The soil used was collected in southern Cali, Valle del Cauca, Colombia, at latitude 3° 21’ 45.979’’N and longitude 76° 30’ 53.277’’W. It was initially dried at ambient temperature, and a disc mill was used to de-flocculate the particles. The particle size distribution was determined by sieving, according to ASTM D6913-17 (13), the sieved material showed a very fine texture with 96.75% of the material passed through sieve number 200. The Atterberg limits were determined according ASTM D4318-17 (14). The soil presented liquid limit of 58% and plastic limit of 34%. The Plasticity Index calculated was 24%. These results were used to classify the soil as a A-7-5 according to ASTM D3282-15 (15).

For the preparation of the alkali-activated binding pastes, GBFS and commercial lime were used as sources of calcium, and two classes of fly ash (FA1, FA2) were used as precursors. The FA1, FA2 and GBFS were previously milled for 2 h; after which the average particle size after the grinding process of the precursors was 37.064 μm, 24.968 μm and 27.729 μm, respectively. The commercial lime had an average particle size of 24.740 μm. A general-use Portland cement (OPC) was also included in the study as reference material. Table 1 shows the chemical composition of the materials used that was determined by X-ray fluorescence (XRF) using a Philips PW-2440 Pro PANalitical MagiX spectrometer, with a maximum power of 4 kW and equipped with a rhodium tube. A NaOH/Na₂SiO₃ combination, in solution form, was used as the alkaline activator.

Two alkali-activated binding pastes (F1L and F2G) were designed using response surface models where the SiO₂/Al₂O₃ and Na₂O/SiO₂ molar ratios were considered as the independent variables. The pastes were cured at 30°C. The uniaxial compressive strength (UCS) after 28 days was defined as the dependent variable. The response surface models were optimized to obtain the best combination of independent variables and maximize the response of the dependent variable, producing the contour plots shown in Figure 1.

F1L (Figure 1a) is composed by 80% FA1 and 20% lime, with optimized independent variables SiO₂/Al₂O₃ = 3.0 and Na₂O/SiO₂ = 0.3, yielding an UCS, after 28 days, of 21 MPa. F2G (Figure 1b) is composed by 80% FA2 and 20% GBFS, with optimized independent variables SiO₂/Al₂O₃ = 3.9 and Na₂O/SiO₂ = 0.32, yielding an UCS, after 28 days, of 23 MPa.

Modified Proctor compaction tests were performed to determine the maximum dry density (MDD) and optimum moisture content (OMC) of each mixture, according to ASTM D1557-12e1 (16). The binding paste was added to the dry soil in pre-determined percentages. The soil-cement mixture was prepared with 13 wt.% of OPC, which was the content required for the compaction test according to the classification of the soil (17). The soils stabilized with alkali-activated pastes were mixed with 20% F1L, 30% F1L and 30% F2G.

The soil and the precursor, both in dry state, were manually mixed until a uniform coloration was obtained, for homogenisation. The alkaline activator, calculated as a function of the precursor content, was then slowly, while still mixing. Finally, additional water was added, to guarantee the pre-determined OMC for each mixture.

The specimens were then made by compacting the material into cylindrical moulds, with 52 mm in diameter and 104 mm in height, and prismatic moulds with 50 × 50 × 150 mm, according to ASTM D1632-17 (18). Once compacted, the specimens were cured in a climatic chamber, with a temperature and humidity of 30 °C and 80%, respectively. UCS tests were performed after 7 and 28 days curing, and flexural tests after 28 days, based on ASTM D1633-17 and D1635-12 (19, 20), respectively, in an INSTRON 3300 universal testing machine.

Wetting and drying tests were performed on stabilized soils after 7 days of curing, following the protocol, outlined in ASTM D559-15 (21). Each specimen was subjected to a total of 12 wetting-drying cycles. One full cycle consisted of immersion in potable water for 5 h and then drying for 48 h at 71 °C. This test was done to calculate the changes in the volumes of the stabilized soil specimens and the mass loss during the wetting and drying cycles. The volume change was calculated by measuring (in saturated and dry conditions) the average height and diameter of one of the specimens. To calculate the mass loss, the second specimen in the dry condition was brushed in each cycle with a steel brush applying an average force of 15 N, passing 20 times over the lateral surface of the body and four times over each face of the specimen.

The compaction curves for each stabiliser type and content used are shown in Figure 2, and the MDD / OMC of all tests is presented in Table 2. The MDD of the nonstabilized compacted soil is 1.56 g/cm³, and the OMC of this sample is 24.67%. The addition of OPC as stabilizer increases the MDD to 1.59 g/cm³ with an OMC of 26.15%; this phenomenon is related to the addition of a material of higher specific weight to the mixture, and the increase in the moisture content is attributed to the additional water the OPC requires to begin the hydration processes (22). In contrast, by stabilizing the soil using alkali-activated cements, in general, the MDD decreases, compared to the nonstabilized soil.

The addition of 30F1L and 30F2G produced slightly different results, with 1.47 g/cm³ and 26.92% for the former, and 1.51 g/cm³ and 26.27% for the latter. FA1 is the precursor of the 30F1L system, and the loss-on-ignition (LOI) (unburned carbon) of this sample is 38.76%, while FA2 is the precursor of 30F2G with an LOI of 20.67%. Additionally, the proportions of calcium sources for both cements are the same, 20% by weight, but are of different nature and composition. 30F1L contains lime, and 30F2G contains GBFS, which theoretically makes the specific gravity of cementitious material 30F2G higher than that of 30F1L (23, 24). For the OMC of the soil stabilized with 30F1L, this mixture has a slightly higher OMC than that of 30F2G; according to Osinubi (2006) (25), lime content increases the OMC of clayey soils.

The MDD is also affected by the content of cementitious material used to stabilize the soil. With the addition of 20F1L, the MDD was 1.50 g/cm³, which is slightly higher than the 1.47 g/cm³ of the soil stabilized with 30F1L (Table 2), and is consistent with reports from certain authors (11, 22). Regarding the materials used, one of the reasons why the MDD decreases between 20%F1L and 30%F1L, is the amount of unburned material in the precursor (fly ash 1 with 38.76% LOI), which results in a mixture with lower specific gravity.

Figure 3 shows the UCS results of the unsaturated and saturated specimens. For the unsaturated conditions (Figure 3a), the reference mixture (OPC) registered the best performance after 7 days (UCS_7D = 4.79 MPa), compared with the values obtained using alkali-activated cementitious materials 20F1L, 30F1L and 30F2G, with UCS_7D of 3.88, 4.21 and 1.52 MPa, respectively. This trend continues after 28 days, with OPC, 20F1L, 30F1L and 30F2G showing UCS_28D values of 6.55, 5.26, 4.45 and 2.43 MPa, respectively. There are two key factors explaining why the performance of the soil treated with alkali activated cementitious material is worse than that of the reference sample (OPC). One factor is the compositions of the main precursor materials (FA1 and FA2), and the other factor is the chemical composition of the soil. As shown in Table 1, the content of unburned carbon (LOI) is quite high in both ashes (FA1 = 38.76% and FA2 = 20.67%). This carbon content may have a negative impact when the mixture of cementitious material and soil is alkali activated because carbon has a large surface area that absorbs a certain amount of the alkaline activator and removes alkali cations, hindering the kinetics of the dissolution reactions of the precursor, condensation and polymerization of cementitious gels (26–28). The chemical composition of the soil (Table 1) may also have an adverse effect on the strength development of the soil stabilized with alkali-activated pastes due to the laterization degree. This is estimated by calculating the ratio of silicon to sesquioxides (SiO₂/(Al₂O₃ + Fe₂O₃)) (22). The soil used in this study has a ratio of 0.96, this means that the chemical weathering process of the soil is quite advanced due to its high amount of iron and aluminium oxides. Certain studies indicate that reactive Fe₂O₃ may precipitate much faster than Al or Si compounds under highly alkaline conditions and generates iron hydroxides or oxyhydroxides that consume the OH^- ions of the alkaline activator, which slows the dissolution of the remaining particles of the precursor material and reduces the number of nucleation sites of the oligomers that form the cementitious aluminosilicate gel. However, other studies suggest that Fe⁺³ acts as nucleation sites since in silicate glasses, Fe⁺³ is known as a network former due to a charge and ionic radius similar to those of Al⁺³ (29–32).

Regarding the soils stabilized with the same type of alkaline cementitious material but at different proportions, in unsaturated conditions, 20F1L has a higher UCS than 30F1L at 28 days of curing (20F1L UCS_28D = 5.26 vs. 30F1L UCS_28D = 4.45 MPa). The reason for this difference is not clear; however, an increase of 20% to 30% in the proportion of cementitious material leads to a slight decrease in the maximum dry density of the compacted mixture (Figure 2). The greater proportion of unburned carbon acts as a “lubricant” between the soil particles and decreases the internal friction between the particles, which affects the resistance to external stresses (33, 34). In contrast, the cementation capability of the 30F2G is lower than that of 30F1L, even though the maximum dry density of the mixture is higher (Table 2). This behaviour is attributed to the effect of the lime on the composition of cementitious material 30F1L. Double charge cations such as Ca⁺⁺ and Mg⁺⁺ can generate nucleation sites quickly without needing the species from the main precursor (FA1) to dissolve, due to the action of the alkaline activator; therefore, the particle agglomeration and cementation of the stabilized soil improves with shorter setting times (35, 36).

Figure 3b shows the results of the UCS tests under water-saturated conditions. The untreated soil completely collapses after a few minutes of immersion in water, and in general, for the different cementitious materials, the compacted specimens maintain their integrity during the saturation period of 4 h before the test, due to the type of bonds that the cementation produces, which are not present in the untreated soil. The untreated soil has a flocculated structure; this condition creates clay bridges by electrostatic attraction, which provides strength to the material, but the strength is lost when the degree of saturation increases. The structure of the stabilized soils has cementation bonds that make the soil stiffer and stronger. When the soil is saturated, the precipitates can dissolve and gradually break the bonds. The loss of such bonds is slower in stabilized soil than in nonstabilized soil.

In the saturated conditions, the soil stabilized with OPC has the best performance at both 7 and 28 days of curing, UCS_7D = 2.5 and UCS_28D = 2.7 MPa, respectively. For the saturated conditions, increasing the proportion of alkaline cementitious material F1L from 20% to 30% impacts the performance of the stabilized soil (20F1L UCS_7D = 1.09; UCS_28D = 1.26 MPa vs. 30F1L UCS_7D = 1.36; UCS_28D = 1.54 MPa) and improves the soil strength with 30F1L, UCS_7D = 19.85% and UCS_28D = 18.18%. As mentioned above, this behaviour can be attributed to the proportion of lime in the alkaline cementitious material 30F1L. Despite the increase in the proportion of unburned carbon in the mixture, the lime immediately reacts with the soil, flocculates, and generates nucleation sites that help cement the soil structure. This effect could change the size and distribution of the pores of the compacted material, thus increasing the tortuosity and decreasing the saturation occurring during the time the sample is immersed in water (37, 38). Regarding the treatment with 30F2G, the soil resists the saturation period much better than the untreated soil, but the strength of this soil is the lowest recorded, either after 7 or 28 days (30F2G UCS_7D=0,66; UCS_28D=0,70 MPa), and the specimens showed substantial deterioration during the saturation process.

The FS of the stabilized soil is directly related to the UCS and MDD. The evolution of the FS over the curing period is moderate, compared with that of the UCS. Figure 4 shows the FS for each mixture of stabilized soil. In general, the FS of the different cementitious materials is between 1/3 and 1/5 of the UCS, except that of 20F1L, which is below this range (1/6 at 0.84 MPa).

Studies by Mandal et al. (2018) (39) showed that soils stabilized with different cementitious materials and compacted at low densities had low FS values. However, in these studies, the effect of cementitious materials and the proportion of these materials in the mixture also affected the FS, together with the MDD. When the same alkaline cementitious material is used, although in different proportions, the FS increases (20F1L; FS = 0.84 MPa vs. 30F1L; FS = 0.99 MPa). Such improvement is due to the overall increase in the number of cemented particles. As mentioned above, a higher lime content increases the agglomeration and cementation capability of the soil with cementitious material 30F1L, improving the ability of this soil to dissipate energy in the bending (38). Furthermore, by changing the type of alkaline cementitious material, even if the same content is used, the FS varies (30F1L = 0.99 MPa vs. 30F2G = 0.46 MPa). This is attributed to the agglomeration and hardening effect of the cementitious material. In these types of soils, 30F1L provides better cementation than 30F2G. Therefore, in addition to MDD and UCS, the nature and cementing capability of the cementitious material is essential to improve the FS (40).

Figure 5 shows the volumetric variation of the soils stabilized with different cementitious materials subjected to wetting and drying cycles. The volumetric variation of the soil stabilized with OPC is quite uniform overall. In the first cycle, it expands 0.56% and contracts -1.90%, compared with the initial volume of the specimen after 7 days curing. The rest of the expansion and contraction measurements are made with respect to the last calculated volume, which means that, after 12 cycles, the reference specimen has an average expansion of 0.59% and average contraction of -0.68%.

The soil stabilized with 20F1L has an initial expansion of 0.74% and contraction of -1.99% after the first cycle. Compared with the soil stabilized with OPC, the change in the volume of the soil with 20F1L is greater in both expansion and contraction and, at the end of the 12 cycles, an average expansion / contraction of 0.84% and -0.93% was determined. The volumetric variation of the soil stabilized with 20F1L is far more irregular than the soil-OPC.

The soil stabilized with 30F1L has an expansion of 0.17% and a contraction of -0.97% in the first cycle and, on average, a volume changes slightly lower than that of the OPC and 20F1L, with an expansion of 0.51% and a contraction of -0.57%. The test of the soil stabilized with 30F2G was discarded because the specimen collapsed completely in the second wetting and drying cycle, and it was not possible to determine the corresponding weight and measurements.

The expansion and contraction limits proposed by Nogami & Villibor (41), for paving tropical lateritic soils without stabilization, are used in the present work as a reference (Table 3). This is in accordance with the Rojas et al. (2008) (42) proposal, based on the lack of standard parameters to assess the volumetric variation of stabilized soils. Overall, based on these data, at the end of the WDT cycles, the expansion and contraction obtained for the different systems can be considered average (0.5 to 3%).

Scanning electron microscope (SEM) images (Figure 6) shows the cementation of soil particles. The soils stabilized with OPC and 30F1L show structures with well-cemented particles that help to obtain a uniform volumetric variation and relatively low expansion and contraction averages, while the soil stabilized with a minor proportion of alkaline cementitious material (20F1L) shows several cracks in the structure. Higher proportions of alkaline cementitious material modified the natural behaviour of the soil, which is naturally very susceptible to significant volume changes under wetting and drying cycles (according to its Unified Classification, ASTM D3282-15). In addition, the cementation observed when stabilizing the soil with 30F2G is poor, which is attributed to the collapse of the sample that prevented the completion of 12 wetting and drying cycles.

The mass loss due to wetting/drying a soil stabilized with different cementitious materials is an important property. Regarding the WDT results of this study, the reference specimen stabilized with OPC at the end of the 12 cycles has a mass loss of 3.86%; the specimen of the soil stabilized with 30F1L has a very similar mass loss of 3.74%, while the soil stabilized with 20F1L loses 10.21% of its mass after 12 cycles (Figure 7). The mass loss calculations were made using the parameters of ASTM standard D 559-15 (21), used for testing soil-cement specimens considering the weight of the water that reacts in the hydration process of OPC.

According to the Colombian standard of the National Roads Institute (43), there are two soil-cement classes depending on the design criteria for the mixture: soil-cement durability (SC-D) and soil-cement durability-resistance (SC-R) (Table 4). Based on the above, clayey soil A-7-5 stabilized with alkaline cementitious material 30F1L could be a viable alternative for pavement design following durability criteria SC-D, since the mass loss of this soil is only 3.74%, lower than the maximum mass loss allowed by the Colombian standards, and the mass loss of this soil is similar to that of the reference soil stabilized with OPC. In addition, the change in volume of this soil under both expansion and contraction is very similar to that of the reference soil stabilized with OPC.