The primary raw material used was a blast furnace slag with a specific gravity of 2900 kg/m³ and Blaine fineness 399 m²/kg. The particle size range, determined through laser diffraction, was 0.1–74 μm, with an average particle size, d₅₀, of 15 μm. The chemical composition of this slag is shown in Table 1.

The MK used was produced in the laboratory by calcination of a kaolin containing minor quartz and dickite impurities, at 700°C in air, for 2 h. The particle size range of the MK was 1.8–100 μm, with a d₅₀ of 13.2 μm and 10% of particles finer than 4 μm. Alkaline activating solutions were formulated by blending a commercial sodium silicate solution with 32.4 wt.% SiO₂, 13.5 wt.% Na₂O, and 54.1 wt.% H₂O, together with 50 wt.% NaOH solution, to reach the specified activator compositions.

Pastes with different GBFS/(GBFS + MK) ratios and with a constant water/solids ratio of 0.23 (where “solids” is defined as GBFS + MK + anhydrous activator) were produced using the activation conditions described in Table 2. The specimens were cast in a cylindrical mold and stored in hermetically sealed containers with a relative humidity of 90% and a temperature of 25 ± 2°C for 28 days.

Pastes were then crushed and sieved to a particle size of 0.074 mm, and exposed to CO₂ using an accelerated carbonation chamber with a CO₂ concentration of 1.0 ± 0.2%, a temperature of 20 ± 2°C, and RH = 65 ± 5%. These accelerated carbonation conditions were selected based on a previous study (Bernal et al., 2012b), where it was demonstrated that comparable carbonation reaction products to those forming after several years of atmospheric exposure are identified in samples carbonated under these conditions. Non-carbonated samples, and samples exposed to CO₂ for 540 h (referred to as carbonated throughout the document) were analyzed through:

X-ray diffraction was carried out using a high-resolution high-throughput synchrotron powder X-ray diffractometer with a system of 12 simultaneous analyzer/detectors, at beamline 11-BM at the Advanced Photon Source (APS), Argonne National Laboratory (USA). The tests were conducted according to the standard protocols for automated data acquisition on the 11-BM beamline. The conditions for the data collection were: continuous scanning of a detector covering an angular range of 20° with a step rate of 0.01°/s, a step size of 0.001° and wavelengths of λ = 0.401738 Å for non-carbonated samples, and 0.458879 Å for carbonated samples. Data have been converted to Q units (Q = 4π sinθ/λ) for presentation here, to enable comparison between data sets collected at different wavelengths.

High-resolution X-ray diffractograms of non-carbonated pastes of the compositions studied here, as a function of the curing time, were previously reported by Bernal et al. (2013b). As the main partially crystalline reaction products, tobermorite-type phases with three different basal spacings (9, 11, and 14 Å) were observed in these cements, most likely with some Al substitution (Myers et al., 2013). A sodium calcium silicate hydrate (Na₂Ca₂Si₂O₇·H₂O; Powder Diffraction File (PDF) # 00-022-0891), and two aluminosilicate zeolite products, gismondine (CaAl₂Si₂O₈·4H₂O; PDF # 00-020-0452), and garronite (NaCa_2.5(Si₁₀Al₆)O₃₂·13H₂O; PDF # 99-000-1300), the last of which was solely identified in specimens containing MK. Variations in the intensities of the reflections of some of these phases were observed as a function of the MK content, which is discussed in detail by Bernal et al. (2013b). Other crystalline phases, including gehlenite (Ca₂Al₂SiO₇, PDF # 00-035-0755), quartz (SiO₂) (PDF #01-078-2315), and calcite (CaCO₃) (PDF# 01-072-1214), attributed to the unreacted slag, were also observed.

Upon CO₂ exposure, samples with GBFS/(GBFS + MK) = 1.0 (Figure 1A) showed the formation of the sodium carbonate trona (Na₂CO₃·NaHCO₃·2H₂O) (PDF #00-029-1447), associated with the carbonation of the pore solution present in the sample (Bernal et al., 2012b), along with minor traces of the sodium-calcium carbonate gaylussite (Na₂Ca(CO₃)₂·5H₂O, PDF# 00-012-0255). This double salt has been identified as a carbonation product in some alkali-activated slag cements (Bernal et al., 2012b, 2013a), although was not observed in a study of carbonated samples produced with a similar slag to that used in this study, but activated with a lower activator dose (Bernal et al., 2010). A notable increase in the intensity of the reflections assigned to calcite was observed, along with reduction in the intensities of the reflections assigned to the Al-substituted C-S-H and C-(N-)S-H type phases. This demonstrates that the exposure of these cements to high concentrations of CO₂ is leading to decalcification of the main binding reaction products, toward formation of calcium carbonates, as has been reported for Ca-rich alkali-activated cements (Palacios and Puertas, 2006). Changes in the reflection assigned to the zeolites gismondine and garronite were not observed upon CO₂ exposure. The fact that traces of C-S-H type phases are still identifiable in the diffraction data for the carbonated samples indicates that the pastes are not yet fully carbonated after 540 h of CO₂ exposure.

Similar results to those observed in samples solely based on slag (Figure 1A) were identified in carbonated samples with GBFS/(GBFS + MK) = 0.9 (Figure 1B). This suggests that the substitution of slag by MK in a low quantity (≥10 wt.%) does not induce notable changes in the phase assemblages of these alkali-activated cements. Conversely, in samples with GBFS/(GBFS + MK) = 0.8, neither trona nor gaylussite were identified as carbonation products, although these pastes were produced with the highest activator Na₂O concentration among the cements studied (Table 2). This might be associated with the higher degree of consumption of alkalis via the activation reaction of MK, reducing the amount of free alkalis in the pore solution of these cements, and consequently supressing the formation of Na-rich carbonates.

In non-carbonated pastes (Figure 2A), the main weight loss is identified below 250°C, and is more notable in pastes with higher MK contents. The weight loss in this temperature range is attributed to the evaporation of loosely bound water in the system, and dehydration of the substituted tobermorite-type (C-(N-)A-S-H) gel (Alarcon-Ruíz et al., 2005) present in these pastes. The temperature at which the weight loss is most rapid (i.e., the minimum in the differential thermograms plotted in Figure 2) is similar for the three samples at 62°C, in good agreement with results previously identified for alkali-activated GBFS/MK systems with high slag content, where the structure of the binders is dominated by the activation of the slag (Buchwald et al., 2007; Bernal et al., 2011a). The fact that more weight is lost by samples with GBFS/(GBFS + MK) = 0.8 might indicate that water is more loosely bound in the reaction products forming in this paste. The small peak observed at 107°C in the data for samples with GBFS/(GBFS + MK) = 1.0 and 0.9 is attributed to the dehydration of the zeolite gismondine (Bernal et al., 2011a), identified as one of the main crystalline reaction products in activated GBFS/MK under the activation conditions used in this study (Bernal et al., 2013b). Gismondine was not observed as a reaction product in pastes with 20 wt.% MK (Bernal et al., 2013b), consistent with the thermogravimetry results presented here.

The peaks between 400°C and 418°C cannot be assigned to any crystalline phase present in these systems. The recent results of L’Hôpital et al. (2016) suggest that a weight loss in this temperature range most likely corresponds to dehydration of alkali-containing C-A-S-H type gels, such as those forming in the pastes assessed here. The peak at 640°C is assigned to calcium carbonate present in the anhydrous GBFS as a result of weathering prior to sample synthesis (Bernal et al., 2010). Negligible weight changes were observed above 700°C.

In carbonated samples (Figure 2B), an initial weight loss band centered at 52°C is identified in all of the samples. Carbonation of alkali-activated slags leads to decalcification of the C-(N-)A-S-H type gels forming in these systems, resulting in the formation of an aluminosilicate hydrate type gel (Bernal et al., 2013a, 2014a), whose water is more loosely bound than in the original C-(N-)A-S-H. This might be inducing the shift toward lower temperatures of this weight loss peak compared to the uncarbonated samples. In specimens with GBFS/(GBFS + MK) ratios of 1.0 and 0.9 a peak at 112°C was observed, and assigned to trona, which is one of the main carbonate product forming in these binders as identified by XRD (Figure 1). Barall and Rogers (1966) identified that the thermal decomposition of trona takes place in two stages, as shown in Eqs 1 and 2. The first decomposition stage leads to weight losses at ~110°C, while decomposition of the sodium bicarbonate leads to weight losses at slightly higher temperatures (~123°C).

A peak centered at 168°C is also identified in pastes with a GBFS/(GBFS + MK) of 1.0 and 0.9, whose intensity reduced with the inclusion of MK in the binders. A less resolved peak has been identified at a similar temperatures in accelerated-carbonated alkali-activated slag cements (Bernal et al., 2013a), and assigned to decomposition of hydrous sodium–calcium carbonates [e.g., gaylussite or pirssonite (Johnson and Robb, 1973)]. In the samples assessed, traces of gaylussite were identified by high-resolution XRD (Figure 1) and, therefore, this weight loss is assigned to decomposition of this phase. In samples with GBFS/(GBFS + MK) = 0.8, peaks associated with decomposition of the sodium carbonate products were not observed, in accordance with the high-resolution XRD results (Figure 1).

In non-carbonated samples solely based on slag (Figure 3A), a band at 1645cm⁻¹ is observed, assigned to the bending vibration mode of the H-OH bonds in the reaction products forming. The band centered at 1414 cm⁻¹ is attributed to the asymmetric stretching mode of the O-C-O bonds of CO32− ions of the calcite present in the unreacted slag. The asymmetric stretching resonance of the Si-O-T bonds (where T can be Si or Al) was observed at 968 cm⁻¹, along with a shoulder at 866 cm⁻¹ assigned to the vibration mode of the Al-O bonds of the AlO4− groups in the reaction products. The Si-O-T band position is characteristic of the environments of silicon tetrahedra (SiO₄) in the chain structure of C-S-H type gels (García Lodeiro et al., 2009). The band centered at 714 cm⁻¹ corresponds to the bending vibration mode of the Al-O-Si bonds, and the band at 648 cm⁻¹ is attributed to the symmetric vibration mode of the Si-O-Si or Al-O-Si bonds. This band corresponds to C-A-S-H type phases, and the zeolites gismondine and garronite forming in these systems, as identified by XRD (Figure 1). This band has also been observed in unreacted slag (Bernal et al., 2011b), and has been attributed to gehlenite. All infrared band assignments follow references (Huang and Kerr, 1960; Farmer, 1974; Gadsden, 1975; Sitarz et al., 2000a,b).

The addition of 10 wt.% MK [i.e., GBFS/(GBFS + MK) = 0.9] led to a reduction in intensity of the O-C-O band centered at 1418 cm⁻¹, consistent with the lower content of slag in these cements. In the region of the Si-O-T vibration modes, two overlapping bands at 1026 and 957 cm⁻¹ were identified. A higher wavenumber for the asymmetric stretching Si-O-T bonds is related to increased polymerization degree of the C-S-H type gels (Yu et al., 1999), but this effect is counteracted when the Al supplied by MK is incorporated into this gel, which shifts the Si-O-T band toward lower wavenumbers (Bernal et al., 2011b). Myers et al. (2015) identified highly polymerized silicate sites Q⁴(3Al) and Q⁴(4Al) in alkali-activated slags via high-resolution NMR spectroscopy, indicating the formation of a disordered aluminosilicate product resembling an alkali aluminosilicate (hydrate) (N–A–S(–H)) type gel, typically observed in MK geopolymers (Williams et al., 2011), along with the C-(N-)A-S-H type gels. Therefore, the two peaks identified in the Si-O-T region of the slag/MK cement spectrum (Figure 3B) cannot be assigned to a specific phase, but most likely correspond to the highly crosslinked C-(N-)A-S-H type gels typically identified in alkali-activated slag cements (Myers et al., 2015), and potentially to a secondary N-A-S-H type gel derived from the activation of both slag and MK, and structurally related to the zeolites present in the samples. The shoulder at 887 cm⁻¹ along with a peak at 706 cm⁻¹ are assigned to the Al-O and Al-O-Si bonds that can be present in C-(N-)A-S-H and/or N-A-S-H type gels.

The addition of 20 wt.% MK [i.e., GBFS/(GBFS/MK) = 0.8], Figure 3C, also led to the shifting of the T-O-T band, which is more symmetric than in samples with lower MK content and observed at 989 cm⁻¹, along with the formation of a well-resolved peak at 875 cm⁻¹ assigned to calcite. The fact that a sole band is identified in this spectrum might suggest that there are less differences either in chemistry and/or structure (e.g., degree of crosslinking) of the C-(N-)A-S-H and N-A-S-H type gels forming in these cements as the MK content increases.

Upon accelerated carbonation (Figure 3A), in cements solely based on slag, a shoulder centered at 1475 cm⁻¹ is identified along with the asymmetric stretching vibration band of the CO32− groups at 1414 cm⁻¹. This secondary carbonate band is assigned to trona (Huang and Kerr, 1960), which was identified by XRD (Figure 1). Between 1100 and 900 cm⁻¹, formation of a secondary band centered at 1130 cm⁻¹ is observed, along with a slight sift of the T-O-T band from 1014 to 1003 cm⁻¹. Accelerated carbonation of alkali-activated slags cements leads to decalcification of the C-A-S-H type gels (Palacios and Puertas, 2006), and higher wavenumbers can be associated with a high degree of polymerization of silicates. This is consistent with the formation of aluminosilicate hydrate type gels, which present bands at similar wavenumbers (García-Lodeiro et al., 2008) to those identified in the carbonated pastes. The sharp peak identified at 871 cm⁻¹ is assigned to calcite, also identified by XRD (Figure 1). A slight increase in the intensity of the band at 714 cm⁻¹, associated with the bending vibration mode of the Al-O-Si bonds, was also observed. This suggests an increase in the fraction of Al-O-Si bonds within this paste, consistent with the formation of N-A-S-H type gel in this cement upon CO₂ exposure (Bernal et al., 2013a).

In carbonated pastes containing 10 wt.% MK (Figure 3B), similar bands to those identified in the formulation solely based on slag were observed. In this case, the T-O-T band shifted from 946 cm⁻¹ in non-carbonated samples to 1008 cm⁻¹ in carbonated specimens, along with the formation of a distinct shoulder at 1140 cm⁻¹. Upon accelerated carbonation of Ca-free alkali-activated cements, changes in the structure of the N-A-S-H type gel have not been identified (Bernal et al., 2013a); therefore, the changes in the position of the T-O-T band are assigned to the decalcification of the C-(N-)A-S-H type gel. The two distinct peaks correspond to N-A-S-H type gels (García-Lodeiro et al., 2008) and the vibration mode identified at 642 cm⁻¹ is assigned to vibrations of rings present in this type of gel (Mozgawa et al., 2002).

In carbonated pastes with a GBFS/(GBFS + MK) ratio of 0.8 (Figure 3C), the band at 1475 cm⁻¹ assigned to trona was not identified, consistent with the XRD (Figure 1C) and TGA (Figure 2B) results. Similar to the pastes with 10 wt.% MK (Figure 3B), two bands at 1001 and 1115 cm⁻¹ are observed in the carbonated samples, consistent with the formation of N-A-S-H type gels. A slight increase in the intensity of the peak at 875 cm⁻¹ assigned to calcite is observed in the carbonated sample.

Non-carbonated samples based solely on slag present a highly dense, smooth, and continuous microstructure (Figure 4A), consistent with the results typically reported for these materials (San Nicolas et al., 2014). Similar features are identified in samples with 10 wt.% MK (Figure 4B), although the surface of this cement seems rougher than that of the slag-only binders.

Carbonation induces notable changes in the microstructures of the binders assessed. In pastes solely based on slag (Figures 5A1,A2), the continuity of the matrix is lost upon accelerated carbonation, so that highly fragmented laminar type structures are identified throughout the sample. A clear separation between each laminar type structure is observed, which might induce an increase of the microporosity in carbonated pastes, favoring the ingress of aggressive agents through the fragmented matrix. In pastes with 10 wt.% MK (Figures 5B1,B2) the formation of laminar type structures throughout the matrix, along with formation of large and continuous particles, is observed. The laminar structures seem more continuous than those observed in cements without MK.

The most notable changes are identified in the carbonated pastes with 20 wt.% (Figures 5C1,C2). This binder shows a continuous and highly porous microstructure, with a mesh type structure, where formation of distinctive crystalline phases was not identified.