The measured LOI-based degree of hydration results for the three cement pastes are provided in Figure 1. As has been observed previously [1, 2], the addition of alkalis accelerates the early age (less than 7 d) hydration, while retarding the hydration achieved at later ages (28 d and beyond). Interestingly, at an age of around 7 d, the three pastes have very similar degrees of hydration as measured by non-evaporable water content. Furthermore, as shown in Figure 2, the TGA scans for 8-d old specimens of the three pastes are basically identical, suggesting that the same amounts of bound water exist in the three pastes, both in the C-S-H gel and other hydration products (removed in the temperature range of 100 ºC to 400 ºC) and in the calcium hydroxide (removed in the temperature range of 400 ºC to 550 ºC). The TGA-measured mass loss between 105 ºC and 1000 ºC was essentially identical (0.176 g /g ignited cement) for all three specimens.
Figure 1: Degree of hydration via LOI technique for w/c=0.40 cement pastes with and without alkali additions. Each data point is the average of two specimens taken from a single cast wafer at the indicated age.
LTC results for the three pastes after 2 d and 8 d of saturated curing are presented in Figures 3 and 4, respectively. The peaks near −15 ºC indicate the presence of a connected or percolated capillary pore system within the three−dimensional paste microstructure . For the 2 d specimens, a prominent peak near this temperature is observed for all three pastes. The two pastes with additional alkalis do indicate a shifting of this peak to lower temperatures by 2 ºC to 3 ºC, as would be expected due to the higher ionic concentrations in their pore solutions. The much smaller peaks at −40 ºC to −45 ºC indicate dense gel pores within the C-S-H hydration product . Here, the two pastes with alkalis have a greater area under these peaks, consistent with their higher degree of hydration after 2 d (Figure 1). The more surprising LTC results are to be found in Figure 4, where only dense gel pores are detected for the two specimens with added alkalis, while the original paste exhibits three percolated pore structures corresponding to capillary pores, open gel pores (peak at −25 ºC) , and dense gel pores . These results imply that the larger open gel and capillary pores in the pastes with alkalis have been depercolated by the hydration products, and at a degree of hydration equal to that achieved in the original paste without additional alkalis, which did not achieve depercolation.
Figure 2: Relative mass of w/c=0.40 cement pastes versus temperature for specimens with and without alkali additions hydrated under saturated conditions for 8 d at 20 ºC.
Figure 3: LTC results for CCRL 140 w/c=0.40 cement pastes with and without alkali additions cured under saturated conditions for 2 d.
This difference in pore space percolation at 8 d for specimens with equal w/c ratios and equal degrees of hydration suggests a difference in either the specific volume or the morphology of the C-S-H gel hydration products. One possibility would be that the presence of substantial numbers of alkali ions in the pore solution and their incorporation into the C-S-H gel nanostructure leads to a gel with a higher total water content (greater degree of swelling). For example, it is well known that the alkali silica gel (similar in chemical composition to C-S-H ) formed in concretes undergoing alkali-silica reaction can be quite expansive. While the TGA results suggest that the non-evaporable (bound) water content of the gels in the three pastes are similar, the evaporable water contents could be different. A more voluminous C-S-H gel would be likely to both reduce later age hydration rates due to spatial limitations and to also result in an earlier depercolation of the capillary pores, with respect to degree of hydration. The second possibility would be that the presence of substantial alkalis results in the formation of a C-S-H hydration product with a different morphology than that formed in low alkali cement pastes. Electron microscopy observations on C3S and portland cement pastes [3, 4] have indicated a greater tendency towards plate or lath−like hydration products with a higher degree of crystallinity in the presence of alkalis. Such a morphology could reduce long-term hydration, due to either a less permeable, more crystalline layer of C-S-H forming on the remaining unhydrated cement clinker particles or reduced diffusion rates within the pore space. The issue of whether hydration product morphology could affect the percolation of the capillary pores is considered next.
Figure 4: LTC results for CCRL 140 w/c=0.40 cement pastes with and without alkali additions cured under saturated conditions for 8 d.
The influence of hydration product morphology on porosity percolation was preliminarily examined using a simple dissolution/precipitation microstructure model. The model is similar to the original pixel-based C3S hydration microstructure model developed at NIST , but with the creation of only one type of hydration product that nucleates and grows as "crystals" in the available water-filled capillary pore space. The original particles dissolve, diffuse, and precipitate and grow these hydration products. The three-dimensional morphology of the reaction products is controlled to be either one−dimensional needles (one pixel by one pixel across oriented along one of the three principal axes), two-dimensional plates (one pixel thick), or random agglomerates. The model was run with an initial "water−to−cement" (or water−to−solid) ratio of 0.40 and an expansion factor of 2.15. The expansion factor is defined as the ratio of the volume of the reaction products to that of the solid reactants; for a typical portland cement, it has a value of about 2.15 [7, 9, 12]. As the reaction proceeds, the percolation of the pore space is periodically monitored using a simple burning algorithm . Typical results for the connected fraction of the capillary porosity as a function of total capillary porosity are provided in Figure 5. Clearly, in the systems where the reaction product morphology is limited to either needles or plates, the capillary porosity is depercolated at a significantly higher total porosity (equivalent to a lower degree of hydration). The formation of either needle or plate structures is seen to be more efficient at depercolating the capillary pore space than the formation of a random reaction product. This simple model thus supports the experimental observation that the capillary pore space depercolates at a lower degree of hydration in the pastes with additional alkalis where plate-like hydration products are apparently preferable. The estimated capillary porosity  for the three cement pastes examined above at an age of 8 d is 22 % to 23 %. In Figure 5, the system based on a random product morphology is still percolated at this total porosity value, while those based on plates or needles are both depercolated, in further support of the experimental observations.
Figure 5: Dissolution/precipitation model results for capillary porosity percolation as a function of capillary porosity for different product morphologies (plates, needles, or random) as indicated in the legend. Results shown are the average of 10 separate simulations for each morphology at a w/c=0.40, with an expansion factor of 2.15.
This difference in percolation should ultimately influence the transport properties and durability of the three cement pastes. For instance, as directly indicated by the LTC results in Figure 4, the pastes with alkalis contain little if any freezable water for temperatures down to about −40 ºC, which should result in an immense improvement in freeze/thaw durability relative to the original low alkali cement paste. Also, as first indicated by Powers et al. , there is typically a dramatic decrease in measured fluid permeability as the capillary porosity depercolates. While higher potassium and sodium contents in cements may not be favored due to ASR and other concerns, similar effects may be achievable by the addition of lithium admixtures conventionally used to mitigate ASR, as lithium ions have been observed to be "the alkali most preferentially incorporated into the cement hydrates" .