The alkali-activated cementitious material was prepared by mixing the solid materials and alkali solution. In terms of the reference sample, the weight ratio between FA and GGBS was 7:3. The solid materials were mixed homogeneously for approximately 1 min before alkali solution was added in the mixture to prepare a paste. FA was replaced by TS with weight ratios ranging from 10 to 70% to investigate its influence on the setting and hardening properties of the paste. The paste was poured into cubic molds with a size of 40 × 40 × 40 mm and cured at the standard condition with temperature of 20±2°C and relative humidity (RH) ≥95%.

The particle size distribution of FA, GGBS, and TS was analyzed using a Mastersizer 2000 laser diffraction analyzer. The specific surface area was measured by Blaine specific surface area tester according to the Chinese Standard GB/T 8074-2008 (GB/T 8074, 2008). The reactivity of these materials was measured following the Standard GB/T 1596-2017 (GB/T 1596, 2017). It was evaluated by the compressive strength ratios of blended cement mortar (70% Portland cement blended with 30% TS) to the Portland cement mortar at 7 and 28 days of curing, respectively. Setting time of the alkali-activated cementitious material was measured at 20±2°C complying with GB/T 1346-2011 (GB/T 1346, 2011). The initial setting time was determined when the penetration depth of the initial setting needle was 36 ± 1 mm. The fluidity of alkali-activated cementitious material was tested according to GB/T 8077-2000 (GB/T 8077, 2000). The paste was poured into a mini-slump cone, and then the cone was lifted to allow the paste to flow freely for 30 s. The diameters of the spread were measured in two perpendicular directions, and the average value was the fluidity. Compressive strength of the hardened paste being cured at 20 ± 2°C with RH ≥95% for 1, 3, and 28 days was measured by a hydraulic universal testing machine following the Standard GB/T 17671-1999 (GB/T 17671, 1999).

The sample at different curing ages was collected and immersed in ethanol for 7 days to stop its hydration. Humid samples were then dried at 45°C in a vacuum for 7 days before the microstructure analysis. The dried samples were ground in an agate bowl by hand, and the ground powders smaller than 0.075 mm were retained for XRD and FTIR testing. The mineral phases of the alkali-activated cementitious materials were analyzed by X-ray powder diffraction (XRD; PANalytical X’Pert Powder). The scanning range of 2θ was from 5 to 70°, and the scanning rate was 2°/min with a step of 0.02°. The functional groups of the hydration products of the alkali-activated cementitious material were characterized by FTIR spectroscopy (Nicolet iS50). SEM (Quattro S) was applied to investigate the morphology of the alkali-activated cementitious material. Before measurement, the dried samples were coated in gold with a thickness of approximately 20 nm to improve the electronic conductivity.

The influence of mechanical grinding on the particle size distribution of TS is shown in Figure 2. Initially, the fineness of the extracted titanium tailings slag was increased with increasing grinding time, followed by an obvious decrease. The minimum particle size of TS was achieved when the grinding time was 15 min. The particle sizes of D₁₀, D₂₅, D₅₀, D₇₅, and D₉₀ were 23.69, 31.91, 48.31, 69.16, and 82.89 μm, respectively.

The reactivity index of the TS in Figure 3 showed a similar tendency that the value increased first and then decreased with increasing grinding time. When the grinding time was 15 min, the reactivity index was greatly improved and the values of 7- and 28-day reactivity index were 42.60 and 81.18%, respectively. However, when the grinding time increased to 60 min, the values of 7- and 28-day reactivity index decreased to 34.65 and 63.75%. The initial increase in reactivity can be attributed to the mechanical grinding of refined particles and optimization of the particle gradation in cement paste. In addition, the glassy structure was destroyed and the surface area of TS particles involved in the chemical reaction increased, which is beneficial to silicon and aluminum dissolving and ion penetration in the cementitious system. In addition, the particle lattice was distorted with breakage of the Si–O and Al–O bonds in the network, the structure became disordered, and the number of charged particles increased, and as a consequence, the reactivity increased. The decrease in the reactivity when the grinding time exceeded 15 min was probably due to the aggregation of fine particles into coarser size by further grinding. With respect to the energy consumption during grinding, therefore, the grinding time of 15 min was applied on TS for the preparation of alkali-activated binder in this study.

It can be seen from Figure 4 that TS played an important role in changing the fluidity of the binder. The initial fluidity (L0 in the figure) of the reference sample (without TS) was 265 mm, and it increased by 5.7% to 280 mm as the content of TS in the alkali-activated materials increased to 40%. An apparent decrease of the initial fluidity was observed when the TS content increased from 40 to 70%. In terms of the fluidity at 30 min (L30 in the figure), a slight increase occurred when TS content increased to 10%, followed by a subtle decrease with increasing TS content to 40%. The value of fluidity at 30 min (without TS) was 270 mm, and it decreased by 5.5% to 255 mm at the TS content of 40%. The substantial decrease appeared as the TS content increased from 40 to 70%. The increase in the fluidity at a relatively lower content of TS was likely due to the smaller specific surface area of TS than FA, which reduced the water demand of the mixture. In this case, the amount of free water and the thickness of water film on the particles increased, and thereby the fluidity of the mixture was improved (Xiang et al., 2019). Compared with that of the reference sample, the alkali-activated materials with 60% TS had the lowest values of the initial fluidity and fluidity at 30 min, which was decreased by 11 and 48% to 240 and 140 mm, respectively. A probable reason was that the increased amount of Ca²⁺ in the TS accelerated the reaction process, and the gelation of the binder was accelerated to form calcium silicate hydrate gel (C-S-H) and calcium aluminosilicate hydrate gel (C-A-S-H) (Yip et al., 2005; Rakngan et al., 2018; Song et al., 2019).

Figure 5 shows the influence of TS content on initial setting time of alkali-activated ternary binders. It can be seen that adding TS can shorten initial setting time evidently. Generally, the initial setting times decreased with the increment of TS content. The setting time of alkali-activated ternary binder decreased sharply as the content of TS increased to 40%. The initial setting time was 185 min without TS, and it decreased by 54% to 85 min at the TS-to-FA replacement ratio of 40%. When the TS content in the binder was higher than 40%, a slight diminishment was observed in initial setting time, and the value was 61 min at 70% TS. That is because the setting time is highly related to the amount of reaction products, such as C-A-S-H or C-S-H gels (Chindaprasirt et al., 2011; Cui et al., 2017). Increasing the amount of TS was equivalent to increase the content of calcium oxide in the ternary system, thus accelerating the formation of C-S-H and C-A-S-H gel, and subsequently shortening the setting time (Nath and Sarker 2014; Nath et al., 2015; Nishanth and Patil 2021).

The effect of TS on the compressive strength of the alkali-activated ternary binder is shown in Figure 6. It can be seen that the replacement of FA by TS generally improved the compressive strength of alkali-activated material, and the compressive strength increased first and then decreased with increasing TS content. The maximum compressive strength of the binder at early ages was achieved as TS content is 40%. Its compressive strength at 1, 3, and 28 days was 18.9, 34.5, and 49.3 MPa, respectively, which was increased by 80.77, 30.72, and 21.18%, respectively, as compared with reference samples without TS addition. This mixture was deemed to have the optimum particle gradation that forms a more compacted structure than the other mixtures, which contributes to the gaining of compressive strength. Apart of the sodium aluminosilicate hydrate gel (N-A-S-H) formed from the reaction between FA and alkali solution, the replacement of FA by TS in the binder system is likely to incorporate more Ca²⁺ ions to produce C-(A)-S-H gel. The combination of C-(A)-S-H and N-A-S-H gels tends to the optimization of pore structure, and resulted in the growth of compressive strength. However, excessive amount of TS was negative to the compressive strength of binder, which is mainly because the reaction between solid materials and alkali solution was accelerated, and subsequent superfluous C-(A)-S-H gel formed in a relatively shorter time. It is likely to initiate cracks in the specimens as a result of the easy loss of gel water, which adversely affected the strength development (Cui et al., 2017).

The XRD patterns of specimens at different ages are shown in Figure 7. It can be seen that the main reaction products were C-(A)-S-H gel with diffused peak ranging from 25 to 35° 2θ and gismondine (d-spacings of 0.427, 0.334, and 0.299 nm) (Chindaprasirt et al., 2012). Mullite (d-spacings of 0.534, 0.287, and 0.269 nm) and quartz (d-spacings of 0.537, 0.342, and 0.338 nm) derived from the raw materials were still present in all the samples at different ages. In addition, calcite (d-spacings of 0.386, 0.303, and 0.283 nm) was also detected, which was probably due to the carbonation of C-(A)-S-H gel. The peak intensity of mullite and quartz decreased with the increase of curing age. Meanwhile, the peak intensity of reaction products (e.g., C-(A)-S-H and gismondine) were stronger and sharper, suggesting the formation of new phase due to the occurrence of reaction between the raw materials and alkali solution. Furthermore, it is found that the peak intensity of mullite and quartz of the samples with addition of TS in Figure 7B became weaker than that of reference sample (Figure 7A), which is ascribed to the decreased amount of FA, while the peak intensity of C-(A)-S-H in the former sample were stronger and sharper. It indicates that the replacement of FA by TS can promote the formation of C-(A)-S-H gels, and reaction process proceeded with the increase of curing age.

Figure 8 shows the FTIR spectrum of the alkali-activated binder with different compositions. The bands at the region of 1,600–4,000 cm⁻¹ were an indication of hydrogen stretching vibration groups (Si–O–H). The vibration at 3,445 and 1,647 cm⁻¹ was attributed to the bending of the hydroxyl band. These bands indicated the presence of bound water molecules from the reaction products. For the reference sample (Figure 8A), the increased vibration intensity in the region of 2,000–4,000 cm⁻¹ indicated the increased amount of C-(A)-S-H over the curing age. It was low at 1 and 3 days, but increased dramatically at 28 days, indicating a relatively low reaction rate at early ages, and the reaction improved at a later age, whereas a slight increase of intensity in the hydrogen bond vibration was observed with TS addition. The bands at the region of 1,400–1,500 cm⁻¹ were the signal of asymmetric stretching vibration of CO₃ ²⁻. In addition, the band position at 872 cm⁻¹ is due to the bending vibration of the C–O bond of carbonates. The bands positioned at 700–950 cm⁻¹ and 400–500 cm⁻¹ were the signals of [AlO₄]⁵⁻ tetrahedron and [AlO₆]⁹⁻ octahedron, respectively.

The most important bands positioned in the range of 750–1,400 cm⁻¹ are related to the stretching vibration of Si–O–T (where T = Si, Al, Ti, and Na etc.) (Liu et al., 2020), as illustrated in Figure 9. The band at 950–970 cm⁻¹ was ascribed to the Q² units of Si–O–Si (Lodeiro et al., 2009). The vibration at about 1,160 cm⁻¹ derived from quartz and mullite disappeared at 3 and 28 days, indicating the consumption of raw materials by chemical activation. A weak band appeared at 1,374 cm⁻¹ in the samples cured for 1 day, which was ascribed to Q⁴ units of Si–O–Si derived from FA. This band shifted to a lower wavenumber (1,340 cm⁻¹) when the samples were cured for 28 days, implying a lower polymerization degree with curing age forward. This phenomenon became more visible as shown in Figure 9B as adding TS to the alkali-activated binder system, suggesting the alteration of the structure of reaction products. Q² units appeared at 977 cm⁻¹ at 28 days of curing for the sample without TS, but it shifted to a higher wavenumber (994 cm⁻¹), indicating the higher degree of geopolymerization of the structure. From the aforementioned, it can be concluded that the substitution of TS to FA caused the promotion of the reaction of binder (Lee and van Deventer 2002).

Figure 10 exhibits the morphology of alkali-activated binder at 28 days of curing. For the reference sample without TS, the products with petal-like shape with the size of 2 μm were observed, and these particles were loosely packing (Figure 10A). When TS was added, plate-like structure was developed, and some unreacted fly ash particles were also observed (Figure 10B). It indicates that the consumption of TS particles was faster than that of FA particles as the binder was activated by alkali solution, providing more reaction products with a denser structure, therefore leading to a higher compressive strength of binder. However, several micro-cracks with approximately 2 μm in width were observed in the sample containing TS. This may be ascribed to the fact that the addition of TS accelerated the reaction process of binder and promoted the formation of micro-cracks. Furthermore, the presence of TS produced a higher amount of C-(A)-S-H gel, which contained more adsorbed water molecules to maintain its shape. The drying shrinkage of the binder occurred after the removal of water molecules during the drying process, and it caused the formation of micro-cracks.