Effects of Cement Treatment on Water Retention Behavior and Collapse Potential of Gypseous Soils: Experimental Investigation and Prediction Models

,


Introduction
Gypseous soils refer to soils containing enough gypsum content to change or affect their engineering properties and are widespread in many regions worldwide (e.g., it covers approximately 33% of Iraq's areas) mainly in unsaturated conditions.They are usually characterized as collapsible and problematic soils.The primary geotechnical issue with these soils is their susceptibility to significant shear strength reduction, volume reduction, and progressive settlements when exposed to water flow caused by rainfall, irrigation, broken water lines, ground water rise, and so on [1][2][3][4][5][6].
Recent studies have shown that the chemical stabilization can enhance the characteristics of collapsible soil [5,[7][8][9].Hayal et al. [3] investigated the effects of nanosilica on the reduction of collapse potential (CP) in gypseous soil and found that even a small amount of nanosilica could reduce CP by up to 91%.However, the addition of further stabilizer can increase the CP.In another study, Ibrahim [4] examined the potential for enhancing gypseous soil by mixing it with cement at various concentrations.The results showed that the reduction in collapsibility was about 87%-92% for a 12% mixed cement to soil ratio.
The previous investigations on gypseous soils have primarily concentrated on their response in either dry or saturated states.Nevertheless, in natural environments, soils undergo variations in water content, and soil layers near the surface remain in unsaturated state during certain periods throughout the year.Recognizing the distinctions in mechanical behavior between saturated and unsaturated soils, the differentiation between these states is considered crucial [10].
Matric suction (the difference between pore air and water pressures (ua−uw)), which relates to the soil-water content, is a critical factor in the mechanical behavior of unsaturated soils.The soil-water retention curve (SWRC), defines the relationship between soil suction and its water content [10][11][12][13].The creation SWRCs enable the geotechnical experts to characterize the suction behavior of unsaturated soil and effectively address the diverse geotechnical engineering challenges related to slope stability, foundation bearing capacity, seepage, permeability, and shear strength properties of unsaturated soil [14][15][16][17].While significant advances have been made in the field of unsaturated soil mechanics in recent years, there is still a notable gap in our understanding of the unsaturated water retention behavior of cementitious stabilized soil, especially in gypseous soil.Despite its common use in various ground improvement projects, this soil type has not been thoroughly studied in terms of its suction characteristics.In this regard, Emery et al. [18] studied the effects of stabilization with cement and lime on the SWRC of clayey soil.Their comprehensive laboratory testing and data analysis revealed that while stabilization with lime reduced the air-entry potential, stabilization with cement enhanced it.
1.1.Objectives and Outline of This Study.As mentioned before, the behavior of gypseous soils in unsaturated conditions and their water retention potential under different matric suctions remain largely unknown.To address this critical knowledge gap, this research aims to deepen our understanding of gypseous soils with different gypsum contents and their water retention potential which serves as a fundamental factor which control the unsaturated soil behavior.It also seeks to advance predictive capabilities through the development or refinement of PTFs for estimating the SWCC of gypseous soils.Moreover, this study evaluates their susceptibility to collapse, and the potential of cement stabilization as a mitigation technique and assesses the impact of cement stabilization on the soil-water retention potential.
To these aims, the study employs single oedometer tests (SOTs) to determine the optimum soil-cement mixture for mitigating the CP of gypseous soils.Additionally, it explores and compare the water retention behavior of pure and cement-treated gypseous soils, utilizing a controlled-suction oedometer apparatus.To facilitate predictive modeling, an evaluation of various existing PTFs is conducted to identify or refine a prediction method that can accurately estimate the water retention curve of gypsum soil.

Soil and Material
2.1.1.Gypseous Soil.The soils used in this study are collapsible sandy gypseous soils.Soil samples were collected from two different locations: Karbala holy city (G1) with 18% gypsum content and Tikrit university site (west of Iraq) (G2) with 66% gypsum content.All samples are representing disturbed soils.The sand replacement method was used to calculate the soil's field dry density, and the values were 1.61 gm/cm 3 for G1 and 1.638 gm/cm 3 for G2.The physical parameters of the soil are presented in Table 1, and the grain size distribution for the G1 and G2 samples is shown in Figure 1.To ensure comparability, both soils were prepared with similar gradations using sieve analysis in accordance with the American Society for Testing and Materials' requirements, both soils are classified as poorly graded sand (SP) based on the unified soil classification system according to ASTM D-422 [38].[39] for the pure gypseous soils G1 and G2, as well as gypseous soils treated with cement "G1 + 1%, 3%, and 5%" and "G2 + 1%, 3%, and 5%", to obtain the optimum moisture content (OMC) and maximum dry density.

Single Oedometer Test (SOT)
. SOTs were performed according to ASTM D5333 [40] to investigate the collapsibility of gypseous soils under saturated conditions.To obtain the optimum soil cement percentage for reduction of CP, the soil was treated with different cement contents (i.e., 1%, 3%, and 5%) for these tests, and the CP results were compared.The soil utilized passed sieve # 4 (4.75 mm), and the size of the samples was 50 mm in diameter and 20 mm in height.
They were prepared at a density of 90% of the maximum dry density with OMC obtained from standard compaction test for pure soils (G1 and G2).The pure soils were tested immediately after preparing the samples, while the treated soil samples were kept in desiccator for 7 days for curing.
During the tests, the soil samples underwent incremental loading at initial conditions until reaching a vertical stress of 200 kPa with a load increment ratio of 1. Subsequently, the samples were soaked in water for 24 hr, and the additional settlement at 200 kPa resulting from the soaking process was measured.The testing continued with further loading and unloading stages.
The CP was calculated using the following equation: Tables 3 and 4 present two criteria usually used to classify the gypseous soils.

Controlled-Suction
Oedometer.In the current study, a controlled-suction oedometer apparatus was used to obtain the SWRC for pure gypseous soils (G1 and G2) and the improved gypseous soils with optimum cement contents (G1 + optimum cement content) and (G2 + optimum cement content).The schematic shape and the control panels of this device are illustrated in Figures 2 and 3, respectively.The unsaturated equipment was adjusted and calibrated as defined by Fredlund and Rahardjo [12,13].Soil was dried in an oven with 45°C and then samples were prepared with 90% of maximum density and OMC, which were obtained from the standard compaction test.As mentioned before, the pure samples were tested immediately after preparing while the samples treated with cement were left for 7 days for curing.The tests were conducted in main wetting path followed by a drying path with volume changes measured throughout.The size of the sample in the oedometer ring test was 5.03 cm in diameter and 1.96 cm in height.
The air and water pressure were set using the data logger, and the change in water volume was measured using the graduated tube.To start the wetting path, an initial 100-kPa suction was applied followed by a step-by-step reduction in matric suction to zero suction.For the following drying path, step-by-step matric suction increment was applied until 100-kPa suction.In each step, the suction was applied until equilibrium was reached, which was determined by monitoring the water level in the graduated tube until no water entered or exited the sample.The water volume change and applied suction were recorded at each step to draw the SWRC.

Test Results and Discussion
3.1.Standard Proctor Test.Table 5 presents the results of adding cement to G1 and G2.The effect of different percentages of cement (1%, 3%, and 5%) on the test results was observed through an increase in density and a decrease in OMC of both soils (1 and 2), as shown in Figures 4 and 5. 6 and 7 illustrate the result of the single oedometer test conducted on G1 and G2 soils, both before and after treatment with cement.The summarized results can be found in Table 6.It can be observed that after adding 3% of cement, the coefficient of collapse decreased from 3.97 to 1.05 for G1, indicating a 72% change in CP.This change represents a transition in the degree of specimen collapse from a moderate level to a slight level, based on the collapse index classification criteria (Ie%-D5333-2003).Similarly, for G2, the coefficient of collapse decreased from 7.04 to 1.28 after adding 5% of cement, resulting in an 82% change in CP.This improvement shifted the CP from troublesome to slight.

Advances in Civil Engineering
In fact, when the gypseous soil samples were fully submerged, the bonds between particles, as well as the gypsum crystals present in these soils, dissolved and brock apart in water; this process weakened the shear strength between the soil grains and led to collapses.However, the treatment of soils by adding cement introduced a different kind of binding that could resist the effects of water.As a result, the CP was reduced when the cement-treated samples were submerged.

The Soil-Water Retention
Curves.In this study, the SWRC of untreated and cement-treated G1 (moderate gypsum content) and G2 (high-gypsum content) soils was investigated in wetting and drying paths using a suction-controlled oedometer device.The differences in SWRC between wetting and drying paths and the observed hysteresis of SWRC is primarily attributed to the ink-bottle effect which is a result of differences in pore-size distribution.Additionally, the generation of trapped air in closed   pores during the wetting process also plays significant roles in these differences [42].Figure 8 illustrates the SWRC for G1 and G2 soils before and after cement treatment, separately for the wetting (Figure 8(a)) and drying (Figure 8(b)) paths.The curves clearly show the influence of cement treatment on both soils and paths.It can be observed that the SWRC curve for treated soil lies mostly above that of untreated soil for both G1 and G2 and for both drying and wetting paths.This can be attributed to the increase in the pore-size distribution index due to cement stabilization.The finer soil matrix in the treated samples exhibits a greater capacity to attract and retain water, leading to a higher position of the SWRC curve.Furthermore, it is noticeable that the effect of cement treatment on the wetting paths is more pronounced than on the drying paths.In other words, the differences between G2 and G2 + 5% in the wetting path are greater than in the drying path, and the same applies to G1.This can be attributed to the solubility of gypseous soil in water.The changes (collapses) in the gypseous soil structure caused by water absorption during the wetting path result in more significant differences between the SWRC of treated and untreated samples compared to the drying path.
To compare untreated and treated samples of G1 and G2 soils separately, their SWRC are shown in Figures 9(a) and 9(b), respectively, for drying and wetting paths.It can be observed that the differences between the wetting and drying curves are more significant in pure samples (G1 and G2) than in the treated soils.This is attributed to the collapsibility potential of untreated gypseous samples during the wetting path.In these untreated samples, the increasing water content during the wetting path gradually dissolves and breaks Advances in Civil Engineering apart the existing gypsum crystals and gypseous bonds between the soil particles.Conversely, in cement-treated samples, the cement bindings can resist the effects of water and remain unchanged during the wetting path.As a result, the wetting path of untreated gypsum soil samples exhibits more significant changes compared to the cement-treated samples.The effectiveness of cement improvement was also observed in the increase of the air-entry value (AEV) for both the drying and wetting paths of G1 (moderate gypsum content) and G2 (high-gypsum content) soils.Table 7 presents the differences in AEV between the treated and untreated samples.
It can be observed that the stabilization with cement resulted in a slight increase in the AEV for both the drying and wetting paths of G1 and G2 soils.This increase can be attributed to the interaction between the cement and soil  Advances in Civil Engineering particles, which enhances the bonding and fills the voids.Consequently, there is an increase in the pore-size distribution index.This indicates that the soil voids are expected to drain water at a slower rate compared to the untreated soil.
It is noteworthy to mention that all the above-mentioned results were more highlighted for G2 soil.This is attributed to its higher gypsum content and resulted CP.

Fitting Curves.
Due to the time-consuming and extensive testing involved in obtaining multiple points on the water retention curve for a specific soil, researchers have developed equations to derive a continuous mathematical representation of the soil-water characteristic curve (SWCC) from a limited number of data points.

Fredlund and Xing Fitting Equation.
In the present study, the equations were proposed by Fredlund and Xing [43].Equation 2 was employed to fit the experimental points obtained through controlled-suction oedometer tests: Where: a: is a related parameter that fits with AEV of the soil measure in (kPa).The point (a, θ(a)) can be used to approximate the inflection point in the SWRC curve.
(ψ i , θ i ): inflection point on the SWRC plot; θ s : is the θ when the suction is zero; n and m: are also fit the parameters; n: is related to the slope of the SWCC, while m is related to the residual water content of the soil; s: is the slope of tangent line (at the inflection point; i.e., (ψ i , θ i ); ψ p : is the intercept of the tangent line and the matric suction axis in (kPa).
Figures 10-13 depict the fitting curves of untreated and cement-treated samples for both gypseous soils based on this equation.The results demonstrate that the water retention behaviors of both untreated and cement-treated gypseous soils were modeled accurately by Fredlund and Xing [43] equation.

Feng and Fredlund Fitting Model and the Simple
Scaling Method.In this study, the Feng and Fredlund [44] SWRC-fitting equation was also evaluated to fit the experimental points obtained through controlled-suction oedometer tests.This method, along with the related scaling method for calculating the main drying and wetting curves from fitted initial drying curve was suggested and used in some previous studies, such as Pham et al. [45] and Johari and Hooshmand Nejad [46].
Following the procedure described by Pham et al. [45] and Johari and Hooshmand Nejad [46], the experimental results of the initial drying paths for pure G1 and G2 soils were fitted and then the main drying and wetting curves were  8 displays the chosen parameters for the fitting curves of Pure G1 and G2.D sL and R sL were set to 2.0 and 0.2, respectively, based on the recommendations from Pham et al. [45] and Johari and Hooshmand Nejad [46] for sandy soils.The obtained results are illustrated in Figures 14(a) and 14(b), where the experimental data points are also included for comparison.
While the initial fitting curve demonstrated acceptable accuracy, it is apparent that the calculated main wetting curve did not align with the experimental results.Notably, the calculated wetting curve is intersecting with both the initial and main wetting curves, which is unacceptable.This variance could be attributed to the fact that in gypseous samples, the increasing water content during the wetting path gradually Advances in Civil Engineering dissolves and breaks apart the bonds between particles, as well as the gypsum crystals present in these soils.As observed before, this unique behavior in gypseous samples has made their behavior different from the regular soil samples.Therefore, this could have led to this different outcome, deviating from the expected results of the mentioned approach.
In an attempt to refine the accuracy of the results, a trialand-error procedure was employed to adjust the values of D sL and R sL , the parameters that control the distance and slope ratio between the two boundary curves on the SWRC curve are shown in Table 8 and Figure 15.Although the resulting wetting curve exhibited improved compatibility with the experimental results compared to the previous curves, it still fell short of the expected accuracy.It is worth mentioning that since the results of the wetting path using this method did not align well with the experimental results, the curves are reported only for the pure G1 and G2 soil samples.
3.5.Assessment of PTFs for SWRC of Untreated and Cement-Treated Gypsum Soil.Laboratory tests such as pressure plate and suction-controlled oedometer are often time-consuming and costly for obtaining SWRC.An alternative is used to utilize PTFs that establishes the correlations between SWRC and basic soil characteristics.In the current study, the applicability of multiple prevalent PTFs proposed by Zapata et al. [31], Scheinost et al. [33], Saxton et al. [34], and Vereecken et al. [47] was assessed and modified to effectively predict the SWRC of gypseous samples with varying gypsum contents, both untreated and treated with cement.Table 9 summarize    Advances in Civil Engineering these PTFs.These modified PTFs were then compared against results from suction-controlled oedometer tests to assess their prediction accuracy (Figures 16-19).
The MSE values obtained from the three abovementioned PTFs are listed in Table 10.Based on the MSE results, all three models provided reliable predictions for the water retention behavior of gypseous soil with different gypsum content levels, including both treated and untreated samples.Among these models, the modified Vereecken et al. [47] model exhibited the highest level of accuracy, as indicated Advances in Civil Engineering 13 by significantly low-MSE values across all soil samples (e.g., 0.000252, 0.00032, 0.000129, and 0.000278).
To provide a comparison of the MSE values considering previous related studies, it is noteworthy to mention that Zapata et al. [31] conducted evaluations on different soils.Their most accurate PTFs yielded MSE values of 0.000431, 0.002581, and 0.004605 for the three different studied soils.Additionally, Shahnazari et al. [21] investigated calcareous soil samples with three different gradation curves, and the most accurate PTFs resulted in MSE values of 0.002761, 0.000180, and 0.000418, which were obtained using the model proposed by Zapata et al. [31].

Conclusions
In this research, the CP of gypseous soils with moderate (G1) and high (G2) gypsum content, as well as their cementtreated samples, was evaluated and compared.The results of SOTs were utilized to assess the CP of untreated and cement-treated samples and determine the optimum  soil-cement mixture for reducing the CP.Additionally, the water retention behavior of treated and untreated gypseous soils was studied using a controlled-suction oedometer, and the effects of cement stabilization on the SWRC in drying and wetting paths were examined.Finally, the performance of widely used PTFs was assessed to discover or modify a reliable method to predict the SWRC of gypsum soil, taking into account both untreated and cement-treated conditions, with satisfactory accuracy.The following results were obtained: (i) Cement stabilization effectively reduced the CP of both G1 and G2 gypseous soils, as indicated by the SOTs results.With the addition of 3% and 5% cement, the coefficient of collapse decreased by 72% for G1 and 82% for G2, respectively.This shift in CP transformed it from moderate to slight for G1 and from troublesome to slight for G2.(ii) The controlled-suction oedometer tests demonstrated that cement stabilization significantly influenced the SWRC.(iii) The SWRC curves of the treated soil, in both the drying and wetting paths, consistently lay above those of the untreated soil, particularly for gypseous soils with high-gypsum content (G2).This can be attributed to the improved bonding and void filling resulting from the soil cementation.As a result, the treated soils exhibited a finer-soil matrix with improved water-retention capacity.(iv) The impact of cement treatment on the wetting paths was more pronounced compared to the drying paths.This can be due to the changes in the gypseous soil structure caused by water absorption during the wetting path, resulting in more significant differences between the SWRC of treated and untreated samples compared to the drying path.(v) The differences between the wetting and drying curves of each specimen were more significant in pure gypseous samples compared to the treated soils.This is attributed to the fact that in untreated gypseous samples, the increasing water content during the wetting path gradually dissolves the gypseous bonds between soil particles.However, the cement bindings in treated samples are able to resist the effects of water during the wetting path, leading to less significant changes in the soil-water retention behavior.(vi) Cement stabilization slightly increased the AEV in both G1 and G2 soils.This is due to the improved pore structure in the treated samples, which increased the pore-size distribution index and subsequently raised the AEV.(vii) The predictability of commonly used PTFs for the SWRC of gypseous soils was assessed, and it was found that the preexisting PTFs, including those suggested by Saxton et al. [34], Zapata et al. [31], and Vereecken et al. [47], provided acceptable predictions for both treated and untreated gypseous soils.Notably, the modified Vereecken et al. [47] model demonstrated the most accurate predictions for all soil samples in this study.

FIGURE 10 :
FIGURE 10: Best fitted SWRC for pure G1 and its equation: (a) drying path and (b) pure G1 wetting path.

FIGURE 11 :FIGURE 12 :
FIGURE 11: The best fitted SWRC for treated G1 and its equation: (a) drying path and (b) wetting path.

FIGURE 15 :
FIGURE 15: Experimental and fitted wetting curve results for pure G1 and G2 (D sL and R sL through trial-and-error).

TABLE 1 :
Physical properties of soils.

TABLE 2 :
Chemical components of cement.

TABLE 3 :
[41]criteria proposed by Jennings and Knight[41]for classification of collapsible soil.

TABLE 6 :
Changes in degree of specimen collapse due to cement treatment (considering ASTM D5333-2003 criteria).

TABLE 7 :
Changing in air-entry values.

TABLE 8 :
Chosen values for the fitting curves parameters for pure G1 and G2 and the calculated results.
Note: * D sL and R sL through trial-and-error.

TABLE 10 :
MSE values for the evaluated PTFs.