Samples which have undergone DEF exhibit a characteristic microstructure after expansion, in which clear rims are seen around aggregate particles, largely filled with oriented ettringite. To some degree the thickness of the rims follows the size of the aggregate particles, although not all rims are continuous or filled and there are additional cracks through the cement paste. The diversity of microstructures seen is discussed by Taylor et al. (2001).
Materials subjected to heat curing have a significantly altered microstructure when compared with materials cured at ambient temperature. Heat curing causes a rapid production of hydrates at early age, and leads to the formation of dense rims around the larger cement particles. Smaller cement particles fully react, often leaving Hadley-grain-like (Barnes et al., 1978) features behind. These features persist to late age, with samples having much coarser and more open microstructures compared to samples cured at room temperature. In many cases, the rims around large cement grains are found to consist of distinguishable concentric rims of hydrates, which have different densities, formed at different stages of the curing process. For example, Famy et al. (2002b) have shown the presence of two or even three toned rims. The material formed during heat curing is brighter in backscattered electron images of polished sections in the scanning electron microscope due to its higher density (Famy et al., 2002c), while hydrates formed subsequent to the heat cure are darker because they have a lower density. These differences in microstructure and density no doubt account for the observation that materials that have been heat cured attain lower ultimate strengths than companion samples cured at room temperature, even where DEF expansion and strength regression do not occur.
At both ambient and elevated temperature, ettringite ends up present in voids in the microstructure, so the presence of large deposits of ettringite is an insufficient indicator for the diagnosis of DEF. The characteristic pattern of rims around aggregate must also be present, along with evidence of the concrete having been heat cured.
Analyses in the SEM can be used to probe the microchemistry taking place in samples. Samples cured at room temperature are found to have ettringite and/or AFm deposits mixed in with the outer C-S-H from early age. In contrast, in heat cured samples, only admixture of AFm is seen, while the Al and S from ettringite that would have formed are sorbed on to the C-S-H. Work by Lewis (Lewis, 1996 summarised in Scrivener and Lewis, 1996) showed that expansive and non-expansive materials could be distinguished by the level of sulfate adsorbed on to the inner C-S-H (measured in the rims present around the largest cement grains).
More recently, Famy (1999) extended this work and showed that, for samples in which expansion took place, there were significant levels of AFm present in outer product C-S-H, along with high levels of sulfate sorbed on to inner and outer product C-S-H. She proposed that the expansion was caused by conversion of AFm present in small pores into ettringite. This led to expansion of the paste and the onset of damage. Subsequently ettringite migrated to the aggregate rims where large visible deposits were found. In order for damage to occur, there needed to be sufficient AFm present in outer product, along with sufficient sulfate adsorbed on to inner and outer product. Taylor et al. (2001) discussed the details of this mechanism in their review. However, this is just one of a number of mechanisms that have been proposed. We will consider three here:
1. Paste expansion in which the cement paste expands, leaving rims around the aggregates which later infill with ettringite either when it forms or more likely by Ostwald ripening from much smaller ettringite crystals which formed earlier embedded in the paste. A number of causes can be postulated for expansion of the paste. If ettringite forms in small pores from AFm as suggested by Famy (1999), then it is possible to generate sufficient pressures to cause expansion. Expansion is possibly delayed after the ettringite formation, but this can be explained by a creep mechanism, or by an osmotic mechanism.
2. Formation of expansive ettringite rims has been proposed by a range of other authors. It has been suggested that this mechanism cannot generate sufficient pressures to cause expansion (Scherer, 1999), since the degree of super- saturation can only be low, and very high supersaturations would be needed to generate sufficiently large forces in such large rims. There is in addition some evidence that the width of the rims around aggregates is a function of aggregate size, which would be expected for a paste expansion mechanism, but not for a mechanism of crystal growth at aggregate surfaces. However, it is possible to imagine a mechanism in which diurnal temperature changes allow stepwise crystal growth and gradual levering apart of surfaces (as happens at the macro scale if expansion joints in concrete become gradually filled leading to damage to the surrounding slabs) so it may be premature to exclude this mechanism. Additionally DEF does often occur when the concrete is damaged by other degradation mechanisms such as ASR ± in these cases only a weak expansive force may be needed.
3. The osmotic mechanism was proposed for expansive ettringite formation by Thorvaldson (1954). This mechanism is implicated in other forms of damage to concrete such as ASR and freeze thaw attack. The transformation of AFm to AFt in the paste leads to consumption of considerable water, which may produce an increase in the ionic strength of the surrounding liquid resulting in osmotic pumping of water into the paste. Interestingly this mechanism easily accounts for the delay in expansion since water transport must take place following the formation of ettringite.
This topic has been highly controversial partly because of a number of lawsuits trying to assign responsibility for problems between relevant parties. Now that legal activity is reduced, consensus seems to be shifting towards the paste expansion mechanism but it may be time for the contribution of osmotic forces (Thorvaldson, 1954) to be re-examined within the context of paste expansion, since there is some evidence that expansion does not immediately follow ettringite formation, despite the fact that most techniques to probe ettringite formation become more efficient with increasing crystallite size (i.e., with ongoing Ostwald ripening). It is also important not to exclude the possibility that the mechanism can vary with circumstances. For example, if a mortar bar suffers from DEF and is then subsequently re-heated and re-exposed to moisture (Famy, 1999, fig. 7.46, p. 207), a very rapid burst of additional expansion takes place. At this stage, there is little constraint, and so formation of ettringite in rims is much more likely to be able to generate expansion. In contrast the onset of initial expansion needs much higher forces, and it is harder to see how these could be generated by simple crystal growth. Recently, Scherer (2004) has suggested that sufficient forces are generated transiently in very small pores fracturing the concrete, and that the constraints on expansion are then much reduced so that subsequent expansion can take place due to smaller forces in larger pores.
In summary then, DEF is a complex process which is still not fully under- stood. It can be avoided most readily by limiting maximum curing temperatures. Aside from DEF, internal sulfate attack caused by the presence of excessive sulfate in concreting materials is generally avoidable by adhering to the compositional limits in standards for cements and aggregates.
