Results and Discussion

Axial stress and expansion were measured on all specimens. Figure 2 and Figure 3 show the axial stress data obtained at two different w/cm ratios, 0.295 and 0.338. Figure 4 and Figure 5 show the results obtained for the expansion of companion specimens that were not confined in the frame.

First consider, in Figure 2 (w/cm = 0.295), the stress vs. time curve for the sample containing nominally non-reactive aggregates. This curve shows only a very small increase in stress level, due to water absorption, as was observed before (Ferraris 1997, Miyazawa et al. 1996). Figure 4 shows the companion expansion vs. time curve, showing little expansion. These results confirm the limestone aggregates were non-reactive under the test conditions.

Figure 2: Stress vs. time measured on the mortar specimens, for w/cm=0.295. One curve is given for a sample with non-reactive aggregates. All the other samples had highly reactive aggregates. Two plain cement samples are listed since one of them broke in the apparatus (broken specimen is marked #1). Results for silica fume and inert filler cement replacement samples are also given.

Figure 3: Stress measured in the mortar specimens at w/cm=0.338. The graph includes mortar with non-reactive aggregates and silica fume, with re aggregates; with reactive aggregates and cement replacement with silica fume or inert powder.

In Figure 2, two curves are shown for mortars made with reactive aggregates with no mineral additions. The curve that shows the lowest maximum stress, and even a loss of stress after 5 d, belongs to a specimen that was severely cracked and, therefore, could not bear any further load. The stress measurement was stopped for the other specimen because the load measured exceeded the capacity of the load cell (about 8 MPa). In Figure 4, the expansion measured for this mixture reached a plateau. Because, the expansion is measured on the vertical axis (axis of symmetry of the cylinder), the specimen may continue to expand in the radial direction while the vertical expansion has essentially ceased, thus the plateau in Figure 4. Therefore, the expansion plateau is considered apparent, rather than an absolute indication that the specimen has ceased expanding.

The stress vs. time curves in Figure 2 for the specimen with silica fume and reactive aggregates shows two phenomena:

• the maximum load and expansion is lower than the mortars without silica fume
• an equilibrium stress is reached after about 25 d.

This observation signifies that silica fume does mitigate the deleterious response of the concrete to the ASR in two ways: the overall stress level is reduced compared to samples without silica fume, and the stress level equilibrates, instead of monotonically increasing. One possible reason for this stress equilibration may be that the amount of dissolved alkalies was depleted. Studies of the pore fluid show that when using silica fume or other pozzolans, alkali concentrations first increase, and then decrease, reaching a low equilibrium value after 20 d to 30 d. In the absence of silica fume or other pozzolans, the alkali concentration increases continuously over time (Duchesne et al. 1994, Dreux et al. 1995). In Figure 2, the stress for the silica fume sample equilibrated after about 25 d. We interpret these results as indicating that the soluble alkali concentrations were reduced to a low value, so that the alkali-silica reaction was greatly reduced or stopped. If the reaction cannot continue, the deleterious effects (expansion and stresses) would be mitigated. This low value of alkali concentration has been attributed to the formation of hydration products such as C-S-H, which physically or chemically traps the alkali ions (Duchesne et al. 1994).

Figure 4: Expansion measured on mortar specimens with w/cm = 0.295. The graph includes mortar with non-reactive aggregates (NR), with reactive aggregates (R); with reactive aggregates and cement replacement with silica fume (R+SF) or inert powder (R+NR). The uncertainty bars represent one standard deviation.

Mentioned in the "Background" section were three other possible ways silica fume mitigates the deleterious effects of ASR. One was that the lowered permeability induced by the silica fume, which is well-documented in the literature, prevents the ingress of water. Recall that the swelling associated with ASR is thought to be induced by the ASR gel absorbing water. Of course, a lower permeability will also tend to prevent mix water from escaping by evaporation, in air-cured specimens. Since in the present study, the specimens were cured and kept in water, it is possible that the lowered permeability of the silica fume specimen did mitigate the ASR expansion by limiting the ingress of water. However, the equilibration of the stress vs. time curve for the silica fume specimen might tend to rule against this, since the specimen will absorb some water over time, if left long enough in the water bath. We plan to let a silica fume sample stay in the load frame in lime water for an extended period of time to study this. It is interesting to note that as shown in Figure 4, the expansion of the silica fume specimen also equilibrated and stayed fairly constant for a period of up to 70 d. This finding would tend to rule against the lower permeability hypothesis, as some expansion should have been seen if the ingress of water were only slowed, not stopped. Also, at this w/cm ratio, 0.295, even the samples without silica fume have been found to have a low value of permeability, but still showed high values of expansion and stress.

Another possibility for the ASR mitigation effect of silica fume is the strengthening of the material due to the strengthening of the interfacial transition zones. However, the main effect of silica fume is to increase the compressive strength of the material. The tensile strength of the specimen would still have been relatively low, so that the local cracking of the matrix induced by ASR expansion would not have been affected. The increase in elastic modulus reported in the literature for silica fume specimens (Dreux et al. 1995) would tend to reduce the expansion to some extent, by making the specimen stiffer, but not to the extent seen in the data.

The fourth possible explanation for the silica fume mitigation effect on ASR was the reduction of alkali concentrations because of the reduction of cement, as silica fume does not contain any appreciable amount of alkalies. This was tested directly by replacing the cement with an inert filler, silicon carbide, using the same mass percentage replacement as for the silica fume. The induced stress and expansion vs. time curves for this scenario are also shown in Figure 2 and Figure 4. The stress vs. time curve was discontinued after a few days, as the load measured exceeded the capacity of the load cell. Up to this point, this curve closely follows that of the sample with no mineral admixtures, which tends to disprove the hypothesis that a reduction in alkalies due to reduced cement content mitigates the ASR stress and expansion.

Figure 3 and Figure 5 contain similar data for a different w/cm (water/solid) ratio, 0.338. The same inferences from the data can be drawn as for the w/cm = 0.295 data, implying that for at least these fairly close values, the w/cm ratio used does not appreciably affect the ASR results.

Figure 5: Expansion measured on mortar specimens with w/cm = 0.338. The graph includes mortar with non-reactive aggregates and silica fume (NR+SF); with reactive aggregates (R); with reactive aggregates and cement replacement with silica fume (R+SF) or inert powder (R+NR). The uncertainty bars represent one standard deviation.

Using SEM imaging, the area fraction of cracks in the aggregates and in the matrix were measured. The specimens were cut perpendicular to the vertical axis and a "disc" was prepared for SEM imaging. Therefore, the cracks that are measured are radial. A point count of SEM backscattered electron images for estimates of area and volume fractions of selected components was made. Fifty-six fields from the approximate center of each specimen were examined at a magnification of 200x, allowing recognition of cracks of less than one µm in width. The point count grid of 25 points provided a point spacing of about 100 µm, with a total of about 1400 points being counted for each specimen. Table 3 shows the results obtained. The error estimates indicate that the specimen with silica fume was different than the other two specimens. The difference appears to lie in the lower volume of cracking within the aggregates of the silica fume specimen. The volume of paste cracking appears to be the same, within the experimental uncertainty, from that of the other two specimens.

Figure 6 and Figure 7 show the SEM image of the cracks in a specimen with no mineral addition and a specimen with silica fume addition, respectively. The A part of both figures shows the aggregates (light gray) and the paste, while the B part of both figure shows the cracks in black and the solid material in white. As indicated in Table 3, the reduction of cracks in the aggregates for the specimen with the silica fume addition can be seen. Since ASR usually involves gel formation and cracking of the aggregates, leading to cracking in the paste, we tentatively conclude that the silica fume specimen has less reaction in the aggregates, leading to less cracking in the aggregates and less overall expansion.

Table 3 -- Estimated relative crack area in the specimens. The uncertainties are one standard deviation.

Figure 6: SEM picture of crack pattern for a specimen with no additions. A) SEM picture in normal light and showing the aggregates in light gray; B) the same picture but with the cracks and pores highlighted in black, with everything else, paste and aggregates being white.

Figure 7: SEM picture of crack pattern for a specimen with silica fume additions. A) SEM picture in normal light and showing the aggregates in light gray; B) the same picture but with the cracks and pores highlighted in black, with everything else, paste and aggregates being white.