To gain further insights into the influence of cement PSD on initial pore size distribution and early age autogenous properties, the four experimental systems were also simulated using the NIST cement hydration and microstructure development model [6,18]. The measured PSDs (Fig. 2) and Bogue composition of the cements were used to create starting microstructures with w/c=0.35. The starting microstructures were then hydrated under both saturated and sealed conditions to simulate the experimental conditions for the chemical shrinkage and autogenous measurements, respectively. Under saturated conditions, all capillary porosity remains water-filled throughout the hydration process, but the model calculates the volume of water which would be imbibed to maintain this saturation (equivalent to the chemical shrinkage) as a function of the achieved degree of hydration. When operated under sealed conditions, the model creates the appropriate volume of empty pores as the hydration proceeds . By calibrating the chemical shrinkage predictions of the model operated under saturated conditions to those observed experimentally, the degree of hydration vs. time for hydration under sealed conditions could be inferred. This allows one to plot the observed autogenous measurements against degree of hydration (a material parameter) as well as against elapsed time, to separate differences in kinetics from those in microstructural features.
Two-dimensional slices from the initial 3-D microstructures for the four cement pastes are provided in Fig. 3. One can clearly observe that the open "pores" between cement particles are much larger in the systems based on the coarser cements. In order to attempt to quantify this microstructural difference, a 3-D spherical adsorption/desorption program [19,20] was used to determine the pore volume accessible as a function of sphere diameter. The algorithm is a rough estimation of adsorption/desorption in a porous media, because it assumes a spherical meniscus shape, while in reality, a variety of 3-D ellipsoidal shapes are possible for the meniscus. Still, the algorithm does provide good insights into differences in the initial pore size distributions of the four systems, as shown in Fig. 4. For adsorption, the algorithm simply determines the pore volume occupied by all of the locations within the capillary porosity where a sphere of a given diameter could be placed. For desorption, a connectivity criteria is added such that these identified pores must be accessible from the exterior surface of the sample by a pathway composed only of pores of this diameter and larger. For a sphere diameter of 1 µm (pixel), a conventional burning algorithm  is used to determine the accessible porosity. For all four w/c=0.35 systems, nearly all of the initial porosity is percolated, since the initial porosity in a w/c=0.35 paste is about 0.53 (porosity= 1./((1./(w/c)/3.2)+1.) where 3.2 is the specific gravity of cement). However, for a sphere diameter of 3 µm, large differences are observed between the four finenesses. The two higher finenesses contain almost no porosity that is accessible for a diameter of 3 µm, while the two coarser ones contain significant pores that are accessible to this sphere diameter. This would suggest that for a fixed amount of created empty porosity (due to self desiccation), smaller pores would be emptied in the finer cement pastes than in the coarser ones. This should result in a faster decrease in autogenous relative humidity, an increased autogenous shrinkage, and perhaps an increase in the eigenstresses existing at an aggregate (or stress sensor) surface. This hypothesis is examined quantitatively in the following results.