There have been two main previous model studies of the percolation characteristics of cement paste phases, using two different models [9,20].
The first of these was on an early version of the present model, with fixed cement chemistry (C3S only), fixed cement particle size distribution (psd), and fixed resolution (1003 pixels). Computational limitations at that time prohibited consideration of higher resolutions, but it was determined that finite size errors and statistical errors were negligible. As will be seen later, the cement psd does not matter very much as long as it is broad. Unfortunately, the particle size distribution that was used in this early work was for only four different sizes of particles. This in itself was not so much of a problem, but the four different sizes had equal numbers, which meant that the distribution was heavily skewed, in terms of particle mass, towards the larger particles. The effect of the cement psd is studied in Section 5, along with the effect of cement chemistry and other changes to the basic model made since the time of the first study (1991). The effect of digital resolution is left to Section 6.
The second study of percolation in cement paste focused on the percolation of the capillary pore space, which is defined as the original water-filled space around the cement particles and their hydration products. The model used was a version of an older model [21], where the particles were represented by collections of continuum spheres, not digital images. Hydration was simulated by the spherical growth of cement particles, along with the generation of CH particles. The particles were allowed to overlap, with some correction made for the common volume between two overlapping particles as they both grew.
Percolation of the capillary pore space was evaluated by digitizing the resulting continuum image at varying resolutions, then studying the percolation properties as a function of resolution. The resolutions studied with this model ranged from 10 µm per pixel to 1 µm per pixel [20] (the resolutions studied in the present paper range from 1 µm per pixel to 0.125 µm per pixel, so that the ranges are complementary).
The percolation threshold of the capillary pore space was found to depend on the resolution used, which is not surprising [20]. The reason is that the microstructure was generated using continuum objects, so one had to go to fine enough resolution until the continuum pore space was properly sampled. In particular, some fraction of pore throats that were narrower than the size of the pixel used would be lumped into the solid phase and not treated as a pore at all. This would tend to decrease the connectivity of the pore space, as connections via these kind of pores would not be counted until the resolution was fine enough. The results seem to agree with this explanation. This model is similar in many respects to a simple overlapping solid sphere model, where the pore space is just the space around a collection of randomly centered and sized solid spheres. A recent study of the elastic properties of this model found that the effect of digital resolution was quite pronounced [13].