Figure 2 compares the heat release measured from isothermal calorimetry for all the mixes. As previously reported, hydration of NaOH activated slag results in the main peak between 1 and 24 h while sodium silicate-activated slag exhibits a relatively long induction period and the main reaction peak only after 2 days (Shi and Day, 1995; Haha et al., 2011; Haha et al., 2012). Nevertheless, similar cumulative heat of hydration is achieved in both cases. In our case, where the slag is activated with a mixture of NaOH and sodium silicate solution, the main hydration peak is observed after 5 h. Hubler et al. (2011) attributed this main peak to be derived from nucleation and growth of C-S-H based on nucleation seeding experiments.

With a decrease in the amount of slag and an increase in the amount of fly ash in the mixes, we observe a later occurrence of the main hydration peak with a lower intensity. In other words, relatively shorter induction periods and narrower peaks are observed with increasing slag content. This is attributed to the higher reactivity of slag over fly ash, particularly to the fast reaction of the very fine slag particles during early hours, which increases silicate concentration and thereby supersaturation with respect to C-S-H (Gruskovnjak et al., 2006). An increase in the induction period with the addition of fly ash is often reported for Portland cement and C₃S (Deschner et al., 2012; Joseph et al., 2017b; Schöler et al., 2017). Mixes with 25% slag (S25F75_0.7 and S25F75_0.52) and without slag (S0F100_0.7 and S0F100_0.52) do not show a distinctive peak within 7 days. The lower heat release with a broader peak and the longer exothermal activity imply slower hydration of the slag-fly ash mixes leading to a lower early-age compressive strength (c.f. Compressive Strength section).

An interesting finding is the earlier occurrence of the main peak of hydration at lower solution-to-binder ratios. A slight decrease in the induction period and a faster occurrence of the main peak of alite hydration are often reported for Portland cement and C₃S systems at low W/C ratios (Thomas, 2007; Hu et al., 2014; Joseph et al., 2017a). The faster occurrence of the main hydration peak in our samples with slag and fly ash may be due to the variations of alkali concentrations in the solution. We explain this based on two assumptions:

1) With the progress of the reaction, there is more consumption of silicate ions compared to OH⁻ ions from the alkali solution;

It is known that the increase in pH of the pore solution accelerates the occurrence of the main hydration peak (Altan and Erdoǧan, 2012; Gebregziabiher et al., 2015) due to the increased dissolution of slag and fly ash, which is strongly dependent on the pH (Song and Jennings, 1999). Although the initial concentrations are the same with different S/B ratios, there are higher amounts of alkalis (c.f. Table 2) and higher water content in the overall system with a higher S/B ratio. Hence, at the same degrees of reaction, the concentration of silicates in the pore solution would be lower for the mix with a lower S/B ratio. In that case, the system’s pH would be higher for mixes with a lower S/B ratio as the silicate ions induce lower pH compared to the OH⁻ ions. Nevertheless, further study is required to explore this issue in more detail.

Figure 3 shows the cumulative heat released measured until 7 days of reaction. Mixes with higher amounts of slag exhibit higher heat release as expected, owing to its faster reaction kinetics. The heat release is almost similar with respect to the change in the S/B ratio, although it is slightly higher for mixes with a higher S/B ratio. This difference is more pronounced for mixes with a higher volume of slag than those with a lower volume of slag, which is attributed to the increased consumption of alkalis at higher degrees of reaction and lower amount of alkalis in the initial mix composition at lower S/B ratios (c.f. Table 2). Nevertheless, it is remarkable that the deviation is almost negligible for all the mixes at least till 1 day of reaction. It should be noted that the heat release within the initial 45 min of reaction has been omitted in the results because this is the time required from mixing to placing the samples in the calorimeter and the time required for the stabilization.

Figure 4 shows the thermogravimetric data and DTG for all the mixes at a 0.7 S/B ratio after 28 days of reaction. In the hydrated alkali-activated slag (S100F0_0.7), two main hydration products are formed as detected in the TGA data: C-S-H (main weight loss at 50°C–200°C) and hydrotalcite-like phase (weight loss at around 200°C and 400°C), which is in agreement with the previous findings (Wang and Scrivener, 2003; Gruskovnjak et al., 2006; Haha et al., 2011). The mass loss from C-S-H consists of the water chemically bound to the structure of C-S-H and physically bound water (Köster and Odler, 1986). The broad peak of hydrotalcite at around 400°C remains almost the same in systems blended with fly ash up to 50%. Increasing fly ash contents beyond this ratio, this hydrotalcite peak becomes weaker. The C-S-H formed from slag hydration is characterized by a higher crystallinity, low Ca/Si ratios, and generally higher aluminum incorporation than that formed in OPC or blended OPC systems (Schneider et al., 2001; Wang and Scrivener, 2003; Gruskovnjak et al., 2006).

From the DTG data, it is noticeable that irrespective of the proportion of slag and fly ash, the profile of the rate of mass loss is similar. While more C-(A)-S-H type reactants are expected from calcium-rich slag-based systems, N-(A)-S-H type reactants are expected to form in low calcium fly ash-based systems (Garcia-Lodeiro et al., 2011). Previous studies also suggest that the dehydroxylation temperature of these C-S-H type gels occurs at similar temperatures (Winnefeld et al., 2010; Ismail et al., 2014). The peak of weight loss in activated fly ash without slag (S0F100_0.7) occurs at around 100°C and is much broader than the one reported by Winnefeld et al. (2010), who attributed this peak to sodium aluminosilicate hydrates. Incorporating the slag and increasing its content, this peak becomes much sharper.

Figure 5 shows the thermogravimetric (TG) data and the first derivative of TG (DTG) for S50F50_0.55 mix after 3, 7, 28, and 90 days of curing. Total mass loss is consistently increasing with an increase in age as expected. Most mass loss occurs before 500°C, indicating a C-(A)-S-H type reaction product (Sakulich et al., 2010). The nature of the peaks does not change over the course of the reaction, and they follow the same trend. As reported previously, alkali-activated slag-fly ash blends can demonstrate the co-existence of the C-A-S-H type gel and geopolymer gel (N-A-S-H) (Wang and Scrivener, 1995; Brough and Atkinson, 2002; Ben Haha et al., 2011; Ismail et al., 2014). The lack of a peak around 600°C–800°C confirms no significant degradation in the samples due to carbonation (Lothenbach et al., 2016).

Figure 6 shows the total bound water content measured from TGA, corresponding to the mass loss between 20°C and 600°C (Lothenbach et al., 2016). With an increasing amount of slag, the bound water content is generally higher because slag exhibits a faster reaction than fly ash. This is noticeable at 3 days when the activated slag (S100F0_0.7) indicates the highest bound water content. However, after 7 days, this situation changes in favor of slag blended with 25% fly ash (S75F25_0.7), showing higher bound water content than activated slag. A striking difference is an increase in the bound water content with an increase in the S/B ratio. This is expected as there is a high amount of alkalis in the system (c.f. Table 2), which will continue to react with slag and fly ash. This difference is more pronounced at later ages due to the increased consumption of the alkalis.

The compressive strength development of the samples listed in Table 2 is plotted in Figure 7. A lower S/B ratio consistently shows higher strengths than the same blends with a higher S/B ratio (Ruiz-Santaquiteria et al., 2012), most likely due to the lower porosity of the system when the S/B ratio is lower. At a similar S/B ratio of 0.7, the strength results are comparable with cumulative heat (Figure 3) and bound water content (Figure 6), particularly till 7 days.

Regarding the sample S100F0_0.7, the strength does not increase significantly after 7 days. On the one hand, an increase in the bound water content is recorded until 90 days. This seems to be due to microcracking induced by shrinkage that can cause a strength loss (Collins and Sanjayan, 2001; Wardhono et al., 2015). On the other hand, there is a continuous increase in the strength of the other samples. In the early ages (3 and 7 days), samples with a higher proportion of blast furnace slag yield a higher strength which could be attributed to the higher reactivity of slag over fly ash. At 28 days, for the same S/B ratio of 0.7, the compressive strengths of S100F0_0.7, S75F25_0.7, and S50F50_0.7 are almost similar. Considering the samples with a similar flow, by 28 and 90 days, the samples with 25% and 50% slag replacement with fly ash (S75F25_0.6 and S50F50_0.55, respectively) have higher strength values than the one with 100% slag (S100F0_0.7). Compared to Portland cement, while the strength of 25% slag (S25F75_0.52) after 28 days is in the same range with CEM I 52.5 with a 0.5 W/C ratio, samples with higher percentages of slag are much superior even at higher S/B ratios. There are no surprises here as similar results are often reported in the literature (Puertas et al., 2000; Nath and Sarker, 2014; Fang et al., 2018).

Figure 8 compares the overall cumulative heat released per gram of binder measured from isothermal calorimetry with the bound water content measured from TGA after 3 and 7 days of curing. A good correlation exists with an R² value of 0.97. The comparison could only be drawn until 7 days due to the lack of calorimetry data beyond 7 days. A remarkable outcome of this correlation is that irrespective of the amount of slag and fly ash in the current systems, the bound water content and heat release is proportional to each other, both being a good indication for hydrate formation. Nevertheless, there is an offset for the trend line amounting to around 56 J. Part of this could be due to the omission of heat release in the first 45 min. This needs to be addressed in detail in future studies by comparing the early age heat release with TGA (<3 days).

Figure 9 compares the compressive strength with the bound water content determined from TGA. While the general trend can be captured, the correlation is slightly poor, with an R² value of less than 0.7, due to the different S/B ratios that lead to differences in initial porosity. This is evident from Figure 6 and Figure 7, as a lower S/B ratio leads to a higher compressive strength but a lower bound water content. Hence, the mixes with the same S/B ratio of 0.7 are compared in Figure 10, which gives much better correlations (R² of ∼ 0.9). The outlier S100F0_0.7 after 90 days of reaction is marked in Figure 10.

To consider the effect of the porosity while varying the S/B ratio, a new parameter λ B W is introduced. It is defined as follows: λ B W = B W / W s , (1) where BW is the bound water content in percentage and Ws is the water-to-solid ratio (W/S). The rationale behind selecting such a parameter is that this could consider the effect of initial porosity to a certain extent. A higher W/S ratio will contribute to a decrease in strength and vice versa. Such a parameter could mimic Schiller’s model (Schiller et al., 1971), where the increase in strength is correlated with the decrease in capillary porosity and initial porosity. An increase in bound water content corresponds to the hydrate formation and thereby reduction in total porosity.

A significant improvement in the correlation with an R² of 0.88 is obtained when compressive strength is compared with λ B W (Figure 11). While the number of hydration products and the amount of initial porosity are important parameters for strength development, other parameters could explain the recorded deviations. One of the major ones can be the microcracking of samples at later ages due to shrinkage, particularly for the mix with 100% slag, as indicated in the previous section. The result of the 90-day strength of S100F0_0.7, which is highlighted with a red circle in Figure 10 and Figure 11, is the most deviated one from the trend lines. Nevertheless, no visible cracks were observed on these samples, assuming there were microcracks. Furthermore, other parameters are not included in this approach, including pore structure (Kumar and Bhattacharjee, 2003; Chindaprasirt et al., 2005) and interfacial transition zone (Ollivier et al., 1995).

TGA is costly equipment, so an alternative route using an oven could be considered. The mass loss could be calculated for samples after arresting the reaction from a laboratory oven by keeping the sample at 600°C by taking the mass before and after placing it in the oven. Precautions need to be made with respect to the holding time, as the temperature displayed in the oven may not be the temperature of the material. Furthermore, there could be errors in measurements due to oxidation at high temperatures if an inert gas such as N₂ is not used (Bernal et al., 2017).

This section compares the compressive strength with heat release from calorimetry till 7 days. Similar to the approach used in the previous section, to consider the effect of the S/B ratio, the parameter λ c a l is introduced: λ c a l = Q / W s , (2) where Q is the cumulative heat released measured from the isothermal calorimeter and Ws is the water-to-solid ratio (W/S).

Figure 12 shows a good correlation between the compressive strength and λ c a l (R² = 0.97). This is expected as a good linear correlation exists between the bound water content and the cumulative heat release (Figure 8). A higher R² for Figure 12 compared to Figure 11 shall not be interpreted as this approach is better than bound water content because, in this case, only measurements until 7 days are compared.