Mechanical and microstructural performance of concrete containing high-volume of bagasse ash and silica fume

In this study, researchers examined the effect of replacing a high-volume of cement with sugarcane bagasse ash (BA) and silica fume (SF). In addition to the control, three binary and three ternary blends of concrete containing different percentages of cement/BA and cement/BA/SF were tested to determine the various mechanical and microstructural properties of concrete. For each mix, eighteen cylindrical concrete specimens were cast followed by standard curing (moist at 20 °C) to test the compressive and tensile strengths of three identical specimens at 7, 28, and 91 days. The test results indicated that the binary mix with 20% BA and ternary mix with 33% BA and 7% SF exhibited higher strengths than all the other mixes, including the control. The higher strengths of these mixes are also validated by their lower water absorption and apparent porosity than the other mixes. Following mechanical testing, the micro and pore structures of all mixes were investigated by performing scanning electron microscopy/energy-dispersive X-ray spectroscopy (SEM–EDS), Fourier transform infrared (FTIR) spectroscopy, thermogravimetric analysis (TGA), and nitrogen (N2) adsorption isotherm analysis. In SEM–EDS analysis, a dense and compact microstructure was observed for the BA20 and BA33SF7 mixtures due to the formation of high-density C–S–H and C–H phases. The formation of a large amount of C–S–H phases was observed through FTIR, where a prominent shift in peaks from 955 to 970 cm−1 was observed in the spectra of these mixes. Moreover, in N2 adsorption isotherm analysis, a decrease in the intruded pore volume and an increase in the BET surface area of the paste matrix indicate the densification of the pore structure of these mixes. As observed through TGA, a reduction in the amount of the portlandite phase in these mixes leads to the formation of their more densified micro and pore structures. The current findings indicate that BA (20%) and its blend with SF (40%) represents a potential revenue stream for the development of sustainable and high-performance concretes in the future.

Several studies have revealed that the flexural strength of concrete and chloride diffusion coefficients decreased while the workability improved with the addition of SCBA 26,34,37,38,41 . Batool et al. 34 argue that there is utmost need to investigate the salient characteristics and influence of locally available SCBA, whose content generally ranges between 5 and 30%, to produce durable and green concrete/ecological concrete 10,42 . However, Ganesan et al. 43 and Le et al. 44 revealed that the slump values decreased at higher SCBA contents. The impact of bagasse ash replacement on concrete workability was also investigated by Venkatesh et al. 45 and Mahmud et al. 46 . Their results show that slump values for all levels of SCBA replacement concrete were higher than control concrete mix values. In addition to the isolated effect of SCBA, a variety of additives, such as GGBFS, nanosilica and SF, have been added in conjunction with SCBA to attain better performance. While studying the adverse impact of dust on the concrete properties in asphalt pavement, the compressive strength improvement was in the order SF mixes > SCBA mixes > FA mixes 47 . The addition of 10% SCBA enhanced the tensile strength by approximately 10%, thus saving costs and improving durability. In another study, the compressive strength of SCC increased significantly and showed resistance to sulfate attack with 30% SCBA + 30% GGBFS substituting OPC 44 . The inclusion of SCBA (10-30%) in mixtures blended with nanosilica (3-6%) decreased the compressive strength 41 . For a 10% SCBA + 10% micro silica mix, maximum compressive and flexural strengths and improved resistance to acid attack were achieved after curing 48 . Moreover, 15% SCBA + 10% SF was recommended by Teja et al. 49 to attain an increase in strength (up to 24%). www.nature.com/scientificreports/ Larissa et al. 50 used SCBA and metakaolin (MK) to produce SCC for exposure at temperatures of 200 to 800 °C and revealed that the extent of concrete deterioration at high temperatures was significantly reduced. Additionally, SCBA and MK at contents up to 40% were less susceptible to elevated temperatures, thus producing fewer cracks and reducing strength losses compared to those at room temperature. By incorporating the SCBA-SF content (10 to 30%), it was concluded that the slump, temperature, and volumetric mass decreased with increasing percentages. Additionally, the compressive strengths at 7, 14 and 28 days decreased by 7.6%, 7.2% and 7.1%, respectively. Moreover, the mixture prepared using a 20% SCBA-SF mix had excellent corrosion performance in contrast to that prepared using only OPC 51 . Khan et al. 52 found that by replacing cement by weight with 10% SCBA and 10% micro silica, the blended concrete exhibited better durability and offered optimal resistance to sulphate and chloride attack. The literature suggests that there exists a lack of research focusing on the usage of SCBA at higher percentages (such as 30 to 40%) in combination with highly reactive materials such as SF without compromising the mechanical properties of the resulting mixture. A comprehensive investigation of the aforementioned performance of SCBA-SF in concrete is scarcely available in the existing literature. Therefore, further development is needed to evaluate the characteristics of fresh and hardened concrete modified to generate efficacious SCM blends to be substituted with highly polluting OPC.
Therefore, the impact of BA and its blend with SF, as a partial replacement of cement, on the mechanical, durability and microstructural characteristics of concrete must be examined. In addition, the physical and chemical characteristics of cement, BA, and SF were determined with X-ray fluorescence (XRF) and X-ray diffraction (XRD) methods. Concrete specimens, namely, control specimens (100% cement), three binary mixes with only BA (10, 20, and 30%) and three ternary mixes with BA along with SF (25% BA + 5% SF, 33% BA + 7% SF and 40% BA + 10% SF), as a partial substitute of cement, were selected. In ternary mixes, a large amount of cement was substituted (up to 50%). Mechanical properties, such as the compressive and split tensile strengths with aging (7, 28 and 91 days), and water absorption (WA) and apparent porosity (AP) after aging for 91 days were evaluated for hardened concrete samples. Eventually, to study the formation of hydration products due to BA and SF replacement in the cement matrix, the isolated effect of BA and blended effect achieved by incorporating SF along with BA on the micro and pore structures of the cement mortar samples were explored by SEM-EDS, FTIR spectroscopy, TGA, and the N 2 adsorption isotherm method.

Materials and methods
Materials. In this study, conventional OPC, complying with ASTM C150, commercially available SF and sugarcane BA processed in the laboratory were used as the main binders for the preparation of concrete specimens. The chemical and physical properties of these binder materials are given in Table 1.
In addition to binder materials, locally available coarse and fine aggregates were used as filler materials in concrete. The particle size analysis curves of both fine and coarse aggregates are shown in Fig. 1. The coarse aggregates blending percentages are 50% (20 mm down) and 50% (10 mm down). According to ASTM C33, the specific gravity and water absorption of fine aggregates were 2.60 and 1.03%, respectively, whereas the fineness modulus was calculated to be 2.54 ( Table 2). The specific gravity and water absorption of coarse aggregates were 2.65 and 0.82%, respectively. www.nature.com/scientificreports/ Burning and grinding of bagasse ash. The chemical properties of BA mainly depend on its source, organic composition, and sintering temperature. The chemical composition of bagasse husks also varies from region to region and depends on climatic conditions 53 . The obtained sugarcane bagasse husk was burned in an electric furnace at different controlled temperatures. To achieve silica-rich BA, several combinations of burning duration and sintering temperature were employed. This is because the incineration time and temperature significantly affect the quality of BA [54][55][56] . Therefore, in this study, sugarcane bagasse husks were burned in a heating kiln (Fig. 2a) at different controlled temperatures of 600 °C for 3 h and 6 h and a higher temperature of 800 °C for an hour. Subsequently, XRF and XRD analyses were performed on the powdered samples to study the effect of different burning temperatures on the chemical and mineralogical compositions of BA. The chemical composition of all three BA samples shows that the sum of the SiO 2 , Al 2 O 3 , and Fe 2 O 3 contents is a maximum (70%), corresponding to a burning temperature of 800 °C. The XRD analysis shown in Fig. 3 demonstrates the amorphous nature of BA due to the silica present in the sample. Based on the chemical composition and XRD results, the BA obtained from burning at 800 °C was selected for grinding to obtain a fine powder. Thus, after cooling under normal air, the BA was subjected to grinding in a rotary ball mill at 15 rpm for 12 h (Fig. 2b). This ground BA was then used as a partial substitute for cement to prepare binary (only BA) and ternary concrete specimens (BA and SF).
Concrete mixture proportions. Including the control, a total of seven different concrete mixture proportions were selected. The control contained 100% cement, the three binary mixes contained only BA as a partial substitute of 10, 20, and 30% cement (BA10, BA20, BA30), and the remaining three ternary mixes contained BA and  www.nature.com/scientificreports/ SF. In ternary mixes, a large amount of cement ranging from 30% (25% BA + 5% SF) to 40% (33% BA + 7% SF) and 50% (40% BA + 10% SF) was substituted. Complete details of the ingredients of each mixture and details of the test specimens are given in Table 3. The primary objective of adding SF to ternary mixtures was to maximize the percent cement replaced with BA without compromising other mechanical and durability-related properties. Therefore, to ameliorate the  www.nature.com/scientificreports/ strength of binary concrete, compared to control concrete (CC), SF is added at 5 to 10% to mixes containing a high percentage of BA (25% or more). A constant water-to-binder (w/b) ratio of 0.35 was maintained for all the mixes. Since the total binder content for each mix was 457 kg/m 3 , and therefore according to w/b ratio the water content of all mixes was kept same to 160 kg/m 3 . However, the targeted workability corresponding to slump values of 120 ± 30 mm was achieved by additionally employing various percentages (wt.% of binder) of a naphthalene-based high-range water-reducing admixture ( Table 3).

Methods. Concrete mixing and preparation of specimens.
A power-driven rotating pan mixer was used to mix the concrete ingredients, in accordance with the guidelines given by ASTM C192. Immediately after mixing and testing to achieve the required slump, cylindrical concrete specimens 100 mm in diameter and 200 mm in height were cast according to the standard method specified by ASTM C39. A total of eighteen specimens were cast for each mix to test three identical specimens with respect to aging (7, 28, and 91 days) to determine the compressive and splitting tensile strengths. After fabrication, the cylindrical molds were covered by plastic sheets and stored at the standard laboratory temperature (20 ± 1 °C) and relative humidity (60 ± 5%) for 24 h. All the concrete specimens were demolded upon completion 24 h after casting and moist-cured in a curing tub until testing on 7, 28, and 91 days. Upon reaching the desired curing period, both the upper and the lower ends of the concrete specimens were smoothly levelled by employing an end surface grinder.
Compressive and splitting tensile strength tests. The compression strength test was conducted on the cylindrical specimens as per ASTM C39 by deploying a 200-ton-capacity universal testing machine (UTM) assembly at a loading rate of 0.2 MPa/s. In contrast, the split tensile test was conducted as per ASTM C496 by deploying a 200-ton UTM at a loading rate of 1 MPa/min. Figure 4 shows the test setup used to measure the compressive and tensile strength of concrete specimens.
Water absorption. In addition to compression and tensile tests, WA tests were also performed on hardened concrete samples of all mixes according to ASTM C948. To perform WA tests, concrete cylindrical samples with a diameter of 100 mm and thickness of 50 mm were obtained by cutting 91-day moist-cured concrete specimens. The cut concrete samples were soaked in water at 21 °C and periodically weighed at a constant interval of 24 h until saturated-surface dry (SSD) weight stabilization was achieved. The SSD weight was deemed stabilize when less than a 0.5% difference in weight occurred between two consecutive readings. The latest yielded weight of the samples is referred to as "B". Subsequently, the mass of the specimens suspended in water was measured to the nearest 0.01 g and referred to as "A". Note that the concrete samples were oven-dried at 100-110 °C and that weights were recorded every 24 h. After attaining a weight loss < 0.5% of the last measured weight, the sample was cooled inside a vacuum desiccator at room temperature. The sample weight recorded at room temperature is denoted "C". Finally, the WA and AP can be calculated by using the following equations: Preparation of paste samples for microstructural testing. Paste specimens of the control mix (100% cement) and the mixes with different contents of BA (binary) and BA and SF (ternary), as partial replacements of cement (by weight), were prepared. For all mixes, a paste of standard consistency was obtained by mixing inside a Hobart mixer. Immediately after mixing, the fresh cement pastes were poured into small plastic containers (diameter of 20 mm and height of 50 mm) for casting. Subsequently, the plastic containers were capped and sealed prior to curing. After 91 days in the cured state, the specimens were dried to inhibit the process of hydration with the solvent exchange method. Finally, the flaky slices alongside the powder specimens were subjected to washing www.nature.com/scientificreports/ using isopropanol for 15 min. Afterwards, the paste samples were dried in an oven for 30 min at 40 °C to remove the isopropanol present in the samples, and the specimens were stored in sealed plastic bags prior to testing.

Scanning electron microscopy/energy-dispersive X-ray spectroscopy (SEM-EDS) analysis.
For all the concrete mixes, SEM-EDS analysis was conducted on hardened fragments in the form of slices of the cement paste with a JSM-IT100 scanning electron microscope. Notably, the solvent exchange method was employed to dry the fragments of the hardened paste using isopropanol. Afterwards, the changes in the morphology and composition of all the specimens were studied.
Nitrogen adsorption isotherm technique. Similar to SEM-EDS, a N 2 sorptiometry study was also conducted on the paste samples of all the studied concrete mixes. N 2 sorptiometry was used to characterize the surface area and pore structure of the powdered specimens (weighing approximately 0.3 g) obtained from the 91-day cured hardened paste with a N 2 sorption analyzer (NOVA2200e, Quanta chrome, USA) at 273 K. During the process, first, the specimens were degassed to remove absorbed contaminants from the open air. Consequently, N 2 adsorption of the specimens was performed at ambient temperature and controlled pressure.
Fourier transform infrared (FTIR) analysis. The different phases of a variety of cement pastes were additionally characterized for all the studied mixes by performing FTIR using a Perkin Elmer Spectrum Two FTIR spectrometer. The powdered samples of all the studied mixes were dried and subjected to an infrared light source, and the infrared spectra were recorded in the wavenumber range of 400 to 4000 cm −1 .
Thermogravimetric analysis (TGA) of cement pastes. TGA was also conducted on all the studied concrete mixes. The dried powdered specimens obtained from 91-day cured pastes were placed inside the ceramic vessel of a thermal gravimetric analyzer. To dry the specimens, they were heated inside the thermal gravimetric analyzer to 20 to 1000 °C at a rate of 10 °C/min, with N 2 acting as the medium under static conditions. Additionally, alumina powder was incorporated as a reference to attain stability at elevated temperatures. Finally, the weight loss of various paste specimens in various temperature ranges was depicted by a comparison plot drawn using the built-in software.

Results and discussion
Influence of BA and SF on the evolution of the compressive and tensile strengths of concrete. Table 4 lists the compressive and splitting tensile strengths evolution results of all the tested concretes at 7, 28, and 91 days. These results are expressed at each age in terms of average of three identical specimens ± the corresponding standard deviation values. Figure 5a shows a comparison of the compressive strength results of CC and concrete where cement partially replaced with different percentages of BA and blends of BA and SF. In binary mixes with BA alone, up to 30% cement (10%, 20%, and 30%) was replaced, while in the ternary mixes with BA and SF, the cement substitution percentage was increased up to 50% (30%, 40%, and 50%) due to the highly reactive nature of SF. The purpose of studying ternary mixes of BA and SF was to compare their results to those of the control and binary mixes to www.nature.com/scientificreports/ determine the maximum replacement percentage of cement that does not compromise the strength. The concrete specimens were all water cured at a standard curing temperature of 20 °C until the age of testing.
The results show that the binary concrete containing 20% BA exhibited highest compressive strengths among all mixes, including those of the CC, irrespective of aging. However, the compressive strength of the other two binary mixes containing BA alone was lower than that of CC at all ages. However, the reduction was more significant for the binary mix containing a low percentage of BA (10% BA) than for the binary mix containing a higher percentage of BA (30% BA). The low compressive strength of the binary mix with only 10% BA is due to www.nature.com/scientificreports/ its lower degree of pozzolanic activity than mixes with high percentages of BA. Although a significant reduction in strength was observed for these binary mixes at early ages, an enhancement at later ages, especially at 91 days, was noticeable due to their late pozzolanic reactions. Moreover, the compressive strength of the binary mix containing 30% BA was comparable to that of CC. To achieve high sustainability, efforts were made to regain the reduction in the compressive strength of binary mixes containing high percentages of BA by adding different percentages of SF. The test results show that no improvement in strength occurred in ternary mixes with an equal percentage of cement substitution (25% BA along with 5% SF) to that of the corresponding binary mix (30% BA). However, an encouraging response was noticed for ternary mixes containing high percentages of BA and SF (33% BA along with 7% SF), where the compressive strength was comparable to that of CC at all ages. The above results demonstrated fast early-age hydration and better packing and filling abilities due to the addition of very fine SF along with the later-age pozzolanic reaction of BA. With a further increase in cement substitution (50%), once again, a significant reduction in strength was observed in the ternary mix containing 40% BA and 10% SF. Note that among all the tested binary and ternary mixes, the lowest compressive strength results were observed for this ternary mix, which is attributed to its high percentage of cement substitution. This is because a high percentage of cement substitution affected both the early hydration and later-age pozzolanic reaction due to the production of less Ca(OH) 2 .
Unlike the compressive strength, all binary and ternary concrete mixes had higher splitting tensile strength than CC irrespective of aging (Fig. 5b), except at an early age (7 days) for the ternary mix with 50% cement substitution. Similar to the compressive strength, both the binary (BA20) and ternary (BA33SF7) mixes showed significantly higher splitting tensile strength than CC, irrespective of aging. Among all the mixes, the highest splitting tensile strength was demonstrated by the ternary mix with 40% cement substitution (BA33SF7).
In summary, the results demonstrated that the compressive strength of binary concrete mixes decreased at both low (10%) and high (30%) cement substitution percentages and that the best compressive strength results were achieved when 20% BA was used. Similar to that of binary mixes, the compressive strength of ternary mixes with low (25% BA + 5% SF) and high percentages of cement substitution (40% BA + 10% SF) decreased, and the best strength results were achieved when 33% BA was used along with 7% SF.
Comparison of the water absorption and apparent porosity of control concrete and binary and ternary concrete mixtures. In addition to the strength characteristics of the studied binary and ternary concrete mixtures, other important properties, such as WA and AP, were also measured and compared to those of CC. For the hardened concretes, WA and AP can be used as indirect indicators for the evaluation of their durability. Figures 6 and 7 show the effect of replacing different percentages of cement with BA alone (10, 20, and 30) or BA and SF (30,40, and 50%) on WA and AP. Irrespective of the amount of cement replaced, all the studied binary and ternary concrete mixes exhibited lower WA than CC (Fig. 6). The lower WA of binary and ternary mixes compared to that of CC can be attributed to the binary and ternary mixes having lower AP values (Fig. 7). The current results indicated that the WA of binary mixes decreases with increasing percentage of BA up to a certain level of cement replacement (20%). For instance, a binary concrete mixture containing 30% BA demonstrated a slightly higher WA than that containing 20% BA (Fig. 6). A similar trend of AP values can also be observed in these binary concrete mixtures (Fig. 7). This can be attributed to the slower rate of the pozzolanic reaction of the concrete mixture with a high percent of cement replaced with BA. However, unlike binary concrete mixtures, a remarkable improvement in terms of lowered WA and AP values can be noted in ternary concrete mixtures with a high percent of cement replaced with BA and SF. The results indicated that the ternary mix with 30% (BA25SF5) cement replaced exhibited lower WA and AP than the corresponding mix (BA30) and other binary mixes (BA10 and BA20). However, a slight increase in WA and AP occurs in ternary mixes with increasing percent replacement of cement from 30 to 40 or 40 to 50%. However, the ternary mix with www.nature.com/scientificreports/ an even higher percent replacement of cement, such as 40% (BA33SF7), still exhibited lower WA and AP than the binary mixes with a relatively lower percent replacement of cement (10, 20 or 30%). This can be attributed to significant pore refinement in ternary concrete mixtures due to the very fine size of SF particles compared to BA. The slightly high values of WA and AP in a ternary mixture (BA40SF10) with a very high percent replacement of cement (50%) can be related to its slower rates of pore refinement and pozzolanic reaction. Moreover, the most existing correlations are developed using the 28 days experimental results and for conventional concrete subjected to standard curing (moistcured under 20 °C) using Type-I cement. However, it is worth mentioning that, other than the characteristic or mean compressive strength of concrete, the variation of the exponent b can also be influenced by several other factors such as the type of cement, water to binder ratio, curing temperature, aging and so on 67 . Although some researchers have considered the effect of different influencing factors on the rate of both properties such as Kim et al. 63 had proposed their correlation according to different binder types, curing, and aging. However, the effects of other influencing factors such as seasonal variations, geometry of specimen, type of concrete using SCMs was not considered. Therefore, it is necessary to evaluate the different existing correlations to determine the splitting tensile strength of different concretes produced in this study with different percentage of BA alone (BA10, BA20, BA30) and BA with SF (BA25SF5, BA33SF7, BA40SF10).

Prediction models to evaluate tensile strength based on compressive strength of con-
As shown in Fig. 8, a comparison of experimental compressive and splitting tensile strength relationship was drawn for various concrete specimens (control, binary and ternary mixes) with aging (3, 7, and 28 days), and compared to existing models used to evaluate tensile strength based on compressive strength [57][58][59][60][61][62][63][64][65][66] . Table 5   The above findings suggest that aging and the binder type do not significantly affect the correlation between splitting and compressive strength of concrete. These findings are in line with those noticed by Kim et al. 63 , where the relationship between splitting tensile strength and compressive strength of concrete was reported as independent of aging, cement type, and curing temperature. Consequently, it is proposed that the Noguchi-Tomosawa model 64 , or the model proposed by De Larrard and Malier 66 be safely used to determine splitting tensile strength of all the tested concretes, within an error range of ± 5% (Fig. 9a,b), except for control and BA20 concretes. The JSCE model 65 can be safely used to determine splitting tensile strength for control as well as BA20 concretes within an error range of ± 5% (Fig. 9c). Furthermore, Table 6 presents the summary of experimental results of splitting tensile strength as well as those calculated based on the prediction equations of closely estimating models of Noguchi-Tomosawa 64 , De Larrard and Malier 66 , and JSCE 65 . A ratio of experimental/predicted splitting tensile strengths data is also presented in Table 6. The minimum and maximum values of this ratio ranged between 0.85-1.05, 0.87-1.11, and 0.97-1.20 with an average value of 0.97, 1.0, and 1.10 for Tomosawa, Larrard and Malier, and JSCE model, respectively. The above ranges of ratio and average values of ratio as unity for Larrard and Malier model (1.0), and close to unity for Tomosawa model (0.97) having an average correction factor of 3%, indicate their high prediction accuracy. On the other hand, a relatively low degree of accuracy observed for JSCE model due to a higher average value of ratio (1.10) having an average correction factor of 10%. Though conservative, the JSCE model be safely used to predict splitting tensile strength of all the tested concretes, however, with slight underestimated values within an error range of ± 15% (Fig. 9c). Since the splitting tensile strength is used as the criterion of crack control, the underestimation of the splitting tensile strength using JSCE model remains safe and conservative with respect to current experimental values.  www.nature.com/scientificreports/

SEM-EDS analysis of control (C), binary (C/BA) and ternary (C/BA/SF) cementitious pastes.
The microstructure of all the mixes used in this study was analyzed by performing SEM-EDS analysis on paste samples, as shown in Fig. 10. The EDS analyses were performed on SEM micrographs at different magnification levels for all mixes overall and at specific points to study the effect of BA and SF on the development and www.nature.com/scientificreports/ enhancement of the microstructure of cementitious pastes. The Ca/Si ratio calculated from EDS analysis for all the mixes is given in Table 7. Previous studies performed on cementitious mixes found that the Ca/Si ratio for the C-S-H phase ranges from 0.5 to 2, while a Ca/Si ratio greater than 2 represents C-H phases present in the paste sample [68][69][70] .
The SEM-EDS results show that the control mix has a higher Ca/Si ratio than all the other mixes containing BA or BA and SF. This may be attributed to the better pozzolanic properties and high fineness of both BA and SF, which causes densification of the paste matrix due to the formation of high-density C-S-H and C-H phases. Such results show the development of the microstructure due to incorporation of BA and its blend with SF in the cementitious matrix. The Ca/Si ratio of mixes with different amounts of BA shows that BA20 has the lowest Ca/Si ratio compared to BA10 and BA30. The results demonstrate that the optimum amount of BA used as a substitute for cement is approximately 20% for the development of stronger and more durable concrete. Other mechanical and microstructural testing results also revealed better performance by mixing 20% BA compared to the other two binary mixes containing only BA. Furthermore, the mixes with blends of both BA and SF (BA25SF5, BA33SF7, and BA40SF10) have the lowest Ca/Si ratios compared to all other mixes. Moreover, the blended mix with 40% cement replacement (BA33SF7) showed enhanced microstructure properties compared to the other two mixes with 30% (BA25SF5) and 50% (BA40SF10) cement replacement. In a similar manner to current study, several research studies in past have also proved an improved performance of concrete due to the addition of SF within a range of 5-10% [71][72][73] . The decrease in the Ca/Si ratio is attributed to the high fineness and reactivity of SF, which causes the formation of high-density C-S-H and C-H phases 74,75 . The formation of conventional high-density C-S-H and C-H phases cause densification and refinement of the microstructure, which would ultimately yield better performance when used in practical engineering applications. Figure 11 compares the FTIR analysis results of the 91-day cured paste samples of the control and all the other binary and ternary mixes.

Fourier transform infrared (FTIR) analysis results of cement pastes.
As illustrated in Fig. 11, the peaks between 900 and 1100 cm −1 are attributed to the vibrations of Si-O bonds in the C-S-H phase 76 . Compared to the control samples, the relative intensity of the (Si-O-Si) band is higher in paste samples with only BA or its blend with SF. The shift in the Si-O band toward a high wavenumber is due to the polymerization of silica. A slight shift (965 cm −1 ) is observed in the binary mix sample containing 20% BA. However, a significant shift (970 cm −1 ) is found in the ternary concrete samples with 5% SF (BA25SF5) and 7% SF (BA33SF7). The shift in the spectra of these binary (BA20) and ternary (BA25SF5 and BA33SF7) mixes indicated the formation of a large amount of high-density C-S-H gels. The formation of more C-S-H gels might be the possible reason for the development of the high compressive strength of these mixes. Additionally, the calcite formed due to carbonation are linked to the peaks at 720 cm −1 , 875 cm −1 , and 1415 cm −1 . www.nature.com/scientificreports/  www.nature.com/scientificreports/ The peak at 3645 cm −1 demonstrates the presence of free OH groups, which indicated the presence of the portlandite phase in all the tested paste samples 77 . This peak is more intense and visible in the control sample than in all other samples containing BA or SF. A relatively small peak is noticed for binary mix samples with different amounts of BA, which is reduced further for samples of ternary mixes with increasing percentages of SF. The peak remained very small or almost diminished in ternary mix samples with large amounts of SF (7% or 10%), which demonstrated their high pozzolanic reactivity due to the consumption of Ca(OH) 2 in the paste samples 78 . The formation of gels of high-density C-S-H and the depletion of Ca(OH) 2 due to the addition of BA and SF are also evident from the SEM-EDS and TGA results.
Nitrogen adsorption results (surface area and pore structure of cement pastes). Figure 12 compares the cumulative nitrogen intrusion volumes versus pore widths of all the tested paste samples subjected to 91 days of standard curing. According to the test results, the paste of the ternary mix with 40% cement replacement (BA33SF7) exhibited the least amount of nitrogen volume intrusion (0.050 cm 3 /g), followed by the paste samples of the BA25SF5, BA20, BA30, BA10, Control and BA40SF10 mixes (Table 8). Despite the high percentage replacement of cement, it can be inferred that the lowest amount of nitrogen intruded in ternary mixes (BA33SF7 and BA25SF5) compared to the control or binary mixes, with the exception of the ternary mix with 40% BA and 10% SF. This is due to the densification of the pore structure, which can be attributed to the incorporation of highly reactive SF, which leads to improved pozzolanic reactivity and refinement of the pore  www.nature.com/scientificreports/ structure 79,80 . Further, both BA20 and BA30 mixtures showed less nitrogen volume intrusion compared to the control samples, which could be explained by the densification of pore structures caused by the pozzolanic nature of BA. Among all the studied mixes, the maximum amount of nitrogen volume intruded in BA40SF10 is due to the presence of unreacted BA in the samples, which might have caused the increased porosity of the paste matrix 81,82 . Figure 13 compares the Brunauer-Emmett-Teller (BET) surface areas of control samples and the samples containing BA and blends of BA and SF. The ternary mix with 40% replacement of cement (BA33SF7) exhibited the largest surface area among all samples. According to past studies, an increase in surface area indicates an improvement in the microstructure characteristics of C-S-H gels 83,84 . In addition to the ternary mix, some binary mixes (BA20 or BA30) containing only BA also showed a slightly large surface area than the control sample. However, a decrease in surface area was observed in a binary mix when a very low percentage of BA (BA10) was incorporated. The above findings demonstrate substantial effects of different BA contents on the formation of pore structures.
A significant increase in the surface area of ternary mixes (BA25SF5 and BA33SF7) is attributed to the formation of highly dense C-S-H phases due to the presence of both BA and SF. This was demonstrated earlier through SEM-EDS analysis, which demonstrates a dense cementitious matrix with fewer pores. Among the ternary mixes, an exception also exists where the addition of a large volume of BA along with SF (BA40SF10) leads to a decrease in the surface area. The decrease in the surface area suggests the presence of unreacted BA in the paste matrix, which causes jamming of hydration products and ultimately leads to the formation of large pores and a more porous system 80 .
Thermogravimetric analysis (TGA) results of cement pastes. Thermogravimetric analysis of cement paste samples was performed to evaluate the effect of BA and SF on the amount of Ca(OH) 2 , which was found from the weight loss between 400 and 500 °C 85 . Figure 14 shows a comparison of the thermogravimetric analysis results obtained for the 91-day cured samples of the control and other binary (BA10, BA20, and BA30) and ternary mixes (BA25SF5, BA33SF7, and BA40SF10). Table 9 shows that the amount of Ca(OH) 2 in the control sample (100% cement) is higher than that in all the other tested samples of binary and ternary mixes. The samples with different amounts of BA (10%, 20% and 30%), used as a cement substitute, show a decrease in the amount of Ca(OH) 2 , which is associated with the pozzolanic reactivity of BA reacting with Ca(OH) 2 to form more C-S-H. The least Ca(OH) 2 was recorded www.nature.com/scientificreports/ in ternary specimens with both BA and SF. This is attributed to the higher pozzolanic reactivity of amorphous silica present in SF, which leads to the formation of more C-S-H by utilizing the C-H phase, thus lowering the C-H content in the solution. The normalized C-H content of all the binary and ternary mixes containing BA and SF was less than that of the control sample, as shown in Table 9. This is due to the increased number of nucleation sites and filling effect of mineral admixtures, as discussed in the FTIR and SEM-EDS analyses. In addition, the ternary mixes with both BA and SF show less normalized C-H than all the other samples, which is attributed to the high fineness and reactivity of SF. Interestingly, the ternary mix (BA40SF10) contains slightly more normalized C-H than BA33SF7. The results also indicate that the ternary mix (BA40SF10) consumes less C-H, which may be due to the incorporation of a large amount of BA in the mix; therefore, some part of the BA remains unhydrated.

Conclusions
In this study, researchers have investigated the possibility of using a high-volume BA and its blend with SF as a partial replacement of cement to produce a sustainable and high-performance concrete. In addition to the control (100% cement), several binary mixtures prepared by partially substituting cement with BA (10%, 20%, and 30%) and ternary mixtures prepared by partially substituting cement with BA/SF (25%/5%, 33%/7%, and 40%/10%) were selected. The influence of high-volume cement replacement with BA alone and BA/SF on the hardened mechanical properties of concrete (compressive and tensile strengths, WA, and AP) was assessed and compared to that of the CC. In addition to concrete specimens, several paste samples for each mix were prepared to scientifically understand the influence of BA and SF on the matrix by performing micro and pore structure analyses using SEM-EDS, FTIR, TGA, and N 2 adsorption techniques.
Following is a summary of major findings of this study: The current results demonstrated that the compressive strength of binary concrete mixes decreases at both low (10%) and high (30%) cement substitution percentages and that the best compressive strength results were achieved when 20% BA was used. Similar to that of binary mixes, the compressive strength of ternary mixes with low (25% BA + 5% SF) and high percentages of cement substitution (40% BA + 10% SF) decreased, and the best strength results were achieved when 33% BA was used along with 7% SF. The higher strengths of these binary (20% BA) and ternary (33% BA + 7% SF) mixes are also validated by their lower WA and AP compared to other mixes.  www.nature.com/scientificreports/ A close agreement between experimental correlation of tensile and compressive strength to those of Noguchi-Tomosawa and the model proposed by De Larrard and Malier suggests that these models be used to reasonably estimate the tensile strength of all concrete mixes containing BA or BA with SF, except the control and the one containing 20% BA where use of JSCE model is recommended. Furthermore, the JSCE model can also be safely used to predict the tensile strength of all the concretes with their slightly underestimated values.
FTIR analysis confirmed the densification of the microstructure with the addition of BA alone or BA and SF. A prominent shift in peaks from 955 to 970 cm −1 was observed in the spectra of BA20 and BA33SF7, which indicates the formation of a large amount of C-S-H phases with a low Ca/Si ratio. The intensities of Ca(OH) 2 peaks (3641-3644 cm −1 ) remarkably decreased with the incorporation of both BA and SF due to their high pozzolanic reactivity, which ultimately led to the formation of more C-S-H phases.
N 2 adsorption analysis also showed that the pore structure densifies with the addition of BA alone or BA and SF to the cement paste matrix. A decrease in the intruded pore volume and an increase in the BET surface area of the paste matrix indicates an improvement in the pore structures of BA20, BA25SF5 and BA33SF7. However, an increase in the pore volume and a decrease in the surface area were observed for the high-volume replacement mixes (BA30 and BA40SF10). This might be due to unreacted BA, which ultimately causes high vascularity of the paste matrix.
The formation of high-density C-S-H and C-H phases was also observed in SEM-EDS analysis with the addition of BA alone or BA and SF, which caused densification of the microstructure. The Ca/Si ratios only slightly decreased with the addition of BA alone, while a significant decrease in Ca/Si ratios was found with the combined addition of BA and SF. Thus, the considerable decrease in the Ca/Si ratio with the incorporation of SF leads to the formation of high-density C-S-H and C-H phases, which causes densification and compaction of the microstructure and ultimately leads to better engineering performance.
TGA reveals a reduction in the amount of portlandite phase in the mixes with BA alone or BA and SF compared to that in the control sample. The decrease in the amount of portlandite phases is an indication of the high pozzolanic reactivity of both BA and SF, which utilizes the C-H phase to form more C-S-H phases and leads to the formation of more densified micro and pore structures. www.nature.com/scientificreports/