Physical, mechanical, and microstructural characteristics of fly ash replaced cement deep mixing columns

The soil samples were collected from Konya, Turkey by open excavation at 6–7 m depth. The soil was classified as high plasticity clay (CH) according to grain size distribution and consistency limits (Table 1). The clayey soil contains 25.0% CaO, 28.0% SiO₂, 7.7% Al₂O₃, and 4.46% Fe₂O₃. The soil particles have a flocculent structure and mostly face-to-face contact (Fig. 1a). Natural clay at the sampling depth has stiff consistency since liquidity index (LI) is between 0 and 0.25 (Sowers 1979). LI is used to scale clay water content to its consistency limits (ASTM D4318 2017):where, w is moisture content, LL is liquid limit, PI is plasticity index (PI = LL-PL), and PL is the plastic limit.

The primary binder used in the slurry is CEM-I 42.5 R type ordinary Portland cement. The cement contains 63.1% CaO, 14.6% SiO₂, 2.78% Al₂O₃, and 4.23% Fe₂O₃. The cement gains high strength in early curing periods and complies with the ASTM C150 (2021). The initial set and final set of the cement are 200 min and 240 min, respectively. The specific gravity of it is 3.2. The Blaine fineness (specific surface area) of cement is 355 m²/kg. The diameter of cement grains ranges from 1–60 μm (Fig. 1b). In addition, the amounts of C₃S, C₂S, and C₃A in the cement are 61.41%, 11.33% and 8.2%, respectively. While C₃S and C₃A enable to gain early strength, C₃S and C₂S cause it to gain final strength.

The secondary binder in the slurry is fly ash and it is obtained from Seyitömer thermal plant (Kütahya, Turkey). The fly ash contains 54.49% SiO₂ (S), 20.58% Al₂O₃ (A), and 9.27% Fe₂O₃ (F). Moreover, reactive SiO₂ and CaO contents are 39.01% and 2.49%, respectively (Türker et al. 2009; Yücel and Erten 2023). It is classified as F class (low lime) fly ash since the amount of SiO₂ + Al₂O₃ + Fe₂O₃ is 84.34% (S + A + F > 70%, ASTM C618-17 2018) and the amount of CaO is less than 18% (ASTM C618-23 2023). The specific gravity of fly ash is 2.13. The Blaine fineness of fly ash is 495 m²/kg. The diameter of fly ash grains ranges from 1–30 μm, and the grains are in a mixed form with spherical and irregular geometries (Fig. 1c).

Modified phosphonate-based superplasticizer additive of CHRYSO Optima 100 (CHRYSO 2023) was used to decrease the viscosity of the slurry containing high volume of fly ash and low water/binder ratio. High volume fly ash materials contain at least 40–50% replacement ratio of fly ash (Malhotra 1990; Park and Choi 2021). The density of the superplasticizer is 1.06 g/cm³ and the pH value is 4.

DM columns were installed by using the wet deep mixing technique. DM equipment consists of a mixing machine and a slurry batch plant. Binder slurry was prepared in the slurry batch plant and then the DM machine was used to inject and mix the slurry with the clayey soil. The DM machine comprises a frame, a mixing head, a drilling shaft, and a control panel (Fig. 2). The mixing head and drilling shaft are fixed to the machine frame by a mechanism of rack gear and a linear guide system. The mixing head can move 170 cm vertically. The rotation speed and the penetration/withdrawal speed of the drilling shaft can be adjusted from the control unit in the range of 0–100 rpm and 0–150 cm/min, respectively.

The slurry batch plant consists slurry pump, batching vessel, and power supply panel. The injection pressure of the transmitted slurry by the slurry pump is 30 bar. The slurry pump transfers the prepared slurry to the drilling shaft and mixing head.

The mixing blade is designed to disaggregate and shear the untreated clay mass during DM column installation. The diameter of the mixing blade, which rotates in one direction, is 30 cm. The upper and lower mixing blades are spaced 10 cm apart and tilted to a horizontal level with an angle of 0° and 45°, respectively. During the DM column installation, the slurry was injected into the clayey soil during penetration (top-to-bottom construction technique). Therefore, there are four injection nozzles at the bottom of the lower mixing blade.

The DM column was installed using a single drilling shaft has a single mixing blade, squirting the binder slurry into the soil, and conducting the mixing stage of soil and slurry only throughout the mixing blade. This installation method is the wet-rotary-end (WRE) technique (Bruce 2000). The installation procedures of columns in the laboratory comprise creating untreated soil mass, preparation of the slurry, setting tooling parameters, and mixing stage of soil.

Design of experiments (DOE) is a statistical procedure that deals with planning, conducting, analyzing, and interpreting controlled tests to evaluate the relationship between factors and outputs. One of the most powerful optimization techniques is the Taguchi method (Taguchi 1988; Taguchi et al. 2004). Taguchi method is used to determine the optimum levels of parameters controlled in the process, and it has been widely used in quality engineering and robust design.

The present study aims to optimize the slurry parameters considering the untreated clay consistency affecting the installation process of DM columns that have maximum unconfined compressive strength (q_u) and ratio of soil improvement (RI). In this geotechnical problem, the Taguchi analysis stages are given below in brief:

A series of tests are required to determine the density, viscosity, and bleeding characteristics of freshly mixed slurry. A pycnometer was used to measure the slurry density. In order to determine the viscosity, a slurry (1500 ml) is poured into a Marsh funnel and is then allowed to flow into a graduated cup. The time for the slurry to fill a volume of 946 mL is measured as the Marsh funnel viscosity (ASTM D6910 2019). The bleeding characteristic of the slurry is determined by introducing the slurry into a 1000-mL graduated cylinder and the volume change in the sample was followed for 2 h (ASTM C940-16 2016).

Immediately after the installation of the DM column, bulk samples were collected from the fresh soil-slurry mixture, then the density and water content of the samples were measured. Then, the DM column was left to cure for 28 days in the natural untreated soil mass. After 28 days, the test tank was opened, the clayey soil adhering around the column was peeled off, and 28-day-core samples were taken from half of the column (Fig. 3c). The rest of the column was bagged and continued curing. At the end of 56 days, core samples were taken again. While the diameter of the core samples was 50 mm, they were cut to 100 mm in length with a chainsaw. The bulk density, moisture content, and unconfined compressive strength (UCS) of all core samples were measured (Figs. 3d-3f). ASTM-D2166 (2016) was followed to carry out UCS experiments. The test specimens were loaded at a constant 1 mm/min speed which corresponds to an axial strain rate of 1%/min and is in compliance with the ASTM standard (0.5–2%/min strain rate). Moreover, the consolidation and permeability characteristics of some selected core samples were determined following the ASTM D2435-04 (2020) and ASTM D5084-10 (2016) standards, respectively.

The micromorphological change of the DM samples was performed by scanning electron microscope (SEM) imaging in a ZEISS EVO LS10 field emission machine. After 28 and 56 days of DM installation, a 1 × 1x1 cm cubic piece was taken from the shear surfaces of UCS test samples and SEM images were obtained by 10,000 magnifications.

During the installation of the DM column, the slurry must have a certain viscosity so that the injection nozzles are not blocked, the slurry can be easily transferred to the ground, and the soil-slurry mixture can be mixed homogeneously.

Mix proportions and physical properties of slurries are presented in Table 6. Marsh funnel viscosity values of the slurries were measured between 26.99 s in design 9 and 69.24 s in design 12. The slurry in Design 9 has no FA (only cement in dry binder), while the SP content is 3% and W/B is 1.4. Due to the absence of FA and the high amounts of SP additive and W/B, the viscosity of the slurry in Design 9 was the lowest. The slurry in Design 12 contains 40% C, 60% FA, 1% SP and the W/B ratio is 0.8. Due to the high FA content in this slurry and the low amounts of SP additive and W/B, the viscosity of the slurry in Design 12 was measured high. The viscosity of the slurry is affected by the amounts of FA (also cement) and SP and the W/B ratio. In the L16 design table, the mix proportions of materials are varied in each design. For a more detailed evaluation, viscosity values are presented in Fig. 4a. Viscosity values of slurries having the same W/B ratio decrease as the FA amount decreases, even if the SP amounts are varied. When the W/B ratio is 0.8 and the amount of FA decreases from 60 to 0%, the viscosity decreases by 50%. When the W/B ratio is 1.0, 1.2, and 1.4, the decrease in viscosity is 21, 14 and 10%, respectively. As the FA replacement ratio increases, the resistance of the grout against deformation under surface tension (viscosity) increases because the specific surface area (Blaine fineness) of FA (495 m²/kg) is 1.4 times greater than cement (355 m²/kg). As the W/B ratio increases, the interaction between the binder particles weakens and the viscosity decreases as the slurry becomes more fluid. Marsh funnel experiments were performed on all combinations of slurries containing 60% FA to examine the effect of SP on viscosity (Fig. 4b). While the SP ratio increases from 0 to 5% and the W/B ratio is 0.8, 1.0, 1.2, and 1.4, the viscosity decreases by 45, 30, 5, and 4%, respectively. SP weakens the interaction between binder particles and water by covering the surface of binder particles. Therefore, the viscosity decreases since the amount of SP in the slurry increases. SP additive reduces the viscosity of all slurries containing only cement or high volumes of FA. In addition, the change in viscosity is insignificant since W/B ≥ 1.2 (Figs. 4a-4b). Considering that the slurry pump of the DM machine can easily pump slurries with a viscosity of 120 s, all slurries satisfy the viscosity criteria since the Marsh funnel viscosity of the water is 26 s.

The bleeding water feature of the slurries is inversely proportional to the viscosity. The lowest and highest bleeding water capacities were measured in design 12 (highest viscosity) and design 9 (lowest viscosity), respectively. After 2 h, most slurries except designs 7, 12, and 14 are unstable because their bleeding water capacities are greater than 10% (Table 6). However, this is not important in DM column installation because the slurry is mixed continuously until it is pumped to the ground. In addition, the time of DM column installation in the laboratory is up to 15 min. During this period, the slurries are stable because the bleeding water of the slurries is smaller than 4.3%. In the field, installation time can be up to 30 min, even in which case the bleeding water capacity of the slurries is less than 10%.

The densities of the slurries vary between 1.322 g/cm³ (design 4) and 1.618 g/cm³ (design 1) according to C, FA, SP, and water contents (Table 6). The slurry in design 4 contains 40% C, 60% FA, and 5% SP, and the W/B ratio of the slurry is 1.4. The densities of C, FA, SP, and water are 3.2, 2.13, 1.06, and 1 g/cm³, respectively. Since the amount of FA, SP, and water per unit volume was at the maximum level in design 4, the slurry density was obtained at least. While there is no FA and SP (only cement as the binder) in the slurry in design 1, the W/B ratio is the lowest level of 0.8. Therefore, the maximum slurry density in design 1 was obtained at 1.618 g/cm³. No correlation could be established between the density of the slurry and the viscosity (or water bleeding capacity) because SP does not change the density substantially, but it affects the viscosity considerably.

Binder content (a_w), volume ratio (VR), total water/binder ratio (W_T:W_B), LI of fresh soil-slurry mixture (LI_mix), and blade rotation number (BRN) values are required for DM column installation (see Table 7). The a_w is the dry mass ratio of binder (total mass of C + FA in Table 6) to untreated clay in the column region. Dry mass of untreated clay is calculated by using initial conditions of soil mass (Table 2) in the test tank since clay volume in the column region (0.057m³), natural unit weight, and moisture content is known. The a_w values change between 14.6% (design 1) to 37.9% (design 14). Binder dosage (BD) is binder amount in unit volume of column, but a_w is the dry mass ratio of binder to clay. VR is the volume ratio of the slurry (total volume of C + FA + SP + W in Table 6) to the column (0.057m³). VR values varies between 22.3% (design 1) to 75.9% (design 15). The W_T:W_B of the fresh mixture is the ratio of the total water mass in the mixture (total water from the slurry and clayey soil) to the binder mass and changes between 2.17 (design 14) and 4.32 (design 4). Immediately after column installation, the water content of the freshly mixed soil-slurry mixture was measured and LI_mix was calculated by scaling it with the consistency limits of the clay (Eq. 1). The LI_mix varies between 0.61 (design 1) and 1.46 (design 5). The notations of a_w, VR, W_T:W_B, and LI_mix are presented in Table 3. BRN is the total rotation number of the drilling shaft for unit length of column during penetration and withdrawn (Eq. 3) (Mizuno et al. 1988). During installation of DM columns, BRN values varied between 252 rev/m (design 1) and 858 rev/m (design 15).where ΣM = total number of mixing blades, N_p, N_w = rotational speed during penetration and withdrawal, respectively, in rpm, and V_p, V_w = velocity of mixing blade during penetration and withdrawal, respectively, in m/min.

The variations in moisture content and bulk density values of unstabilized clay, freshly mixed soilcrete, and 56-days-cured core sample (hardened soilcrete) give the time-dependent physical property of DM columns considering mix proportions (Fig. 5). The moisture content values of the unstabilized clay having LI values between 0.4 and 1.0 vary between 35 and 52%. Depending on the installation parameters, the moisture content of the freshly mixed soilcrete increases by 16.3% in design 1 and 70.9% in design 15 (an average of 31.9%) compared to the unstabilized clay. The moisture content of the 56-days-cured core samples decreases by an average of 7.4% compared with the fresh soil-slurry mixture. During the curing period, descending free water amount chemically adheres to the cementation products because free water is utilized in the hydration/pozzolanic reactions between cement, fly ash, and clay. Bulk densities of fresh and hardened soilcrete samples vary between 1.602–1.811 g/cm³ and 1.634–1.828 g/cm³, respectively.

The mixing efficiency of the untreated soil and the slurry was investigated to decide the minimum number of mixing by visual observation, which is the judgment that fresh soilcrete had a creamy appearance during the installation of the column. In designs 1, 2, 7, and 8, a homogeneously mixed fresh soilcrete was reached after 6 mixing processes (three times penetration and withdrawn of mixing blades). The LI of untreated soil was 0.4 or 0.6 and the volume ratio (VR) varied between 22.3 and 39.5% in these designs. The volume ratio of the slurry to the clayey soil in the column region gives the VR (Table 7). A homogeneous mixture of soil and slurry was observed after 4 mixing processes (two times penetration and withdrawn of mixing blades) in designs 3, 10, 12, and 15. In these designs, LI was 0.4 or 0.8 and VR varied between 32.1 and 75.9%. In the other designs, a homogeneous mixture was observed after 2 mixing stages (one time penetration and withdrawn of mixing blades) since LI was 0.8 or 1.0 and VR varied between 37.1 and 68.3%. 2 mixing stages consist of mixing the slurry with the clayey soil by the mixing blades through they are penetrated from the ground surface to the bottom of the column and withdrawn to the ground surface again.

Observations during column installation showed that the time to achieve a homogeneously mixed fresh soilcrete depended on the untreated soil and slurry parameters. The soil was sheared, crumbled and clay lumps were formed during the first mixing process. Clay lumps were not formed in the liquid consistency clay (LI = 1), but they were in stiff consistency with large sizes in clays had a LI of 0.4. During the first mixing process, the binder slurry could not interpenetrate to the clay lumps immediately. The slurry came into contact with the clay lumps better and the dimensions of them decreased since the mixing number increased. A homogeneously mixed fresh soilcrete was obtained easily in soft clays but a longer time of mixing stage was required in stiff clays. It is concluded that the minimum quantity of mixing process to reach a homogeneously mixed fresh soilcrete depends on the LI of untreated soil and VR of slurry since the RS of the drilling shaft and slurry outlet pressure are constant. Therefore, a key parameter (LI∙VR) is defined to decide minimum number of mixing processes. If LI∙VR is greater than 0.35, 2 times of mixing is sufficient. 4 times of soil mixing is necessary when LI∙VR is between 0.2 and 0.35. If LI∙VR is smaller than 0.2, a minimum of 6 times mixing is required to obtain a homogeneous mixture (Fig. 6).

The mechanical properties of each column, which were the unconfined compressive strength and the soil improvement ratio, were estimated at the end of the 28-day and 56-day curing periods (Table 8). The ratio of the average column strength (q_u) to the shear strength of the unstabilized clay (q_u,s) was the soil improvement ratio (RI). The number of core samples taken from each column is 10–17 in 28 days and 12–20 in 56 days (26–36 from each column in total). The average column strength was obtained by taking the average of the strengths of the core samples taken from each column during the relevant curing period.

The q_u values of DM columns varied between 491.2–3269.8 kPa at 28 days and 641.2–3837.1 kPa at 56 days. The ratio of maximum core strength to minimum one (q_u,max/q_u,min) was between 1.2 and 2.5. The coefficient of variation (COV) values varied between 0.11 and 0.38. The COV is the ratio of standard deviation to the mean of all core strength data in each DM column. The higher COV value represents a greater level of dispersion around the mean since the lower COV value represents the strength data distributed close to the mean. In field cases of the DM method, the COV value for the strength variability changed from 0.34 to 0.79 (Filz and Navin 2006). The geological conditions of the site might be highly variable in the field, but DM columns were installed in a homogeneous clay mass in the laboratory. Therefore, the strength variability of DM columns had lower COV values in the laboratory than in field cases. COV values obtained in the laboratory demonstrated that the mixing efficiency was high, and also soil and slurry were mixed highly productively along the column length.

Minimum and maximum RI values obtained at design 1 (1415% in 28 days and 2173% in 56 days) and design 14 (47,388% in 28 days and 55,610% in 56 days), respectively. The lowest and highest strengths were in designs 4 and 14, respectively. In design 4, the strength is minimum since the BD is minimum (200 kg/m³) and the levels of other parameters are maximum (FA = 60%, W/B = 1.4, LI = 1). In design 14, the BD is maximum (425 kg/m³), the FA ratio is 20% and the W/B ratio (0.8) is minimum. In addition, the W_T:W_B ratio is maximum in design 4 (4.32) and minimum in design 14 (2.17). When total water amount in the mixture (design 4) is high, the clay minerals lose their electrostatic attraction force and the void ratio also increases due to the excess water in the mixture. Consequently, the clay-clay contact surfaces where cementation or pozzolanic products can form are reduced and the strength decreases (Lorenzo and Bergado 2004, 2006). In contrast, the strength increases when the total amount of water in the mixture (design 14) decreases. However, W_T:W_B ratio is not only parameter but also BD and FA affect strength. The relationship between q_u, BD, and W_T:W_B is given in Fig. 7a. In general, strength increases as BD (or a_w) increases and W_T:W_B decreases. Many scientists have reached similar results (Miura et al. 2001; Filz et al. 2005). Although the strength increases when BD increases, the effect of cement and FA on the strength can be determined as a result of optimization studies. FA decreases the strength when used more than a certain amount even if BD is high (as in Design 16). The total amount of water in the mixture is used both in hydration reactions and is necessary to obtain a homogeneous soil-binder mixture (Porbaha et al. 2000; Miura et al. 2001; Lorenzo and Bergado 2004). There is also a relationship between q_u, BD, and LI_mix. The highest strengths were obtained LI_mix = 1.25∙LI (Fig. 7b). The highest strength in cement DM columns was obtained when LI_mix is equal to LI (Lorenzo and Bergado 2004, 2006). However, LI_mix increases up to 1.25 in FA replaced cement DM columns because FA has a higher water absorption capacity (specific surface area) than cement. In addition, RI increases when both conditions are met as LI_mix increases and W_T:W_B decreases (Tables 7 and 8). In conclusion, column strength is affected by all of the design parameters (BD, FA, SP, W/B, LI, W_T:W_B, LI_mix).

BRN value is another parameter that may affect column strength. Average BRN values are 335 rev/m (designs 1–4), 460 rev/m (designs 5–8), 585 rev/m (designs 9–12), and 710 rev/m (designs 13–16) (Table 7). Previously, it was concluded that an increment of BD increases column strength. The increase in column strength has also been attributed to the increase of BRN since the average of it increases from 335 to 710 rev/m while BD increases from 200 to 425 kg/m³. In the present study, a homogeneous soil-slurry mixture is obtained at least 252 rev/m of BRN value (design 1) for cement DM columns. In addition, the minimum BRN value for 20% FA replaced cement DM columns is 306 rev/m (design 2). The BD is minimum (200 kg/m³) in these designs. However, the VR of slurries are 22.3% and 27.1% in designs 1 and 2, respectively. This means that the slurry volume in design 2 is 22% greater than design 1. BRN of design 2 is also 22% greater than design 1. Therefore, BRN depends on the BD and VR since the rotation speed of the drilling shaft is constant. Increments of BRN (also BD and VR) lead to increase mixing efficiency and to obtain a homogeneous soil-slurry mixture easier (Larsson 2005; Larsson et al. 2005; Topolnicki 2010). FHWA (2013) recommends a minimum BRN value of 360 rev/m for the cement DM method as a result of field studies, but Topolnicki (2010) suggests that the minimum BRN value should be 430 rev/m. However, FHWA (2013) and Topolnicki (2010) did not provide any information regarding the volume ratio (VR) or binder dosage (BD) for which the minimum BRN value applies. In this study, BRN was in a linear relationship with VR. The BRN value can practically be calculated with BRN = 11.3VR (Fig. 8). This relationship can be modified and used to determine appropriate installation parameters to obtain a homogeneous soil-slurry mixture in field applications. The laboratory type slurry batch plant has a pumping capacity of 90 lt/min but the slurry outflow rate from injection nozzles is 20 lt/min. In the field, the binder plant delivers a slurry of 80–150 lt/min (BS EN 14679 2005) but slurry outflow rate from injection nozzles depends on the machinery parameters (number of drilling shaft, mixing blade, and injection nozzles).

Despite differences in untreated clay and slurry properties in DM column installation, column strengths in some designs such as 7 (1598.8 kPa) and 8 (1519.5 kPa) for 28 days and 8 (2463.0 kPa) and 16 (2429.6 kPa) for 56 days were very close. For this reason, q_u and RI results were used in statistical evaluations as the response characteristics.

Evaluation of response characteristics to fit any distribution is performed by multiple probability graphs since there are two curing periods for each response (Fig. 9). Multiple probability graphs are obtained by Minitab software. In these graphs, the probability percentage of response values in the data set which is less than or equal to it plots along a fitted distribution line (middle straight line). In addition, the probability boundaries of each response (curved lines) for the 95% confidence intervals (CI) could be displayed. How well the data follow a particular distribution is decided by Anderson–Darling statistics (AD and p-values) (Anderson and Darling 1952). If the AD value is small and the p-value is greater than 0.05 (for 95% CI), the dataset (test results) is considered to fit the specified distribution. The data of q_u response used in the statistical analysis follow a normal distribution because smaller AD values (0.172 in 28 days and 0.160 in 56 days) are obtained. Moreover, the p-values greater than 0.05 (0.914 in 28 days and 0.935 in 56 days) (Fig. 9a). The RI data fit the lognormal distribution (AD = 0.466–0.530, p-value = 0.218–0.149) (Fig. 9b). Probability graphs show that the distribution of q_u and RI values obtained at 28- and 56-days curing times is significant.

Permeability tests were carried out for DM columns of design 4, design 14, and optimum design. Designs 4 and 14 belong to the columns with the lowest and highest strength, respectively, in the L16 experimental design. Permeability tests were performed on three samples of each for the upper, middle, and lower regions of the columns, and the average permeability coefficient (k) values are presented in Table 10.

The surrounding soils of DM columns in 4, 14, and optimum designs were a liquid consistency (LI = 1). The permeability coefficient of soil was 0.529∙10^–8 m/s. Regardless of the installation parameters, the permeability of the columns was lower than the surrounding soil. In addition, the permeability coefficient of the columns decreases (about 35–50%) and becomes more impermeable as the curing time increases. This result shows that the hydration reactions continue and the voids between the clay particles decrease over time.

The permeability of the column in Design 14 is 100 times smaller than in Design 4 since the strength in Design 14 (BD = 425 kg/m³) is 6–7 times higher than in Design 4 (BD = 200 kg/m³). The permeability of the column in the optimum design is 30% less than that of design 14. The installation parameters for design 14 and optimal design are the same, except for the FA replacement ratio (20% for design 14 and 40% for optimal design). The interesting result here is that the permeability decreases even though the 28-day strength is smaller at the optimum design (2814.5 kPa) than at design 14 (3269.8 kPa). FA decreases permeability even if it decreases the strength in the early curing period. Meanwhile, FA fills the voids and reduces permeability since FA is a finer material than other materials. In the optimum design, an almost impermeable column is installed since the permeability coefficient of the DM column is quite low (< 10^–11 m/s).

The permeability coefficient of cement DM columns in the field varies between 10^–8-10^–9 m/s (O’Rourke and McGinn 2004). In this respect, more impermeable columns can be installed with FA-replaced cement DM columns.

The one-dimensional compression curves of hardened soilcrete belongs to design 4, design 14, and optimum design are presented in Fig. 12 In design 4, the consolidation curve exhibits yielding after a certain stress which closes to the strength of hardened soilcrete since the 28-day and 56-day strengths of design 4 were 491.2 kPa and 641.2 kPa, respectively. The consolidation curves of design 14 and optimum design continue linearly with low slopes and do have not a yield point because the strengths of hardened soilcrete are well above the maximum stress in the consolidation test. In addition, the compression index (C_c) of the hardened soilcrete decreases as the curing time increases for the same design. Moreover, if the strengths of the hardened soilcrete are ordered from low to high as design 4, design 14, and optimum design, C_c decreases as the strength increases. As a result, the C_c values of the DM column in the optimum design were obtained at the end of 28 and 56 days as 0.055 and 0.048, respectively, which means that the compressibility of the column is very low.

The microstructure of the hardened soilcrete was examined with SEM images according to curing time and design parameters. In the microstructure of soilcrete samples of design 4, CSH gels, ettringite needles, and large voids between solid masses are observed. The strength of the DM sample in design 4 is low, compression index and permeability coefficient are high since BD is minimum (200 kg/m³), FA (60%) and W_T:W_B (4.32) are maximum. Moreover, FA particles appear not to be hydrated due to the high amount of FA in the mixture (Figs. 13a-b).

In design 14, CSH and CAH gels, CH plates, and ettringite needles are observed in the microstructure of the DM samples. All FA particles were hydrated. The presence of CH plates in this design indicates that they cannot transform into CSH or CAH gels because there is not enough SiO₂ and Al₂O₃ in the column microstructure. In other words, if there was more FA in the mixture (20% FA in design 14), CSH and CAH gel formation would increase. This situation is evidence of 40% FA in optimum design. In addition, the strength is high, and the compression index and permeability coefficient are low since the samples have fewer voids and a compact structure (Figs. 13c-d).

CSH gels, CAH gels, and ettringite needles were observed in the microstructure of the hardened soilcrete in the optimum design. During the curing period of 56 days, the amount of ettringite needles and voids decreased, as well the amount of hydration gels increased (Figs. 13e-f). In the SEM image of this design, CSH gel formation on the surface of the fly ash and clay particles was detected because of the ongoing hydration reactions on the 28th day (Fig. 13e). The highest strength, lowest permeability, and lowest compressibility are obtained in optimum design since BD is maximum (425 kg/m³), FA amount is optimum (40%), and W_T:W_B is minimum (2.20).

SEM images clearly show that when the curing time increases, the voids between solid masses decrease as well the amount of hydrated gel increases (Fig. 13).

CSH and CAH gels are the main cementation products responsible for improving the mechanical properties of cementitious stabilized materials (Wu et al. 2022). Ettringite needles are formed because of the natural process of hydration reactions of cement-based materials (reaction of C₃A, C₄AF, CSH₂, and CaSO₄ in cement). Gypsum (CSH₂) and lime (CaSO₄), which cause the formation of ettringite, are necessary to regulate the setting time of cement (Soomro et al. 2023), but in the presence of sulfate-rich groundwater, ettringite expands and causes deterioration (expansion, cracking, and mass loss) in the internal structure of the cement-based materials (Puppala et al. 2018; Akula and Little 2020). Sulphate-resistant cement (C₃A < 5% and C₃A + C₄AF < 25% according to ASTM C150 (ASTM C150 2021)) should be used in construction works where there is sulphate-rich groundwater that will increase the ettringite formation.

The study clearly shows that optimum slurry parameters (binder dosage of 425 kg/m³, FA replacement ratio is 40%, SP content is 3%, water/binder ratio is 0.8) maximize the strength and minimize the permeability and compressibility characteristics of FA-replaced cement DM columns. The feasibility of FA-replaced cement DM method is determined by comparing the slurry cost with the cement DM method (Table 11). For 1 m³ soil stabilization with DM method, the slurry has only cement (425 kg) in cement DM method but 255 kg cement (60%) and 170 kg fly ash (40%) in FA-replaced DM method. Super plasticizer and water masses can be determined as 12.75 kg and 340 kg, respectively, for both methods (see Tables 3 and 11). In the FA-replaced cement DM method, binder cost and slurry cost are reduced by 33% and 28%, compared to the cement DM method.