Next: Summary and Conclusions Up: Main Previous: Model Description

Results and Discussion

Plain Cement Paste Transition Zone

Fig. 2 provides a composite image of middle (right) and surface (left) slices for plain C₃S paste before and after hydration. The transition zone is evident in both the pre and post hydration images as the C₃S particles are unable to pack near the aggregate surface, a phenomena best observed in the surface slice prior to hydration. Even after extensive hydration, the transition zone remains inhomogeneous in comparison to the bulk paste, containing large pores and CH crystals, and relatively little C-S-H hydration product. As mentioned in the introduction, this inhomogeneity is due to both the initial packing and the one-sided growth effects.

Figure 2: Composite image of middle and surface slices for plain C₃S paste of w/c=0.45 before and after hydration (Black represents porosity, yellow represents C-S-H, blue represents CH, magenta represents aggregate, and red represents C₃S).

These qualitative observations are supported by the quantitative analysis of the various phase fractions as a function of distance from the aggregate surface, as shown in Fig. 3. As the interface is approached, large increases are noted in the CH and porosity concentrations, while correspondingly large decreases are noted in the C₃S and C-S-H phase fractions. Here, the simulation is in general agreement with observations of the transition zone in real concretes [2,3,4,5]. The largest change across the transition zone occurs in the volume fraction of the C-S-H phase, which exhibits a difference of 0.28 between its minimum and maximum volume fraction values. Since C-S-H, in combination with unhydrated cement particle cores, is generally accepted as being the major load-bearing phase in cement paste, it is straightforward to speculate that the interfacial zone in concrete may be the weak link in this composite due to the low amount of C-S-H and cement it contains, and not simply because of its porosity. If this is the case, increasing the amount of C-S-H in the transition zone may produce a higher-strength concrete.

Figure 3: Plot of phase fractions as a function of distance from aggregate surface for neat C₃S paste at 77% hydration.

Effect of Inert Mineral Admixture on Transition Zone

Ten percent by weight of an inert mineral admixture was added to a plain C₃S cement to investigate its effect on the transition zone. The admixture consisted of fine one-pixel particles. The simulation for this system was carried out using two different hydration conditions. In case one, the admixture was completely inert, so that there was no interaction with the diffusing species. In case two, the admixture was chemically inert (no pozzolanic reaction), but the diffusing CH species were allowed to nucleate onto admixture surfaces, forming solid CH crystals. Fig. 4 shows a composite image of the surface slices for this system before (left) and after hydration (middle-case 1, right-case 2), for both cases. As can be seen in the Fig. 4, the two cases produce greatly different microstructures in the interfacial zone. Prior to hydration, slightly more of the inert admixture is located within the transition zone, as can be seen quantitatively in Fig. 5, which shows the various phase fractions as a function of distance from the aggregate surface.

Figure 4: Composite image of surface for C₃S paste containing 10% inert filler with w/s=0.45 before and after hydration. Left-before hydration, middle-case 1, right-case 2. Black represents porosity, yellow represents C-S-H, blue represents CH, purple represents inert mineral admixture, and red represents C₃S.

Figure 5: Plots of phase fractions as a function of distance from aggregate surface for C₃S paste containing 10% inert mineral admixture at 77% hydration.

In the first case, when the admixture is totally inert, its presence has little effect on the final phase fractions after hydration in comparison to those shown for the plain C₃S paste (Fig. 3). In fact, because some of the cement has been replaced by an inert filler, the porosity is even higher than in the plain paste at the same degree of hydration. The CH distribution is slightly more homogeneous in the paste containing the inert admixture, as the admixture particles that are packed near the aggregate surface do slightly reduce the initially porosity and thus the volume of CH formed in this zone. In the second case, the large number of nucleation sites provided by the filler results in much more CH forming very near to the aggregate surface. Because of this excess CH formation, the porosity gradient is slightly reduced. It is unknown whether these large amounts of finely dispersed CH in the transition zone are capable of supporting a load or would simply act as weak links, as is apparently the case with plain paste concrete. The model does not address any possible orientation of the CH crystals, which could also play a major role in the strength of the interfacial zone.

Overall, the simulation results suggest that at a w/s ratio of 0.45, 10% inert filler addition would be detrimental to concrete performance, specifically in terms of its effects on the interfacial zone. Detwiler and Mehta have observed about a 10 reduction in 28-day compressive strength for concrete containing 10% carbon black at a w/s ratio of 0.5 [8]. Interestingly, at lower w/s ratios (0.25 and 0.35), their concretes containing carbon black had slightly higher (2-4%) strengths than those of ordinary portland-cement concretes. At these lower w/s ratios, the cement particles are more closely packed and the fine carbon black particles may bridge gaps between hydration products or provide additional sites for CH nucleation. This refinement of overall microstructure could lead to increases in compressive strength, which make up for the lower cement content. At higher w/s ratios, the cement particles are spaced farther apart and as indicated by the simulation, these admixture particle packing effects may become insignificant. Strength would then be reduced due to a lower initial cement content which would result in a higher porosity after hydration.

These simulations do not address other issues such as bleeding. If the transition zone in plain portland-cement concrete is also caused by bleeding effects, mineral admixtures that only act as inert fillers may still enhance the performance of the concrete by reducing the amount of bleeding in the fresh mix via capillary forces.

Effect of Pozzolanic Mineral Admixtures on the Transition Zone

Table II lists the five different pozzolanic mineral admixtures studied via simulation. Fig. 6 provides a composite picture of the initial particle packing for the pastes containing the large (bottom two images) and small (top two images) admixture particles. The ability of the smaller admixture particles to pack more closely to the aggregate surface is clearly evident. However, as shown by the results for the inert admixture, only filling in the interfacial zone with these particles may be inadequate to improve microstructural homogeneity. This is where the pozzolanic reactivity of the mineral admixture becomes important. Because the reaction of a pozzolanic admixture with CH produces secondary C-S-H having a greater volume than the original solid reactants, the effect of the presence of the mineral admixture in the interfacial zone may be the production of a more homogeneous microstructure and a better bond between paste and aggregate.

Figure 6: Composite image of middle and surface slices of initial particle packings for C₃S pastes containing 20% small(top) and large (bottom) admixture particles with w/s=0.45 Black represents porosity, tan (orange) represents pozzolanic material, magenta represents aggregate, and red represents C₃S.

All of the pozzolanic admixtures investigated were found to improve the integrity of the interfacial zone in simulated concrete. Fig. 7 shows images of surface slices taken after

Figure 7: Composite image of surface slices of various pastes (plain and with admixtures, w/s=0.45) after 77% hydration. Black represents porosity, tan (orange) represents pozzolanic C-S-H and admixture, yellow represents C-S-H, blue represents CH, and red represents C₃S. Top, left to right: neat paste, 20% small fly ash, 20% large fly ash. Bottom, left to right: 20% large condensed silica fume, 20% small reactive fly ash, 20% small condensed silica fume.

hydration. In Fig. 7, the top,left to right, represents: neat paste, 20% small fly ash, and 20% large fly ash. The bottom, left to right, shows: 20% large condensed silica fume, 20% small reactive fly ash, and 20% small condensed silica fume.) Clearly, the pozzolanic reaction of the admixture particles that produces secondary C-S-H results in substantial decreases in the CH in the interfacial zone and in the bulk paste, and in a more homogeneous spatial distribution of total C-S-H. The reduction in CH in the interfacial zone is consistent with experimental observations on concretes containing silica fume, fly ash, or blast-furnace slag [12,26]. However, the images in Fig. 7 also show that the different mineral admixtures have different degrees of effectiveness in modifying interfacial zone microstructure. For instance, in the presence of large (agglomerated) silica fume particles, some isolated CH is still produced even though the assigned pozzolanic reactivity and silica fume content are sufficient to consume all of the CH. The

large size of the admixture particles results in a more uneven distribution throughout the microstructure, so that in some localized areas, there are no silica surfaces with which the CH diffusing species can react. Therefore some solid CH is still formed. This is in agreement with the experimental observations of Rodger et al. who found the removal of CH to be greatly accelerated by a reduction in silica particle size [23,27].

The quantitative phase fraction vs. distance from aggregate surface plots for the concretes incorporating the fillers listed in Table II are provided in Fig. 8. For these plots, the primary C-S-H

Figure 8: Plots of phase fractions as a function of distance from aggregate surface for various pastes at 77% hydration.

and pozzolanic C-S-H are plotted separately, with the latter also including any unreacted mineral admixture particle cores. This separation of the two kinds of C-S-H, while easily accomplished using the model, would be extremely difficult to achieve experimentally as pozzolanic C-S-H is probably virtually indistinguishable from primary C-S-H in a back-scattered electron SEM micrograph [23]. Several points are worth noting in Fig. 8. First, the presence of the pozzolanic admixtures did not remove the porosity gradient present in the transition zone. Even with pozzolanic material present, the porosity still increases as the interface is approached. This behavior has been observed experimentally [5]. The most significant reduction in this gradient is achieved by the 20% "silica fume" addition and in general, at a constant reactivity, small admixture particles are superior to large ones in decreasing this porosity gradient. "Small" of course means small with respect to the average cement particle size. This analysis addresses only the overall porosity fractions and not the size of individual pore necks. In the images in Fig. 7, there is some evidence of a reduction in average pore size in the presence of the pozzolanic material, particularly with the smallest sized admixture particles.

In Fig. 8, the pozzolanic C-S-H gradient is seen to be greatly influenced by the filler particle size. Because the large admixture particles can not pack efficiently near the aggregate, less pozzolanic C-S-H is produced as the interface is approached. This is similar to the gradients observed for primary C-S-H in the plain and mineral admixture-modified pastes as the C₃S particles also cannot pack closely to the aggregate surfaces. Conversely, since the small admixture particles can pack more closely to the aggregate, there is actually an increase in admixture volume concentration as the aggregate is approached (Fig. 5), which translates into a greater fraction of pozzolanic C-S-H being produced in this region as well. For smaller admixture particles, the more reactive materials have larger positive pozzolanic C-S-H gradients, as each admixture volume unit that is present is more effective at filling-in the volume surrounding it with the pozzolanic C-S-H reaction product.

Goldman and Bentur have directly related the compressive strength increase of concrete containing silica fume to the reduction in CH achieved due to the pozzolanic reaction [15]. In Fig. 9, the fraction of CH vs. distance from the aggregate is plotted for the five admixtures listed in Table II along with that observed for the plain paste. All of the admixtures listed reduce the CH content substantially, with 20% fine silica fume being far superior to the others. Based on CH reduction alone, simulation results suggest that due to its high pozzolanic activity, silica fume would enhance the strength of concrete much more than an equal weight of fly ash, in general agreement with experimental results [8,28].

Figure 9: Plot of fraction CH as a function of distance from aggregate surface for neat and filled C₃S pastes with w/s=0.45 at 77% hydration.

Although correlation has been observed between strength enhancement and CH reduction, it is unlikely that the latter event is the direct cause of the former. A reduction in CH alone is not sufficient to increase the strength of concretes. For example, a CH reduction could be achieved by leaching of the CH out of concrete, but this would surely not increase strength. The true cause of the strength enhancement may be more related to an increase in C-S-H rather than a reduction in CH. Thus, both CH reduction and strength enhancement would be caused by the same phenomena, the pozzolanic reaction, and therefore would naturally be correlated.

To support this hypothesis, Fig. 10 shows a plot of the fraction of total C-S-H + C₃S as a function of distance from the aggregate surface for the different concretes. The total C-S-H fraction includes any unreacted admixture remaining in the concrete, as well as the primary and pozzolanic C-S-H fractions.

Figure 10: Plot of total C-S-H + C₃S + mineral admixture fraction as a function of distance from aggregate surface for plain and admixture-modified C₃S pastes with w/s=0.45 at 77% hydration.

Total C-S-H and C₃S have been combined together based on the assumption that these phases form the primary load-bearing matrix in the cement paste. The plots for total C-S-H only are similar to those in Fig. 10, but with slightly lower phase fraction values. The contribution of CH in the bulk paste to strength is being neglected in this analysis. Relevant statistics for the sum of the "C-S-H" and C₃S phases are provided in Table III. For purposes of comparison, statistics obtained for a 10% fine silica fume (reactivity = 2.08) admixture have also been included. The first column of Table III lists the volume fraction for the combined C-S-H and C₃S phase in the bulk paste, far from the interfacial zone. The second column lists the minimum volume fraction of this phase, found near the aggregate surface, and the third column lists the minimum volume fraction as a percentage of the bulk value. A completely uniform microstructure would have a value of 100% in the third column. Based on a weak link theory for compressive strength, it is assumed that compressive strength should correlate with the difference between the bulk and minimum load-bearing phase volume fractions, as this is a measure of the strength of the interfacial zone. This results in the following ranking in terms of strength: 20% CSF > 10% CSF 20% small reactive FA > 20% large CSF > 20% small FA > 20% large FA 10% Inert Plain Paste.

The results indicate that both admixture particle size and reactivity affect the hydrated microstructure, particularly in the interfacial region, consistent with the suggestions of Goldman and Bentur [15]. The results also suggest that decreasing the size of fly ash particles, whether or not accompanied by an increase in pozzolanic activity, would result in improved performance in concrete containing fly ash. Experimental results supporting this hypothesis have been reported based on the separation of fly ash into different size classifications and the subsequent observation that the finest fly ash provided the greatest strength enhancement in mortar [28,29].