Creep and Long-Term Properties of Alkali-Activated Swedish-Slag Concrete

High-MgO-content BFS, type Merit 5000 (MEROX, Oxelösund, Sweden) was used in this study. The chemical composition determined by a PANalytical-Zetium XRF spectrometer and the physical properties of the slag are shown in Table 1. The mix proportions of the concretes studied and compressive strength results are shown in Table 2.

All concretes and pastes contained ${450\mathrm{kg}/\mathrm{m}}^3$ of BFS. The water-to-binder ratio ($\mathrm{w}/\mathrm{b}$) was 0.45 for all concrete mixes and 0.36 for pastes. Powdered SC provided by CEICH SA Poland, and liquid SS provided by PQ Corporation were used as alkali activators. The SS as-received had a modulus ($M_s{=\mathrm{SiO}}_2/{\mathrm{Na}}_2\mathrm{O}$ as a mass ratio) of 2.2 with 34.37% by weight ${\mathrm{SiO}}_2$, 15.6% by weight ${\mathrm{Na}}_2\mathrm{O}$, and a solid content of 49.97% by weight. The $M_s$ was lowered to 1.0 for use in AAS mixes by an addition of sodium hydroxide (SH) pellets from PQ Corp., 98% purity with 76.31% by weight ${\mathrm{Na}}_2\mathrm{O}$. No plasticizing admixture was used. The total dosage of the solid part of the alkali activator was 10% by weight of the binder weight; similar materials were used previously by Humad et al. (2018). The Jehander Heidelberg Cement Group, Sweden, provided granite aggregates. The total aggregate content ($0–8\mathrm{mm}$) was ${\mathrm{1,663}\mathrm{kg}/\mathrm{m}}^3$ with 80% fine content and a specific gravity of ${\mathrm{2,700}\mathrm{kg}/\mathrm{m}}^3$. Powdered SC and liquid SS were dissolved into the mix water $24\mathrm{h}$ before the concretes were mixed. The mixing procedure included 1 min mixing of all dry materials, the addition of water with dissolved activators, followed by mixing for another 3 min using a rotating pan mixer of the type Zyklos-ZZ75HE.

After mixing, the slump of the AAS concrete mixes was determined following ASTM C143. The initial and final setting times of the AAS pastes were determined following ASTM C191 using a Vicat apparatus. The specimens for compressive strength testing were cast in alkali-resistant polymer $100\text{-}\mathrm{mm}$ cube molds. Immediately after casting, the samples were sealed against evaporation using plastic bags. Two types of curing procedures were applied: (1) sealing of specimens and storage at $20°\mathrm{C}\pm 2°\mathrm{C}$ and $40\%\pm 7\%$ RH until testing; and (2) specimens were sealed and heat cured at 65°C for $24\mathrm{h}$, followed by storage (sealed) in lab conditions until testing.

Compressive strength values were obtained as the average of three $100\text{-}\mathrm{mm}$ cubes at 3, 7, 28, 180, and 360 days following the standard SS-EN 12390-3 and using a Toni Technik compressive strength machine. The loading rate was set to $10\mathrm{N}/\mathrm{s}$.

For drying shrinkage testing, three concrete cylinders with a diameter of 100 mm and height of 200 mm were produced for each mix. The specimens for drying shrinkage measurements were sealed only until final setting ($2–3\mathrm{days}$) in lab conditions, or for $24\mathrm{h}$ (heat curing in oven), then kept unsealed in lab conditions. The plastic foil wrapping the cylinder molds was removed when the samples solidified and the drying shrinkage measurement could be initiated.

For creep testing, six cylinders were made for each mix. Three of these six replicates were not loaded but used for shrinkage measurements, and the others were loaded for creep measurement. The applied load was set to 40% of the 28-day compressive strength of the cylinders (Table 2). The creep strains were determined by subtracting the measured drying shrinkage strains on the nonloaded specimens from the strains measured on the loaded specimens. All samples were stored in laboratory conditions at a temperature of $20°\mathrm{C}\pm 2°\mathrm{C}$ and $40\%\pm 7\%$ RH; the samples were sealed for 28 days, and then the sealing was removed. Strain values were recorded using an electronic manual strain gauge, DEMEC type (Mayes Instruments, Vansittart Estate, UK) (Fig. 1). Two pairs of stainless-steel studs were glued with epoxy resin on the specimen surfaces to give a gauge length of 100 mm. The creep test rig consisted of two steel plates connected with adjustable steel rods. Cylinder surfaces were ground flat, and samples were placed in stacks of three in the loading rig (Fig. 2). The load was applied at the age of 28 days using a NIKE hydraulic jack with a maximum oil pressure of 70 MPa. After the desired load was achieved, nuts on the connecting threaded rods were locked. On the next day, the load values were checked and adjusted. Creep and shrinkage values were recorded every day for the first 2 months and every week for the following 2 months. Later values were recorded once a month. The shrinkage strain ${\epsilon }_{sh}$ value was calculated based on an average value of six replicate readings following Eq. (1) [ACI Committee 209 (ACI 2008)]:where $L_i$ = measured distance between pair of gauge points in micrometers; and $L_0$ = initial distance between pair of gauge points measured immediately after demolding.

The actual creep ${\epsilon }_c$ values were calculated following ASTM C512, where the drying and the autogenous shrinkage were extracted [Eq. (2)] [ACI Committee 209 (ACI 2008)]:where ${\epsilon }_c$ = creep strain; ${\epsilon }_t$ = total strain including creep and shrinkage; ${\epsilon }_{sh}$ = shrinkage strain; $L_{i,t}$ = distance measurement in creep test; $L_{0,t}$ = initial distance measurement in creep test; $L_{i,sh}$ = distance measured in shrinkage test; and $L_{0,sh}$ = initial distance measured in shrinkage test.

The carbonation depth was determined after 12 and 24 months of storing the shrinkage specimens (cylinders $100\times 200\mathrm{mm}$) under laboratory conditions ($20°\mathrm{C}\pm 2°\mathrm{C}$ and $40\%\pm 7\%$ RH). Phenolphthalein Deep Purple Indicator was sprayed onto the split concrete surfaces. When the pH value exceeds 9.5, the indicator changes the binder matrix color to dark magenta, while areas having a pH of less than 9.5 retain a natural concrete color because the indicator is colorless.

XRD analysis was conducted on 7- and 28-day-old powdered paste samples using a PANalytical Empyrean XRD unit using Cu $\mathrm{K}\alpha$ radiation, a step size of $0.0262°2\theta$, and a total scanning time for each sample of $16\min$. The raw data were evaluated using HighScore Plus software. A scanning electron microscope (SEM) type JSM-IT100 (JEOL) was used to study the microstructure of 24-month-old samples. Carbonated, semicarbonated, and noncarbonated areas were analyzed using a QUANTAX energy dispersive X-ray spectrometer (EDX) produced by Bruker with ESPRIT version 2 software. The samples for the SEM studies were impregnated with resin and polished using progressively finer grades of a diamond spray. The porosity was analyzed based on backscattered electron BSE-SEM images using ImageJ software. To remove the noise, a median filter with a two-pixel kernel was applied. Then threshold sensitivity analysis was performed to segment the porosity from the rest of the sample. The so-called overlay method was applied according to the procedure of Wong et al. (2006a). The method makes it possible to determine the upper gray-level threshold of porosity based on the cumulative brightness histogram of BSE images (Wong et al. 2006a, b; Yio et al. 2016). Based on the aforementioned calculation, the global threshold values were estimated for each image. Afterward, the images were binarized and the porosity $P$ was calculated aswhere $A_p$ = pore area in image (white pixel area in binary segmented image); and $A_{tot}$ = total area of sample (sum of all pixels—black and white).

Fig. 3 shows the effect of activator type on workability and setting times, as discussed in a previous publication (Humad et al. 2018), and similar trends were also obtained by other researchers (Rajesh et al. 2013). The final setting times of Mixes SS10, SC10, and $\mathrm{SC}5+\mathrm{SS}5$ were 27, 43, and $52\mathrm{h}$, respectively. Previous studies indicated that the presence of calcite ${\mathrm{CaCO}}_3$ tends to decrease the viscosity and to elongate the setting time (Shi and Day 1995; Puertas et al. 2006; Bernal et al. 2015).

Fig. 4 shows the compressive strength development of all mixes at ages of 3, 7, 28, 180, and 365 days. The measured 3-day compressive strength results were higher for the heat-cured specimens due to a more extensive dissolution of the BFS and accelerated formation of binding phases. The 28-day-old SS- and SS/SC-activated slag mixes showed higher compressive strength values than the SC-activated mix, and the SS-activated mix showed the highest strength values with both heat and ambient curing. Similar trends have also been observed by others (Wang et al. 1994; Bakharev et al. 1999; Humad et al. 2018), in connection with the higher extent of reaction. Prolonged curing times tend to enhance the strength development under ambient curing for SS activator owing to the strongly alkaline pore solution, and a similar result was obtained by others (Bernal 2016). However, after 6 months of sealed curing, the mixes containing SC as activator (SC10, $\mathrm{SC}5+\mathrm{SS}5$) showed a reduction in the compressive strength for both heat-treated and non-heat-treated specimens (Fig. 4). A similar trend was observed in PC pastes containing ${\mathrm{Na}}_2{\mathrm{CO}}_3$ (Janotka 2001) and in heat-cured AAS concretes activated with SS (Bakharev et al. 1999). Decreasing the compressive strength in some AAS mortars has also been observed by other researchers (Bernal 2016), and in some cases the compressive strength stopped developing at about 90 days owing to a lack of moisture required for slag hydration (Collins and Sanjayan 2001). Here, the concrete mixes activated with SS and cured in both regimes (Mixes SS10 L, SS10H) showed only a slight progression in compressive strength values between 6 months and 1 year, but no decrease. The densities of the AAS concretes ranged between 2,220 and ${\mathrm{2,300}\mathrm{kg}/\mathrm{m}}^3$, and the modulus of elasticity calculated at 28 days increased consistently at higher compressive strengths (Table 3).

The drying shrinkage values measured after 1, 3, 6, and 12 months are also shown in Table 3. The AAS concretes showed rather high values in comparison with the body of literature data for PC concretes, consistent with earlier results (Häkkinen 1993; Collins and Sanjayan 2000a; Reddy and Tilak 2015; Ye and Radlińska 2016). The majority of the heat-cured specimens had a lower ultimate measured drying shrinkage, mostly because a significant part of the shrinkage is likely to have developed during the first $24\mathrm{h}$ after casting ($24\mathrm{h}$, 65°C in an oven) before the start of measurement. Additionally, between the first and third months the measured shrinkage was more than double in the heat-cured versus the lab-cured samples. These results could be related to the higher amount of crystalline phases produced at higher temperature, as revealed in the XRD analysis (Fig. 5). Increasing the amount of crystalline phases due to that increased temperature accelerates the densification of the solid phases within the binder, which leads to a more porous microstructure (Shi et al. 2006). The higher initial rate of hydration tends to retard the next steps in the reaction process as dense reaction products block the availability of slag grain surface for future dissolution, creating a nonuniform distribution of the hydration products and more porous microstructure (Helmuth and Verbeck 1968; Shi et al. 2006). In SC-activated slag pastes more calcite and less gaylussite (Ga) and hydrotalcite-group minerals (HT) were also detected in the heat-cured samples than in the lab-cured one (Fig. 5).

Considering the development of shrinkage over time, $80\%–90\%$ of the 1-year shrinkage values were developed during the first 28 days in the case of the non-heat-treated samples, while it took 4 months for the heat-cured specimens to develop the same percentage of their 12-month shrinkage. For comparison, in a typical PC concrete containing 80% by volume of aggregates, having a water-to-cement ($\mathrm{w}/\mathrm{c}$) ratio of 0.45 and stored at a RH between 40% and 45% the drying shrinkage values would be $0.5–1.2\mathrm{mm}/\mathrm{m}$ after 1 year (Neville and Brooks 2010). All of the AAS concrete specimens produced in this study, heat-cured and non-heat-cured, showed higher drying shrinkage values than this ($1.81–3.48\mathrm{mm}/\mathrm{m}$). This higher drying shrinkage of the AAS has been linked by other researchers to the structural incorporation of alkali cations in $\mathrm{C}–\mathrm{A}–\mathrm{S}–\mathrm{H}$, which caused the collapse of $\mathrm{C}–\mathrm{A}–\mathrm{S}–\mathrm{H}$ layers under drying conditions (Ye and Radlińska 2016), also altering the binder matrix microstructure (Ismail et al. 2013). Typically, AAS contains a larger amount of mesopores than PC concrete, leading to the generation of higher tensile forces during drying, which leads to a higher shrinkage (Häkkinen 1993; Collins and Sanjayan 2000a). The small maximum aggregate size (8 mm) used in these concretes may also be a contributing factor, as is the high binder content of ${450\mathrm{kg}/\mathrm{m}}^3$ (Table 2), particularly when considering that the BFS used here is less dense than PC, so the high binder content on a mass basis becomes particularly notable if converted to a volume basis.

The measured creep values after subtraction of the drying shrinkage are shown in Fig. 6. The instantaneous creep strain (initial strain) was measured immediately after the application of the compression load (Table 3). In addition, creep coefficients and specific creep values were calculated and are shown in Figs. 7 and 8. The creep coefficient was calculated as the ratio of the ultimate creep strain (total creep strain minus drying shrinkage strain) to the initial strain, while the specific creep is the ratio of the creep strain (total creep strain minus drying shrinkage strain minus initial strain) per unit stress [ACI Committee 209 (ACI 2005)]. All concretes activated with SS, which, as described earlier, had the highest drying shrinkage, showed the lowest creep and creep coefficient and the lowest specific creep but the highest instantaneous creep strain. The measured creep values were higher for the heat-treated samples compared with ambient-cured ones, especially for the SS-activated mix (Fig. 6). Similar results were obtained previously for both alkali-activated fly ash concrete and PC concretes (Neville and Brooks 2010; Collins and Sanjayan 1999; Bazant and Li 2008, Wallah and Rangan 2006).

The calculated creep coefficients were nearly identical for heat-treated and non-heat-treated concretes when SS and the combination of SS and SC were used as activators. From these coefficients, the creep strain was approximately two times greater than the initial strain in SS-activated concretes and nine times greater than the initial strain when the combined SS–SC activator was used. Despite the similar 28-day compressive strength values measured for laboratory- and heat-cured mixes, excluding mixes activated with SC, the heat-cured samples showed a higher instantaneous creep (initial strain). This can be related to the more porous binder matrix formed in heat-cured samples. On the other hand, the lab-cured concrete activated with 10% SC showed a higher creep coefficient than the comparable heat-cured samples. The calculated specific creep, which takes into account the compressive strength of the concrete when load is applied (the converted strength to cylinder), was higher for heat-cured samples (Fig. 8). The highest ultimate specific creep was calculated for SC-activated heat-cured concrete. All results can be directly linked to the compressive strength at 28 days, where the mix with a high compressive strength showed a low creep value. In the case of lab-cured mixes activated with 10% SS (Mix SS10L) 90% of the measured creep in 24 months developed during the first 55 days, and 230 days for the heat-cured samples (Mix SS10H). Concretes activated by 10% SC, which showed higher creep, achieved 90% of the ultimate measured creep after about 130 days for both curing types (Mixes SC10L and SC10H). Mixes activated with the combination of SS–SC, for both curing types (Mixes $\mathrm{SC}5+\mathrm{SS}5\mathrm{L}$ and $\mathrm{SC}5+\mathrm{SS}5\mathrm{H}$), developed 90% of the total creep after about 160 and 220 days, respectively. All observed trends related to creep can be linked to the microstructure of the binder matrix. Investigation of the concretes by SEM showed significant microcracking (cracks filled with resin), as well as coarsening of the porosity caused by carbonation, and the extent of these effects differed between the various concrete mixes used in the present study. The heat treatment produced a bright rim of hydration product deposited on the surface of the partially hydrated slag particles. Conversely, in lab-cured samples, at later ages, the hydration product deposited was darker in grayscale value (i.e., had a lower atomic number density), especially in the AAS mix containing SS as the activator.

High creep values were related to the microcracking evident in the binder matrix (Figs. 9–11). Furthermore, after storing the AAS concrete cylinders in the lab environment ($20°\mathrm{C}\pm 2°\mathrm{C}$ and $40\%\pm 7\%$ RH), the specimens showed a significant degree of carbonation. Carbonation of AAS concrete is caused by the reaction between ${\mathrm{CO}}_2$ from the air and the alkaline ${\mathrm{Na}}_2\mathrm{O}$ in the pore solution, which causes uptake of carbonate or bicarbonate anions by the decalcification of the $\mathrm{C}–\mathrm{S}–\mathrm{H}$ (Bakharev et al. 2001). The carbonation depths determined by application of phenolphthalein indicator to 12- and 24-month-old samples increased with time (Table 3 and Figs. 9–12). Sealing with plastic bags provided rather good protection against carbonation (Fig. 13). The 24-month-old concretes activated with 10% SC, under both curing regimes, showed carbonated and semicarbonated regions with highly cracked areas (Fig. 9). A higher porosity of the binder matrix in the carbonated region was observed in Mixes SS10H and SC5+SS5H [Figs. 10(b) and 11(b)]. The calculated total porosity based on segmentation of SEM images is shown in Table 4. The differences can be directly related to the formation of secondary products, which in the case of concretes based on high-MgO slag are likely to include calcite ${\mathrm{CaCO}}_3$, huntite ${\mathrm{CaMg}}_3({\mathrm{CO}}_3{)}_4$ natron ${\mathrm{Na}}_2{\mathrm{CO}}_3·10{\mathrm{H}}_2\mathrm{O}$, thermonatrite ${\mathrm{Na}}_2{\mathrm{CO}}_3·{\mathrm{H}}_2\mathrm{O}$, and gaylussite ${\mathrm{Na}}_2\mathrm{Ca}({\mathrm{CO}}_3{)}_2·5{\mathrm{H}}_2\mathrm{O}$, which have been identified as carbonation products of the various binding phases present in AAS paste (Bernal et al. 2014). Alkali-activated high-MgO slag was reported by others to have a lower carbonation depth than others slags with low MgO content (Bernal et al. 2015). The trend was associated to the absorption of ${\mathrm{CO}}_2$ by the hydrotalcite-type minerals present, which hindered the process of carbonation of the $\mathrm{C}–\mathrm{A}–\mathrm{S}–\mathrm{H}$ (Bernal et al. 2015). In the present study, EDX analysis showed that noncarbonated regions of the 28-day-old AAS concrete contained O, C, Ca, Si, and Al and small amounts of Na and Mg. In contrast, the carbonated regions had lower amounts of Ca, Si, Al, O, Na, and Mg but significantly increased the amount of carbon, which is consistent with the formation of calcium carbonate. The formation of calcium carbonate was also confirmed by an increased $\mathrm{Al}/\mathrm{Ca}$ ratio and a decreased Si/Ca ratio observed in the carbonated matrices (Fig. 14). Carbonated $\mathrm{C}–\mathrm{S}–\mathrm{H}$ is also identified as weaker, resulting in a higher creep, similar to the results obtained by others (Zhang et al. 2014; Nguyen et al. 2014). Concrete with higher $\mathrm{Si}/\mathrm{Ca}$ and $\mathrm{Al}/\mathrm{Ca}$ ratios showed higher creep, similar to the results observed by Nguyen et al. (2014) when the creep was measured with micro- and nanoindentation. The heat-curing procedure appeared to limit the carbonation (Figs. 10–13). Carbonated regions showed a higher extent of deterioration of the microstructure (Table 4 and Fig. 15).

Significantly lower creep strains and creep coefficients than those presented here have been measured for other types of concretes, including high-volume fly ash PC concretes and alkali-activated concretes based on fly ash (Malhotra and Mehta 2002; Wallah and Rangan 2006). The lower creep values of alkali-activated fly ash concretes were related to what those authors called a block-polymerization concept that assumes only partial dissolution of silicon and aluminum from fly ash. Consequently, remnant fly ash particles could act as microaggregates and, thus, reinforce the binder matrix (Pacheco-Torgal 2014). However, this concept does not appear to be valid in the case of alkali-activated systems based on slag, which may be related to the irregular/nonspherical shape of slag particles or differences in the binder microstructure between the $\mathrm{C}–\mathrm{A}–\mathrm{S}–\mathrm{H}–\mathrm{based}$ binders in AAS and the low-Ca alkali aluminosilicate ($\mathrm{N}–\mathrm{A}–\mathrm{S}–\mathrm{H}$) gels that form in alkali-activated fly ash concretes. In addition, the cracked microstructure (only cracks filled with resin, ignoring cracks formed owing to a SEM vacuum) and a maximum aggregate size of only 8 mm could contribute to the higher creep values measured in the present study.