Low Temperature Calorimetry

Initially, LTC analyses were conducted on only the most well hydrated (182 d and beyond) samples of the different cement pastes. Typical results for the w/c=0.35 and w/c=0.435 pastes are provided in Figs. 9 and 10, respectively. "Sealed/resat" refers to a specimen that was cured under sealed conditions, then resaturated (typically for 1 d) prior to the LTC measurement.

As can be seen in the figures, a single LTC scan produces a graph of heat flow versus temperature that consists of a baseline and one or more peaks at specific temperatures. Each peak corresponds to water freezing in a set of pores with an entryway pore diameter determined by the freezing point depression. ²⁴ Adopting the naming convention of Snyder and Bentz,¹⁰ the three most commonly occurring peaks are referred to as corresponding to capillary pores freezing at about −15 ºC, open gel pores freezing between about −25 ºC and −30 ºC, and dense gel pores freezing between about −40 ºC and −45 ºC. While only a coarse approximation, using the techniques outlined by Fagerlund,²⁴ these three freezing temperatures would correspond to pore radii of about 10 nm, 6 nm, and 4.5 nm, respectively. These pore radii are provided only as rough guides, as the freezing point depression is influenced by the ionic concentration of the (freezing) pore solution in addition to the size of the assumed cylindrical pore entryways.¹⁴

Figure 9- LTC results for CCRL cement 152, w/c=0.35 pastes, cured for 228 d.

As proposed by Snyder and Bentz,¹⁰ the presence or absence of the −15 ºC peak apparently indicates the percolation state of the capillary porosity within the hydrating cement paste. A percolated set of capillary-size pores will dramatically increase the permeability of the material ²⁵ and also directly influence its freeze/thaw durability, as the water in these coarser connected pores will be the first (and perhaps the only water) to freeze during exposure to low temperature weather conditions. In general, lower w/c pastes require less hydration to achieve depercolation of capillary pores than higher w/c pastes. For ordinary portland cement pastes with w/c>0.5, depercolation may not be possible even at complete hydration because sufficient capillary porosity remains to form a percolated network through the three-dimensional microstructure. ^{25, 26} For the w/c=0.35 pastes (Fig. 9), the only indication of a highly percolated capillary pore system was found for a specimen first cured under sealed conditions for 228 d and then resaturated (for 3 d) prior to the LTC measurement. This specimen exhibited peaks corresponding to both capillary and open gel porosity that were not observed for any of the three originally applied curing conditions. The most likely explanation is that the w/c=0.35 specimen cured under sealed conditions developed a connected (but empty) network of these pores, which became water-filled (and only then detectable by LTC) during resaturation. This network of connected pores would likely include the extremely coarse pores visible for the specimen cured under sealed conditions in Fig 5(a). The question remains as to whether these connected coarse pores were present throughout the hydration under sealed conditions or whether they depercolated initially due to hydration and later reopened and repercolated due to shrinkage of the C-S-H during self-desiccation, similar to the depercolation/repercolation observed by Bager and Sellevold during the drying/resaturation of well hydrated cement pastes.^{12, 13} Answering this question required further LTC experiments described below.

For the w/c=0.435 pastes, shown in Fig. 10, in general and as would be expected, a coarser percolated (open gel and capillary) pore structure was present for all three curing conditions, as indicated by the peaks at −15 ºC and −25 ºC. Surprisingly, the specimen cured under saturated conditions exhibited the most evidence of percolated capillary pore (small peak at −15 ºC) and percolated open gel pore structures. The sealed/saturated curing and the sealed curing specimen that was resaturated after 214 d both exhibited a much larger volume of the dense gel pores, suggesting that the capillary pores and many of the open gel pores had depercolated under these curing procedures and were thus only detectable as part of the dense gel pores (similar to the well known ink-bottle effect in mercury intrusion porosimetry). This is rather surprising in that it suggests that in terms of depercolating the capillary porosity to reduce the permeability and increase the durability of the cement paste, sealed curing may be superior to saturated curing, at least for the thin w/c=0.435 paste specimens examined in this study. To understand why this could be so, it is helpful to introduce a simple four-particle model for the hydrating cement paste microstructure.

Such a model is provided in Fig. 11, which shows four cement particles hydrating under either saturated or sealed curing conditions. For saturated curing, all of the capillary porosity remains water-filled and accessible to the growing hydration products. When curing for maximum strength, this may represent the ideal curing, as the degree of hydration will be maximized and no large empty self-desiccation pores will form. For sealed curing, such empty pores will indeed form in the largest available water-filled spaces and the growing hydration

products will be limited to forming in the remaining water-filled pore spaces. This will tend to concentrate the hydration product formation in the smaller pore entryways between particles, where it will be the most efficient in depercolating the connected capillary pore network, and thus this may be the preferred curing procedure for optimizing performance with respect to permeability and durability. Note that under sealed curing conditions in Fig. 11(b), the four pore entryways between particles are disconnected by the hydration products, while they remain connected under the saturated curing conditions in Fig. 11(a). Previously, both Swayze ²⁷ and Powers²⁸ have suggested that sealed curing for a short period of time prior to application of curing water could be beneficial for both short term (thermal cracking²⁷) and long term (frost action²⁸) concrete durability in some cases. The simple model in Fig. 11 would support the experimental observation that for the w/c=0.435 pastes, the capillary and open gel porosities are depercolated to a greater extent for the sealed and sealed/saturated curing procedures than for the saturated curing. Obviously, this did not hold true for the w/c=0.35 pastes, where a coarse percolated pore network developed in the paste cured under sealed conditions for 228 d (Fig. 9). Further LTC experiments were conducted to elucidate the differences between these two w/c pastes.

Cement pastes were prepared as before with w/c=0.35 and LTC scans were conducted at more frequent intervals beginning at an age of 1 d. Saturated, sealed, and sealed/saturated curing conditions were again investigated, but the amount of sealed curing time before resaturating the specimens was varied, unlike the constant 7 d value used in the first experiments. Results are shown in Figs. 12, 13, and 14 for early (1 d to 3 d), middle (4 d to 7 d), and late (14 d to 50 d) ages, respectively. At the earliest ages of 1 d and 2 d shown in Fig. 12, the dominant feature in the LTC scans for both saturated and sealed curing conditions is a large peak corresponding to connected capillary porosity. For both ages, the area under this peak is considerably less for the sealed specimens, in agreement with the self-desiccation occurring in these specimens relative to the water imbibition occurring in those cured under saturated conditions. At 3 d, in Fig. 12 one can clearly observe the formation of the open gel and dense gel connected pore structures at the expense of the connected capillary pore system, suggesting that the capillary pore entryways are being filled by gel hydration products as shown in Fig. 11(b).

As the hydration proceeds to 4 d, 5 d, and 7 d, the capillary pore peak is no longer observed in any of the LTC scans in Fig. 13. At 4 d, a large open gel porosity peak exists; this peak then gradually decreases, converting over to the dense gel pore structure, as the pore entryways are gradually filled by dense gel instead of open gel hydration products.¹⁰ In each case in Fig. 13, the sealed/resaturated specimens exhibit LTC scans quite similar to those obtained on the specimens cured under saturated conditions. The specimens that were resaturated after 7 d of sealed curing (bottom of Fig. 13) indicate that upon continuing hydration to 30 d, the open gel pores depercolate and nearly all water-filled porosity is detected as part of the dense gel network. These LTCs scans would suggest a less permeable, more durable cement paste microstructure. But, this is not yet the end of the story.

Proceeding to the later age hydration results shown in Fig. 14, perhaps surprisingly, the coarsest capillary pore peak reappears. For the w/c=0.35 paste, once the capillary porosity depercolates, self-desiccation will occur to some level in the hydrating microstructure, even in the specimens cured under saturated conditions. Significant autogenous shrinkage stresses will be created; the smaller the pores being emptied by self-desiccation, the greater the autogenous stresses and strains created in the three-dimensional microstructure.²³ Thus, these stresses and their accompanying shrinkage strains can be much greater in the w/c=0.35 cement paste, as it initially contains smaller pores than the w/c=0.435 cement paste, due to the closer packing of the cement particles. ^{15, 23} While lower w/c pastes would be expected to have higher values of elastic modulus, in practice, the autogenous stresses are higher, so that the measured autogenous shrinkage strain increases as w/c decreases.²⁹ As the gel shrinks, the pore entryways that were first depercolated by the gel growing during hydration may now become repercolated by the gel shrinking (like a wet sponge drying around a golf ball) during internal drying. This would result in the microstructure changing from one similar to that shown in Fig. 11(b) to one similar to that shown in Fig. 11(a), assuming that the cement particle cores remained in their original (fixed) locations. These observations are consistent with those of Bager and Sellevold concerning the drying and resaturation of room temperature cured cement pastes.^{12, 13} They observed that upon drying and resaturation, coarse connected capillary pores reappeared in the cement paste microstructures. While they imposed reduced relative humidities on their specimens via external drying, here, the internal relative humidity is reduced autogenously due to the chemical shrinkage and self-desiccation accompanying the cement hydration reactions. ^{23, 30, 31} It should be noted that for sealed curing conditions, microcrack formation could also be contributing to the observed repercolation of the open gel and capillary pores.

Figure 12- LTC scans for w/c=0.35 cement pastes cured for 1 d (top), 2 d (middle), and 3 d (bottom). Note that the y-axis scales are different. Sealed/resat specimen indicates first the time of sealed curing (3 d), followed by the time at which the LTC measurement was performed (4 d) after subsequent (re)saturated curing.

Figure 13- LTC scans for w/c=0.35 cement pastes cured for 4 d (top), 5 d (middle), and 7 d (bottom).

Figure 14- LTC scans for w/c=0.35 cement pastes cured for 14 d (top), 30 d (middle), and 50 d (bottom).

It is interesting to consider the ramifications of this repercolation of the capillary pore network for the distinct cases of sealed and saturated curing. In sealed curing, as the gel shrinks and the pores reopen, they will basically remain filled with air (or water vapor). As the shrinkage continues and the autogenous stresses intensify with any further hydration, the pores will become even more connected, as exemplified by the presented sealed/resaturated LTC scans for the 14 d, 30 d, and 50 d sealed curing regimens in Fig. 14. Thus, by 228 d as shown for the sealed/resaturated specimen in Fig. 9, both the open gel and capillary pore structures are highly percolated. Conversely, for saturated curing of the small specimens used in this study, as the pores repercolate, the permeability of the paste will increase and water will be more easily imbibed from the exterior to refill the empty pores. With continuing hydration and water (re)absorption (and possibly swelling) by the shrunken gel, the capillary pores may depercolate once again, starting the depercolation/repercolation cycle all over again. In this manner, the behavior of the material would be somewhat analogous to a pump, as a suction potential would be periodically created and satiated within the hydrating microstructure, all the while "pumping" in external curing water. Thus, the w/c=0.35 specimens cured for 228 d under saturated conditions exhibit primarily only a dense gel porosity peak in the LTC scan (Fig. 9) and there are very few coarse pores present in the hydrated microstructure (Fig. 5(c) and Table 2).

Whether this behavior would hold for w/c significantly lower than 0.35 is unknown at this time. With the initially much denser particle parkings in these lower w/c cement pastes, the initial depercolation of the capillary porosity may not be easily overcome by subsequent shrinkage of the C-S-H, so that the coarse capillary pore network may not repercolate. Furthermore, the shrinkage stresses will be even higher than in the w/c=0.35 cement paste, so that the C-S-H and other hydration products may shrink tightly (like shrink wrap) around the unhydrated cement particles, effectively preventing further hydration of the particles even if additional curing water were to be imbibed into the microstructure.²⁸ As with drying shrinkage, a portion of this autogenous shrinkage is likely irreversible, as new chemical bonds will be created within the C-S-H nanostructure.³² In this case, the best curing approach might be to avoid the self-desiccation altogether or at least reduce it by the use of internal curing,^3-5 especially considering the increased propensity for extensive microcracking and possibly macrocracking damage in extremely low w/c mixtures cured under non-saturated conditions. The best efforts to create a low permeability material will be totally compromised if the pore structure is only depercolated between percolated cracks.

As always, numerous complicating factors arise when considering the extension of these experimental results on small cement paste specimens in the laboratory to the curing of concrete structures in the field. One such factor is the presence of interfacial transition zones (ITZs) in concrete that are not found in the cement paste specimens. These ITZs have a higher effective w/c than the bulk cement paste and generally contain more and larger capillary pores. Thus, they will often be the first to empty during self-desiccation³³ and could easily constitute their own percolated network of capillary pores, compromising the low permeability of the bulk cement paste. With this in mind, it is not surprising that the permeability of concrete is often two orders of magnitude higher than that of comparable cement paste.³⁴ The microstructure of the ITZ can be densified, however, by the addition of silica fume or other fine pozzolanic or hydraulic materials,³⁵ so that this weak link may be effectively removed.

he shrinkage/swelling of the C-S-H gel seems to have major implications for the depercolation/repercolation of the capillary porosity in hydrating cement paste. Further support for this can be found in the LTC scans shown in Fig. 15, for a w/c=0.4 CCRL cement 140 paste cured under saturated conditions, with and without the incorporation of additional alkali sulfates in the mixture. CCRL proficiency sample 140 was basically a low alkali portland cement, containing 0.093 % Na₂O and 0.186 % K₂O per unit mass of cement; 0.76 % Na₂SO₄ and 0.93 % K₂SO₄ per unit mass of cement were added to the mixing water to create the mixture with additional alkali sulfates. After curing for about 100 d, clearly the paste with the added alkalis exhibits a pore structure that is substantially more depercolated than the paste with no additional alkalis. While it is known that alkalis generally accelerate early hydration of cement paste, they could also result in the formation of either a more voluminous or different morphology C-S-H hydration product,³⁶ due to the incorporation of potassium and sodium ions into the gel structure. While a depercolated capillary pore system can be beneficial from a low permeability and durability standpoint, it may be detrimental in terms of reducing the ultimate degree of hydration and strength achievable in the hydrating cement pastes. This would be consistent with the general observation that while additional alkalis generally accelerate the early hydration of cement paste, they often lead to reductions in the achieved degree of hydration and strength at later ages. ³⁶

To summarize, based on the microstructural observations from hydrating cement pastes, optimum curing is a function of w/c ratio, as well as whether curing is to optimize strength or permeability/durability. The following is a preliminary set of recommendations based on the present study for ordinary portland cement-based materials:

• very low w/c (≤ 0.35): internal curing may be a necessity to avoid self-desiccation, autogenous stresses and strains, and their accompanying damage to the hydrating cement paste microstructure,
• low w/c (0.35 to 0.4): internal or external curing may be a viable option, but "saturation" of the hydrating cement paste should be maintained as long as possible,
• intermediate w/c (0.4 to 0.45): sealed/saturated curing may be a viable option when curing for durability; otherwise, external saturated curing should be adequate,
• high w/c (≥ 0.45): it will likely be very difficult to depercolate the pore structure regardless of the curing practice employed.

In the latter case, if w/c is close to 0.45, it may be possible to depercolate the pore system by the addition of modest amounts of silica fume or other fine fillers. These materials further refine the pore structure and lead to an earlier depercolation of the capillary pore structure, as evidenced by the LTC results of Villadsen,³⁷ for example. In fact, as illustrated by comparing the results shown in Fig. 16 (for a 0.435 water-solids ratio (w/s) cement paste with 20 % limestone substitution for cement) with those in Fig. 10, even the substitution of fine limestone filler for cement,^{38, 39} may result in the formation of a depercolated capillary pore system after extensive (213 d) curing. Thus, in general, addition of silica fume or other fine materials that aid in depercolation of the pores would shift all of the above recommendations to higher w/c. For example, in these blended systems, internal curing may be a necessity for all w/s<0.40. In higher w/c residential concretes, as suggested by Hooton*, the best approach to increasing durability may be to add a moderate amount of these fine particles and pay appropriate attention to curing details.

*Hooton, R.D., personal communication, Dec. 2004