**2. Materials and Methods**

### *2.1. Experimental Parameters*

## 2.1.1. Colloidal NS

Four commercially available colloidal NS sols were used in this study. The compositions and average physical properties provided by the manufacturer of the NS sols are shown in Table 1. Figure 2 shows their spherical shape.


**Table 1.** Properties of NS sols.

**Figure 2.** Scanning electron microscopy (SEM) image of NS-75.

#### 2.1.2. Dynamic Light Scattering

The PSD of NS was obtained by DLS using a Zetasizer Nano ZS (Malvern Instruments Ltd., Malvern, UK). Measurements were performed using disposable polystyrene cuvettes at room temperature (25.0 ◦C). The dispersant (water) properties were set to default values of viscosity: 0.8872 cP and refractive index: 1.330. The material absorption coefficient was set to 0.001 and the refractive index to 1.50 [53]. These values were kept constant throughout all measurements. Samples were thermally equilibrated for 30 s and the angle of detection was set to 173◦ (Non-Invasive Backscatter (NIBS)). Three measurements were performed for each sample at 11 runs per measurement, 10 seconds per run. The measurements reported in the results section are the mean of the corresponding three measurements. The general purpose algorithm and the method of cumulants were used for analysis. All measurements were taken within one hour of sample preparation. Raw materials were stored at room temperature.

#### 2.1.3. Zeta Potential

ZP was measured using the Zetasizer Nano ZS with the dip cell accessory. The same material parameters were used as in the particle size measurements. Four measurements were performed for each sample at 30 runs per measurement. The voltage selection was set to automatic, with values between 1 V and 20 V. ZP measurements were taken from the same sample as the corresponding DLS measurement.

#### 2.1.4. Sample Preparation

The parameters for testing the four NS sols from Table 1 included varying solid concentration by adding deionized (DI) water, varying pH value by adding 0.1 M aqueous potassium hydroxide (KOH) or 0.1 M hydrochloric acid (HCl), varying amounts of calcium nitrate (Ca(NO3)2·4H2O), and adding

UHPC-like synthetic pore solution (PS) as outlined by Schrofl et al. (9.72 g Ca(NO3)2·4H2O dissolved in 148.5 g of 0.1 mol/L aqueous KOH solution) [54]. An overview of the test matrix is shown in Figure 3 highlighting the four test parameters. The concentration of NS in Figure 3 is by weight percent. It was observed that PS causes irreversible agglomeration. The two ingredients of the PS were isolated to further understand the system.

NS sols at 2% solid content were chosen to investigate the effect of pH value, concentration of Ca(NO3)2, and ionic concentration of the PS. The alterations and concentrations are summarized in Table 2. The synthetic PS contained 0.1 M of KOH and 0.4 M of Ca(NO3)2·4H2O in accordance to Reference [54]. KOH6, Ca8, and PS10 correspond to the molar concentration of the as received synthetic PS. Each successive concentration step down was half of the prior concentration. The pH value of the samples was measured using pH paper in increments of 0.5. DI water, aqueous KOH, aqueous Ca(NO3)2, and pore solution were filtered to 0.1 μm using Anotop (GE Whatman, Marlborough, MA, USA) syringe filters before mixing with NS.

**Figure 3.** Experimental design DI: Deionized. AR: As received.



### 2.1.5. CryoSEM

Cryogenic scanning electron microscopy (cryoSEM) was used to visualize the NS particle dispersion after mixing with cement paste to be able to determine the technique's efficacy for samples of this type. Cement paste was dropped on a 2 mm rivet holder and shock frozen by immersing into a liquid nitrogen slush. Afterwards, samples were fractured, etched, and sputter-coated with Au/Pt in a Leica EM MED020 with QSG10 (Leica Microsystems Inc., Buffalo Grove, IL, USA). Samples were then transferred using the Leica EM VCT100 (Leica, Wetzlar, Germany) and imaged with an FEI Nova NanoSEM 450 (FEI, Hillsboro, OR, USA). The accelerating voltage was 10 kV and the working distance was 5.5 mm.

A cement paste sample was mixed using white Portland cement (w/c ratio 0.3) (Type I, ASTM C150, Lehigh White Cement Co., Waco, TX, USA) and NS-75 (1.5% bwoc) [26]. The cement paste was mixed in a LabRAM Mixer (Resodyn Acoustic Mixers, Butte, MT, USA) at an intensity of 50% for a total mixing time of 300 s. The time between initial hydration and freezing was 1 h.

### **3. Results**

#### *3.1. Diluting NS*

All four NS sols were altered and measured for particle size distribution (PSD) and zeta potential (ZP). The effect of the concentration of NS particles was first studied to isolate the potential influence of the number of particles in a sol. All sols except NS-75 in their as-received state were found to be polydisperse, with a PDI greater than 0.5. Interestingly, the PSD for NS-20b changed as it was diluted. It exhibited a bimodal distribution at higher concentrations and at concentrations below 8 wt % solid, it exhibited a monomodal distribution, as shown in Figure 4, indicating a better dispersion quality. At 2 wt %, the z-average was 24.7 nm (standard deviation (SD) = 0.4), which is close to the provided value of 20 nm by the manufacturer. This suggests that at higher concentrations, the particles tend to weakly agglomerate. By reducing the concentration, there is a greater time and distance between particle collisions, resulting in a better dispersion quality.

**Figure 4.** Selected particle size distribution (PSD) of dilution of NS-20b demonstrating the transition from bimodal to monomodal distributions.

NS-20a was too polydisperse for DLS until it was diluted to a concentration of 0.2 wt %. The z-average was 68.3 nm (SD = 1.5), about three times the supplied value. NS-5 was too polydisperse for quantitative results throughout all concentrations. It is hypothesized that these two sols were polydisperse because of the increased specific surface area of the smaller particle diameters. The increased surface energy was not strongly influenced by the reduction in concentration. The complete results are reported in the Supplementary Materials.

The ZP as a function of solid concentration is shown in Figure 5. In the cases of the two sodium-stabilized NS sols, NS-20b and NS-75, the size and ZP trends were directly correlated. As the size decreased, indicating a better dispersion quality, the ZP increased, indicating a higher stability. On this plot, a lower, more negative value indicates a greater ZP. At a concentration between 2% and 4%, the sols experienced the largest ZP, indicating the highest stability.

**Figure 5.** Zeta potential of NS sols through a range of dilution.

Based on the experimental results, there exists an ideal solid concentration for the greatest stability in DI water. Based on the PSD and ZP measurements, this can be concluded to be about 2 wt % solid NS. This conclusion can only be made for NS-20b and NS-75, as the two other sols were not suitable for DLS due to their high polydispersity and produced inconclusive results.

#### *3.2. pH Adjustment*

The pH of the NS sols was adjusted from about 2 to 12 at 2 wt % solid using HCl and KOH. Throughout pH adjustment, the average size of the sols was remarkably consistent. For the two sodium-stabilized sols, the difference between the maximum and minimum particle size was 7.5 nm (NS-20b) and 7.4 nm (NS-75). For the ammonia-stabilized NS-5, the difference was 5.4 nm. The non-stabilized NS-20a experienced a difference in size of 8.9 nm.

NS-20b had its minimum size at a pH of 8: 23.0 nm (SD = 0.4). The maximum magnitude of ZP was at a pH of 9: −43.3 mV (SD = 3.3). It is not surprising this was observed near its supplied pH value of 9 since the sol is optimized by the manufacturer to experience the greatest stability at this pH. From Figure 6 it is apparent that the average size reaches its greatest value at the ends of the pH scale. The other three sols also generally experienced the largest sizes at the ends of the pH spectrum.

However, the ZP trends were not as apparent. Qualitatively, the ZP was lower at lower pH and higher at higher pH, as is loosely observed in Figure 6. One reason is because the HCl used to lower the pH compressed the electrical double layer. Compared to KOH, the HCl had a negative effect on the ZP. For NS-20b, a reduction in ZP was measured but agglomeration did not occur as would be expected from DLVO theory, exhibited in Figure 6. Similar trends were seen for NS-5 and NS-75. The complete data is presented in the Supplementary Materials.

In a similar experiment, KOH was also added in concentrations relative to the PS to isolate its effect. At the two lowest concentration, KOH1 and KOH2, NS-20b exhibited some agglomeration on the order of magnitude of 3000 nm, while at higher concentrations it experienced a better dispersion quality. NS-20a had a polydisperse size distribution at 2% solid, however, with the introduction of KOH, the dispersion changed to monodisperse distribution. It reached an agglomeration state of about 40 nm, or about twice the manufacturers supplied value, which remained fairly consistent during significant pH adjustment. The addition of KOH altered the electrical double layer and resulted in a metastable state, where slight agglomeration occurred to yield the average size of 40 nm. ZP measurements of NS-20a experienced a similar trend, remaining at about −32 mV. NS-5 exhibited a similar trend of reaching a metastable agglomeration state, although it took an increased amount of KOH to reach a monodisperse distribution. The initial agglomeration state was about twice the supplied manufacturer's value and approached three times that value at the highest concentration of KOH.

Above pH 11, silica was expected to start dissolving. Effects of dissolution were not apparent. Several samples were observed with DLS over a period of 96 h at a pH above 12. The average particle size remained consistent, indicating that the particles were not dissolving. The remarkable stability of the NS sols through a wide range of pH adjustment and over periods of time at high pH suggests that the manufacturer's proprietary stabilization method is fairly robust, which should be useful for achieving uniform dispersion in a concrete mix.

**Figure 6.** pH adjustment of NS-20b with 0.1 M HCl and 0.1 M KOH; (**a**) displays z-average while (**b**) reports zeta potential.

#### *3.3. Calcium Nitrate*

Calcium nitrate was added in several concentrations to isolate its effect in the PS at 2 wt % solid. For NS-20b and NS-5, the sols were able to tolerate the addition of calcium nitrate up to Ca4. Above Ca4, there was significant agglomeration and DLS could not provide meaningful data. The other two sols, NS-75 and NS-20a, had a fairly constant size throughout the range of Ca2<sup>+</sup> concentrations. From the average size alone, it is apparent that Ca(NO3)2 has a greater influence on agglomeration than KOH. After 24 h, all NS-Ca(NO3)2 samples were observed to sediment to the bottom of the cuvette, indicating coagulation had occurred. It was hypothesized that the particle size and stabilizing ion would have an influence on the agglomeration with the addition of Ca2<sup>+</sup>. However, no conclusions could be drawn based on the PSD.

ZP measurements provide greater insight into the stability of these samples. All four sols experienced a reduction in the magnitude of the ZP and switched from an initially negative ZP to a positive ZP. NS-75 at Ca8 had a ZP of 29.2 mV (SD = 12.3), a difference of almost 78 mV without Ca(NO3)2. Therefore, it can be concluded that Ca<sup>2</sup><sup>+</sup> significantly changes the surface charge of the silica particles. This happens more rapidly than the K<sup>+</sup> ions due to the higher valency of Ca2+. These ions offset the surface charge and reduce the electrostatic repulsive forces. The bridging of particles with calcium is hypothesized to be the main cause of agglomeration. Small concentrations (<Ca2) of Ca(NO3)2 resulted in large changes to the ZP. These results were consistent to those of Iler [41] and the hypothesized mechanisms of agglomeration can be reported with confidence. These findings illustrate the importance of DLS and ZP which provide different information on the same sample.

#### *3.4. Synthetic Pore Solution*

Synthetic PS was added in several concentrations to study the effects of the two previous parameters combined. The concentration of the PS was reduced by dilution with DI water. Again, this was added to the 2% solid.

NS-20b and NS-75 both experienced agglomeration beginning at PS4, with significant agglomeration above that concentration. When KOH and Ca(NO3)2 were combined to create the synthetic PS, its propensity to cause agglomeration was more significant than either component alone, as predicted by Plank et al. [55]. The effects of increasing pH and the introduction of two cations, K<sup>+</sup> and Ca2<sup>+</sup>, accelerated the agglomeration. The magnitude of the ZP was greatly reduced and the reduction occurred quicker with PS compared to either component alone. It is interesting to note that in no case was the ZP positive, as was observed with the addition of Ca2<sup>+</sup> only. Before ZP could approach 0, significant agglomeration occurred. These results were expected from the hypotheses of Iler [41] and Tadros [46] and can report these mechanisms with confidence.

PS10 is the pore solution concentration that would be equivalent in pH and ionic concentration to an actual UHPC environment. It is apparent from Figure 7d and Table S13 that NS stability in an actual concrete environment is far from being achieved. Since this pore solution only contains water, KOH, and Ca(NO3)2 it can be inferred that the Ca(NO3)2 is the main contributor to NS agglomeration, due to its bridging effect between silica particles.

**Figure 7.** ZP and size measurements for NS-20b: (**a**) solid concentration varied with DI water; (**b**) potassium hydroxide concentration varied with DI water; (**c**) calcium nitrate concentration varied with DI water; (**d**) pore solution concentration varied with DI water.

#### *3.5. NS Visualization*

An important aspect of understanding dispersion and agglomeration of NS in a true cementitious environment is to directly observe what happens in a concrete sample. This can often be challenging as NS can lead to an increased amount of C-S-H gel and therefore making individual NS particles indistinguishable. Observation of NS was explored by using cryoSEM, as shown in Figure 8. Since the hydration was stopped after 1 hour, some NS was unreacted and visible. No NS particles were observed to be completely separate from another NS particle and the NS appears to be spread out over about half the cement grain. This data was not used quantitatively but proved that this will be a viable method to explore the effects of cementitious pore solution on the dispersion of NS. Future data analysis could include calculating the number of particles per unit area to quantify the dispersion quality. Based on direct observation, the effects of superplasticizers on the dispersion of cement, silica fume, and NS can be investigated. Ongoing research will utilize cryoSEM to visualize the effects of superplasticizers tailored specifically for improved dispersion of NS.

**Figure 8.** CryoSEM image showing unreacted NS on a cement grain.

#### **4. Conclusions**

In this research, the stability of NS in a concrete environment has been studied. The optimal concentration for the best dispersion quality of colloidal NS in a cementitious environment is 2 wt % solid. This concentration allows for sufficient screening of short-range van der Waals attractive forces. The supplied NS sols are fairly pH-robust since they tolerate a wide range of pH changes without agglomerating. This is especially the case with the addition of KOH. The magnitude of the zeta potential remained high and this helped to improve the dispersion quality. On the contrary, Ca2<sup>+</sup> ions are the main contributor to the agglomeration of NS in pore solution. Its higher valency further compresses the electrical double layer, causing the ZP to shift and thus results in a positive surface charge. This neutralization of negative sites results in a mosaic of positive and negative surface charges, leading to bridging between silica particles. By analyzing the size and ZP measurements, NS-20b experienced the best dispersion quality compared to the other three sols. This is due to its intermediate surface area, compared to higher (NS-5) and lower (NS-75) surface areas. The smaller particles were more susceptible to agglomeration due to the higher surface energy. Achieving stability in full strength PS, representing UHPC, is far from being accomplished. NS sols of similar stabilization methods and size will be of greater focus for future studies to obtain stability in PS. Methods to neutralize the effects of Ca2<sup>+</sup> will be specifically targeted to improve the dispersion quality in the presence of PS. Finally, cryoSEM proved to be a valuable tool for cementitious material characterization in the fluid state. The ability to stop hydration and image the microstructure in the fluid state will

be critical to understanding the dynamics of NS interaction, and thus their dispersion quality in a cementitious environment.

**Supplementary Materials:** The following material is available online at http://www.mdpi.com/1996-1944/12/12/ 1976/s1.

**Author Contributions:** Conceptualization, D.H. and K.W.; investigation, D.H. and J.M.; formal analysis, D.H.; writing—original draft preparation, D.H.; writing—review and editing, D.H. and K.W.; supervision, K.W.; funding acquisition, K.W.

**Funding:** This research was supported by the National Science Foundation CMMI Grant 1454574, CAREER: Understanding Behavior and Properties of Nano-Sized Particles in Cement-Based Materials.

**Acknowledgments:** This work was performed in part at the Bioscience Electron Microscopy Facility of the University of Connecticut with Xuanhao Sun with the support of NSF Grant #1126100. The authors would also like to acknowledge the support in terms of material used for experiments from the following companies: Nalco, Lehigh White Cement Company, LafargeHolcim, and Chryso.

**Conflicts of Interest:** The authors declare no conflict of interest.
