Fly ash (FA), ground granulated blast-furnace slag (GGBS), and silica fume (SF) were used as precursors for SH-UHPGC production. The commercial fly ash and GGBS are provided by Green Island Cement Co. Ltd., Hong Kong, and silica fume is purchased from mainland China. According to X-ray fluorescence (XRF) tests, the fly ash was classified as Class F according to ASTM C618-19 (ASTM, 2019), with 52.4% SiO₂, 25.8% Al₂O₃, and 8.4% Fe₂O₃, GGBS contained 44.2% CaO, 32.1% SiO_2, and 14.1% Al₂O₃, while silica fume contained over 95% SiO₂. The three types of precursors (i.e., FA, GGBS, and SF) have D ₅₀ particle sizes of 13.47 μm, 10.83 μm, and 0.56 μm, respectively. Fine silica sand was used as the aggregates in SH-UHPGC, which have an average diameter smaller than 300 μm and water absorption of 0.8%. Figure 1 presents the morphological patterns of these materials under scanning electron microscopic (SEM). Fly ash particles were mostly spherical while GGBS particles were angular, and silica fume was in a much smaller size. Two types of solid alkaline activators were adopted in this study, i.e., sodium metasilicate (Na₂SiO₃-anhydrous) and sodium carbonate (Na₂CO₃-anhydrous). Both were in analytical grade with purity over 99.5%. Besides, liquid waterglass purchased from Kowloon Sodium Silicate Factory Ltd. containing 27.7% SiO₂, 8.7% Na₂O, and 56.8% H₂O was also used as the alkaline activator in this study. Finally, straight copper-coated steel fibers with a length and diameter of 13 mm and 200 μm, respectively, were used as reinforcements to realize the tensile strain-hardening behavior of UHPGC.

The mix proportion of SH-UHPGC was adapted from the authors’ previous work (Lao et al., 2022), wherein the precursor and aggregate contents were optimized by the particle packing theory to achieve the ultra-high compressive strength. With the same waterglass content, different ratios of Na₂CO₃ and Na₂SiO₃ were adopted as the variables (i.e., pure Na₂CO₃, hybrid Na₂CO₃ and Na₂SiO₃, and pure Na₂SiO₃), and the mix proportions of SH-UHPGC are summarized in Table 1. Specifically in this study, the fly ash-to-GGBS ratio was fixed at 1:4, and the water/precursor ratio was 0.26. Besides, the Na₂O/precursor ratio was fixed at 6%, and 2.0% (by volume) steel fibers were added. In the table, the Mix ID “CaSb” was used to represent the mixtures with different sodium silicate/carbonate ratios, wherein C and S represent Na₂CO₃ and Na₂SiO₃, respectively, and a and b represent their percentages. It should be noted that for C50S50 in Table 1, the mole ratio of Na₂CO₃ anhydrous and Na₂SiO₃ anhydrous was 1:1.

Before the SH-UHPGC preparation, the solid activators, waterglass, and water were mixed to form a uniform alkaline solution, which was cooled down until the room temperature was reached. For the SH-UHPGC production, precursors and sand were firstly dry-mixed for 5 min, followed by the adding of alkali solution and the continuous stirring for another 10 min. Finally, steel fibers were added and the mixture was further mixed for 5 min. After all the mixing steps were completed, the mini-slump tests were conducted to measure the flowabilities of different mixes, which were measured as 135 mm, 162 mm, and 185 mm for C100S0, C50S50, and C0S100, respectively. The flowability of SH-UHPGC decreased with higher ratio of Na₂CO₃. After that, the fresh slurry was cast into cubic and dumbbell molds, and sealed with plastic films to avoid excessive water loss. It should be noted that demolding was conducted after 7 days as the C100S0 specimens took a much longer time to set and get hardened. Then, in order to accelerate the reaction and achieve high strength, the demolded samples were sealed with plastic films and cured at 80°C in an oven for 3 days. After heat curing, the specimens were dried at room temperature (23°C) for 2 days until further tests.

The compressive strengths were measured by three 50 mm × 50 mm × 50 mm cubes under the loading rate of 1.0 MPa/s. The direct tensile stress-strain relationships of SH-UHPGC were tested by three dumbbell specimens (Figure 2) under a displacement-controlled rate of 0.5 mm/min (Wu et al., 2018; Wu et al., 2020). The tensile strain of the middle area (with 80 mm gauge length) was measured by two symmetrically arranged linear variable differential transformers (LVDTs). For digital image correlation (DIC) analysis, one side of the dumbbell specimen was sprayed with black and white spackles (Huang et al., 2017; Li et al., 2021), and this side was continuously photographed by a digital camera at an interval of 3 s during the test.

For the matrix characterization, the release of the reaction heat during the first 7 days was recorded by an isothermal calorimeter (Calmetrix I-Cal 4,000). For each mix, approximately 100 g paste sample (excluding sands and steel fibers) was stirred outside the machine for 3 min ahead of time, after which they were put in the isothermal calorimeter for 7 days for the reaction heat measurement. Then, the hardened pastes were collected and cured in the same way as described in Section 2.2, and the fragments from the inner regions were cut, fixed in epoxy, polished to obtain a smooth surface, freeze-dried at −80°C for 4 h, and coated with a gold sputter for backscattered electron (BSE) analysis (Tescan VEGA3). In the BSE test, a magnification of 1,200 times was adopted. Additionally, for Fourier-Transform Infrared Spectroscopy (FTIR) tests, fragments of the pastes were collected and pulverized into powders (smaller than 75 μm). The spectra from 400 cm^-1 to 2000 cm^-1 with a resolution of 4 cm^-1 were recorded under ATR (Attenuated Total Reflection) mode.

The compressive strengths of SH-UHPGC are presented in Figure 3. All the mixes exhibited compressive strengths over 130 MPa. In the aspect of the different activators used, the compressive strength of the mix with sodium carbonate (i.e., C100S0, 135.8 MPa) was significantly lower than the mix with sodium silicate (i.e., C0S100, 179.0 MPa). Considering that the former mixture took a much longer time to set during sample preparation (up to 7 days), pure sodium carbonate showed a lower activation potential and cannot achieve rapid strength development of SH-UHPGC. Interestingly, C50S50 (186.0 MPa) with hybrid Na₂CO₃ and Na₂SiO₃ showed the highest strength among the three mixes (37.0% and 3.9% higher than those of C100S0 and C0S100, respectively). This phenomenon indicates the positive effect of combining Na₂CO₃ and Na₂SiO₃ in the alkali-activation and the mechanical performance improvement of SH-UHPGC. The mechanism of this phenomenon will be discussed in the following sections based on reaction kinetics.

The 7-day reaction heat release curves of the SH-UHPGC pastes are plotted in Figure 4. Generally, the reaction heat release can be divided into five stages as proposed by Shi and Day (Shi and Day, 1995): 1) initial stage, 2) induction stage, 3) acceleration stage, 4) deceleration stage, and 5) steady-state diffusion stage. Such five-stage heat release procedure was observed in the mixture containing Na₂SiO₃ (i.e., C0S100 and C50S50), while for C100S0, the acceleration and deceleration stages were not clearly observed in the early 7 days.

To make a clearer observation, the early-age heat release procedures of different mixes in the first 10 hours are presented in Figure 4A. For the initial stage associated with the preliminary dissolution of the precursors and the initial reaction within the first hour, the heat release rate was higher as the Na₂CO₃ ratio increased. A possible reason could be that at the very early stage, the Ca²⁺ released from the precursors immediately combined with CO₃ ^2- to form CaCO₃ polymorphs or sodium-calcium carbonate phase (pirssonite, hydroxysodalite, and gaylussite) (Bernal et al., 2016). But for C0S100, no such phenomenon was observed due to the absence of CO₃ ^2-.

After the initial stage, C100S0 maintained an extremely slow heat release rate until the end of the test (168 h). In comparison, C50S50 showed a significant induction stage, which appeared as a concave between the initial peak and the acceleration stage in the heat release rate curve. For C0S100, however, such induction stage was less obvious, which was the major difference between C50S50 and C0S100. The induction stage corresponds to the progressive dissolution of the precursors and the initial condensation and precipitation of the reaction product. The reason may lay in the comparatively lower PH value of the hybrid activators and the fixation of Ca²⁺ ions by CO₃ ^2- in C50S50 at the early stage, which made the release of Ca²⁺ from the precursors slower, reduced its concentration in the pore solution and consequently impeded the precipitation of reaction products. However, no such phenomena occurred in C0S100 due to the absence of CO₃ ^2-.

Then, after the acceleration stage, C50S50 and C100S0 peaked almost at the same value in the heat release rate curve. After 48 h, C50S50 exhibited a higher heat release rate than that of C0S100 as the difference between their cumulative heat release curves gradually became closer. The reason for the above phenomenon is that in C50S50, the CaCO₃ polymorphs formed at the initial and induction stages, could act as nucleation seeds (Tan et al., 2019), promote the reaction and facilitate the formation of the reaction products. From the end of the 7-day detection, it could be forecasted that the cumulative heat release of C50S50 may exceed that of C0S100 under further curing. In comparison, the heat release rate of C100S0 showed a gradually increasing trend over time, indicating that the alkali-activation reaction would gradually take effects after the CO₃ ^2- was continuously consumed by the Ca²⁺ release from the precursors (Bernal et al., 2015).

The microstructures of SH-UHPGC pastes with different activators observed from BSE tests are shown in Figure 5. As can be seen in all the mixes, a large number of unreacted GGBS and fly ash particles with different sizes remained in the pastes, and the dark-grey region encapsulating the unreacted particles represented the space-filling gels generated from alkali-activation. For C100S0, obviously, the number of the unreacted raw precursors seemed to be the largest, and the microstructure of the generated gels was loose and heterogeneous with some evident flaws, which is the reason for the lowest compressive strength of this mix as presented in Figure 3. In comparison, both C50S50 and C0S100 presented denser and more uniform microstructures without significant flaws and voids. Therefore, the C50S50 and C0S100 mixes with Na₂SiO₃ had a higher reaction degree than C50S50 and the reaction products both presented excellent space-filling effects as compared to C100S0.

To further illustrate the reaction conditions of SH-UHPGC pastes with different activators, the reaction degree was calculated based on the BSE images as recommended by Scrivener et al. (2016). For each mix, 15 BSE images with the dimensions of 170 μm × 230 μm were used for calculation, and the obtained results are shown in Figure 6. As observed from the figure, the reaction degrees of C50S50 and C0S100 were much higher than that of C100S0, which coincided well with the reaction heat results in Figure 4. Interestingly, a higher reaction degree was observed in C50S50 (40.8%) than that in C0S100 (35.1%). As discussed in Section 3.2, the CaCO₃ polymorphs, as initial products from Na₂CO₃ activation, could play as nucleation seeds and facilitate the gel formation of C50S50, which thus presents a higher degree of reaction than C0S100 under further curing. In comparison, C100S0 exhibited a reaction degree even lower than 20% (i.e., 17.9%) and was almost half that of C0S100. Therefore, such a low reaction degree of C100S0 could not provide sufficient reaction products to complete the space-filling procedure. In order to present the relationship between reaction degrees and compressive strengths of SH-UHPGC, these two indices are plotted together in Figure 6 as well. Obviously, the compressive strength showed a positive relation with the reaction degree, indicating that the adjustment of the reaction degree of the matrix is important for tailoring the compressive strengths of SH-UHPGC.

FTIR was employed to investigate the chemical composition differences of the reaction products in different mixes, and the tested results are presented in Figure 7. Baseline correction was done on all the spectra. In the figure, the intensity of the characteristic bands at around 1,421–1,470 cm^-1 corresponds to the asymmetric stretching vibrations of ν³ C-O bonds in CO₃ ^2-, and the intensity at around 876 cm^-1 corresponds to the out-of-plane bending vibration of ν² C-O bonds in HCO₃ ⁻ (Nedeljković et al., 2018), which both tended to increase as the sodium carbonate ratio increased. Here, it is noted that the weak signals occurred in C0S100 are attributed to the unavoidable carbonation during the sample preparation. The signal of ν⁴ C-O bonds in CO₃ ^2- and the bending of Al-O-Si could have overlapped at around 712 cm^-1. The peak centered at around 449 cm^-1 was assigned to the bending vibrations in the Al-O octahedrons (Cao et al., 2020). The main peaks at around 970 cm^-1 are designated to the Si-O-T (T denotes Si or Al) bonds in silicate gels, whose intensity was highest in C50S50, followed by C0S100 and C100S0 in sequence. This result is in accordance with the findings in reaction heat and BSE observations, which may further validate the enhanced alkali-activation reaction degree in SH-UHPGC with hybrid Na₂CO₃ and Na₂SiO₃ activators.

The tensile stress–strain curves of SH-UHPGC with different matrices are presented in Figure 8. Obviously, strain-hardening behaviors were observed in all the mixes (i.e., tensile strength could further increase after the first cracking strength, together with the increase of tensile strain). Unlike ECC materials showing multiple stress drops in tensile responses (Yu et al., 2020; Xu et al., 2022a; Li et al., 2023; Xu et al., 2023), the tensile stress-strain curves of SH-UHPGC were very smooth, indicating the excellent crack width control ability of steel fibers (Lao et al., 2022). To further analyze the tensile performances of SH-UHPGC, their tensile strengths and strain capacities are summarized in Figure 9. From the figure, the highest tensile strain capacity (0.44%) and tensile strength (11.9 MPa) were achieved in C50S50, indicating the excellent tensile performance of the mix using hybrid Na₂SiO₃ and Na₂CO₃ activators. Considering that the tensile strength of strain-hardening cementitious (geopolymer) composites is highly dependent on the fiber/matrix bond (Lao et al., 2023), the highest tensile strength of C50S50 can be attributed to the highest reaction degree of the matrix (as presented in Figure 6). In comparison, the other two mixes showed similar tensile strengths, and C100S0 presented the lowest tensile strain capacity (0.37%), indicating that Na₂SiO₃ showed better activation effect than Na₂CO₃ in producing SH-UHPGC.

The DIC analysis was performed on a subset radius and subset spacing of 30 pixels and 3 pixels, respectively. The local strain value was calculated based on a strain radius of 3 pixels. Figure 10 presents the DIC strain fields of SH-UHPGC at different strain levels under direct tension. Here, four constant strain levels (i.e., 0.10%, 0.20%, 0.30%, and 0.40%) and the ultimate strain were adopted for analysis. From the figure, for all the SH-UHPGC mixes, multiple cracks were observed at all the presented strain levels, and the cracking became more saturated as the strain level increased. It should be highlighted that very significant multiple cracking can be observed for all the developed SH-UHPGC even at a very low tensile strain level (e.g., 0.1% or 0.2%). This phenomenon is quite different from those of cement-based strain-hardening UHPC and high/ultra-high-strength ECC, which only showed few cracks with the tensile strain lower than 0.2% (Huang et al., 2021b; Huang et al., 2021c; Huang et al., 2022b; Zhu et al., 2022). For the developed SH-UHPGC, the steel fibers used can effectively narrow the crack widths (i.e., below 30 μm). However, the tensile crack width of ECC/SHCC is commonly 60–150 μm (Ding et al., 2018; Yu et al., 2019a; Huang et al., 2019). Thus, at the same tensile strain level, the crack number of SH-UHPGC would be much larger than that of ECC/SHCC, leading to the pronounced multiple-cracking behavior of SH-UHPGC.

Due to the limitation of the digital camera used, the maximum resolution of the captured photographs was not enough for the analysis of crack width, and thus no visible cracks could be found in the photographs at different strain levels. It indicated that the crack widths of the developed SH-UHPGC should be smaller than 30 μm, as 1 pixel = 30 μm in the captured photographs. The actual value of the crack width of SH-UHPGC in this study remains unknown, and it should be further investigated in the following work.

Although geopolymer is generally regarded as a green material owing to its clinker-free feature, the use of conventionally adopted alkaline activator (e.g., Na₂SiO₃) still shows a heavy impact on the environment from life-cycle assessments (Habert et al., 2011). Therefore, it is of great significance to evaluate the environmental impact and economical potential of replacing Na₂SiO₃ with Na₂CO₃ in the production of SH-UHPGC.

The embodied carbon, embodied energy, and costs of raw materials and the produced SH-UHPGC are summarized in Table 2 and Table 3, respectively. In addition, the results of SH-UHPGC with different matrices are presented in Figure 11 for a more distinctive comparison. It can be found in Table 2 that Na₂SiO₃ shows higher embodied carbon, embodied energy, and material cost than Na₂CO₃. In consequence, when the replacement ratio of Na₂SiO₃ by Na₂CO₃ increased from 0% to 100% to produce SH-UHPGC, the embodied carbon decreased from 793.4 kg CO₂/m³ to 619.6 kg CO₂/m³ and the embodied energy decreased from 7,375.3 MJ/m³ to 6,945.8 MJ/m³, while the material cost changed very little. Therefore, the Na₂CO₃-based SH-UHPGC shows better sustainability and is more eco-friendly than the Na₂SiO₃-based one. Also, among all the raw materials of SH-UHPGC, it is mentioned that the use of steel fibers made heavy contributions to the three indices (i.e., 55.6%–71.3% of the total embodied carbon, 76.1%–80.9% of the total embodied energy, and 56.6%–56.9% of the total material cost). In this aspect, it is meaningful to seek for alternative fibers to realize tensile strain-hardening behavior of greener and cheaper UHPGC in the future.

Based on the results in the above sections, an overall assessment was conducted considering mechanical performances, environmental impacts, and material costs. A radar graph was used to present the results (i.e., Figure 12). Here, because lower embodied carbon, embodied energy, and material cost are desirable for the practical applications of UHPGC, their reciprocals were used in the six-dimensional presentation. For easy comparison, all values are normalized by the corresponding value of C0S100. Among the three mixes, C100S0 was the most environmentally friendly, but presented the poorest mechanical performances. In comparison, C50S50 showed the most distinguished tensile strain capacity, tensile strength, and compressive strength, and a moderate environmental impact, demonstrating the superiority of this mix. For C0S100, although its mechanical performance is also excellent, this mix showed the highest embodied carbon, as well as comparatively high embodied energy and material cost. Therefore, the use of hybrid Na₂SiO₃ and Na₂CO₃ is a promising method to achieve the best overall performance of the developed SH-UHPGC.