While Powers first inferred depercolation of the capillary porosity in hydrating cement paste via measurements of permeability [8], the depercolation can also be observed based on chemical shrinkage measurements on pastes of various thicknesses [21] or via low temperature calorimetry scans [22-24]. A cooling scan in an LTC experiment is basically equivalent to a mercury porosimetry intrusion scan [22], but with the advantage that no drying of the specimen is required. Generally, a percolated water-filled capillary pore structure is indicated by a peak around −15 ºC [24]. Thus, the presence or absence of this peak can be used to infer a percolated or depercolated capillary pore structure, respectively.
Figure 4 provides representative LTC scans of a w/c=0.35 cement paste cured under saturated or sealed conditions, along with scans on the pastes cured under sealed conditions after 24 h or more of resaturation. The w/c=0.35 is low enough that depercolation of the capillary pores would be expected to occur during the first week of curing [8]. For the various scans in Figure 4, basically three different peaks are observed corresponding to percolated capillary pores (freezing at about −15 ºC), open gel pores (freezing at -25 ºC to −30 ºC), and dense gel pores (freezing at −40 ºC to −45 ºC) [24]. For saturated curing, the capillary pores are observed to depercolate between 3 d and 4 d of curing. Initially, a similar depercolation is observed for the capillary pores in the specimens exposed to sealed curing conditions. After a few weeks of curing, the open gel pores generally also depercolate, so that only pores accessible via the dense gel pores are detected via LTC. Specimens cured under sealed conditions are consistently seen to have smaller peaks for the capillary and open gel pores, as these larger pores are the first to empty due to chemical shrinkage and self-desiccation [11, 25].
Figure 4: LTC results for CCRL cement 152 specimens (w/c=0.35, 20 ºC) at various ages, cured under saturated (top) and sealed (middle) conditions and for specimens cured under sealed conditions for the indicated number of days and then resaturated (bottom) [11].
Interestingly, resaturation of the sealed specimens reveals a change in the percolation of the capillary pores that is not observed for the specimens cured under saturated conditions. For sealed conditions, while the capillary pores do initially depercolate, by 14 d, when resaturated, the specimens exhibit a repercolated set of capillary pores (bottom plot for Figure 4), most likely due to the autogenous stresses and strains placed on the three-dimensional microstructure due to self-desiccation [25]. Similar effects have been observed by Bager and Sellevold upon exposing well-hydrated cement pastes to drying/resaturation [22]. Drying, whether external or internal (self-desiccation), results in the creation of a percolated set of capillary pores (or perhaps microcracks). As shown by the scanning electron micrographs in Figure 5, the specimens cured under sealed conditions definitely contain a set of large capillary pores relative to those present in the paste specimens cured under saturated conditions. Thus, a plausible explanation for the behavior observed for the sealed/resaturated specimens is that the autogenous shrinkage of the C-S-H gel reopens the entryways of the previously depercolated capillary pore network. Still, it can not be ruled out that some microcracks could also participate in either repercolating the capillary pores or creating their own percolated network of porosity.
Figure 5: Ninety-two day (92 d) segmented scanning electron microscopy (SEM) images for CCRL cement 152, w/c=0.35 cement paste specimens cured under sealed (left) and saturated (right) conditions [11]. Unhydrated cement particles are white, calcium hydroxide is light grey, C-S-H and other hydration products are dark grey, and capillary pores are black. Images are 384 µm by 512 µm.
While sealed curing appears to be detrimental in terms of microstructure (specifically pore structure) development for w/c=0.35 pastes, for w/c=0.435 pastes, as illustrated by the LTC cooling scans in Figure 6, it may actually be beneficial. As self-desiccation occurs during sealed curing, the largest water-filled pores in the three-dimensional microstructure will empty first [11, 25]. Since cement hydration products will generally not precipitate and grow in air (or water vapor)-filled pores, hydration product formation will tend to be concentrated in the remaining smaller pores and pore entryways, where it should be more effective in depercolating the (water and vapor-filled) capillary pores [11], as supported by the experimental results in Figure 6. Thus, if curing to minimize transport and maximize durability, for an intermediate range of w/c (e.g., 0.4 to 0.45), some type of sealed/saturated curing could be superior to maintaining saturated conditions throughout. Such a seemingly counterintuitive concept is not new, having been suggested by both Swayze [26] and Powers [27] over 50 years ago.
Figure 6: LTC results for CCRL proficiency cement 152 specimens (w/c=0.435, 20 ºC) cured for 214 d [11].
Recently, it has been demonstrated that the percolation of the capillary porosity in hydrating cement paste can also be influenced by the level of alkalis in the cement paste [11, 13]. In the presence of sufficient alkali ions, the C-S-H has a tendency to form lath or plate-like nanostructures, with a higher degree of crystallinity [28]. Simple three-dimensional microstructure models have indicated that hydration products forming as needles or plates, as opposed to a random geometry, can be more efficient at depercolating the capillary pore space between the original cement particles [13]. In Figure 7, LTC cooling scans are provided for a set of hydrated cement pastes with and without additional alkalis, all of which have achieved nominally the same degree of hydration after 8 d of curing at 20 ºC [13]. While the paste with no additional alkalis clearly contains both percolated capillary and open gel pore structures, only dense gel pores are identified in the pastes with either alkali sulfate or alkali hydroxide additions. For both additions, the same molar quantities of potassium and sodium ions (per unit mass of cement) were added to the mixing water and dissolved completely prior to the addition of the cement. This example illustrates one potential method for engineering the nanostructure of the dams formed during cement hydration.
Figure 7: LTC results for CCRL cement 140 specimens (w/c=0.40, hydrated under saturated conditions for 8 d at 20 ºC) with and without alkali additions (sulfates or hydroxides) [13]. All specimens have achieved basically the same degree of hydration, as measured using LOI [13].