DEF is a fairly weakly expansive mechanism, and so is likely to be less severe in dense flaw-free pastes and concretes. This is clearly seen by the significant effect of the constraint present around cement paste on the onset and magnitude of expansion. Grattan-Bellew et al. (1998) showed that, for ASR-inactive quartz aggregate, the rate of expansion was inversely proportional to the mean aggregate size. Lawrence (1995) tested a range of mortars made from limestone and silica sands, observing that replacement of the coarse fraction of the siliceous sand with limestone was particularly effective in limiting expansion, presumably because the bond to limestone was much better. Lower expansions were also seen for cement pastes than for the corresponding concretes (Yang et al., 1996). Indeed Diamond and Ong (1994) showed that the use of an ASR- active aggregate induced ASR damage around the aggregate, weakening the concrete and making it more susceptible to DEF expansion. Thermal cycling, which might also be expected to cause micro-cracking at interfaces, has also been shown to increase the rate of expansion, although not significantly changing the time of onset or the overall expansion amount (Fu et al., 1997).
While much data have been obtained detailing the expansion of mortars and concrete, fewer data are available regarding strength development in systems suffering from DEF expansion. Lewis (1996) tested mortars made from a range of cements (w/c ˆ 0.5 and sand/cement ˆ 3) and measured compressive strength and weight gain on wet storage against expansive behaviour. Expansion data were measured on 16mm by 16mm by 160mm bars, while the other properties were measured from 25mm cubes, so there is some expected variation between samples, particularly as different storage containers were also used, implying that samples may have experienced different alkali levels in the storage solutions (see below). At 20ëC, all mortars were non-expansive and reached compressive strengths of approximately 50 MPa with a weight gain at early age of approximately 1%. Some samples cured at 90ëC did not expand; these samples also gained approximately 1% mass and reached around 30±40 MPa. The weight gain was attributed to consumption of water by hydration, and the lower ultimate strength from heat curing to a coarser microstructure. Mortars made from some cements expanded at 90ëC and these gained weight and strength initially in a similar manner to the non-expansive mortars. However, drops in strength and significant additional weight gains took place coincident with expansion.
The timing and duration of heat curing is important. Very long (>24 h at 90ëC) heat curing leads to delayed or reduced expansion (Lawrence, 1995) presumably because the components needed for the delayed ettringite formation chemically react to form other products (for example hydrogarnets) and are no longer available. Similar results were found by Famy (1999). It is also found that delaying the heat curing process leads to much reduced expansion (Famy et al., 2002b), and it is postulated that this is because the bulk of the C-S-H is then formed at 20ëC and only C-S-H formed at 90ëC can store sufficient aluminate and sulfate to promote subsequent DEF expansion.
It is not possible to predict the expected behaviour of a material solely from the composition of the cement used, although composition clearly is one important factor as demonstrated by Kelham (1996a). For example, high levels of alkali destabilise ettringite relative to AFm more than lower levels do. The ratio of aluminum to sulfur is important in determining the expected final phase assembly. However the silicate components of the cement are also very important since it appears that, prior to ettringite reformation, sulfur (and possibly aluminium) must be stored by adsorption on to the C-S-H gel component of the hydrates. DEF appears to be associated with reactive cements which give high early strength (Kelham 1996a,b), suggesting that substantial amounts of C-S-H gel must be formed during the heat curing in order to sorb sufficient sulfate to allow subsequent formation of sufficient quantities of ettringite within the outer product C-S-H gel to cause DEF expansion (Taylor et al., 2001; Famy, 1999).
Some workers have suggested that slow release of sulfates from clinker sulfate phases may lead to late formation of ettringite and expansion. However, a study by the PCA of the solubility of sulfate in a range of North American cements (Klemm and Miller, 1996) showed that, in 33 cements tested, nearly all of the sulfate was released while the concrete remained plastic. Thus slow release of sulfate was unlikely to be a major cause of DEF. Indeed slow release of sulfate might also be expected to cause expansion at room temperature but results by many investigators (e.g., Kelham, 1996a,b; Yang et al., 1996, 1999; Famy, 1999) do not show such expansion.
Another important factor in DEF is the storage environment. As in the case of ASR, very high relative humidities are needed for expansion to occur. Famy et al. (2001) showed that wet storage led to a more rapid expansion than storage above water. This is easily explained, since ettringite has a very high water content and is unstable at low relative humidities. Famy et al. (2001) also showed that the ionic strength of the storage solution was important. Storage of bars in solutions of high alkali content reduced or delayed expansion. They postulated that, if high alkali levels were maintained in the pore fluid, then ettringite was destabilised and its reformation delayed. An alternative explana- tion of this phenomenon can be envisaged if DEF involves an osmotic mechanism since a high external ionic strength would impede osmotically induced expansion.
This sensitivity of DEF to the storage environment is important practically since, in many studies of DEF, bars cast at various times and ages have been stored in a common water bath with the alkali concentration gradually increasing as pore fluid from the various bars leached into the water. Thus the storage conditions varied for bars cast at different times. Testing for expansion should always be carried out with care taken over the storage conditions. For example, all bars could be stored in their own containers, each with a fixed quantity of water. The sensitivity to alkali leaching might also explain why DEF has been a particular problem in environments such as railroad sleepers and abutment walls, where water flow and leaching can take place.
