Comparison of Experimental and Computer Model Data

For the tritiated water diffusion studies, the conventional C-S-H was assigned a relative diffusivity of 0.0025 (as used previously [17]) and the pozzolanic C-S-H a value of 0.0005 (5 times lower, as suggested by the nanostructural models). In all simulations, the capillary porosity was assigned a relative diffusivity of 1.0. All model relative diffusivity values were then converted to absolute values by multiplying by the diffusivity of tritiated water in bulk water at 23 ºC, D_HTO, taken to be 2.24 x 10^-9 m²/s [27]. In this series of tests [20], silica fume replaced 6 % of the cement on a mass basis, and diffusion coefficients were determined from a steady-state experiment.

A comparison of the simulated values and those measured experimentally is provided in Table 3. For the systems without silica fume, the model values are within 25 % of the experimental ones measured by Delagrave et al. [20], a good agreement. The model value for the 0.2 w/c ratio cement paste of Matte and Moranville is about half of the measured value. Two possible reasons for this discrepancy are the 2-day 90 ºC curing regime applied to the experimental specimen and that the tritiated water diffusion test may have been conducted using solutions that were not saturated with respect to calcium hydroxide [22] (so that some leaching may have occurred during the measurement) unlike the tests performed by Delagrave et al. [19]. Both of these effects would tend to coarsen the pore structure, increasing the relative diffusivity of the specimen in accordance with the observed value relative to the model prediction. For the systems with 6 % silica fume replacement [19], the model values are within a factor of two of the experimental ones, a reasonable agreement.

 

Table 3: Data for Tritiated Water Diffusion in Cement Pastes

w/c [Ref.]	% CSF replacement	Exp. D x 10-12 m2 /s	Model D x 10-12 m2/s
0.45[20]	0	9.9	7.8 ± 0.3 a
0.45[20]	6	3.8	1.98 ± 0.02
0.25[20]	0	0.63	0.76 ± 0.01
0.25[20]	6	0.16	0.338 ± 0.002
0.20[23]	0	2.0	1.11 ± 0.01

^a Standard deviation for model diffusivity computed in each of the three principal directions.

For chloride ion diffusion (assuming a diffusivity of 1.81 x 10^-9 m²/s for chloride ions in bulk water at 20 ^oC [8]), good agreement with the experimental unleached diffusion data of Jensen [22] could only be obtained when the pozzolanic C-S-H was assigned a relative diffusivity of 0.0001 (25 times less than that of the conventional C-S-H). Once this value was used, however, as can be seen in Table 4, excellent agreement was observed between the experimental and computer model values for silica fume additions up to 10 %. For the 20 % silica fume addition, however, the experimental value was still far below the model value, perhaps suggesting further modification (densification) of the C-S-H nanostructure at very high levels of silica fume addition. An alternative possibility would be a further reduction in the diffusion coefficient due to the presence of unsaturated capillary porosity, despite the efforts to maintain saturation of the sample. The fact that the hydrated model system in this case has a capillary porosity of less than 2%, however, would suggest that this possibility is quite remote.

 

Table 4: Data for Chloride Ion Diffusion in w/c=0.3 Cement Pastes

% CSF addition	Exp. D x 10-12 m2/s	Model D x 10-12 m2/s
0	1.4	1.44 ± 0.01 a
3	0.7	0.80 ± 0.01
6	0.15	0.182 ± 0.001
10	0.05	0.0805 ± 0.0005
20	0.008	0.0769 ± 0.0002

^a Standard deviation for model diffusivity computed in each of the three principal directions.