3. Results and Discussion

Five readings of the surface tension were performed for each of the two liquids. For distilled water, the mean value of the measured surface tension was 0.0765 N/m with a standard deviation of 0.0003 N/m. For the solution containing 6 % SRA, the mean value was 0.0325 N/m with a standard deviation of 0.0005 N/m, a reduction of about 57 %.

It might also be expected that the lower surface tension would result in a greater drying rate for the bulk solution containing the SRA [10]. This hypothesis is confirmed in the results presented in Figure 1 which plots the cumulative mass loss for the two solutions, as well as for the organic SRA itself. The initial mass loss is substantially greater for the solution containing the SRA than for distilled water. Naturally, one might expect that the same trend would be observed in cement pastes, i.e., cement pastes with the SRA will dry at a faster rate than those without it. But, is this indeed observed...?

  

Figure 1: Mass loss of bulk solutions vs. time. Filled and hollow data points indicate the two replicates examined for each of the three solutions.

In fact, the exact opposite behavior is observed in the cement paste specimens examined in this study, as shown in the mass loss curves presented in Figure 2. Here, relative water mass loss indicates the fraction of the water initially present in the specimen which has been removed due to drying. Clearly, the presence of the SRA is now reducing the mass loss rate. Some insight into why this is the case is provided in the concurrent X-ray absorption results. The results plotted in Figures 3 and 4 show the differences in X-ray counts over time, relative to the setting time of about 3 h. This count difference is plotted against the location sampled by the X-ray system, with the bottom of the specimens being located at position 0 mm and the top exposed surface at 8 mm. As the cement paste dries out, the X-ray counts increase (resulting in a positive difference relative to any earlier time), as there is less material to absorb the X-rays passing through the specimen. As has been observed previously [5,6], for the paste without the SRA (Figure 3), a uniform drying is observed, with the largest pores throughout the paste thickness emptying first, followed by the next largest, etc. However, for the paste with the SRA (Figure 4), instead of drying uniformly, at early times, a relatively sharp drying front is formed at the exposed surface of the sample. It is hypothesized that the initial drying at the surface concentrates the SRA within the remaining surface pore solution so that it is more likely to continue to dry out as opposed to "drawing" water from the higher surface tension pore solution beneath it. After 13 h or so of drying, this front begins to disappear and ultimately, a uniform drying profile is achieved.

  

Figure 2: Mass loss of cement paste specimens (w/c=0.35) vs. time. Filled data points correspond to specimens exposed in the X-ray environmental chamber and hollow data points to those exposed in the laboratory.

  

Figure 3: Differences in normalized X-ray counts vs. position for w/c=0.35 cement paste without the SRA. Differences are relative to readings at 3 h.

  

Figure 4: Differences in normalized X-ray counts vs. position for w/c=0.35 cement paste with the SRA. Differences are relative to readings at 3 h.

Further insight into why the two systems exhibit different drying rates can be obtained from a simple three-dimensional microstructure model [11]. An initial three-dimensional microstructure (100 µm by 100 µm by 100 µm) consisting of digitized non-overlapping spherical particles (minimum diameter of 3 pixels (3 µm)) dispersed in water (w/c=0.35) was created. Empty pores were then created under two conditions: 1) empty porosity was created anywhere within the microstructure by emptying the largest (spherical) pores first regardless of their location, and 2) empty porosity was created by emptying the largest pores from the top surface inward to simulate the presence of a sharp drying front. Two-dimensional images from these three-dimensional microstructures are provided in Figures 5 and 6. In the latter case, one can easily suppose that the drying rate would be reduced by the presence of a "boundary" layer at the top surface of the specimen. Water molecules that evaporate must first diffuse through this tortuous boundary layer before reaching the external environment, resulting in a reduced mass transfer coefficient. In the former case, because water is quickly and easily "rearranged" throughout the microstructure (due to its high permeability), no boundary layer is observed to form at the top surface. Here, the drying rate can remain relatively high as water molecules evaporate directly from the surface into the external environment during an extended constant rate drying period [12]. Indeed, in Figure 2, it is observed that the drying rates for the two systems are similar for the first few hours and then diverge substantially as the paste with the SRA enters its "decreasing rate drying period" [12] well before the paste without the SRA. More details on the influence of microstructure on the constant rate and decreasing rate drying periods in porous materials can be found in the recent publication of Coussot [12]. It should be noted that for higher w/c ratio systems, if a layer of bleed water forms at the specimen surface (due to settling, etc.), it would then be expected that the drying rates would be controlled by the evaporation kinetics of the bulk solutions; in this case, the SRA might still accelerate the drying process, at least until the removal of solution from the settled microstructure is initiated.

  

Figure 5: Two-dimensional image from three-dimensional microstructure for the case of uniform drying. Red represents the solid particles, white water, and black empty pore space. Image size is 100 µm by 100 µm.

  

Figure 6: Two-dimensional image from three-dimensional microstructure for the case of drying from the top surface inward. Red represents the solid particles, white water, and black empty pore space. Image size is 100 µm by 100 µm.

It is also of interest to examine the drying kinetics and profiles for layered composite systems. The mass loss measurements for both possible geometric configurations were provided in Figure 2. The drying profiles for the case of paste with the SRA on top of paste without the SRA and vice versa are provided in Figures 7 and 8, respectively. When the paste containing the SRA is placed on top and exposed directly to the drying environment, it is seen to function as a very effective curing layer, significantly reducing the bulk mass loss from the specimen (Figure 2), and particularly minimizing the water loss from the paste without the SRA that comprises the bottom layer of the system shown in Figure 7. During the early hydration, in addition to the water loss from its top surface, the SRA paste also contributes water to the curing of the no SRA paste underneath it, as evidenced by the negative counts difference values for positions between 0 mm and 10 mm and times up to 13 h in Figure 7. The higher surface tension in the pore solution in the no SRA paste allows it to "pull" water from the lower surface tension pore solution in the layer above to replace that "lost" due to chemical shrinkage and self-desiccation. Of course, as the water is pulled from the top layer to the bottom one, some of the SRA will be pulled along with it. This is indicated by the "drying front" progressing deeper and deeper into the composite specimen at longer drying times (> 8 h). Perhaps this preferential water movement due to surface tension differences can be used to advantage in the design of future curing systems and methodologies.

  

Figure 7: Differences in normalized X-ray counts vs. position for cement paste with the SRA on top of paste without the SRA. Differences are relative to readings at 4 h.

When the paste without the SRA comprises the top layer of the composite, it first pulls water from the layer underneath it and later (13 h) dries out itself, ultimately resulting in a rather uniform drying profile, as shown in Figure 8. Similar behavior has been observed previously [5,6] for layered pastes where either the w/c ratio or the particle size distribution (PSD) of the cement varied between the two layers. In those cases, water was always observed to move preferentially from the coarse pore structure to the finer one. Here, preferential water movement is observed from the lower surface tension pore solution to the higher one. Because water is maintained at the top surface of the composite specimen for a longer time than in the case where the paste with the SRA is on top, the constant rate drying period is maintained for a longer time and this specimen loses much more water than the one where the layers are reversed (see Figure 2).

  

Figure 8: Differences in normalized X-ray counts vs. position for cement paste without the SRA on top of paste with the SRA. Differences are relative to readings at 4 h.

Further studies were conducted with the SRA for systems cured under sealed conditions. To first verify that the SRA might influence internal relative humidity development and autogenous deformation for specimens cured under sealed curing, the cement pastes prepared with the ultrafine cement (w/c=0.3) were examined; fine cements with a low w/c ratio tend to maximize the observed internal RH reduction [13]. Figure 9 presents the internal relative humidity development over the course of 11 d for pastes prepared with and without a 2 % SRA addition. For ages beyond one day, the relative humidity within the paste with the SRA is seen to be higher than that in the paste with no SRA. At earlier times, the SRA actually slightly decreases the RH, most likely due to a dilution effect (Raoult's Law [10]). If it assumed that the pore structure in the two cement pastes is basically the same, the differences in RH can be related to differences in surface tension via the Kelvin equation [10]:

$$ln(RH)=\frac{-2 \gamma V_m}{rRT}$$ (1)

where $\gamma$ is the surface tension of the pore solution in N/m, V_m is its molar volume in m³ /mol, r is the pore radius of the largest water-filled pore in m, R is the universal gas constant (8.314 J/(mol$\cdot$K)), and T is the temperature in Kelvin. Using this equation, an acceptable prediction (shown in Figure 9) of the results for the system with the SRA was produced when it was assumed that initially the surface tension of the cement paste pore solution in the system with the SRA was 74 % of that of the paste without the SRA and that the SRA is totally absorbed by the hydration products by 400 h of hydration (with a linear absorption rate during this 400 h). Certainly, these assumptions need to be verified in future studies, but suffice it to say that the SRA definitely affects the internal RH development in pastes cured under sealed conditions. Because the RH is linked to the stresses created within the microstructure, this also means that the SRA should influence autogenous shrinkage and early age cracking. Examining this hypothesis comprised the last portion of the experimental program. In addition to the RH measurements, after the 264 h (11 d) of hydration, the specimens were analyzed to determine their non-evaporable water content. For the specimens without the SRA, a non-evaporable water content of 0.126 g H₂O/g cement (standard deviation of 0.001) was determined. For the specimens containing the SRA, after correcting for the 0.02 mass fraction of SRA, a value of 0.125 g H₂O/g cement (standard deviation of 0.001) was measured. According to these results, the SRA has no effect on the achieved degree of hydration for these cement paste specimens cured under sealed conditions for 11 d.

  

Figure 9: Differences in internal relative humidity development for cement pastes (w/c=0.3) with and without the SRA cured under sealed conditions at 25 ºC. Dashed line is experimental data for the system with no SRA, solid line is experimental data for the system with the SRA, and dotted line is the predicted result for the system with the SRA.

The final portion of the study considered the early age properties of mortars prepared with and without the SRA. Here, a base system with a w/s ratio of 0.35 containing 8 % silica fume was used to provide a system where the autogenous shrinkage is significant. This base system can be contrasted against the same mixture containing 2 % of the SRA to see if the autogenous shrinkage is indeed reduced. The measured results for these two systems for internal relative humidity and autogenous deformation are provided in Figures 10 and 11, respectively. The presence of the SRA is seen to substantially decrease both the reduction in relative humidity and the measured autogenous shrinkage. During the first three days or so, the relative humidity in the system with the SRA is substantially higher than that in the system with no SRA. This translates into a system which undergoes almost no autogenous shrinkage during the same time period. After this, the two RH curves approach one another, and then as the silica fume reaction accelerates, they diverge once more. Even as the pore size within the hydrating cement/silica fume paste continues to be reduced, the presence of the SRA continues to result in a substantially higher RH and less autogenous shrinkage. In the long term, the SRA reduces the autogenous shrinkage by more than a factor of two, which should translate to a lower susceptibility to early age cracking for the mortar containing the SRA. Particularly, the reduction observed during the first 100 h is critical, as it is during this period that the hydrating cement paste is weakest and most susceptible to cracking.

  

Figure 10: Differences in internal relative humidity development for cement mortars (w/s=0.35) with (--) and without (- - -) the SRA cured under sealed conditions at 30 ºC.

The compressive strengths of mortar cylinders (60 mm in diameter and 120 mm in height) with and without the SRA, cured under sealed conditions at 30 ºC, were also measured. In each case, three replicates were tested for each mixture and curing age. At 7 d, compressive strengths of 47.9 MPa (standard deviation of 7.0 MPa) and 56.8 MPa (standard deviation of 6.6 MPa) were measured for the systems with and without the SRA, respectively. At 28 d, the strengths were 70.4 MPa (standard deviation of 5.5 MPa) and 61.3 MPa (standard deviation of 7.0 MPa) for the systems with and without the SRA, respectively. Thus, while there seems to be some early age strength loss due to the SRA, after 28 d of sealed curing, it actually appears to increase the compressive strength of these low w/s ratio mortars containing silica fume. However, considering the standard deviations achieved in the strength testing, these differences are not statistically significant, but the results do indicate that the SRA is not detrimental to 28 d strength for these particular mortars. Conversely, for concretes with w/c=0.42 and cured at 100 % RH, Brooks and Jiang have observed approximately a 28 % reduction in 28 d compressive strength for specimens with an SRA (1.5 % by mass of cement addition) vs. control specimens prepared without the SRA [14]. For concretes with and without silica fume and a w/s ratio of 0.35, Folliard and Berke observed a 6 % to 8 % reduction in 28 d strength for specimens containing a 1.5 % addition of the SRA, cured at 20 ºC and 100 % RH [15]. In the present study, the use of sealed curing at 30 ºC (along with the low w/s ratio and the addition of silica fume) is the most likely reason for the apparent strength enhancement produced in the specimens with the SRA. For example, it could well be that the maintenance of a higher internal RH, due to the presence of the SRA, extends the period of time during which the silica fume reacts pozzolanically with the calcium hydroxide formed from the cement hydration, leading to higher degrees of reaction and higher strengths.

  

Figure 11: Differences in autogenous deformation for cement mortars (w/s=0.35) with (--) and without (- - -) the SRA cured under sealed conditions at 30 ºC.