For the specimens cured under saturated conditions, for those LTC scans that exhibited a discernible peak near −15 ºC, the peak height (peak value - baseline value) was measured as a simple direct indication of the amount of (percolated) freezable water. The direct measurement of peak height was used as opposed to more conventional but complex approaches such as computing the peak area convoluted with the heat of fusion of water as a function of temperature, due to the generally large uncertainties in the appropriate values to use in the latter approach.25 This peak height value was then directly compared to the volume fraction of percolated pores (with a voxel resolution of one micrometer) as computed using a burning algorithm8 in the CEMHYD3D computer model. Because the LTC results are reported on a mass basis and CEMHYD3D provides a measure of the percolated capillary porosity volume fraction, a mass to volume basis conversion that also included the extra water that was imbibed into the saturated specimens due to chemical shrinkage was performed. Comparison plots are provided in Figures 12 to 16 for the various w/c and curing temperatures employed in this study. In Figure 14, model results are presented for three separate executions of CEMHYD3D (with three different starting w/c = 0.35 microstructure representations) to provide some indication of model variability for the simulated percolated capillary porosity volume fraction. In general, reasonable agreement is observed between CEMHYD3D predictions and the values measured in the LTC scans, particularly for connected pore volume fractions above 0.15. Greater fluctuations would be expected at the lower connected pore volume fractions as the critical percolation/depercolation transition is approached, 9 as indicated by the "noise" in the CEMHYD3D curves in Figures 15 and 16, for example.
Figure 12. Comparison of LTC-estimated and CEMHYD3D predicted percolated capillary porosity volume fractions vs. time for w/c = 0.25 cement paste hydrated under saturated conditions at 20 ºC.
Figure 13. Comparison of LTC-estimated and CEMHYD3D predicted percolated capillary porosity volume fractions vs. time for w/c = 0.25 cement paste hydrated under saturated conditions at 40 ºC.
In Figure 15, further justification for the modification of the local C-S-H distance (locations) from n = 8 to n = 3 for the 40 ºC curing is observed. With the former fixed value of n = 8 voxels at all temperatures, the time (and degree of hydration) of the w/c = 0.35, 40 ºC depercolation would be vastly underpredicted, while the predictions with n = 3 are in better agreement with the experimental LTC measured peak heights.
Figure 14. Comparison of LTC-estimated and CEMHYD3D predicted percolated capillary porosity volume fractions vs. time for w/c = 0.35 cement paste hydrated under saturated conditions at 20 ºC. Model results for three different microstructures (created and hydrated with different random number seeds) are shown to provide some indication of model variability.
Figure 15. Comparison of LTC-estimated and CEMHYD3D predicted percolated capillary porosity volume fractions vs. time for w/c = 0.35 cement paste hydrated under saturated conditions at 40 ºC.
Finally, the measured LTC peak heights were used in a similar manner to calculate the later age "damaged" (re)connected pore volume fractions for the specimens cured under sealed conditions. The damaged porosity was simply taken to be equivalent to the peak height of the peak near −15 ºC in the LTC scans. The results, summarized in the graph in Figure 17, indicate that damaged porosities of several percent volume fraction are possible and that the severity of the autogenous damage appears to increase with either a decrease in w/c or an increase in curing temperature. This analysis is heavily dependent on the assumptions utilized to interpret the LTC data. Furthermore, it addresses only the change in connectivity of the capillary porosity due to self-desiccation (internal drying) and not the likely accompanying change in pore size. Both would be expected to contribute to an increase in the permeability of the cement paste, perhaps even on the order of the factor of 70 increase previously observed by Powers and colleagues. 2