The materials used in this study were: Portland cement type III produced by Argos Colombia; an aqueous dispersion of nanosilica particles produced by BASF Chemicals (NS); powdered pyrogenic silica produced by Glassven (PS); and powdered silica fume produced by BASF Chemicals (SF).

Physical and chemical characterizations of the raw materials were performed to obtain their chemical composition, particle size distribution, particle morphology and specific surface area. Chemical composition was obtained by X-Ray fluorescence using an ARL 8680s Total Cement Analyzer equipment, following the ASTM C114-03 standard in B₄Li₂O₇ pills. Specific surface area (SSA) was obtained in a Micrometrics Gemini V 2380 surface area analyzer, using N₂ adsorption and the BET method. Particle size of SF and PS was measured by dynamic light scattering (DLS), using a Mastersizer equipment from Malvern. Particle size of NS was measured by laser diffraction using a Zetasizer equipment from Malvern.

Morphology of the particles was characterized using Transmission Electron Microscopy (TEM) images acquired in a TECNAI 20 Twin microscope from FEI, using an acceleration of 200 kV. Each sample was dispersed in ethanol (C₂H₅OH) and sonicated for 30 min in an ultrasonic bath prior to imaging. After sonication, an aliquot of the sample was dropped in a Fomvar 300 cupper mesh and coated with carbon. Electrophoretic mobility values were determined using ζ-potential measurements in a Zetasizer nanoZS equipment from Malvern. The pH of the dispersions was varied from neutral to basic using ammonium hydroxide (NH₄OH) in deionized water. Each sample was placed in a disposable capillary cell stabilized at 25 °C, ζ-potential was determined 3 times for each pH value.

Cement pastes were produced using 0%, 5% and 10% mass substitutions of cement by silica. Setting time and normal consistency were performed for each paste following ASTM C191 and ASTM C187 standards respectively. Pastes for rheology were mixed manually using water-to-cementing material (w/cm) ratios of 0.45, 0.50, 0.55 and 0.60. Nomenclature and detailed composition of the studied samples are presented in Table 1. Flow curves were obtained 5 minutes after the first contact between cement and water using a Bohlin Instruments C-VOR 200 rheometer. A vane-in-cup geometry with a 25 mm diameter vane tool in a smooth cylindrical cup with a 1 mm gap was used. Each paste was pre-sheared at a shear rate (γ˙) of 10 s⁻¹ for 60 seconds, subsequently, shear rate was increased to 250 s⁻¹ at a 20 s⁻¹/min rate, maintained for 3 minutes at this value, and decreased to 0 s⁻¹ also at a 20 s⁻¹/min rate. Both during the shear rate increase and decrease 50 evenly distributed shear stress (τ) points were acquired. The test had a duration of 28 min, which added to the initial 5 minutes of sample preparation, gives a total duration of 33 min after the first contact of cement with water.

Physical, chemical and mineralogical characterization results obtained for the three studied silica are presented in Table 2 and Figure 1. It was confirmed that NS, SF and PS are mainly composed of amorphous silica, finding only minor crystalline minerals in SF probably from its production process. Regarding their surface area and average particle size, high SSA values were found for all silica samples, while NS presented a d₅₀ in the scale on nanometers and SF and PS presented a d₅₀ in the scale of micrometers. It should be noticed that the obtained SSA values for SF and PS are not compatible with the particle sizes found, especially for PS. Density (ρ) of the cement used was 3.17 g/cm³. Density of the silicas are presented in Table 2. Average particle size from BET result was computed as d_BET = 6000/(SSA·ρ). The agglomeration factor (F_AG) of the silicas was defined as F_AG = d₅₀/d_BET (16).

TEM images were used to understand the incompatibility between average particle size and surface specific area of SF and PS. One typical image of each silica studied is presented in Figure 2. NS was found to be composed of spheres with individual particle sizes compatible with the d₅₀ value measured by laser diffraction. SF and PS were found to be composed of agglomerated particles with individual sizes much smaller than the d₅₀ measured by dynamic light scattering. This indicates that particle size experiments were able to capture the average size of the agglomerates, while SSA experiments were able to capture the surface area of the individual components. The F_AG results confirmed that while NS particles and be considered completely dispersed, the SF and PS particles can be considered highly agglomerated (17).

A morphological difference was also identified among NS, SF and PS. While NS and SF were found to be made of spherical particles, PS was found to be made of connected chain-like elements, which correspond to silica droplets of pyrogenic origin (4).

The tendency to agglomerate of the three silicas at different pH values was studied through ζ-potential measurements. The obtained results are presented in Figure 3. Elevated pH values were used to emulate the typical alkaline environment generated by cement hydration, which depending on the hydration degree, alkalis, calcium silicates and CaO concentrations, can reach values around 12.5 in the pore solution (18). The ζ-potential results are separated in two areas, one under -25 mV, where the colloidal suspension of silica particles in water can be considered stable, and above -25 mV where it can be considered unstable (19). These areas are separated by a horizontal solid line in Figure 3.

NS, SF and PS present a clear tendency to decrease (absolute value) their ζ-potential with the increase of pH but without reaching the isoelectric point. NS had the lowest ζ-potential, always in the stable zone for the entire pH range studied. On the other hand, ζ-potential values for PS and SF approached or surpassed -25 mV at pH values of 12 or higher. This indicates that in the typical environment generated by cement hydration, PS and SF can be considered metastable or unstable with a tendency to agglomerate. SF was found to be prone to agglomeration at a pH of 7, this helps to explain why particle size microscopy images showed particle agglomerates rather than individual ones, since both these tests were carried out in deionized water, which has a typical pH of approximately 7.

Water demand and setting time of Portland cement pastes were considered two important properties that can be modified by NS, SF and PS additions and might lead to misinterpretations of the rheology results if not taken into account. Water demand was studied through normal consistency tests, while setting time was studied using Vicat´s needle. Results are presented in Table 3. It was found that water demand increased proportionally to the amount of silica blended in the paste, being PS the silica that demanded more water for both percentages studied, followed by NS and SF. It should be noticed that the increase in water demand appears to be also proportional to the SSA of each silica, as previously demonstrated by the authors (8). The porous nature of PS should also be taken into account, since it is capable of absorbing water, further increasing the water demand of the pastes blended with it.

Regarding setting time, for all silica studied, the initial and final setting times increased when compared to the reference paste, this as a direct consequence of the increase of w/cm (being cm the cementing material, in this study the sum of cement and silica), necessary to attain normal consistency. Additionally, when comparing the 5% and 10% solid substitutions of NS and SF, it can be seen that higher amounts of silica led to a decrease in setting time: this is associated with nucleation effects activity (6). For PS this was not the case, indicating that any nucleation effects from PS could not overcome the delaying effect of the additional water in the mix. Overall, initial setting times for all pastes are much higher than the total testing time of each rheological experiment (33 minutes).

Rheological experiments were carried out in cement pastes following the proportions presented in Table 1. A Bingham model τ = τ₀ + μ₀, where τ₀ is yield stress and μ₀ is plastic viscosity, was fitted to the linear descending portion of each flow curve. Fitting was carried out only in the linear part of the descending portion, in this case between 100 and 250 s⁻¹ for all samples. A typical flow curve with its corresponding fitted Bingham model, and a typical instantaneous viscosity versus shear rate curve, are presented in Figure 4. Pastes containing 5%PS at w/cm 0.45, and 10%PS at w/cm 0.45, 0.50 and 0.55, could not be placed in the vane-in-cup geometry due to their low workability. All testes pastes presented a viscoelastic shear-thinning behavior, as expected.

All the plastic viscosity (μ₀) and yield stress (τ₀) values obtained are presented in Figures 5 and 6 respectively. For the reference paste, i.e. without any silica blend, it was found that μ₀ and τ₀ decreased with the increase of water in the mixture, as expected. More water in an unit volume of paste increases the mean distance between particles, decreasing their number of contact points and attractive forces (decreasing τ₀), facilitating sliding motions between them (decreasing μ₀) (20). This result was used as reference to understand the effects of NS, SF and PS.

NS and SF were found to decrease μ₀ with respect to the reference paste in the entire w/cm range studied, being NS always more effective in terms of reducing plastic viscosity. This was found to be true for both 5% and 10% substitutions. This behavior, known as “ball-bearing” effect, is typical of spherical air bubbles and spherical mineral additions, which allow particles to move past each other more easily (21). The increase in packing density, expected to be induced by the solid substitutions of cement by NS and SF, also plays an important role on the decreasing in μ₀ (20). On the other hand, PS was found to increase μ₀ due to its open porous structure, which potentially absorbs a portion of the mixing water (22) and decreases the effective amount of water in the liquid medium. Only the 5%PS paste at w/cm of 0.60 presented a μ₀ value similar to that of the reference sample with w/cm=0.60, indicating that for this water content the pores in PS might have saturated and more water was available in the matrix. On the other hand, 10% PS mix only was rheologically measured for w/cm=0.6 and the corresponding plastic viscosity was much higher than reference: 0.11 Pa.s versus 0.03 Pa.s. In this case, the effective water was very low and the plastic viscosity was indeed much higher than that obtained for the reference mix with w/cm=0.45.

Regarding yield stress, NS and SF were found to have from little to no impact over τ₀, following the overall trend imposed by the variation of w/cm ratio. PS was found to increase slightly τ₀ probably due to the amount of water potentially absorbed by it, which increases the contact points between particles, this leads to a less pronounced effect than the one over μ₀, which is more dependent on the amount of water in the media (20).