To correctly model the influence of silica fume on cement paste diffusivity, one must begin with the nanostructure of the C-S-H. Only when an appropriate value(s) is used for the diffusivity of the gel can the cement paste diffusivity be predicted accurately. It has been known for many years that the pozzolanic C-S-H formed from the reaction of silica fume and calcium hydroxide (CH) has a different Ca/Si molar ratio and a different water content than the gel formed from conventional cement hydration [9,10,11]. However, until recently, the question remained open as to whether the underlying nanostructure of pozzolanic C-S-H differed from that of conventional C-S-H. Based on a series of adsorption/desorption measurements on cement pastes with 10 % and without silica fume, Baroghel-Bouny [12] has deduced that the pozzolanic C-S-H has a porosity of only 19 %, in comparison to the value of 28 % for the conventional C-S-H estimated by Powers and Brownyard [13, 14].
Previously, a two level structural model for the C-S-H has been developed based on partially overlapping spheres [15]. At the macro level, the model consists of partially overlapping spheres 40 nm in overall diameter, with an interparticle porosity of 7.6 %. The remainder of the 28 % porosity is present at the micro level, where the model consists of partially overlapping spheres 5 nm in diameter, in accordance with small angle neutron scattering measurements [16]. For the macro level, based on an electrical analogy [15], a formation factor of about 230 is computed. Here, the formation factor is defined as the ratio of the diffusivity of an ion in bulk water to that in the water-filled porous composite material. This value of 230 compares reasonably well with a value of 400 used previously to provide agreement between cement hydration model and experimental diffusivity data [17]. If one maintains the sphere diameter at 40 nm, but decreases the porosity to about 5.1 % (5.1=7.6*19./28.), the model formation factor increases to about 1080, approximately a factor of five times higher than the original gel value. A more direct and quantitative elucidation of the differences in nanostructure between pozzolanic and conventional C-S-H may be possible in the future using either NMR spectroscopy [18,19] or low temperature calorimetry [19].