**1. Introduction**

Ultra-high performance fiber-reinforced concrete (UHPFRC) is a construction material that has excellent mechanical properties, durability, and flowability. Its development was attributed to the achievement of optimal packing density by the use of ultrafine particles such as silica fume (SF) and quartz powder, the prevention of brittle failure by the inclusion of the high volume of short fibers, and advances in the technology of chemical admixtures [1,2]. The demand for this commercial material has increased in earnest since the beginning of this century, mainly related to innovative and sustainable structures such as building facades, infrastructures, and protection or explosion-proof facilities [3,4]. In these circumstances, practical factors such as price competitiveness, sustainability, accessibility of raw materials, and worker safety are becoming increasingly important concerns. With regard to the sustainability of this material, one major challenge was to reduce its enormous content of Portland cement (up to 1000 kg/m3) or unnecessary unhydrated cement [5,6]. Therefore, the incorporation of various supplementary cementitious materials (SCMs) such as limestone powder, calcined clay, fly ash, ground granulated blast furnace slag, steel slag, and metakaolin has been attempted in order to replace a portion of the cement [7–13]. Considering the practical aspect in particularly, cement replacement by limestone powder has been regarded as one of the most effective solutions [14–19].

To enhance UHPFRC's market competitiveness, it is also important to reasonably determine the type and content of silica fume (SF), considering economic efficiency and usability. SF is one of the most widely used ultrafine particles today, but, at the same time, it is also one of the most expensive construction materials due to its limited supply [20]. Therefore, inevitably, the more the material is incorporated, the higher the material cost of concrete [21]. After pioneering studies on the optimization of packing density by the use of ultrafine particles [1,2,22], it has been generally agreed that the optimal ratio of SF to the weight of cement is 25% in UHPFRC [6,23]. However, this ratio is at least two to three times higher than other types of concrete [20], and this gap is even more overwhelming when compared based on the content per unit volume of concrete [24].

Although SF makes a significant contribution to UHPFRC's material price, the use of alternative materials must be approached cautiously because SF is one of the most important raw materials affecting almost all material properties of UHPFRC. The ball bearing effect, which acts as a lubricant by the combination of SF and superplasticizer, and the micro filling effect that fills the space between fine particles or refines the capillary pore structure contribute decisively to exhibiting the unique characteristics and superior performance of UHPFRC [25]. The yield stress of fresh UHPFRC is reduced by the ball bearing effect, which guarantees its superior self-compacting ability [23]. The micro filling effect makes cement paste extremely compact and increases the interfacial bond between aggregates and the paste [26,27]. SF is also responsible for making the fiber matrix interface area; as an example, the interfacial bond strength within the steel fiber matrix is significantly affected by the content of the ultrafine particles [28]. SF particles that can be adsorbed on a cement particle up to millions also affect the chemical reaction of UHPFRC [29]. By providing a site for nucleation of calcium silicate hydrate (C-S-H), it accelerates cement hydration and directly participates in secondary hydration (i.e., a pozzolanic reaction) [30,31]. In particular, the latter not only reduces the content of portlandite, which negatively affects the strength and durability of the cement paste, but it also forms additional C-S-H, which improves these properties [32]. For these reasons, although it has been previously reported that SCMs composed of amorphous silica can replace the chemical roles of SF [33], it has been suggested that, with these materials, it is difficult to completely replace all the complicated roles of the SF in UHPFRC, especially in terms of the physical effects carried out by spherical particles with submicron size [9,24,25]. This further supports the extreme difficulty in finding materials that can replace SF in UHPFRC.

As another alternative, the use of densified SF can be considered, and this can be a practical compromise as it has been the type of product commonly used in the concrete industry [29,34]. When collected in a silo as a produced form, SF originally has a density of 125 to 150 kg/m3 [35]. However, an additional process for densification of ultrafine particles proceeds due to the improvement of convenience in handling and transportation (also related to material cost) and workers' health problems caused by fine dust (also related to labor cost) [36]. When air is blown into the silo, the particles roll and agglomerate in tens of μm to several mm, so that its density increases in the range of 480 to 720 kg/m3 [37]. Although undensified SF with a density of less than 350 kg/m3 is also used in limited amounts for special applications such as those involving pre-bagged products and grouts [38], it is not a widely used form for concrete production due to its high price, caused by low demand and the inconvenience involved with transportation and storage, as well as dust generation in the workplace [37]. However, despite disadvantages in usability and price competitiveness, undensified products have been preferred as suitable SFs in the production of UHPFRC because they can guarantee better dispersion than a densified product. Indeed, improved homogeneity compared to other types of concrete was one of the basic principles in the development of UHPFRC, and it has been considered that homogeneous dispersion of ultrafine particles is a prerequisite prior to particle size and specific surface area (SSA) of ultrafine particles [1]. For this reason, despite the practically great advantages, the feasibility of the densified product and its effect on the material properties of UHPFRC have been barely investigated.

When studying concrete with SF, its agglomeration tendency should be taken into account since SF particles do not exist in the form of independent nanoparticles but almost always as lumps [34,39]. This situation has been commonly reported in concrete containing coarse aggregate despite partial crushing of the lumps during the mixing process [40,41]. The agglomeration phenomenon is no exception in the SF manufactured as a collected state [42]. In other words, undensified SF can also not be free from agglomeration issues, so it is difficult to guarantee its complete dispersion within UHPFRC [43]. Moreover, due to the nature of the agglomeration, it is also considered very challenging to accurately measure the particle size or size distribution of SF [41]. Although ultrasonic dispersion or sonification before measurement is somewhat effective [44], it did not show completely satisfactory results because it could not affect the strong electrical attraction of the whole particles [41,45–47]. In general, laser diffraction (LD) and dynamic light scattering (DLS) are used for particle size analysis of powder materials, with advantages such as short analysis time and good reproducibility [48,49]. These techniques are suitable for particles with sizes between 0.5–3500 μm and 0.5–3 to 5 μm, respectively [49,50]. In this context, it has been suggested that it is reasonable to exclude the results of >1 μm range that are clearly larger than the size of separated SF particles while including the ultrasonic dispersion process [34,35]. Similarly, DLS techniques can be considered to characterize silica nanoparticles as they can naturally exclude large fractions of agglomerates depending on the measurement range of the device [25,32,51].

The main purpose of this study is to investigate the feasibility of using densified SF in order to improve the usability and price competitiveness of UHPFRC.

#### **2. Materials and Methods**

Three types of SF products with different manufacturers, manufacturing methods, and ages were prepared and then their effects on the material properties of the concrete were experimentally examined. First of all, to characterize SF products, chemical analysis by X-ray fluorescence (XRF), morphological analysis by high-resolution field emission scanning electron microscopes (FE-SEM), individual particle size calculation by the image processing technique, particle size distribution measurement by DLS, and SSA measurement by the BET method were carried out. Additionally, the performance of UHPFRC with different SF products was evaluated by measuring slump spread and mechanical properties such as compressive and flexural tensile strengths. Finally, X-ray diffraction (XRD) and thermogravimetric (TG) analyses were performed to compare the hydration mechanism due to the different form of SF on the properties of the concrete.

#### *2.1. Preparation of UHPFRC Samples*

For the experiment, UHPFRC was manufactured using Portland limestone cement, SF, quartz powder (SiO2 > 97 wt %), quartz sand (SiO2 > 90 wt %), water, polycarboxylate (PCE)-based superplasticizer, and steel fiber (Φ0.2 mm × 13 mm, tensile strength > 2500 MPa), as in previous studies [16,17]. The mineralogical composition of the used cement is monoclinic alite (46.79 wt %), triclinic alite (6.94 wt %), monoclinic belite (2.19 wt %), orthorhombic belite (1.38 wt %), anhydrite (0.61 wt %), aluminate (0.88 wt %), gypsum (1.09 wt %), dolomite (12.36 wt %), calcite (21.38 wt %), and amorphous content (6.38 wt %) measured by quantitative XRD analysis. The results of particle size analysis of the raw materials by LD are shown in Figure 1. Considering the measurement range of the equipment used (Mastersizer 3000, Malvern Panalytical, UK), the size distributions of cement, quartz powder, and quartz sand were measured. The size of quartz powder was in the range of 1 to 20 μm; moreover, its sizes were located within the range of the particle size of cement (0.4 to 100 μm), confirming that the quartz powder can properly perform its primary role in UHPFRC, i.e., filling the space between cement particles [1].

**Figure 1.** Particle size distribution of raw materials by laser diffraction.

Three different SF products (labeled as SF1\_U, SF2\_U, and SF2\_D) were prepared as experimental variables. SF1\_U is a high quality, undensified SF product available worldwide. It is named Microsilica-Grade 940U (Elkem, Oslo, Norway) and its bulk density is 200 to 350 kg/m3 according to the manufacturer's specifications. Around three years had passed since the purchase date of the product, but when estimated based on the date of manufacture, a much longer period should have elapsed considering import, distribution, and storage by domestic dealers. SF2\_U and SF2\_D are undensified and densified SF products manufactured in the same factory (POSCO, Pohang, Korea) and have bulk densities of ~180 and ~500 kg/m3, respectively. These domestic products, which do not require shipping or long-term transport and storage, had not been in our possession more than one year from the date of manufacture. As shown in Table 1, analysis results of chemical composition by a wavelength dispersive XRF spectrometer (S4 PIONEER, Bruker, Germany) verify that all products met the purity requirement of SF (>96 wt %) suitable for UHPFRC [23]. One reason that purity is important is that the inclusion of impurities, especially unburned carbon, has serious adverse effects on the hydration, mechanical properties, and workability of UHPFRC [24]. The results of XRF analysis also confirm that all SF products have the ability to function as highly reactive pozzolans and that their chemical compositions can be excluded from experimental parameters.


**Table 1.** Chemical composition of silica fume samples.

<sup>1</sup> Components with less than 0.1% of content (Cl, P2O5, ZnO, MnO, CuO, Ru, Pd); <sup>2</sup> Loss on ignition.

UHPFRC samples were prepared based on the mix proportions shown in Table 2; this was done with a 5-L planetary mixer. SF was blended with quartz sand at low speed (140 rpm) for 5 min to break up the lumped, ultrafine particles by collision with hard grains. This is possible because the collisions between particles enable dispersive movement, which causes collapse and dispersion of the agglomerated particles [52]. The remaining powders (cement and quartz powder) were then blended for another 5 min after being placed into the mixer. After 10 min of the dry blending process, water and superplasticizer were poured into the mixer and low speed mixing was maintained until the mixtures were agglomerated. As shown in Figure 2, the agglomeration of the mixture proceeded in two stages, during which local agglomeration of the relatively small grains was followed by the formation of one large agglomerate due to the agglomeration of the small grains.


**Table 2.** Mix proportions of UHPFRC (by weight of cement).

<sup>1</sup> Solid contents; <sup>2</sup> By volume of UHPFRC.

**Figure 2.** Manufacturing process of UHPFRC: (**a**) formation of small grains by local agglomeration; (**b**) agglomeration of small grains into large lumps; (**c**) dispersion after solid-suspension transition.

When mixing is maintained for a certain period of time after the liquid materials are added, a fluid bond is formed between the particles so that the materials in the mixer begin to agglomerate, as shown in Figure 2a. The interparticle force between particles increases due to the surface tension of the water and the capillary pressure inside the bond, which requires the mixer to increase its power to further disperse the particles; eventually, dispersion efficiency increases [53]. When all the ingredients in the mixer were agglomerated, the power of the mixer reached its maximum, as shown in Figure 2b. Immediately after this, a transition occurred between the solid aggregate and the dispersed suspension; at this time, the particles inside the aggregate were evenly distributed and the space between the particles was filled with liquid. Once this transition occurred, the capillary force disappeared so that the power consumption of the mixer was drastically reduced. Consequently, the mixture began to have self-compacting abilities, as shown in Figure 2c. After the steel fibers were added into the mixture, which had been changed to the suspension state, they were mixed at a high speed (285 rpm) for 2 min as the last step of the manufacturing process.

The freshly prepared UHPFRC was filled in the prepared mold. The concrete was poured slowly at one end of the mold when casting prismatic specimens to exclude the effect of fiber orientation. All specimens were cured in a sealed state at 20 ◦C on the first day. The next day, heat treatment was applied. In other words, the specimens were cured at 90 ◦C for 2 days after demolding, as the standard heat treatment method. From the third day to the experimental day, specimens were cured in a chamber set at 20 ◦C and 60% relative humidity (RH). In addition, the samples without heat treatment were also prepared. In this case, the specimens were exposed to air drying conditions (20 ◦C and RH 60%) from the next day after casting.
