The effects of chemistry, w/c ratio, and cement particle size distribution on the percolation aspects of the NIST cement hydration model were studied. The role of cement particle size distribution was significant. The effect of w/c ratio was the same as that found before: all percolation curves for different w/c ratios for a given cement collapsed on a single curve when plotted against the volume fraction of the phase of interest. Portland cement chemistry beyond simple C3S cements did not have a significant effect on percolation quantities, except for those of the C-S-H phase. This was explained from the relative production of C-S-H in the portland cement hydration vs. that of the C3S.
The three major sources of error present for every digital, random model - finite size error, statistical fluctuation, and digital resolution - were studied. It was found that finite size error and statistical fluctuation were not important for the cement hydration model as it is usually used (1 µm per pixel, 1003 size, about 2000 particles). Digital resolution did have an important effect on the percolation quantities and the transport properties of diffusivity and permeability. However, intrinsic length scales present in cement paste, which include the typical size of cement particles that very quickly dissolve and hydrate, the thickness of the first C-S-H layer, and the size of the largest C-S-H pores, along with the values of the set point for real materials, restrict the resolutions that will give realistic values of experimental quantities for the model.
Further detailed experimental work on cement paste should aid in determining the "right" resolution to use in the model. It does seem that this "right" resolution must be near 1 µm per pixel, as comparison with numerous experimental results appear to point to that resolution. This is much different than the case of finite difference or finite element computations, where the "right" resolution is as fine as possible [13,14,33].