Introduction

The main purpose of this manual is to give guidance on the selection and, when necessary, the modification of natural or raw clays from which durable pottery stoves and stove liners can be fabricated. It is intended for use by project technicians who can then advise potters and producers of the best possible mixes. The tests are simple, but we recommend that they are carried out in a laboratory. Basic mathematics is the only skill required to do the calculations and a meticulous approach to 'note-taking' helps to build up a long-term record of clay and its characteristics.

Stoves are ideally suited for manufacture by experienced rural potters or in small factories producing, for example, bricks and tiles. Here, the fabrication practices are likely to be suitable for stoves but the clay body mixes are certain to require some modification. Stoves are subjected to more severe service conditions than bricks and tiles, or even cooking pots.

To ensure good sales, stoves should not fail in use from thermally-induced stresses and they should be sufficiently strong to withstand the mechanical loads and knocks to which they are subjected. To make them durable is likely to require more than the use of a suitable clay body mixture. The design and the shape are also important, as is the care taken in their manufacture. We include in this manual some comments on design and fabrication but as this is a very inadequately researched area we can only offer the most basic suggestions.

A research project was carried out at Sheffield University by A. Gaspe under the supervision of P.F. Messer to investigate why pottery stoves or liners made at some locations failed in service through thermally-induced stresses, while those made at other locations did not. The project was funded by the Overseas Development Administration of the UK Government and administered by ITDG in Rugby, UK. The results are presented in the following chapters.

Finally, for those with some scientific background, we provide some information on the factors determining the strength of ceramic materials and the factors affecting the development of thermally-induced stresses.

Factors affecting strength and thermally-induced stresses

The strength, σf, of a ceramic is the tensile stress (force/unit area) at which the material breaks. It depends on the Young's modulus of elasticity, E, of the ceramic, its effective surface energy for fracture initiation, γi, and the size of the fracture-initiating flaw, c, in the following way:

σf = constant x ([Eγi/c)0.5

The constant depends on the shape, position and orientation of the fracture-initiating flaw.

The product Eγi determines the toughness of the ceramic. The fracture toughness, Kic, is given by:

Kic = (2Eγi)0.5

Flaws are regions from which the material is missing and through which mechanical load or force cannot be transmitted. Flaws concentrate the stress around their peripheries. A spherical pore is a flaw but, because of its rounded shape, the maximum stress is only twice the average value as shown in Figure 1. The maximum stress does not change with the size of the spherical pore. Rounded pores, even large ones, are unlikely to be fracture-initiating flaws on their own. To be severe, a flaw needs to be either a sharp crack or a pore linked to a sharp crack. This is because the stress next to the crack tip is magnified by a factor very much greater than two.

Inclusions such as quartz grains are often partially or wholly detached from the matrix by cracks or by associated pores or fissures (a fissure is an elongated pore which concentrates stress by more than a factor of two). Therefore, inclusions often act as quasi-pores (see Figure 3), which can be linked with sharp cracks to become fracture-initiating flaws.

When the mechanical load on a material is increased, the stress within the material is also increased. Adjacent to a flaw, the stress will reach the very high value required to pull the atoms apart, whilst the average stress is at a modest level. When the bonds between the atoms are ruptured, the flaw grows in size at the crack from the flaw tip and propagates across the material causing the material to split into at least two parts.

We know from the examination of fracture surfaces that large pores – both of rounded and elongated shape – and inclusions cause fracture. We know theoretically that these flaws must have had a sharp crack-like feature associated with them to make them severe flaws. The equation for strength tells us that the strength of a ceramic decreases as the flaw size increases.

The terms E and γi in the equation both depend on the effective porosity in the ceramic; that is, on the volume function of pores and quasi-pores. Both E and γi decrease as the effective porosity increases.

For E, this occurs because pores and quasi-pores cannot support and transmit a force. Consequently, the material around the pore or quasi-pore is more highly stressed, as shown in Figure 1. A material is therefore stretched more for a given mechanical load when it is porous than when it is fully dense. A porous material therefore has a lower value of E.

Energy is required to fracture any material. Part of the energy is required to form the new surface. Energy is expended in other ways, such as heating the material. The total energy requirement to form each unit area of fracture surface is γi. During fracture a propagating crack is attracted towards pores and quasi-pores lying close to its path, i.e. it takes the path of least resistance. If it intersects the pores and quasi-pores, less surface has to be created when the material is fractured. Consequently the presence of pores and quasi-pores lowers the value of γi.

For a flaw of fixed size, type, position and orientation, the strength will decrease as the effective porosity is increased (E, γi and Kic decrease).

When an object is heated non-uniformly, the temperature varies with position throughout the object. As materials expand when they are heated, the object expands but by different amounts throughout its volume because of the temperature variation. The part which is at the highest temperature wants to expand the most. Its expansion will be constrained by a neighbouring part which is at a lower temperature and wants, therefore, to expand less. The hotter part is constrained to be smaller than it wants to be, whilst the cooler part is forced to be larger. This is how thermal stresses arise. The hotter part is in compression, whilst the cooler part is in tension.

Thermally-induced stresses depend on the thermal expansion coefficient (TEC), the moduli of elasticity, such as E, and the temperature gradient. The stress will increase when any of these factors is increased.

The presence of pores does not significantly affect the TEC, although, as already discussed, they reduce E and the other elastic moduli.

Quartz particles undergo a contraction on being cooled through 573°C as they transform from the β to the α form. Also, the TEC of α-quartz is high, so that on cooling after the transformation quartz undergoes a large contraction. Quartz particles contract more on cooling than the surrounding pottery matrix and this generates stresses which cause large quartz particles (greater than 40µm diameter) to detach themselves from the matrix to form quasi-pores. Small quartz particles, below 10µm, may not break away. Those that remain attached increase the TEC of the pottery, whilst those that are detached do not affect the TEC below 573°C.

Consequently, thermally-induced stresses can be reduced by increasing the effective porosity.

Thermally-induced stresses may also be reduced by designing the object so that it can easily change its shape on being heated. To illustrate this, consider the cylindrical tube, shown in section in Figure 2a, which is heated internally so that the inner surface is hotter than the outer surface.

The tube will develop a compressive hoop stress around the inner surface and a tensile hoop stress around the outer surface. If the tube is slit along its length to produce a very thin gap of a negligible thickness, as shown in section in Figure 2b, the shape of the tube and the gap will change when the tube is heated internally. If the gap were to be brought back to its original width and shape, by applying forces to the tube, the tube would be stressed as in Figure 2a. When the tube is allowed to change shape (Figure 2c), the thermally-induced stresses will be much reduced in magnitude and, in some cases, will be absent.

The tendency of stoves and stove liners to fail from thermally-induced stresses

The research project at Sheffield University investigating why pottery stoves or liners made at some locations failed in service through thermally-induced stresses examined ten clays from various locations throughout Asia and Africa which were being used to make stoves and liners. The tendency of the stoves and liners made from these clays to fail in service from thermally-induced stresses was assessed by Tim Jones of ITDG. He reported their behaviour as varying from very good to very bad. An attempt was made to correlate this observed behaviour with characteristic properties of the clays.

A correlation was found between the clay/non-clay ratio (C/NC ratio) and the tendency of the stoves and liners to fail in service. Each natural clay was considered to be composed of two parts. The material below 2µm was considered to be the clay component and that above 2µm the non-clay component. The C/NC ratio is simply the ratio of the proportions by weight of the two component parts. Stoves and liners made with natural clays with a value for the C/NC ratio of below 1 had good in-service performance, whereas those with a ratio of more than 1 performed badly. The performance worsened as the value of the ratio increased.

The C/NC ratio is a very incomplete description of a natural clay; many different clays would have the same value ratio. In addition, the processing operations used by various stove-making groups are bound to have differed in certain respects. For example, the drying stage may have been carried out with different degrees of care. Consequently, it is very surprising that a correlation was found. This indicates that the proportion of clay in the starting material has a very strong effect on important characteristics of the material while variations in processing, within sensible limits, are less significant.

The question arises as to why the ratio strongly affects the ability of a stove or liner to withstand the thermal stresses met in service. The reasons considered to be important are given in the next section.

The effect of the clay/non-clay ratio on stove behaviour

Drying shrinkage and strength after firing

It is well known that drying shrinkage is strongly affected by the C/NC ratio. With a higher proportion of clay mineral in the raw material, it will shrink more during drying. The tendency for splits to occur will increase with greater shrinkage because shrinkage occurs non-uniformly with non-uniform drying. This causes stresses to develop, which can be relieved by the formation of fissures or splits. Either numerous small fissures or fewer, larger fissures may develop. The larger fissures will not heal during firing. Hence, fired stoves or liners made of material with a high value for the C/NC ratio may contain larger fracture-initiating flaws than those made with material having a lower C/NC ratio. The former would exhibit lower fired strength.

This point may be illustrated by comparing some results obtained by Auke Koopman in Sri Lanka with results obtained in Sheffield for the effect of quartz sand additions on the strength of the fired bodies.

In Koopman's work, test-pieces produced plastically and these shrank on drying. It was found that the strength first increased and then decreased as quartz sand was progressively added.

In Sheffield, test-pieces were made by pressing moist powder and these did not shrink on drying. Starting from material that contained practically no quartz, it was found that the strength decreased as quartz was progressively added. The decrease is considered to result from the quartz grains creating quasi-pores in the fired materials when they break away from the matrix on cooling from the firing temperature (see Figure 3). The consequent decrease in load-bearing area and fraction surface to be formed would reduce the fracture toughness of the material and, for a fixed flaw size, decrease the strength.

The initial increase in strength observed in the Dutch work must, therefore, be explained by a reduction in the size of the fracture-initiating flaws as quartz was first added. This could have arisen from the expected reduction in drying shrinkage with increased quartz content.

Effective porosity, strength and thermally-induced stresses

As the C/NC ratio decreases, the effective porosity will increase because of the quasi-pores associated with quartz grains and other inclusions.

As already discussed, the fracture toughness and the thermally-induced stresses will be affected by a change in effective porosity. Although fracture toughness will decrease as the C/NC ratio decreases, the strength of plastically-formed objects may increase. An increase in strength coupled with the reduction in the magnitude of the thermally-induced stress that will accompany a decrease in the C/NC ratio should make stoves made with low values of the ratio more able to withstand the service conditions.

The anisotropy of shrinkage and residual stress development

It is well known that when clay-containing bodies are formed plastically, the clay particles become aligned. Clay minerals have particles which are plate-like in shape. During plastic forming, the material is sheared. Slip occurs in slip bands, consisting of clay particles and water. In these, the clay platelets line up with their large plane faces parallel so that they can easily slip over one another. Between the slip bands, the more isometric quartz and feldspar particles form a band in which little or no deformation occurs. These regions also contain clay and water, but the clay particles are arranged between and around the coarser particles and are not preferentially aligned with the slip bands. This is illustrated in Figure 4. Fewer slip bands and/or narrower slip bands should be formed as the concentration of clay in the mixture is reduced and the material becomes less workable.

When a plastically-formed body dries, the shrinkage that occurs is anisotropic. That is, it varies with direction such that shrinkage in the direction perpendicular to slip bands is greater than that parallel to the slip bands. The reason for this is readily explained: there are more water films between clay particles in the direction perpendicular to the slip bands than in the parallel direction.

When plastically-formed bodies are fired, the shrinkage is again anisotropic. This is illustrated in the graphs shown in Figure 5. One reason for anisotropic firing shrinkage is that there is a different number of spaces between clay particles in directions perpendicular and parallel to the slip bands. Another reason is that the dehydroxylation of the clay particles causes them individually to shrink anisotropically.

If a sheet of plastic clay body is produced by rolling or extrusion, the clay platelets are preferentially aligned such that their large plane faces are parallel to the large plane surfaces of the sheet. Drying and firing shrinkages will be greater through the thickness of the sheet. However, this anisotropic shrinkage can occur without macroscopic stresses developing in the sheet. This is because the shrinkage through the thickness does not affect shrinkage in perpendicular directions and vice versa.

Consider a tube produced from a plastic body by extrusion or jollying. The large plane surfaces of the clay particles are now aligned tangentially to the cylindrical surfaces in which they lie within the tube (Figure 6). The radial shrinkage of the tube during drying and firing should be greater than the circumferential shrinkage. However, in this case, shrinkage in a radial direction cannot occur without causing circumferential shrinkage and vice versa. It is easy to deduce that shrinkage in this case leads to the development of stress, such that there is a circumferential tensile hoop stress around the outer surface of the tube and a compressive hoop stress around the inner surface. The presence of the stresses affects the shrinkage that occurs.

Stresses may be partially relieved by distortion of the ceramic article during drying or firing and by developing fissures that may subsequently act as fracture-initiating flaws.