to reduce their total surface area and, thus, surface energy. The necessary diffusion can occur entirely in the solid phase or can involve the liquid phase, which can drastically accelerate the process. It is possible and quite common to take advantage of the process of sintering to join smaller ceramic parts into a larger and often more complex- shaped unit. This process is called ‘‘sinter bonding.''
Sinter bonding tends to be almost exclusively restricted to the joining of ceramics, even though it is possible with metals. The reason is that the process is quite natural for ceramics, which are often synthesized into even the simplest forms by powder processing, including sintering. It is thus an integral step in ceramic part production already. Again, the joints between small parts are created by diffusion, often, but not necessarily, during partial melting. Primary chemical bonds of the same type as found in the parent ceramics (i.e., ionic or covalent or mixed) are formed. The growth of grains (i.e., individual, uniquely oriented crystals in the aggregate) across the initial interface between the abutting pieces obliterates the interface if the process is done properly. This so-called sinter bonding can occur directly, without the aid of any intermediate material such as a glassy frit or a slurry of the powdered crystalline ceramic or mixed ceramics, or may require the use of such an intermediate material.13
SHS or CS Welding or Brazing of Ceramics
Self-propagating high-temperature synthesis (SHS) and combustion synthesis (CS) are two modes of an exothermic brazing. They differ in that SHS occurs progressively as a reaction front sweeps through the volume of powdered reactants, while CS occurs all at once (often with near explosiveness). The processes are capable of creating joints by causing the reactions to occur in situ between joint elements. To work, the appropriate reactants are packed between the solid joint elements, the entire sandwich assembly is held under unidirectional squeezing pressure, and the reaction is triggered. 610 Chapter 12 Joining of Ceramics and Glasses (Such reactants are often a powdered elemental metal and an oxide, as in the aluminothermic reaction between powdered Al and powdered Fe3O4 that underlies the thermite welding process.) Once complete, the newly formed product phase bonds one joint element to the other. As examples, Ni3Al can be bonded to Ni3Al by causing powdered Ni and Al to react in situ to produce a Ni3Al bonding layer, or two pieces of graphite can be joined by reacting powdered Ti and graphite in situ to form a bonding layer of TiC. While not widely used yet with ceramics, the processes of pressureassisted SHS and CS offer potential.
Burkin (1987). Production of Aluminium and Alumina West Sussex, John Wiley and Sons.
Choate, W. and J. Green (2006). Technoeconomic assessment of the carbothermic reduction process for aluminum production. TMS Light Metals, San Antonio, TX.
Grjotheim, K. and H. Kvande (1993). Introduction to Aluminium Electrolysis. Düsseldorf, Aluminium-Verlag.
Grjotheim, K. and J. Welch (1988). Aluminium Smelter Technology. Düsseldorf, Aluminium-Verlag.
Russell, A. S. (1981). "Pitfalls and pleasures in new Aluminum process development." Metallurgical Transactions B 12(2): 203-215.
Schwarz, H. G. (2008). "Technology diffusion in metal industries: driving forces and barriers in the German aluminium smelting sector." Journal of Cleaner Production 16(S1): 37-49.
Stroup, P. T. (1964). "Carbothermic Smelting of Aluminum." Transaction of The Metallurgical Society of AIME 230(4): 356-372.
Thonstad, J., P. Fellner, et al. (2001). Aluminium Electrolysis. Düsseldorf, Aluminium-Verlag.