This study considers pastes obtained by alkali-activation of fly ash (FA) and blast-furnace slag (BFS), see Table 1 for the chemical compositions. Their particle size distribution can be found on Supplementary Figure S1 and Supplementary Figure S2. Two activators are considered: a sodium silicate solution (Na₂O + SiO₂) exhibiting a molarity of 10M with 18% (in mass) Na₂O, 28.5% SiO₂ and 53.5% H₂O, and a sodium hydroxide solution 8M (NaOH). The density of the sodium silicate solution is 1.37 g/ml. The two alkaline solutions were mixed together beforehand. The silica modulus, defined as the mass ratio between SiO₂ and Na₂O (M_s = SiO₂/Na₂O), varies between 1.04 and 1.58.

Eight different alkali-activated slag-fly ash (AASF) pastes were prepared according to the European Standard EN 196-1:2016 (CEN, 2016). X. Dai et al. (Dai et al., 2020) reported that an increase in the fly ash content extended the setting times while an increase in the slag increases and speeds up the structural build-up. To benefit from both advantages, each composition contains 50% (in mass) slag and 50% fly ash. The solution-to-binder ratio (S/B) and silica modulus (M_s) differ, see Table 2 (where W/B is the water-to-binder ratio). The reference mix is labeled SB0.55Ms1.44, implying a solution-to-binder ratio S/B of 0.55 and a silica modulus M_s of 1.44. Firstly, four SB compositions (SB0.47, SB0.50, SB0.64, and SB0.70) with varying S/B ratios but reference M_s = 1.44 (equal to the reference composition) are studied. S/B ratios from 0.47 to 0.70 are studied because it is well-known that S/B is a crucial mix parameter that changes the mechanical properties and shrinkage of AAMs (Nedeljković et al., 2018). Secondly, three Ms compositions (Ms1.04, Ms1.24, Ms1.58) with varying M_s but reference S/B = 0.55 (equal to the reference composition) ratio are tested. All the tests are performed in sealed conditions at a curing temperature of 20°C. The results for those different compositions are compared to the ones of an ordinary Portland cement (CEM I 52.5) paste with a water-to-binder ratio (W/C) of 0.5 to have a familiar composition (Delsaute, 2016).

Figure 2A shows the measured compressive strength of AASF pastes with different S/B respectively, after 2 and 7–days. Very remarkably, there is no clear trend of the measured strength with respect to S/B. The largest 7-day strength amounts to roughly 66 MPa and is obtained for the composition with S/B = 0.50. The evolution of the compressive strength between 2 and 7 days for the different compositions is surprisingly different. For SB0.55Ms1.44, SB0.70, and SB0.47, the compressive strength at 7 days is roughly twice as much as the value at 2 days. For SB0.64, the corresponding strength increase factor amounts to four, while for SB0.50 the factor amounts to three. The observed influence of S/B can be explained by two counteracting phenomena: the quantity of water and the quantity of activators, which both increase with increasing S/B. Increasing the amount of activators has a positive effect on the strength because they promote the formation of strength-giving C-N-A-S-H gel. Water, in turn, has a negative effect on the strength because more water will induce a larger initial volume of pores resulting in a lower strength, a well known fact from modeling and experimental studies on AAM and OPC mixes (Taplin, 1959; Feret, 1892; Königsberger et al., 2018a). If S/B is small, the positive effect of the alkali content results in an increase in strength. The positive effect of increasing the alkali content, which is a result of increasing the S/B ratio, outweighs the negative effect of the increased porosity, which is another result of increasing S/B. When the S/B ratio becomes larger than 0.50, the latter effect becomes dominant. (Wang et al., 1994).

Regarding the influence of the M_s modulus, depicted in Figure 2B, the test results highlight that the compressive strength at 7 days decreases when the M_s modulus is increased, which is in accordance with results of Hojati and Radlińska (2017). The authors highlight that a higher M_s modulus implies a lower pH. This results in a rapid initial dissolution and formation of reaction products on the surface of slag and fly ash particles, hindering further reaction and preventing efficient strength development (Hojati and Radlińska, 2017).

The results are compared to an OPC paste with W/B of 0.5. At 2 days, the OPC paste outperforms all the AASF pastes. However, at 7 days, AASF compositions with S/B ≤ 0.55 already exhibit a higher compressive strength. This shows that such pastes continue to react significantly after 2 days as already observed by many authors previously (see e.g. (Feret, 1892; Cartwright et al., 2014; Lee et al., 2014; Hojati and Radlińska, 2017)).

Figures 3A,B, and Figures 3C,D, present the results of heat flow and cumulative heat release as a function of the equivalent age at 20°C, respectively. Concerning the heat flow, two peaks are observed. The initial one corresponds to the early dissolution of the FA and BFS and occurs already during the mixing and is thus not recorded entirely (Fang et al., 2018; Prabahar et al., 2021; Vafaei et al., 2020). The second one corresponds to the acceleration of the formation of the C-N-A-S-H gel (Fang et al., 2018; Prabahar et al., 2021; Vafaei et al., 2020), the period in between is called “induction period” (Fang et al., 2018). Interestingly, the second peak occurs later (the induction period lasts longer) and is more pronounced for pastes with higher S/B and higher M_s. The cumulative heat release (Figure 3C,D) is, in general, relatively close for all pastes after the second heat flow peak and amounts to roughly 140 J/g after an equivalent age of 1 week (168 h).

The slightly higher cumulative heat release for the pastes with S/B ≥ 0.64 may be explained by the increase in activator solution leading to a more significant reaction. The 7-day strength, at least for the mix with S/B = 0.64, is much higher than the 2-day strength, indicating that pastes with high S/B indeed gain more strength (relative to their 2-day strength) between two and 7 days, compared to the mixes with lower S/B ratio, which is in line with the observed heat evolutions. The absolute strength values are, however, still small, given the large initial water content, as mentioned before. The observed delay of the second peak for pastes with high M_s, i.e. for pastes with more sodium hydroxide but less sodium silicate underlines that the sodium silicate leads to a faster reaction, as corroborated by earlier results (Gao et al., 2015).

The heat release of AAMs pastes is much lower than the heat release of OPC paste. This is already well known for blended cements (Moon et al., 2018; Ballim and Graham, 2009). In fact, increasing the amount of fly ash and slag reduces the heat of hydration of the paste, given that typical enthalpies of slag and fly ash in blended cement pastes are around 530 J/g and 285 J/g, respectively (Han et al., 2017), compared to the roughly 500 J/g of OPC (Bentz, 1997).

The ultimate heat release Q_∞ is estimated next, based on extrapolation of the cumulative heat release Q(t). This way, the degree of reaction α(t) can be defined as α(t) = Q(t)/Q_∞ (Joseph et al., 2018; Delsaute and Staquet, 2017). By representing the cumulative heat release as a function of the inverse of the square root of the age, a linear trend is observed after the equivalent age of around 20 h. When time tends to infinity, the inverse of the square root of the age tends to be zero. Thus, Q_∞ is defined as the intersection of the fitted straight line with the vertical axis (Medaets and Lambrette, 2014). An example of the method is illustrated in Figure 4 for the composition SB0.55Ms1.44. The resulting ultimate heats for all eight compositions are listed in Table 3. The resulting ultimate heat release increases when S/B increases while Q_∞ increases when M_s decreases (except for Ms1.24).

Next, AutoShrink test results are studied. Notably, autogenous strains, (excluding strains from thermal deformations, as explained in Section 2.3.2) are set to zero at the beginning of the test, around 1.5 h after the mixing of the pastes, which is the time needed to start the test. Figures 6A,B show the development of the autogenous shrinkage of AASF pastes with different values of S/B and M_s, respectively, over a period of 2 weeks as a function of the equivalent age. Most notably, shrinkage strains for all mixes are rather large, they amount to 3,800–5,000 μm/m after 2 weeks. In general, a paste with a high S/B (and thus a high W/B) exhibits a late onset as well as a reduced magnitude of autogenous shrinkage. The reason is that high W/B induces larger pores, resulting in lower capillary pressures and thus in smaller autogenous shrinkage (Lee et al., 2014). In other terms, densification due to lower W/B, implying a reduction and a refinement of the pores, leads to larger capillary pressures and thus larger shrinkage The autogenous shrinkage slightly increases when the M_s modulus decreases. A smaller M_s also means more NaOH in the composition. As for the results obtained with the compressive strength and the heat release, this higher amount of NaOH leads to a higher pH in the solution and may be at the source of the higher autogenous shrinkage. Similar results were found by Ballekere Kumarappa et al., 2018.

The autogenous shrinkage of AASF pastes is much higher than that of OPC pastes (Jensen and Hansen, 1999). The chemical shrinkage and finer pore structure are certainly responsible for the change in magnitude (Cartwright et al., 2014; Prabahar et al., 2021; Vafaei et al., 2020). For OPC, some swelling may be observed (as is the case for our mix) just after setting, while there is no swelling for the AASF. In fact, the formation of ettringite in OPC during and after setting causes internal pressure inside the matrix and thus at the macro-scale, swelling (Carette et al., 2018). In AASF paste, no ettringite is formed which may be the reason for the absence of swelling.

As explained in the introduction section, the development of the autogenous shrinkage is associated to the reaction process (formation of finer or more refined capillary pores (Prabahar et al., 2021; Vafaei et al., 2020)) and thus the heat release. In Figures 6A,B, the development of the autogenous shrinkage is plotted according to the heat release. It is observed that the autogenous shrinkage evolved linearly with the heat release for all eight compositions. Thus, the autogenous shrinkage strains ε a s t may be readily fitted according to: ε a s t = S ∙ Q t − Q 0 Q ∞ − Q 0 (5)

With S as proportionality parameter, Q 0 is the heat release at the start of the development of autogenous shrinkage, and Q ∞ as ultimate heat obtained from isothermal calorimetry measurements as explained in Section 3.2. The parameters S and Q 0 are obtained by least-square fitting of the observed relations in Figures 6A,B and are listed in Table 5. The rather excellent agreement of the linear fit with the measurements (quadratic correlation coefficients above R ²>96%, see Table 5) may indicate that there is one major reaction occurring that induces autogenous strains when slag and fly ash react with both NaOH and sodium silicate. Also, for the same strain, there is more heat release when the S/B is higher. And for the same heat release, there is less strain for a higher S/B ratio. For most of the compositions, the material parameter S (which is proportional to the slope between the autogenous strain and the heat release) increases when the S/B ratio decreases, see Table 5. When the fitting parameter S is studied as a function of M_s, a maximum emerges at M_s = 1.38. This means that the composition with a lower S/B ratio and the compositions with M_s = 1.38 may exhibit a lower cracking risk: for the same autogenous strain development, less heat is released. However, for a proper analysis of the cracking risk, stiffness, and strength developments would be required (Serdar et al., 2020).

This section discusses the thermal expansion coefficients (CTE), which are obtained by subtracting the autogenous strain from the measured total strains. Figures 7A,B present the results of the (linear) CTE for each composition as a function of the equivalent age. Interestingly, the CTE increases rapidly until it reaches a peak between 40 and 100 h at magnitudes of 40 μm/m/°C, followed by a moderate decline. The evolution of the CTE and the magnitude of the CTE are both sensitive with respect to S/B and M_s, as detailed next. In general, the CTE starts increasing sooner and the peak of the CTE is higher when S/B and M_s are lower. The ultimate CTE (at ages of 2 weeks and later), however, seems to increase with increasing S/B. Water has a higher CTE than the solids building up the microstructure. This means that when there is less water, the CTE decreases. This explains the fact that CTE of AAM pastes decreases as the W/B ratio decreases (Sellevold and Bjøntegaard, 2006). Also, with higher water content, the E-modulus of the paste is lower and thus the pressure applied by the water for a temperature change causes more significant volume changes. (Königsberger et al., 2018b).

Compared to an OPC paste with linear CTEs at mature ages amounting to 20–30 μm/m/°C (Königsberger et al., 2018b), the CTE for all measured compositions is significantly larger, an experimental result that raises the question whether traditional steel reinforcements may even be used with the tested AASF compositions, given the expected mismatch of thermal expansion coefficients. Moreover, the CTE evolution is slightly different, see Figure 7A. For OPC, the CTE increases and then remains almost constant, but for an AASF paste, the CTE increases and then it decreases.

The evolution of internal relative humidity (IRH) is presented as a function of the equivalent age in Figures 8A,B for compositions with different S/B ratios and M_s moduli, respectively. For all compositions, the IRH remains rather high, namely between 94% and 100%. To further analyze the results, the measured IRH is corrected by the relative humidity (RH) above the activator solution, which is therefore measured separately, see Table 6 for the results. The RH above the activator ranges from 92% to 98% which is expected because of the ions dissolved in the water and is thus less than 100%. A slight increase from 92 to 96% is observed when M_s increases, while the dependence on S/B is less clear. We define the IRH change ΔIRH as the IRH of the composition from which the RH above the activator solution is subtracted. The resulting evolutions of ΔIRH are presented in Figures 8C,D for compositions with different S/B ratios and M_s moduli, respectively, and discussed next.

At the beginning of the test (between 1 and 2 days), an increase in the IRH (and ΔIRH) is observed for all compositions, which is in line with literature results and is explained by the progressive reduction of the ion concentration in the pore solution during the reaction of slag (Li et al., 2020b) and fly ash (Hu et al., 2020). Thereafter, the IRH either continues to increase or decreases slightly, resulting from several phenomena (Hu et al., 2020), 1) the consumption of pore water leads to a decrease in the IRH (self-desiccation), while 2) the release of water during the polymerization of aluminum and silicate monomers leads to an increase in the IRH, 3) the continuous change of the ion composition and concentration of pore solution also affects the IRH. Our results show that the magnitude and duration of the initial increase depend on the composition. It is faster and yields a pronounced peak for pastes with S/B ≤ 0.55, for which the IRH slowly decreases after the peak, see Figure 8C. This probably is a consequence of the dominating self-desiccation mechanism. For compositions with S/B ≥ 0.64, the IRH continuously increases during the entire measurement interval of 2 weeks, which suggests that self-desiccation is less significant. The RH difference ΔIRH is significantly larger for pastes with small silica modulus M_s, as depicted in Figure 8D. It amounts to roughly 8% for the paste with Ms = 1.04 but is close to zero for the paste with Ms = 1.58.

Next, we attempt to correlate the IRH results to autogenous strain and CTE evolution, see Figures 9A,B as well as Figures 9C,D. In general, the higher ΔIRH of the paste is, the larger are the autogenous strains, see Figure 9A. This may suggest, that despite the minor changes of IRH (or ΔIRH), it can still be, at least very qualitatively, correlated to the shrinkage, suggesting that self-desiccation is an important phenomenon when explaining the high shrinkage observed in these pastes. More pronounced self-desiccation effects on IRH evolution might, in fact, be masked by the changing ion composition and concentration in the activator solution. The IRH does, however, not correlate with the measured CTE evolutions, see Figures 9C,D.