The geopolymer matrix was produced by alkaline activation of GGBFS, FA, and silica fume, with the chemical compositions of these precursors summarized in Table 1. Industrial-grade NaOH pellets with a purity of 98% ± 1% and a sodium silicate solution were used as alkaline activators, where the sodium silicate solution contained 8.3% Na₂O, 26.5% SiO₂, and 65.2% H₂O by weight. Natural sand obtained from the Xiang River, characterized by a maximum particle size of 2.36 mm, was utilized as the fine aggregate. Furthermore, straight steel fibers with a length of 13 mm and a diameter of 0.12 mm were incorporated into the mixture, providing a tensile strength of 2,500 MPa to the composite matrix.

Based on previous study (Okoye et al., 2016), the mixture proportions were selected to achieve ultra-high-strength geopolymer concrete with adequate workability. Specifically, a W/B ratio of 0.32 was chosen to minimize capillary porosity while ensuring sufficient paste volume for flow. Furthermore, an activator modulus (Ms) of 2.0 with a Na₂O content of 7% was adopted to strike a balance between the dissolution rate of the precursors and the condensation rate of the geopolymeric gel; this combination ensures high mechanical strength without inducing the flash setting often associated with higher alkalinity systems. A GGBFS/FA mass ratio of 4:1 were adopted. Silica fume was incorporated at dosages of 0%, 5%, 10%, 20%, and 30% by mass of the binder. The steel fiber content was fixed at 2 vol%, and the sand-to-binder mass ratio was set to 1.0. The detailed mix proportions are summarized in Table 2.

The alkaline activator was prepared 24 h in advance and stored in an airtight container to ensure chemical stability prior to use. For UHPGC preparation, the binders and sand were first dry-mixed for 3 min at low speed to achieve a uniform distribution of the precursors. The activator was then gradually added and mixed for 3 min at low speed, followed by 1 min at high speed to facilitate the alkali-activation reaction. Subsequently, steel fibers were evenly introduced using a 10 mm steel sieve and mixed for an additional 5 min at low speed to ensure random orientation and prevent fiber balling within the matrix.

The mixtures were cast into 40 × 40 × 160 mm³ steel molds and Φ10 × 150 mm cylindrical molds, ensuring the precise geometric formation of the specimens for subsequent mechanical testing. The specimens were then kept in the molds and pre-cured for 24 h in a controlled environment, where an ambient temperature of 20 °C and a relative humidity of 65% were maintained. During this pre-curing stage, the molds were covered with plastic films to prevent moisture loss and inhibit the premature evaporation of the alkaline solution, which is critical for the initial stage of geopolymerization. Subsequently, based on previous research (Li et al., 2018), the specimens were demolded and subjected to steam curing at 80 °C for 24 h.

Figure 1 illustrates the effect of silica fume content on the flowability of the fresh mixtures. The fresh properties were significantly improved by the incorporation of silica fume, as the flowability first increased from 188 mm to 260 mm with the addition of 20% silica fume, followed by a decrease to 235 mm when the dosage reached 30%. Furthermore, although the incorporation of steel fibers resulted in a decrease in flowability, the differences were less significant, particularly when the silica fume content was below 10%. For instance, the flowability of mixture ZA2 was 228 mm, representing a decrease of only 2 mm compared to the fresh mixture A2, while the variation trend for mixtures with steel fiber remained consistent with those without steel fiber, demonstrating the feasibility of utilizing silica fume to improve the workability of fiber-reinforced systems. These results may be attributed to the lubricating effect of silica fume, because when silica fume is added, the packing density of the solid particles increases and more solution is released from the interstitial spaces to act as a lubricant, thereby significantly enhancing the overall flowability.

A lubricating effect, often referred to as the “ball-bearing” effect, is facilitated by the spherical morphology and glassy surface of FA particles, leading to a reduction in both water demand and internal friction. Conversely, a slight elevation in yield stress is induced by a high FA content, thereby resulting in decreased fluidity due to the increased particle-to-particle contact and the higher volume fraction of solids within the matrix. It has been noted in previous research (Wu et al., 2018) that the influence of FA on the workability of alkali-activated slag/FA systems is intricate; while a reduction in plastic viscosity can be induced by the ball-bearing effect, the overall rheological balance remains complex. Regarding the addition of SF, an enhancement in fluidity is observed at low replacement levels, which is likely attributable to the spherical shape of SF particles providing additional lubrication. However, as the SF content is further increased, a higher water requirement is necessitated by the significantly larger specific surface area of SF, thus leading to reduced flowability at a constant W/B ratio.

The influence of silica fume on the initial and final setting times of fresh UHPGC mixtures is depicted in Figure 2. It is observed that the incorporation of silica fume retards the hardening process. Specifically, as the silica fume dosage increases from 0% to 30%, the initial setting time prolongs from 114 min to 192 min, while the final setting time increases from 143 min to 225 min. Previous investigation has suggested that ground granulated blast furnace slag, which is rich in CaO, exerts a dominant influence on the setting times (Yang et al., 2018). In the present study, the reduction in ground granulated blast furnace slag content, resulting from the addition of silica fume, leads to the prolongation of setting times. Under the high alkalinity of the activator, the cleavage of Ca-O bonds in the slag particles typically promotes the hardening process through the formation and precipitation of primary C-S-H gels. Conversely, the dissolution of silica fume in the alkaline solution consumes hydroxide ions, which reduces the alkalinity of the system and consequently decelerates the reaction rate of CaO.

Figure 3 illustrates the effect of silica fume on the compressive strengths of samples both without and with steel fiber. The incorporation of silica fume is observed to significantly influence the development of compressive strength, where compressive strengths initially increase and subsequently decrease with rising silica fume content. In samples without steel fiber, the highest compressive strength of 110.7 MPa is achieved at a silica fume dosage of 5%. However, further increases in silica fume content result in a considerable reduction in compressive strength, suggesting that an excessive dosage disrupts the optimal particle packing or consumes available alkalinity required for effective geopolymerization.

Through steel fiber reinforcement, the average compressive strength of all samples exceeds 120 MPa, as shown in Figure 3. The variation trend of compressive strengths for samples with steel fiber was similar to that observed in samples without steel fiber. The maximum compressive strength of 157.0 MPa was achieved at a silica fume dosage of 5%, which was slightly higher than that of the sample without silica fume by 0.32%. However, once the silica fume content exceeded 5%, the compressive strength decreased sharply to 130.6 MPa, representing a substantial reduction of 16.8%. With a further increase in silica fume beyond 10%, the reduction in strength continued, further confirming that the excess silica fume introduces defects rather than reinforcing the matrix. This phenomenon is likely due to the formation of more air voids with the continuous incorporation of silica fume in the presence of fibers, as illustrated in Figure 4. These voids induce stress concentration and weaken the bond strength between the geopolymer matrix and the fibers.

Figure 7 illustrates the relationship between compressive strength and water absorption porosity of UHPGC with varying silica fume contents. It is important to note that this assessment quantifies the macroscopic permeable porosity and does not capture the pore size distribution typically characterized by MIP or NMR. The porosity of all mixtures remains below 20%, regardless of silica fume dosage; however, the incorporation of silica fume generally leads to an increase in porosity. When 5% silica fume is added, the porosity increases slightly from 15.40% to 15.82%, while the compressive strength increases from 98.8 MPa to 110.7 MPa. With a further increase in silica fume content from 5% to 10%, the porosity rises markedly from 15.82% to 19.84%, accompanied by a reduction in compressive strength. When the silica fume content exceeds 10%, a slight decrease in porosity is observed, which is consistent with the corresponding strength development. The increase in porosity induced by silica fume addition is likely associated with the formation of air voids. The slight reduction in porosity at silica fume contents of 20%–30% may be attributed to changes in the reaction products, which are examined in the subsequent sections.

Figure 8 presents the XRD diffractograms of the raw materials (GGBFS, FA, and silica fume) and UHPGC pastes with different silica fume contents. GGBFS exhibits a dominant amorphous phase located between 25° and 40°. Its primary crystalline phases are calcite and aragonite, with a minor presence of quartz. The occurrence of calcite is likely attributed to chemical weathering on the surface of GGBFS (Ismail et al., 2014). For FA, the main crystalline phases are quartz, mullite, maghemite, and hematite, while its amorphous phase is observed in the range of 15°–30°. In comparison, silica fume contains a substantially higher proportion of amorphous components than both GGBFS and FA, characterized by a broad diffuse hump spanning from 10° to 40°.

The presence of quartz, mullite, aragonite, and calcite in the UHPGC pastes is attributed to the residual unreacted crystalline phases from the raw materials (Gao et al., 2015), while no portlandite phase was observed, indicating a distinct reaction pathway compared to cement-based systems. The primary diffraction peak of the reaction products was identified as C-A-S-H gel, located at approximately 29.5°, where the incorporation of silica fume led to variations in the intensity of these phases mainly due to changes in the relative contents of GGBFS and FA. Specifically, the intensity of crystalline phases such as quartz, mullite, and calcite decreased as the silica fume content increased, and conversely, no new crystalline phases were formed upon the addition of silica fume. However, it is noted that the intensity of the C-A-S-H peaks slightly increased at a silica fume dosage of 5%, whereas it considerably decreased at 20% and 30% dosages, suggesting that the formation of C-A-S-H gel can be promoted by a low content of silica fume but is inhibited at higher replacement levels.

Figures 9a,b show the FTIR spectra of GGBFS, FA, silica fume, and UHPGC pastes with different silica fume contents. The bands observed between 1200 cm^-1 and 950 cm^-1 in GGBFS and FA are assigned to the asymmetric stretching vibration modes of Si–O–T (T: Si or Al) bonds. This band is centered at 1012 cm^-1 in GGBFS and 1093 cm^-1 in FA. The asymmetric stretching mode of the O–C–O bonds in calcite is located at 1460 cm^-1, which is consistent with the XRD observations. A strong absorption at 676 cm^-1 is likely attributed to the incorporation of gypsum during the milling process (Ismail et al., 2014). For FA, the symmetric stretching band at 839 cm^-1 is ascribed to the presence of quartz. The shoulder between 570 cm^-1 and 560 cm^-1 is due to the presence of mullite, corroborating the XRD results. In the silica fume spectrum, a prominent band from 960 cm^-1 to 1320 cm^-1 is observed, attributed to the asymmetric stretching modes of Si–O bonds. Additionally, two strong absorption bands at 805 cm^-1 and 477 cm^-1 correspond to the Si–O–Si symmetric stretching and O–Si–O bending vibrations, respectively.

After activation, the main absorption band of the hardened pastes is located at approximately 1010 cm^-1, which is attributed to the asymmetric stretching vibration of Si–O–T (T: Si or Al) in the C-A-S-H gels, consistent with the XRD results. This band shifts toward higher wavenumbers, increasing from 1010 cm^-1 to 1041 cm^-1 as the silica fume dosage rises from 0% to 30%, indicating an increased degree of polymerization and a higher Si/Al ratio within the gel framework. While previous studies indicated that a low dosage of nano-silica (3%) was insufficient to alter the gel structure, the addition of 5% silica fume in this investigation successfully induced structural changes, where the dissolution of silica fume in the alkaline activator increases the molar modulus of the solution by breaking Si–O–Si bonds. This process subsequently leads to a higher concentration of silicate species available for the geopolymerization reaction, thereby modifying the molecular configuration of the resulting reaction products. This process promotes the formation of a Si-rich structure, resulting in the observed blue shift of the main absorption band. Furthermore, according to the literature (Torres-Carrasco and Puertas, 2017), utilizing Si-rich precursors often results in the formation of gels with higher silicon and lower aluminum/calcium contents, characterized by higher wavenumbers and lower mechanical strengths. Conversely, the relative decrease in GGBFS content and the increase in silica fume both lead to a reduction in the Al/Si ratio. This variation, influenced by the phase overlapping of raw materials and reaction products, drives the main absorption bands toward higher wavenumbers (Zhang et al., 2019).

Previous research has indicated that the deconvolution of FTIR spectra can be used to investigate the reaction products of alkali-activated binders with reasonable reliability (Ren et al., 2021; Mo et al., 2024; Li J. et al., 2025; Tian et al., 2026). Figure 10 depicts the deconvolution results for UHPGC pastes with different silica fume contents between 800 and 1300 cm^-1. To facilitate the interpretation of the spectral features, the specific assignments of the deconvoluted FTIR bands, including their vibration modes and corresponding structural units, are summarized in Table 4. The relative areas of the deconvoluted component peaks are illustrated in Figure 11. The incorporation of silica fume leads to notable variations in the relative areas of the deconvoluted peaks. With the addition of 5% silica fume, the largest relative area of the band at approximately 1000 cm^-1, corresponding to the Q² site of the main C–A–S–H gel band, is observed. This indicates that a moderate silica fume content promotes the formation of C–A–S–H gels. However, with further increases in silica fume content, the relative area of the Q² band decreases, suggesting a reduction in the proportion of these gels. In addition, the bands located at 1032, 1089, and 1135 cm^-1, which are associated with sodium-bonded silicate-based gels and correspond to the Q³ and Q⁴ sites, are also influenced by silica fume incorporation. When the silica fume content exceeds 5%, the dissolution of excess amorphous silica increases the activator modulus Ms of the pore solution. This chemical shift kinetically favors the formation of highly polymerized sodium-bonded silicate gels characterized by Q³ and Q⁴ species rather than the strength-giving C-A-S-H gel. Simultaneously, the elevated Ms reduces the alkalinity required to dissolve the fly ash particles, evidenced by the increased prominence of the unreacted mullite band at 1183 cm^-1. Notably, when 5% silica fume is incorporated, the relative area of this band reaches its minimum, indicating that FA participation in the reaction is enhanced at this dosage. Furthermore, the continuous decrease in the relative area of the band at approximately 875 cm^-1, attributed to calcite, reflects a gradual reduction in the relative GGBFS content as silica fume replaces part of the binder.

In PC binder systems, silica fume exerts both physical and chemical effects on the properties of UHPC, including lubrication, nucleation, particle filling, and the promotion of hydration. These effects arise from its extremely fine particle size and high pozzolanic activity, which collectively contribute to the remarkable enhancement of UHPC performance. However, based on the above results and discussion, the incorporation of silica fume in UHPGC systems does not produce an analogous improvement, as the alkali-activation mechanism differs significantly from the hydration processes typically found in traditional cementitious matrices.

Regarding the effect of silica fume on fresh properties, the flow behavior of UHPGC is significantly enhanced by the lubricating effect of silica fume. Its incorporation increases the packing density of the mixture, making more solution available for lubrication between particles. This suggests that silica fume plays a key role in improving the workability of UHPGC while requiring relatively less water compared to traditional geopolymer binder systems. Additionally, due to the high alkalinity of the activators, silica fume dissolves rapidly but does not provide a nucleating effect on reaction kinetics. As a result, the hardening process of UHPGC is delayed because of the increased Ms. of the solution, which slows down the initial polycondensation rate. Therefore, in the absence of effective and compatible chemical admixtures for geopolymer binder systems, silica fume can be utilized to enhance rheological properties and retard setting times of UHPGC (Conte and Plank, 2019).

Regarding the hardened properties, the degradation of mechanical performance at silica fume dosages exceeding 5% is governed by the interplay between physical porosity and chemical geopolymerization. It is acknowledged as a limitation of this study that microstructural imaging (e.g., SEM) was not performed to directly visualize the interfacial transition zone between the steel fibers and the matrix. Consequently, the specific mechanism governing the fiber bond degradation remains partly inferential. However, the quantitative water absorption porosity tests (Figure 7) provide decisive macroscopic evidence for the structural deterioration. As the silica fume content increased from 5% to 10%, the porosity exhibited a sharp increase from 15.82% to 19.84%. This trend aligns with the visual observation of macro-voids in the cross-sections of high-dosage specimens shown in Figure 4. This increase in porosity is attributed to the rheological threshold; the excess silica fume significantly increased the viscosity of the fresh paste, entrapping air bubbles that could not be eliminated during vibration. Consequently, these voids acted as stress concentrators, leading to the observed reduction in both compressive strength and flexural toughness. Therefore, unlike in PC-based systems where silica fume acts as a filler to densify the matrix, in this geopolymer system, high dosages of silica fume introduce physical defects that outweigh its potential chemical benefits.

From a chemical perspective, the deconvolution analysis of FTIR spectra further elucidates the optimal dosage of 5%. At this dosage, the highest relative area of the Q²(1) band (associated with C-A-S-H gel formation) was observed, indicating a synergistic promotion of the reaction extent. However, beyond 5% dosage, the dissolution of silica fume increased the Ms of the alkaline solution, which kinetically hindered the activation of fly ash and GGBFS. This is evidenced by the decrease in the Q² band area and the simultaneous increase in the Q³ and Q⁴ bands (Figure 11), suggesting the formation of Si-rich gels with lower mechanical binding capacity. Thus, the optimal 5% dosage represents a critical balance point where the formation of binding gels is maximized before the adverse effects of increased porosity and hindered reaction kinetics take dominance.

In summary, the incorporation of silica fume in UHPGC significantly increases setting times, which can be advantageous for extending workability in large-scale applications. However, the silica fume dosage optimal for UHPC (20%–35%) is unsuitable for UHPGC, as it causes a decline in mechanical performance by increasing porosity and hindering the reaction kinetics of the geopolymer precursors. Considering the high cost of silica fume, such high dosages are also unnecessary from an economic perspective. This study found that an appropriate silica fume content of 5% achieves a good balance of workability and superior mechanical properties in UHPGC, representing the optimal threshold for enhancing the C-A-S-H gel network without compromising the density of the matrix.