*2.2. Experimental Methods*

For morphological analysis of raw materials, microscopic observation was performed using FE-SEM equipment (JSM-7800F Prime, JEOL Ltd., Tokyo, Japan). Powder samples were fixed on a carbon tape bonded holder and the upper surface of the holder was coated with carbon to prevent the charging effect. The holder was then inserted into the microscope and images of the particles were captured at a magnification of up to ×100,000. Additionally, the SEM images of SF were used to obtain its size distribution based on individual spherical particles. Since the results can significantly vary depending on the number of particles or images, the image processing technique was performed according to the method specified in ISO 13322-1 [54].

As another method for obtaining the particle size distribution of SF, an analysis using DLS was performed. To improve dispersion efficiency, the sample was sonicated for 5 min in an ultrasonic bath (53 kHz and 200 W) and then the size distribution was measured at a set temperature (25 ◦C) by the device (Zetasizer Nano ZSP, Malvern Panalytical, UK). Moreover, SSA of SF samples was determined by the BET method, for which the nitrogen adsorption desorption isotherms were recorded using a surface area analyzer (Autosorb IQ-MP/XR, Quantachrome Instrument, Boynton Beach, FL, USA).

To measure the material properties of UHPFRC, a series of experiments was conducted on workability, compressive strength, and flexural tensile strength. Workability of the fresh concrete was evaluated by the use of a flow table test without shock [55]. Compressive strength tests were performed according to ASTM C109 [56]; on the 28th day, cube specimens with one side of 50 mm were loaded by a universal testing machine and the average of the three specimens was determined as the strength value. On the same day, flexural tensile strength was measured by the three-point bending test method specified in ISO 679 [57]. For this, line loads were applied to three prismatic specimens (40 mm × 40 mm × 160 mm) by the machine.

XRD and TG analysis were performed to investigate the effect of SF on the hydration reaction of UHPFRC. To improve the precision of the analysis results, paste samples (excluding the materials that do not contribute to the chemical reaction, such as fine aggregate and steel fibers) were additionally prepared on the same day as the UHPFRC samples. On the 28th day, the hydrated paste was crushed and ground into a powder. Thereafter, the hydration reaction of the samples was stopped using isopropanol and diethyl-ether, according to the solvent exchange method [58,59]. After the pretreatment, the powders were placed on sample holders and then mounted on an X ray diffractometer (SmartLab SE, Rigaku, Tokyo, Japan). XRD patterns were collected by Cu·Kα<sup>1</sup> radiation (λ = 1.5406 Å) under the established conditions, such as voltage: 40 kV; current: 40 mA; step size: 0.02◦, and scanning speed: 1◦/min. The patterns collected between 5◦ and 70◦ were analyzed by SmartLab Studio II software (Rigaku, Tokyo, Japan), equipped with an NIST inorganic crystal structure database and crystallography open database.

TG analysis was performed to quantitatively estimate the consumption of portlandite due to the pozzolanic reaction. Around 50 mg of powder samples was weighed in an alumina made sample holder which was placed on an analytical instrument (SDT 650, TA Instruments, New Castle, DE, USA). Under an environment where nitrogen gas was introduced at the rate of 100 mL/min, the samples were heated from 30 to 1000 ◦C at the rate of 10 ◦C/min. The weight loss with increasing temperature was recorded and the derivative thermogravimetric (DTG) was presented in the graph along with the weight loss to clearly identify the sudden weight change in a specific temperature range.

#### **3. Results and Discussion**

#### *3.1. Observation of Agglomerated SF Particles*

Figure 3 shows SEM images of cement and quartz powders at various magnifications. As depicted below, these μm sized particles do not aggregate together but exist separately. The cement is composed of particles with a wide range of sizes from ~1 to ~50 μm (Figure 3a–c). In the case of quartz powder, the large particles were around 20 to 30 μm and the small ones were around several μm (Figure 3d–f). Compared to the cement, quartz powder showed a narrow particle size range because the crushed rock was sieved within a specific size range during its manufacture. The observation of particle size by SEM tended to be consistent with the result of LD-based analysis, as shown in Figure 1. This implies that particle size analysis for completely separated powders can ensure good reliability.

**Figure 3.** SEM images of cement (**a**–**c**) and quartz powder (**d**–**f**) with various resolutions.

Prior to SEM observation of the SF particles, visual inspection was performed as shown in Figure 4. Although the sample SF1\_U is an undensified product, many lumps large enough to match the size of the coarse aggregates were observed. This was probably due to the agglomeration during a long period of shipment, transportation, and storage. Moreover, during this period, a vertical load was applied by the other products stacked on top of it. On the other hand, such large lumps were not found in the sample SF2\_U, which should have been less affected by the factors related to aggregation after manufacturing. As expected, the sample SF2\_D consisted of globular lumps due to the densification process during manufacturing. The size of the lumps was up to several mm, and, unlike irregular ones in the sample SF1\_U, they were almost perfectly spherical. This means that, once lumps are formed during the densification process, they are difficult to agglomerate thereafter, and thus this type of product has advantages for quality control.

**Figure 4.** Visual observation of SF samples as a state used for manufacture of UHPFRC: (**a**) SF1\_U, (**b**) SF2\_U, and (**c**) SF2\_D.

The true nature of these lumps can be viewed in detail by the use of SEM images (Figure 5). Figure 5a shows dust-like lumps that were observed. As mentioned in the introduction, the spherical nanoparticles never existed independently. The size of the lumps was various, and large ones of several tens of μm were also seen. The magnified images are presented in Figure 5b,c, in which myriad nanoparticles were agglomerated together to form the lumps. It has been reported that such lumps can consist of up to several tens of millions of nanoparticles [41]. As shown in Figure 5d–f, where individual nanoparticles are observed, it was confirmed that their size is tens to hundreds of nm based

on separated spheres. In particular, some particles are not separate spheres but rather they appear to melt and stick together (yellow arrows in Figure 5d,e). Another interesting discovery was that of broken particles (white arrows in Figure 5f), indicating that the SF particles also have the shape of a hollow sphere, like other spherical SCMs such as fly ash.

**Figure 5.** SEM images of SF1\_U samples with various resolutions: agglomerated lumps (**a**–**c**) and spherical nanoparticles (**d**–**f**) inside lumps.

It is generally known that SF particles are mostly (>95%) composed of nanoparticles [27]. However, it should be noted that a significant part of the SF particles has a form of linked spheres, regardless of the densification process. Spherical nanoparticles that are connected as a result of sintering or fusion have been previously observed [39]. As also found in this study (Figure 5d,e), consequently, general forms of SF particles are aggregates of several spheres rather than independent spheres [37]. This type of connection is completely different from the agglomeration due to the densification process because it is a very strong and irreversible bond. When high purity quartz is converted to silicon at a high temperature (~2000 ◦C), SiO2 vapor is generated; the vapor oxidizes and condenses at a low temperature while forming spherical particles with a size of 100 to 200 nm, but when they come into contact with each other in the molten state, primary aggregation occurs [60]. Due to this, dozens to hundreds of spheres are connected, and, typically, the size of these clusters has been reported to be 500 to 800 nm [41]. Such clusters formed by primary aggregation cannot be separated unless they are broken. For this reason, there has been confusion about the effective size of SF in all cases: immediately after manufacture, at the point of use, and after being incorporated into concrete [35].

Figure 6 shows the particle size distribution of SF samples by DLS and image processing techniques. By comparing the results of SF2\_U (DLS) and SF2\_D (DLS), the effectiveness of the technique (DLS with ultrasonic treatment)—which excludes the effect of the densification process—can be verified. As evident from visual observation (Figure 4), there was a significant difference in the degree of aggregation between the two samples, but, nevertheless, there was almost no difference in particle size distribution, as shown in Figure 6. In both samples, the main peak was formed between 70 nm and 1 μm. However, in the size of the peak formed between 1 and 10 μm, the densified sample was larger than the undensified sample due to the contribution of large, aggregated particles. For this reason, SF2\_D showed a higher value than SF2\_U in the average particle size, as shown in Table 3. Meanwhile, by comparing the two yellow and blue curves, differences can be observed between the two products classified by the manufacturer. Regarding the proportion of the particles with sizes of >400 nm, the SF1\_U sample had a higher value than SF2\_U and SF2\_D, which also resulted in a difference in

average particle size (Table 3). This is presumably due to an increase in the agglomeration of small particles during the manufacturing process or thereafter. Moreover, the agglomeration phenomenon that had become stronger over a long period of time could have lowered the dispersion efficiency by the ultrasonic treatment.

**Figure 6.** Particle size distribution of SF samples by DLS and SEM-based image processing techniques.

**Table 3.** Average particle size and specific surface area of SF samples.


Outcomes of the difference between the image processing and DLS techniques can be confirmed by comparing the two results of SF1\_U (SEM) and SF1\_U (DLS) (see gray dotted line and blue solid line in Figure 6). As expected, compared to DLS, the image processing technique formed the particle size distribution in a small size range, since it did not reflect the agglomeration phenomenon at all. Both methods have their own limitations. In particular, DLS can overestimate the particle size of SF because it measures hydrodynamic size (not a static or dry particle size) [61]. Nevertheless, this technique can be considered more suitable than the SEM-based image processing technique to measure the size distribution of SF particles. This is because effective particle sizes have more significant meanings than those that are independent due to their agglomerating nature [22]. Furthermore, the use of a high-resolution microscope is time consuming and expensive, and only a limited number of particles included in the 2D image are considered in the results. Indeed, results from previous studies have suggested that the actual size observed by SEM is unable to reflect the realistic particle size distribution because SF is always present in a clustered or agglomerated state [25,41].

The SSAs of the samples are shown in Table 3. The measured value was 22.9 to 25.3 m2/g and the deviation between the minimum and maximum was around 10%. This is consistent with the results of the previous study in that SSA of SF was in the order of 20 m2/g, despite diversity in type or manufacturer [41]. Moreover, there was no significant difference among the three samples of SSA, which is one of the most important parameters for the chemical effect of SCM on cement-based materials [32,45]. Unlike other natural or industrial byproduct-based SCMs, one of the advantages of SF is very low quality variation in terms of chemical composition, particle size, and SSA, which can vary by factory and manufacturer [62]. In this aspect, the measurement results in this study showed this benefit well.

#### *3.2. Material Properties of UHPFRC with Various Types of Silica Fume*

The measured flow diameters were 270, 260, and 255 mm for the fresh UHPFRC samples containing SF1\_U, SF2\_U, and SF2\_D, respectively. According to the international standard on UHPFRC, their workability class is Cv, meaning viscous UHPFRC with self-compacting ability [63]. When comparing the two UHPFRC samples containing SF2\_U and SF2\_D, there was almost no difference, although the flow diameter was slightly higher than when undensified SF was used. If other conditions are the same, as the SF is more evenly dispersed in UHPFRC, the yield stress is further reduced due to the ball bearing effect [23], thereby increasing the self-compacting ability [52]. This implies the possibility that there was no significant difference in the degree of dispersion of SF particles inside the two fresh concrete samples. Meanwhile, between the two samples containing undensified SFs, when SF2\_U was used, the flow diameter was 10 mm lower than the result of SF1\_U. Regarding the decrease in flowability, both the dispersion efficiency of SF and the difference in SSA can be complicated because, as the SSA increases, water demand and friction between the particles increase [48]. Nonetheless, overall, there was no significant difference in workability among the samples, suggesting that there was no significant difference in the SSA and the dispersion efficiency of SF in the concrete samples.

The experimental results of compressive strength and flexural tensile strength are presented in Figure 7a. Above all, every sample exhibited a very high compressive strength of around 200 MPa. Compared to the sample [UHPFRC], SF2\_U showed the highest strength. [UHPFRC] SF1\_U was ~10 MPa lower but [UHPFRC] SF2\_D was ~5 MPa lower. This is an important result for the feasibility of using densified SF in UHPFRC because it confirms the negative effects that can be caused by the densification process of SF, which are negligibly small in the mechanical properties of UHPFRC. In addition, the concrete's excellent strength can be ensured by using this commercially optimized product. Based on the results of [UHPFRC] SF2\_U, the difference in concrete strength with the same type of product from another company ([UHPFRC] SF1\_U) was more pronounced than with the other type of product from the same company ([UHPFRC] SF2\_D). Various material properties affect SF concrete performance, including chemical composition, SSA, particle size or size distribution, the amount of silanol groups (Si-OH) on the particle surface, and post production age [64]. This suggests the diversity of properties related to SF1\_U and SF2\_U which affect the performance of UHPFRC. On the other hand, in terms of the difference in such properties of SF2\_U and SF2\_D, factors other than the densification process can be ignored. For this reason, the feasibility of using densified SF in UHPFRC can be proposed. Results from the flexural tensile strength test further support this suggestion. When compared to 45.1 MPa of the sample [UHPFRC] SF2\_U, the strength of sample [UHPFRC] SF2\_D was only 1.1 MPa (or 2.4%) lower. However, the strength of the sample [UHPFRC] SF1\_U was 8.1 MPa (18%) lower than this. The flexural strength results were consistent with those of compressive strength, as confirmed by the clear linear correlation in Figure 7b. In summary, the parameters such as the densification process and the manufacturer or storage period affected the compressive strength and the flexural tensile strength with the same tendency. More importantly, the former parameter had little effect on both compressive and tensile properties of UHPFRC.

**Figure 7.** Results of strength tests: (**a**) compressive and flexural tensile strengths; (**b**) the relationship between the two strengths.

Direct measurement of the dispersion degree of SF in cement-based materials is extremely challenging [34]. Thus, indirectly, mechanical properties can be considered as an indicator for evaluating this. To date, the benefits expected from the use of undensified SF (such as improvement in dispersion efficiency in concrete and thereby improvement in mechanical properties) have been reported. This can be explained by enhancements in the role of SF such as the micropore filling effect, pozzolanic reaction, and the provision of nucleation sites. However, comparisons between different types of SF products have been mainly investigated in cement composites other than UHPFRC, such as the concrete and mortar types that contain coarse aggregates or have relatively high *w*/*c* [48,65]. Moreover, results from previous studies confirmed that a large amount of SF lumps (>10 μm) existed in such cement composites [34,39]. In practical situations, the only way to disperse SF products inside concrete or mortar is with mechanical crushing by a mixer, which is also one of the most effective methods [66]. When mixing concrete, crushing and shearing actions are transferred to the lumps of SF particles by a mixer and the dispersion efficiency is highly dependent on the mixture composition and mixing method. However, it has been consistently reported that densified SF cannot be satisfactorily dispersed in the cement composites, which is completely different from UHPFRC in terms of the mixing method and procedure [66–69].

When UHPFRC is manufactured, SF particles should be effectively dispersed by a suitable mixing method and a superplasticizer; otherwise, fundamental principles such as the optimization of mixture composition or packing density are meaningless and, in the end, the excellent performance of the concrete cannot be guaranteed [23]. The manufacturing process of this concrete includes one unique state in which all the ingredients in the mixer are agglomerated like flour dough (Figure 2b). At this point, when the mixer's power consumption reaches its maximum, a strong shearing action is applied to the agglomerate so that the components, including the SF particles, can be effectively dispersed. Moreover, in the presence of steel fibers, additional dispersion is possible in the subsequent high-speed mixing process. Although the perfect dispersion of individual SF particles might be impossible [39], the lumps formed by the densification process are weakly connected and thus can be easily crushed by agitation. Therefore, UHPFRC's unique formulation and manufacturing process can be effective in removing reversible aggregation by the densification process. As a result, there may not be a significant difference between densified and undensified products in dispersion efficiency in this concrete.

Along with compressive strength and durability, excellent crack resistance performance is one of the main features of UHPFRC, which can contribute to the construction of innovative and sustainable concrete structures [70]. This is the reason that the strength under flexural or tensile loading is an important property of this concrete [71]. Regarding this, one notable result was that, unlike the compressive strength, which increased due to the promoted pozzolanic reaction, there was no noticeable change in flexural tensile strength despite the increase in temperature (60 to 90 ◦C) and duration (up to 4 days) during heat treatment [72]. This is because the properties of steel fibers (aspect ratio, shape, surface treatment, etc.) and the distribution state of the fibers inside the concrete (volume ratio, direction, degree of dispersion, etc.) have a decisive effect on flexural or tensile properties [73–76], but the difference between these factors could be neglected in this study. Apart from such promotion of the pozzolanic reaction and the change in fiber parameters, the effect of the mixture composition on tensile strength has also been reported. In the study by Chan and Chu, the interfacial bond strength of the fibers increased proportionally with the contents of undensified SF (between 0 and 30 wt % by cement), which was attributed to the improvement in friction and resistance force of the fibers as the interface became denser [28]. In addition to these previous results, those from this study revealed that the type of SF also affects the tensile properties of UHPFRC, and more importantly, there is no significant reduction in the strength despite the use of densified SF.

#### *3.3. Hydration Reaction of Heat-Treated UHPFRC with Various Types of Silica Fume*

Figure 8a,b show the results of XRD and TG analysis of heat-treated paste samples at 28 days. Typical phases identified in the XRD pattern of this concrete are quartz, calcite, cement clinkers,

ettringite, and portlandite, but the last two phases were not detected in this study. The absence of ettringite at 28 days confirms that it was decomposed at an elevated temperature (>70 ◦C), and delayed ettringite formation did not occur during the subsequent 25 days of curing [77]. In particular, the cause of the latter is the extremely compact microstructure and low *w*/*c* of UHPFRC; the effect of heat treatment makes this condition even more dramatic because elevated temperature curing (or heat treatment) significantly promotes both cement hydration and pozzolanic reaction. These further fill capillary pores while consuming a limited amount of water [78]. However, despite the heat treatment, unhydrated cement clinkers still existed (as shown in Figure 8a). This is the result of an insufficient amount of water to fully hydrate the cement and the lack of space for further forming hydration products [79,80]. Additionally, these conditions do not provide water and space for the reformation of ettringite [81], and, because of its excellent watertightness, external water cannot penetrate into the concrete, even under water curing conditions [80]. The weight loss and DTG curves shown in Figure 8b clearly confirm that ettringite was not formed after decomposition. This is because sudden weight loss at around 105 ◦C is associated with dehydration of ettringite, and a sharp peak on the DTG curve due to this loss could not be observed [58]. Therefore, the gentle peak formed between 40 and 300 ◦C can only be attributed to the dehydration of C-S-H.

**Figure 8.** Analysis of crystal phases and hydration products of heat-treated paste samples: (**a**) XRD patterns at 28 days (Q: quartz, C: calcite, C3S: tricalcium silicate); (**b**) TG and DTG curves at 28 days.

Another notable result from XRD and TG analysis is the absence of portlandite. The presence of portlandite was clearly confirmed in the previous study [72], in which all conditions other than the cement (i.e., raw materials, mixing method, and curing program) were the same as with this study. Additionally, in other studies, although the pozzolanic reaction was significantly promoted by the heat treatment at 90 ◦C, portlandite was not completely consumed [31,82,83]. In the study by Korpa et al., which includes in-depth discussions of the phase development of UHPFRC conditions based on XRD and TG analyses, the presence of portlandite and hence the ongoing progress of the pozzolanic reaction after 28 days was reported [84]. In UHPFRC, this crystal has been reported to be completely consumed by the pozzolanic reaction that was promoted by the heat treatment at >200 ◦C [85]. Unlike previous studies mentioned, the cement used in this study contained a significant amount of limestone powder, which was likely to contribute to the complete consumption of portlandite. The cause can be explained as follows: substitution of limestone powder instead of cement at a given amount of water increases the effective *w*/*c* (0.3 in this study). Additionally, since portlandite is produced by the primary cement hydration, a decrease in the content of primary hydration products due to the cement dilution effect can deplete portlandite even when the amount of accessible water is increased. The previous result also supports this, in that the portlandite content in UHPFRC decreased proportionally with the increase in the content of limestone powder [17].

However, even when a significant portion (up to 74%) of cement was replaced with limestone powder, portlandite could not be completely consumed without heat treatment [15,17]. This is because the reactivity of SF at room temperature is significantly low, and thus, even if a small amount of portlandite is formed, the role of SF related to consumption of this crystal is bound to be limited. The activation energy of SF required to participate in the pozzolanic reaction has been reported to be approximately 80 kJ/mol [86]. In this regard, heat treatment can greatly accelerate the reactivity of SF. It has also been reported that the solubility of amorphous silica increases proportionally with the temperature of water or solution [87]. Pfeifer et al. noted that the reaction degree of SF in UHPFRC was around 5% under 28 days of the ambient curing condition (20 ◦C), but the degree rapidly increased to around 45% (around nine times) when the standard heat treatment was applied to the concrete [31]. They also reported that an increase in effective *w*/*c* significantly improves the proportion of SF participating in the chemical reaction (around three times when *w*/*c* increases from 0.2 to 0.4). For this reason, even under the conditions in which a small amount of portlandite is formed due to the use of limestone powder, complete consumption of portlandite in UHPFRC would not be possible without heat treatment.

In addition to reducing the content of the cement, the use of limestone powder in UHPFRC has several advantages; autogenous shrinkage is alleviated by the increase in effective *w*/*c* [17,18] and the initial hydration is accelerated due to the provision of nucleation site for the formation of C-S-H [16]. Furthermore, the complete consumption of portlandite found in this study can provide additional advantages regarding the use of Portland limestone cement in UHPFRC. The pozzolanic reaction by SF or the consumption of portlandite has been known to persist for several years [88], and even after heat treatment, the compressive strength of UHPFRC increases continuously for 6 to 8 years due to this chemical reaction [89]. Indeed, an inversely linear relationship between the compressive strength and portlandite content has been reported in UHPFRC. This can be explained by the formation of additional C-S-H and refinement of the pore structure and removal of portlandite which has a morphologically undesirable effect on the strength of concrete and hardness of the interfacial transition zone [72]. Moreover, in terms of durability, this crystal can cause an expansion reaction with ions penetrated from the outside, thereby leading to cracking. Consequently, since the pozzolanic reaction has a decisive influence on the long-term strength and durability of UHPFRC [24], the complete consumption of portlandite shown herein suggests that the maximum mechanical properties and durability of this concrete can be guaranteed within 28 days. The stabilization of long-term properties is certainly an advantage when considering only practical usability because it can provide a reliable design strength for informing the work of practitioners and structural engineers [23].

In addition to the reactivity of SF, the chemical composition, particle size distribution, SSA, and degree of dispersion are all important factors in the chemical reaction of concrete [37,90]. In this regard, the results in Figure 8, in which no portlandite was detected in any of the samples, confirm that there might be no difference between densified and undensified SF products in the hydration and pozzolanic reaction of UHPFRC. This also implies that, between them, there was no difference in the degree of dispersion. This can be more evident if the same result is confirmed under the situation in which portlandite is not fully consumed.

#### *3.4. Feasibility of Using Densified Silica Fume in Field Casting UHPFRC*

An important application type of UHPFRC is field casting concrete (e.g., overlaying of concrete decks or slabs, jacketing of beam or columns, and filing material for precast concrete segments) [4,91–94]. In this case, heat treatment is not practically applied. The use of densified SF can be more essential for this type. This is because, in general, powder materials are continuously input and mixed into a mixer at an outdoor construction site; in this environment, the use of undensified SF can further deteriorate the limitations such as workplace dust generation, transportation, and the storage of raw materials. Thus, to fully examine the practical feasibility of densified SF, it is also necessary to conduct an investigation of UHPFRC cured without heat treatment. Moreover, when heat treatment is not

applied, the mechanical properties of UHPFRC can change more sensitively depending on the type of SF, because, as mentioned earlier, the physical role (i.e., micro filling effect) rather than the chemical role (i.e., the pozzolanic reaction) can significantly contribute to the properties. In this case, a decrease in concrete performance due to a decrease in the dispersion efficiency of SF can be more pronounced.

Figure 9 shows the results of TG analysis of the paste samples cured under ambient conditions for 28 days. To exclude factors other than the densification process, paste samples containing SF products from the same company were compared. Analyses of the results verify that the type of SF classified by this process does not affect the hydration reaction of UHPFRC, regardless of heat treatment. However, unlike results from the heat-treated samples (Figure 7b), a sharp peak was formed around 105 ◦C. This clearly confirms the presence of ettringite. The integration result of this sharp peak area was 5.76% and 5.75% in the SF2\_U and SF2\_D samples, respectively, confirming that there is no difference in the main hydration products (C-S-H and ettringite). This also indicates the possibility that there was no difference in the chemical reaction related to the formation of portlandite. Moreover, the result that no protrusion or peak was formed between 150 and 200 ◦C on the DTG curve shows that it is extremely difficult to form a new phase (hemi or mono carboaluminate) by direct reaction of calcite in the UHPFRC condition, regardless of the heat treatment. This is consistent with the results of previous studies [15–17].

**Figure 9.** TG and DTG curves of ambient-cured paste samples at 28 days.

When observing the peak at 400 ◦C on the DTG curve, there was a difference in the sizes of portlandite peaks between the two samples. The quantitative analysis results are presented in Table 4. The contents in the table were determined according to the tangential method along with the normalization method based on the weight at 550 ◦C [58]. The portlandite content of the sample with densified SF was 1.1 wt %, which was 0.21 wt % higher than the other sample of 0.89 wt %. However, based on the weight ratio of calcium oxide, most of the differences in the portlandite content are attributed to the formation of calcite (difference in CaO, Portlandite = 0.16; difference in CaO, calcite = 0.12). This also confirms that there was almost no difference between the two samples in the degree of pozzolanic reaction. The contribution of carbonation to the differences in portlandite content can also be visually shown in the blue and red shaded areas on the DTG curve. Since there was no difference between the two SF samples in the chemical reaction, the difference in the physical role (i.e., micro filling effect) caused by the dispersion efficiency can be considered as another potential factor affecting the mechanical properties. However, there was also no difference between the two samples in the compressive strength of UHPFRC at 7 and 28 days (Table 4). This demonstrates that the densified

SF does not have any negative effect on either the physical or chemical roles of SF in the concrete, compared to the undensified SF.


**Table 4.** Portlandite and calcite contents and compressive strength of ambient-cured UHPFRC.

SF products are only manufactured in limited regions of the world. Therefore, the products are inevitably transported and stored for a long time in a stacked state. Compared to undensified products, densified products are much freer from further agglomeration and quality changes after being manufactured [20]. The agglomeration by the densification process is reversible so that agglomerated SF particles can be effectively dispersed by an optimized mixing method, which is practically the only way this can be achieved. Along with this, the dispersion efficiency also depends on the chemical admixture used (performance or content), mixture composition, and water-to-powder ratio [37]. In particular, when a mixture of high volume powders is lumped and stiff like flour dough and a high mixing energy is applied to the mixture by a suitable mixer (e.g., twin shaft, planetary, or intensive mixer, etc.), the SF agglomerate can be effectively broken; eventually, their dispersion efficiency can be greatly increased [29,60]. In addition, the inclusion of PCE-based superplasticizer under these conditions and the application of high-speed mixing makes the homogeneous dispersion of powders including SF more effective. The composition of UHPFRC is characterized by a very low water-to-powder ratio and a very high content of PCE-based superplasticizer; this causes the dry materials to clump together in the mixing process. This is a clearly unique feature of UHPFRC that differs from other types of cement-based materials, and thus it can provide conditions for efficiently dispersing densified SF (equivalent to undensified products in terms of dispersion efficiency). Indeed, as all the results of this study consistently indicated, there was no notable difference between densified and undensified products in terms of the material properties of UHPFRC. Therefore, densified SF can be used to manufacture UHPFRC. In other words, although using this commercially optimal type of product, a compressive strength of >200 MPa can be achieved within 28 days, without disadvantages in workability and tensile properties.

### **4. Conclusions**

This study was undertaken with the hypothesis that the composition and mixing method of UHPFRC are both unique and thus this condition is effective for dispersing the reversibly aggregated particles in densified SF. The UHPFRC is characterized by very high content in powders, PCE-based superplasticizer, and short steel fibers. Furthermore, a high-speed mixing process is included under all these conditions. In particular, since SF affects all important chemical and physical properties of the concrete (such as microstructure, hydration reaction, self-compacting ability, tensile and compressive strengths, and durability), the difference in dispersion degree should greatly affect the performance of the concrete. Experimental results on the comparison of undensified and densified SF demonstrated the validity of our hypothesis:

• Visual inspection and SEM image analysis confirmed that SF is composed of spherical nanoparticles, but, regardless of the type of SF product, they existed in the form of agglomerated lumps and the sizes of large ones reached several millimeters. The particle size analysis based on SEM images

formed the size distribution in a smaller range compared to the results obtained by the DLS technique. The difference between the two techniques was attributed to the link of nanoparticles at a high temperature, the densification process or the agglomeration of nanoparticles thereafter, or the difference in dispersion efficiency during the ultrasonic treatment.


**Author Contributions:** Conceptualization, S.-H.K. and S.-G.H.; methodology, S.-H.K. and S.-G.H.; investigation, S.-H.K.; writing—original draft preparation, S.-H.K.; writing—review and editing, J.M.; visualization, S.-H.K.; supervision, J.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Korea Agency for Infrastructure Technology Advancement (KAIA), grant number 20NANO-B156177-01, and the Korea Agency for Infrastructure Technology Advancement (KAIA) was funded by the Ministry of Land, Infrastructure and Transport.

**Acknowledgments:** The Institute of Engineering Research at Seoul National University provided research facilities for this work.

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