The original electric arc furnace steel slags (SS for short) were obtained from one steel plant in Shandong Province, China. The EAFs were ground and sieved to less than 200 μm, which is shown in Figure 1A. According to the results of Ca-Looping (CaL) process (Miranda-Pizarro, et al., 2016; Tian, et al., 2014) and fast absorption of CO₂ by steel slag (Dong, 2008; Stolaroff, et al., 2005), the hot-stage carbonation of SS was proposed to carbonate the free calcium oxide in SS. The carbonated EAF slags (CSS for short) were obtained by the raw EAF slags heated at 650°C in a 99.9 vol% of CO₂ stream atmosphere under the atmospheric pressure, which is shown in Figure 1B. The reference cement (RC for short)was the 42.5-grade cement provided by Fushun Cement Co. Ltd. in Liaoning Province. Table 1 lists the chemical composition of SS, CSS, and RC determined by XRF. XRD and TG-DSC analysis of original EAF steel slag and carbonated EAF slag are shown in Figure 2. The results indicated that the most relevant mineralogical phases, including dicalcium silicate (Ca₂SiO₄), the RO phase, srebrodolskite (Ca₂Fe₂O₅), and Brownmillerite (Ca₂(Al,Fe)₂O₅) stayed almost the same after hot-stage carbonation process, while the diffraction peak of hatrurite (Ca₃SiO₅) decreased during hot-stage carbonation, confirming the results of other studies on EAF slag (Wu et al., 2015). The carbonation efficiency (hereinafter referred to as EOC) was widely used to estimate the carbonation performance of various building materials or solid wastes by TG-DSC analysis (Chang and Fang, 2015; Liu and Wang, 2018). Mass change of calcite decomposition of the carbonated EAF slag samples was 1.13%, and the EOCs of the EAF slag samples were 3.45% after hot-stage carbonation process.

Three different carbonation procedures were designed to investigate the effects of CO₂ on the performances of the produced EAF slag binders, as listed in Table 2. As indicated in Table 2, C2C represents the carbonation applied both during the pretreatment of EAF slags and after the demoulding of steel slag paste binders, A2C indicates the carbonation applied only the post-curing treatment of the demoulding paste binders, and A2A refers to the control samples that were pretreated and cured without carbonation. The hot-stage CO₂ pretreatment was applied during the pretreatment of EAF slags (SS) by importing the high purity CO₂ gas into the tube reactor (T = 650°C, 99.9 vol% CO₂ gas stream, and the atmospheric pressure), producing the carbonated EAF slags (CSS). The blend consisted of 80% SS/CSS and 20% RC was obtained by premixing the raw materials in a tumbling mixer. The water, and blend used to fabricate a paste specimen were mixed at a constant water-to-blend ratio of 0.4:1. The specimens were cast as follows: the dry-mixing of the blend were mixed for 3 minutes to ensure uniformity in the dry mix, after which water was added to the mixture, which was then mixed for a further 5 minutes. The paste binders were then cast into a 30 mm × 30 mm × 30 mm mold, prepared for the mineralogical and microstructural characterization. The pastes were demolded after curing in a standard curing chamber (T = 20 ± 2°C, RH ≥ 90%) for 24 h; and then A2A paste specimens transferred back into the curing chamber until 28 days curing age, the A2C paste and C2C paste specimens were placed immediately into a sealed carbonation chamber of 99.9 vol% CO₂ concentration, 0.1 MPa CO₂ pressure, and a temperature of 25 ± 2°C for 24 h, respectively. The pastes were broken into pieces with an iron hammer and immersed in absolute ethyl alcohol for 24 h to stop the further reaction. The pieces for mineralogical analysis, sequential leaching, and surface atomic concentration analysis were grounded into powders with an agate mortar. The pieces for microstructural analysis were dried in a vacuum oven with silica gel at 50°C for 24 h.

XRF analyses of the reference cement and steel slags were conducted on the equipment of ARLAdvant’X Intellipower 3,600. The f-CaO contents of the steel slags were determined using the ethylene glycol method according to the China Industry Standard 4,328-2012. The f-MgO contents of the steel slags were determined using the ammonium chloride-ethylene glycol method (Ma, et al., 2017; Ma, et al., 2020). The carbonation efficiency, formation, and decomposition of CaCO₃ of the steel slag binders were analyzed by TG-DSC analysis (NETZSCH STA 449 F3) under N₂ atmosphere at a heating rate of 10°C/min from 35°C to 1,200°C. The minerals of the obtained samples were identified by Bruker D8 Advance XRD equipment with Cu Kα (λ = 0.15406 nm) ray from 10° to 70°. The Fourier transform infrared spectra (FTIR) were recorded on an infrared spectrometer (Nicolet IS5). The morphologies and microstructure of the binders were observed by scanning electron microscopy (SEM) coupled with energy dispersive spectroscopy (EDS) (Hitachi S-4,800). The pore size distribution were analyzed by a BET analyzer (Micromeritics ASAP2460). The Mastersizer 2,000 laser granularity meter was used to measure the particle size distribution. The surface atomic concentration and chemical states of the obtained samples were determined by a XPS spectrometer (Thermo Fisher K-alpha). The procedure of Tessier sequential extraction was that the sequential leaching characteristics of the metal combined in the paste binders were conducted according to the methods (Tan et al., 1997; Bacon and Davidson, 2008) shown in Table 3. The metal concentrations of the leaching solution were tested by the ICP-OES (inductively coupled plasma optical emission spectrometer, Optima 7000DV).

The fraction of the trace elements (chromium, zinc, vanadium, and titanium) in the sequential leachates is presented in Figure 3. The accumulated active phase fraction of chromium (the sum of exchangeable chromium and chromium combined with carbonates) in the A2A, A2C, and C2C samples was 8.79, 7.62, and 2.06%, respectively. Meanwhile, the accumulated inactive phase fraction of chromium (the sum of chromium combined with Fe/Mn oxides, sulfide compounds, and the residue) in the A2A, A2C, and C2C samples were 91.21, 92.38, and 97.94%, respectively. Cr of the EAF slag paste binders gradually transformed into the inactive phases with the increase of the carbonation efficiency, which was similar to the accumulated leaching results of vanadium, and titanium. In contrast, zinc of the EAF slag paste binders appeared the opposite trend.

The chromium oxides in the EAF slags existed in the form of non-hydrolyzable spinel phases or hydrolyzable silicate phases (Alejandro et al., 2011), and the zinc mineral compounds existed in zinc oxide and zinc ferrite (franklinite, ZnFe₂O₄) (Machado et al., 2006; Chen et al., 2011). The leaching of chromium, and zinc from in the Cr-bearing spinel phases, Mg-bearing spinel phases, and zinc ferrite phase were considered to be limited or even negligible (Zhao et al., 2019). Theoretically, the phases MgCr₂O₄, and ZnFe₂O₄ were carbonated by one static carbonation process. According to the thermodynamic data of the carbonatable mineral phases and the van’t Hoff equation, Δ G θ is the standard Gibbs free energy change, to determine the effect of the reaction factors and to judge the direction of the reaction. Based on the equilibrium equation of CO₂ and the mineral phases, the Δ G θ of the carbonation of MgCr₂O₄ (−22 kJ/mol) were less than that of Ca₂SiO₄ (−116 kJ/mol), CaO (−135 kJ/mol), and MgO (−61 kJ/mol) at 298 K (Wang et al., 2020), and the Δ G θ of the carbonation of ZnFe₂O₄ (−14 kJ/mol) were also less, which indicated that these phases were stable and carbonated just to a less extent. The carbon efficiency was increased to that of C2C, which indicated that some parts of MgCr₂O₄ and ZnFe₂O₄ were involved in the carbonation. The separated Cr₂O₃ and Fe₂O₃ were hydrolyzed as one primary phase during the leaching (Li and Tsai, 1993).

The TG-DSC curves of the uncarbonated area samples from A2A, carbonated area samples from A2C, and C2C after 1 day of curing age are shown in Figure 4. There were weight loss steps ranging from 25°C to 200°C, from 420°C to 460°C, and from 400°C to 750°C in the A2A curve, which corresponded to the removal of bound water and physically absorbed water, the decomposition of Ca(OH)₂, and the decomposition of calcite (Wang et al., 2018). The A2C and C2C samples showed one remarkable weight loss step ranging from 400°C to 950°C, corresponding to the breakdown of amorphous calcite (from 400°C to 700°C) and crystallized calcite (from 700°C to 950°C) (Jeong et al., 2017). The carbonation efficiency (hereinafter referred to as EOC) was widely used to estimate the carbonation performance of various building materials or solid wastes (Chang and Fang, 2015; Liu and Wang, 2018). The EOCs of the A2A, A2C, and C2C samples were 3.3, 52, and 67%, respectively. The maximum carbonation efficiency of the binders was achieved by the simultaneous application of hot-stage CO₂ pretreatment and accelerated carbonation curing routes.

The XRD patterns of the three samples are shown in Figure 5. The A2A-28 days sample mainly consisted of the hydration product of the calcium silicates, such as some amorphous C-S-H gels, that cannot be detected though XRD; all of hydration products are prone to take part in the carbonation reaction and mainly produce carbonation products, such as calcite and monocarbonate (Wang et al., 2019). There were diffraction peaks of calcite were in the XRD patterns when the A2 paste was subjected to the accelerated carbonation curing treatment (Yu and Wang, 2011). The peaks of calcite were more prominent when the steel slag was used instead of the A2 EAF slag mixture under hot-stage CO₂ pretreatment and the paste was put into the accelerated carbonation. These results suggested that Ca-bearing minerals were carbonated, which corresponded to the results of the phenolphthalein indicator testing and thermodynamic calculation (Wang et al., 2019). The calcite existed in the crystalline and amorphous forms, and the content of CaCO₃ cannot be obtained by XRD analysis, which was demonstrated by the TG-DSC study.

XPS analysis results are shown in Figure 6 and Table 4 to demonstrate the chemical state of the elements in the A2A-28d, A2C-1d, and C2C-1d test samples. The Si 2p peaks of the A2A-28d, A2C-1d, and C2C-1d test samples were respectively located at 99.89, 100.95, and 104.43 eV, which indicated that the binding energy of Si 2p increased after the carbonation treatments. The loss of non-bridging oxygens (Si–O–Ca moieties etc.) tended to increase the Si 2p binding energy and related silicate polymerization (Black et al., 2003b). The carbonation reaction increased the consumption of the calcium, leading to a decrease in the Ca/Si ratio, and ultimately forming a Q4 silicate (Black et al., 2007) and calcite (Richardson, 1999). The researchers (Black et al., 2003a; Black et al., 2006; Okada et al., 1998) found that the binding energy of Si had contributed to that of silicate tetrahedral polymerization. The researchers (Black et al., 2004) discovered that there was a strong negative correlation between the Si binding energy and Ca/Si ratio in the Ca-bearing and Si-bearing hydrates phase. The Si 2p peaks of the A2C and C2C samples shifted to higher binding energy, which was slightly broader than that of A2A and consistent with the EOC trend. The broadening of the peak indicated the silicate structure with a higher disorder. There were Ca 2p3/2 and Ca 2p1/2 peaks in the Ca 2p XPS spectra. The changing trend of the Ca binding energy peaks resembled that of Si, and the centre order of the Ca binding energy peak was roughly determined as C2C > A2C > A2A, which suggested that the enhancement of carbonation reaction led to the increase of the Ca binding energy. The C2C sample had much broader Ca 2p binding energy peaks than A2C and A2A, which was due to the differences in the diversity of the products (monocarbonate, calcium carbonates, or Ca and Si-bearing hydrates of the C2C sample). The C 1s peaks of the A2A-28d, A2C-1d, and C2C-1d test samples at lower binding energy were respectively located at 283.10, 282.98, and 286.48 eV, which was assigned to as “adventitious carbon” (Black et al., 2008). The C 1s peaks at higher binding energy (only at 287.88 and 291.68 eV in the A2C-1d and C2C-1d samples) were attributed to carbonate species by combining with the Ca 2p spectra. The carbonation of the A2C-1d and C2C-1d samples was slight, and there was one carbonate peak just about visible in the XPS spectra (Gonzalez-Elope et al., 1990). There were binding energy peaks of Cr²⁺, Cr³⁺, and Cr⁶⁺ in the Cr 2p XPS spectra, and the valence analysis results are listed in Table 4. According to the literature (Wang et al., 2015), the steel slag had the characteristic peak of Cr, such as Cr²⁺, Cr³⁺, and Cr⁶⁺, and the valence states of Cr²⁺ and Cr³⁺ were the main occurrence forms. When pH ≥ 11, most of Cr⁶⁺ exists in the form of CrO₄ ^2–, when pH < 1.2, most of them exist in the form of Cr₂O₇ ^2–; under alkaline conditions, Cr³⁺ mainly exists in the form of Cr(OH)₃, and under acidic conditions, Cr³⁺ exists in the form of ions (Wei et al., 2010). Cr⁶⁺ partly transformed into Cr³⁺ due to the influence of other elements and partial pressure during the treatment (Silva and Monteiro, 2007). The changing trend of the Cr 2p binding energy peaks resembled that of Si 2p after carbonation treatment, the fraction of Cr³⁺ in the C2C sample was far greater than that of the A2A sample, which indicated that carbonate formation, mineralogical changes and the decrease of pH (lower pH) promote the leaching of one oxidation state of Cr or different leaching rates of Cr valences, corresponding to other research results (Wu et al., 2016). XPS is a general semi-quantitative analysis method for analysing the valence states of Cr element, only reflecting the surface atomic characteristics. The valence distribution of Cr can be studied by simplifying the composition of the slag and other analysis methods.

Figure 7 shows the FT-IR spectra of the uncarbonated area in the A2A-28d sample and carbonated area in the A2C-1d and C2C-1d samples. There were asymmetric stretching (υ₃) vibration peaks of [CO₃]²⁻ at 1,424.25, 1,427.5, and 1,448.86 cm⁻¹ in these samples (Chang et al., 2019). The out-of-plane bending (υ₄) peaks of the C-O bond in calcite (Vagenas et al., 2003) were located at 875.14, 873.19, and 873.52 cm⁻¹. The in-plane bending peaks (υ₂) of the C-O bond in the aragonite and calcite (Chang et al., 2016) were centered at 714.01 and 711.6 cm⁻¹. The carbonation promoted the formation of crystalline and amorphous calcite. The asymmetric stretching peaks (υ₃) of the Si-O bond in the silica gel (Ukwattage et al., 2015) were located at 989.94, 1,042.09, and 1,043.34 cm⁻¹, and the out-of-plane bending (υ₄) of the Si-O bond (Moon and Choi, 2019) were detected at 516.10, 568.7, and 576.29 cm⁻¹, which indicated the formation of silica gel during carbonation. The peaks at 3,640–3,645 cm⁻¹ were assigned to the stretching vibration of the O-H bond, while the peaks at 3,420–3,450 cm⁻¹ were contributed to the vibration of the H₂O bond in the hydration products (Lukas, 1976). Compared with the uncarbonated sample of A2A, the bands of the carbonated samples at 870–880 cm⁻¹ decreased, but the bands at 510–580, 700–750, 1,000–1,100, and 1,400–1,500 cm⁻¹ increased, which represented that the calcium silicates gradually reacted with CO₂ to form calcites and silica gels during the carbonation.

Figure 8 shows SEM-EDS images of these samples. According to the A2A and A2C images, the iron-containing substances (such as RO phase and brownmillerite) were involved in the carbonation reaction just to a less extent, and the uncarbonated calcium silicates core surrounded the Ca-leached and Si-rich layers (Quaghebeur et al., 2015). Physical encapsulation of the carbonate, which generated under such a carbonation ratio, reduced the hydrolysis of the primary mineral phases and then immobilized the metal, such as chromium (Spanka et al., 2016).

The total pore volume, and average pore size of the original slag, carbonated slag, A2A-28d, A2C-1d, and C2C-1d steel slag-based binders are shown in Figure 9. By comparing the pore size distribution curves of the samples (as shown in Figure 9), the proportion of 2.5–10 nm mesopores of sample A2C and C2C is higher than that of sample A2A, which would result in higher performance. The average pore sizes of the three samples were 17.44, 7.22, and 6.77 nm, respectively. The pore size was reduced in the blocks, which was caused by the newly formed calcites and silica gels covering the pores of the matrix and precipitates joining together (Pizzol et al., 2014; Muller and Scrivener, 2017). The above findings fully vindicated that both carbonation processes reduced the pore area and porosity of the reaction matrix, thus promoting strength.