Hydro-mechanical coupling and microstructural evolution mechanism of expansive soil under full suction range

Abstract

By combining the pressure plate, saturated salt solution–vapour equilibrium, and WP4C instrument methods, a soil-water characteristic curve (SWCC) for expansive soil at all suctions during drying‒wetting cycles was obtained, and expansive soil hysteresis and volume changes were analysed. Microstructural scanning electron microscopy (SEM) and pore size distribution (PSD) tests were conducted on samples along the wetting path to study microstructural evolution of expansive soil in canal embankments at all suctions. The expansive SWCC at all suctions exhibits hysteresis due to the nonuniform pore distribution, contact angle hysteresis, and trapped air. The energy required for the suction balance differs between the wetting and drying paths. The pore volume of saturated expansive soil is the largest, and the volume decreases with increasing matric suction. Same-void-ratio expansive soils have different degrees of saturation, indicating a significant coupling effect between saturation and void ratio (specific volume) with changes in matric suction. Expansive soils have dual-pore structures with inter/ inter- and intra-aggregate pores. The boundary between the inter/intra-aggregate pores is 0.2 μm. The intra-aggregate pore volume distribution is consistent at all suctions. During initial wetting, the dominant pore diameter decreases and the macropores close. During intermediate wetting, the pore distribution remains relatively unchanged. During final wetting, the pore distribution changes and aggregate expansion deformation and soil expansion occur. Due to differences in the liquid phase properties of macro- and micropores, the pore structure and distribution of expansive soil undergo significant changes with variations in hydromechanical properties.

Introduction

Expansive soils are distributed worldwide and are encountered in engineering projects, such as water conservancy, transportation, and radioactive waste disposal, as impermeable compacted materials^1–3. The South-to-North Water Diversion Middle Route canal passes through expansive soil regions over a total length of 346 km, with strong expansive soil accounting for 6.1%, medium expansive soil accounting for 24.6%, and weak expansive soil accounting for 69.3% of these soils. The construction of this project involves a significant amount of expansive soil. Influenced by factors such as rainfall levels, canal water level changes, and seasonal variations, expansive soil embankments continuously undergo drying and wetting cycles. Different drying‒wetting cycles cause changes in the suction and water content within the soil, leading to significant swelling and shrinkage⁴. Research results show that drying‒wetting cycles increase the cumulative permanent deformation of expansive soil and reduce its elastic modulus under cyclic loading⁴, causing severe deformation damage to many engineering structures^5,6.

Both the macro- and microcharacteristics of the soil are highly important for studying macroscopic deformation mechanisms and establishing microscopic mechanical models of soil^7,8. The macroscopic mechanical behaviour of expansive soil is closely related to its microstructure^9,10. For example, granular expansive soils typically have larger intergranular pores⁹, whereas compacted expansive soils usually have a dual-pore structure dominated by intragranular pores^11,12. Under drying‒wetting cycles, the microstructure of expansive soil significantly changes, and its mechanical properties change accordingly. Conversely, changes in pore structure can alter the hydromechanical characteristics of expansive soil¹³. The response of expansive soil suction to drying‒wetting cycles exhibits significant hysteresis due to the control of its micropore structure¹⁴. Lots of studies have been carried out to quantitively evaluate the water retention hysteretic effect⁸. The water retention characteristics of expansive soil vary under different suction and water content conditions. There is a strong coupling relationship between hydromechanical properties and pore structure in unsaturated expansive soil, which is the basis for establishing the connection between macromechanical and micromechanical properties^15,16. Terzaghi¹⁷and Casagrande¹⁸were the first to recognize the influences of the soil microstructure on macroscopic mechanical soil properties. Lambe¹⁹and Seed and Chan²⁰indirectly confirmed the dependence of the mechanical behaviour of compacted clay on the microstructure through experimental testing. In recent years, advancements in microstructural testing techniques have allowed for the multiscale characterization of expansive soil, including the detailed evaluation of particle or aggregate arrangement, distribution, and connectivity^21–24. However, the relationships between the microstructures and unsaturated hydromechanical properties of cohesive soils are complex, and many controversies remain regarding the coupling between hydromechanical effects and pore structure²⁵. Further experimental data and advanced theoretical explanations are needed.

The mercury intrusion porosimetry (MIP) method is commonly used to measure the pore size distributions of powdered and bulk materials with open and interconnected pore structures^26,27. This method utilizes the nonwetting characteristics of mercury, which is forced into the sample as the pressure gradually increases. The pressure increases from low to high, and mercury sequentially enters the pores. This method has the advantages of high efficiency, high accuracy, and a wide range of pore detection (from a few nanometres to several hundred micrometres), making it a widely used technique for detecting the pore size distributions (PSDs) in porous media. Cai et al.²⁸conducted MIP tests of compacted Guilin red clay samples with a dual-pore structure under different suction conditions and obtained the variations in the pore size distribution curves with suction. Fei et al.²⁹predicted the soil‒water characteristic curve on the basis of the PSD obtained from MIP tests. Simms and Yanful^30,31conducted drying tests of four soil samples via MIP tests and established a soil‒water characteristic curve model considering the pore distribution.

Scanning electron microscopy (SEM) involves a focused high-energy electron beam to scan a sample, exciting various types of physical information through the interaction between the beam and the material. This information is collected, amplified, and reimaged to characterize the microscopic morphology of the material³². Rosenqvist³³was the first to apply SEM technology to study the soil microstructure. Tovey and Sokolov³⁴were the first to quantitatively analyse electron micrographs of the soil structure. Sergeyev et al.³⁵used SEM to analyse clay samples of different ages and origins and reported that clay mainly has honeycomb, skeleton, matrix, turbulent, and laminar structures. Mu et al.³⁶explored the microstructures of sandy loess before and after collapse via SEM and pore and crack image recognition analyses, revealing the collapse mechanism of sandy loess from a microscopic perspective.

The above studies mainly focused on changes in pore distribution curves under mechanical paths and local hydromechanical paths^37,38, whereas relatively few scholars have focused on the changes in soil microstructure under full suction paths^39,40. To study the influences of suction changes on the microstructural changes in expansive soil, it is necessary to conduct quantitative and qualitative analyses of microstructural images of expansive soil and explore the changes in the fabric, pore structure, and particle characteristics of expansive soil under the full suction range. In this work, focus was placed on weak expansive soil from a filled canal embankment in the South-to-North Water Diversion Project. By combining the pressure plate method, saturated salt solution–vapour equilibrium method, and WP4C instrument method, the SWCC of expansive soil under the full suction range during drying‒wetting cycles was obtained. The hysteresis and volume change characteristics of expansive soil during drying-wetting cycles were analysed. Specimens equilibrated at different target suction levels along the wetting path were selected for SEM and PSD tests to study the evolution of the micropore structure in canal embankment expansive soil under the full suction range. The research mainly focused on the explanation of void change and hysteresis behavior under full suction range through systematic experimentation from microstructural aspects. It clearly demonstrates that the dual-pore structure with a boundary at 0.2 μm is the key factor controlling the hydro-mechanical behavior of expansive soil and reveals its three-stage evolution pattern during wetting: pore closure, stability, and significant expansion. The research results further reveal the intrinsic relationships between the microstructures and hydromechanical properties of expansive soil, providing both theoretical insights into the microstructure-property relationships and practical support for engineering applications. For instance, in the South-to-North Water Diversion Project, this study offers experimental evidence for improved numerical modeling of soil-water characteristics and informs the design of mitigation measures like impermeable liners and soil reinforcement to address slope instability caused by cyclic wetting-drying.

Test materials

The expansive soil samples were taken from a filled canal section in Dengzhou that is part of the South-to-North Water Diversion Middle Route Project, as shown in Fig. 1. The embankment fill height is 10–17 m, and the fill soil is classified as ‘weak expansive soil’ according to Chinese standards “Technical code for buildings in expansive soil regions (GB 50112 − 2013)”, typically corresponding to a free swell ratio between 40% and 65%. The test material was weak expansive soil from the embankment of this filled canal section. The expansive soil sample and particle size distribution curve are shown in Fig. 2. The clay mineral content of the expansive soil sample was 49.27%, and the quartz and feldspar content was 50.73%. The basic physical properties of the expansive soil are shown in Table 1, where the specific gravity of the soil particles was G_s=2.7, the natural dry density was ρ_d = 1.62 kN/m³, the natural water content was w = 18%, the maximum dry density was ρ_d = 1.65 kN/m³, the optimum water content was w_o=21.5%, the liquid limit was w_L=49.8%, the plastic limit was w_P=22.8%, and the plasticity index was I_P=27. The free expansion rate is 42.22%, According to the recommendations of the International Society for Soil Mechanics and Geotechnical Engineering (ISSMGE), soils with a free expansion rate between 40% and 70% should be classified as weakly expansive soils. Therefore, the soil involved in this article is weakly expansive soil. To study the microstructural changes in expansive soil in canal embankments under the full suction range, a batch of remoulded ring samples with a dry density of 1.62 g/cm³, a water content of w = 18%, a diameter of 61.8 mm, and a height of 20 mm were prepared via static compaction. The initial water content and dry density values of the remoulded samples were the same as those of the the undisturbed samples and were located on the dry side of the optimal water content. The samples compacted on the dry side of optimum typically exhibit a more aggregated fabric with greater heterogeneity in pore size distribution compared to those compacted at or wet of optimum, which was relevant for studying the evolution of the dual-pore structure.

Fig. 1.

Sampling location of expansive soil.

Fig. 2.

Expansive soil samples and particle size distribution curves.

Table 1.

Basic physical properties of expansive soil.

Specific gravity Gs	Natural dry density ρd/kN/m3	Natural water content w/%	Maximum dry density ρd/kN/m3	Optimal water content wo/%	Liquid limit wL/%	Plastic limit wP/%	Plasticity index IP	Free expansion rate (%)
2.7	1.62	18	1.65	21.5	49.8	22.8	27	42.22

Test methods

The pressure plate method, saturated salt solution–vapour equilibrium method, and WP4C instrument method, three different suction control and measurement techniques, were combined. Some of the remoulded expansive soil samples with a dry density of 1.62 g/cm³ and a water content of w = 18% were placed in a vacuum saturator to prepare saturated samples for drying tests, whereas other soil samples were placed in an oven to prepare dried samples for wetting tests. The SWCC of expansive soil under the full suction range during drying‒wetting cycles was obtained, and the changes in the void ratios under the three suction control techniques were measured. Samples along the wetting path were selected for SEM and MIP tests.

The pressure plate method was used to control the suction range of drying‒wetting cycles from 1 to 480 kPa. Low-suction samples were subjected to suction balance using the pressure plate method. The samples were placed in the pressure plate apparatus for suction balance tests, with the air pressure for the drying path applied at 1→5→10→20→40→80→160→320→480 kPa (the wetting path is the reverse). At each pressure level, the sample was weighed, and the data were recorded after no more water was discharged. This process was repeated for each pressure level. After the final pressure level was reached, the sample was weighed, placed in an oven for 24 h, and weighed again. For the pressure plate method (1–480 kPa), equilibrium at each step was determined when water discharge ceased. The duration varied from 2 to 3 days for lower suctions (1–80 kPa) to 7–10 days for higher suctions (160–480 kPa). The water content, void ratio, and degree of saturation values corresponding to different matric suction levels were calculated on the basis of the recorded data at each pressure level.

The drying/wetting paths in the high-suction range were controlled via the vapour equilibrium method using saturated salt solutions in sealed desiccators. Different salt solutions provided specific relative humidity (RH) environments corresponding to different suction levels according to thermodynamic principles⁴¹. The specific solutions used were K₂SO₄ (3290 kPa), NaCl (38000 kPa), MgCl₂ (149510 kPa), and LiBr (367540 kPa). The desiccators were sealed with Vaseline to ensure a stable humid environment and placed in a temperature-controlled room maintained at 20 °C ± 1 °C to minimize temperature fluctuations. The samples were periodically weighed at intervals of no less than 7 days. A specimen was considered to have reached suction equilibrium when its mass change between two consecutive weighing intervals, spaced at least one week apart, was less than 0.1% of its total mass. Based on preliminary tests and this criterion, the equilibrium time for samples in the high suction range was found to be between 75 and 86 days. To ensure full equilibrium, all samples in this study were equilibrated for a minimum duration of three months before proceeding with subsequent tests or microstructural analysis. After three months, the samples were weighed and placed back in the chamber. If the weight did not change compared with those seven days prior, the sample was considered to have reached the suction balance. If the weight changed, the sample was placed back in the chamber for further equilibration. After the samples reached the suction balance, they were dried, and data were evaluated to obtain SWCC data for expansive soil samples in the suction range of 3 to 367.5 MPa.

The WP4C instrument method was used as a supplementary testing method. This method involves the use of the chilled-mirror dew point technique to measure the water potential by balancing the liquid water in the sample and the vapour phase water in the sealed sample chamber headspace and by measuring the vapour pressure in the sample chamber headspace. The relationship between the water potential of a sample and the air vapour pressure satisfies the Kelvin equation. The drying gradient for saturated expansive soil was set to 22%→20%→18%→14%. The samples were dried to the required water content in a 20 °C indoor environment. After the required water content was reached, the samples were wrapped in plastic wrap and placed in a humidity chamber for 24 h to allow uniform internal moisture distribution. The samples were then placed in sample cups for measurement.

The void ratios of the soil samples were obtained by measuring the height and radius values of the samples after suction using a Vernier calliper (three measurements in different directions were averaged) to obtain the current sample volume v. The samples were then dried and weighed to obtain the mass m_s. The void ratio was calculated via the formula e = v*m_s/G_s−1.

The MIP test was conducted using the AutoPore 9600 V2.03 MIP device with an intrusion pressure range of 0–61,000 PSI (0–420.60 MPa). Pore sizes as small as 10^− 3 μm were measured. Due to its high surface tension, mercury could overcome the internal pressures of pores, allowing mercury to enter the tiny pores of expansive soil samples and fill the pore space. The sample was placed in a sealed container, and gradually increasing pressure was applied to inject mercury into the sample pores. When mercury could no longer fill the pores, equilibrium was reached, and the intrusion volume and applied pressure were recorded.

For the MIP and SEM tests, when the expansive soil samples reached equilibrium at the target suction, 1 cm×1 cm samples from the centres of the standard ring samples were taken and freeze-dried according to the procedure proposed by Delage et al. for the MIP and SEM tests⁴². In the MIP test, the pressure interval was 10 s for pressures below 0.21 MPa and 120 s for pressures above 0.21 MPa. Importantly, sufficient time must be allowed for the contact angle to reach a quasistatic state during the MIP test, and a constant contact angle had to be used for mercury pressure and pore size conversion.

Test results and analysis

Soil–water characteristic curves under the full Suction range

The SWCC of expansive soil is shown in Fig. 3. Figure 3(a)-(d) show the water content‒suction, void ratio‒suction, and degree of saturation‒suction relationship curves under full suction drying‒wetting paths, respectively.

Fig. 3.

Soil–water characteristic curves (SWCCs).

As shown in Fig. 3, the soil‒water characteristic curve (SWCC) of expansive soil under the full suction desiccation/rewetting path exhibits a significant hysteresis effect. Specifically, at the same suction level, only a small portion of the pore space retains water during the wetting process, whereas relatively large pore spaces retain water during the desiccation process. At the same saturation level, the matric suction in the desiccation curve is significantly greater than that in the wetting curve.

As shown in Fig. 3(a), when suction is approximately larger than 100 MPa, the main drying and wetting branch expressed by water content converged together. Otherwise, a significant hysteretic effect is observed. Specifically, at the same suction level, only a relatively small portion of the pore space retains water during the wetting process, whereas relatively large number of pore spaces retain water during the desiccation process.

The reasons behind this hysteresis phenomenoninclude four types: (1) the nonuniform distributions of pore size and shape in expansive soil, which can be explained via the “ink bottle” capillary model theory; (2) the hysteresis effect of the contact angle during wetting and drying process, where the hygroscopic contact angle is smaller than the desiccation contact angle; This difference arises because of the roughness and heterogeneity of the soil particle surfaces. The contact between expansive soil particles and water is in a metastable state, and the local minimum Gibbs free energy indicates that the advancing and receding contact angles on the heterogeneous surface are different. (3) the influence of entrapped air bubbles in expansive soil is observed. As shown in the wetting path in Fig. 3(c), when the matric suction decreases to zero, the measured saturation is less than 1 because the presence of closed air bubbles during the wetting process blocks the flow paths, prolonging the time needed to achieve matric suction equilibrium. Moreover, the high gas pressure within these closed bubbles affects the accuracy of the matric suction measurements. Conversely, closed air bubbles occupy spaces that should be occupied by pore water, thereby reducing the measured water content and degree of saturation. Therefore, when obtaining the soil‒water characteristic curves for expansive soils, it is essential to extend the time for matric suction equilibrium to allow for the sufficient expulsion of air bubbles. Additionally, multiple samples should be measured simultaneously to reduce errors and eliminate the impact of closed air bubbles on the soil‒water characteristic curve. and (4) different spatial connectivity of pores during drying and wetting process. However, within a specific suction range, it is difficult to precisely determine which factor is primarily responsible for this hysteretic phenomenon. This also depends on the water retention mechanism. Within a low suction range, the water retention mechanism is mainly governed by the capillary mechanism, which is controlled by the Young-Laplace equation, where the suction value is inversely proportional to the pore diameter. The ink bottle effect and the entrained air may have a greater impact on this hysteretic phenomenon. Within a high suction range, adsorption plays a dominant role, and the influence of the pore structure can be ignored. Usually, in this case the main wetting and drying branch will converge together, as shown in Fig. 3(a) when suction exceeds 100 MPa. In the water content‒suction relationship curve (Fig. 3(a)), the desiccation curve is positioned above the wetting curve.

For expansive soil, as pore structure of soil would keep changing along the wetting-drying path, it would affect the water retention behavior of expansive soil significantly, as shown in Fig. 3(b). With the suction increases, the void ratio decreases. This indicates that the soil volume shrinks with the loss of the pore water, and pore structure changes occur. When suction is smaller than 500 kPa, the wetting and drying path are almost the same; Then the extent of slopes of drying and wetting curve becomes different, drying and wetting curve starts to deviate, the void ratio of expansive soil along wetting path is larger than that of specimen along drying path. When suction is larger than 100 MPa, void ratio no longer decreases and keeps almost constant with the variation of suction. At the end of the drying process, the free water between the soil particles in the sample completely evaporates, the particles are in contact with each other, the sample is in the densest state, and the void ratio no longer changes. Therefore, the shrinkage of micropores has been considered to compensate for the expansion of macropores, leading to much less pronounced shrinkage behavior of the total volume. In the void ratio‒suction relationship curve (Fig. 3(b)), the wetting curve is consistently higher than the drying curve. Additionally, during the drying process, expansive soil shows more noticeable changes in porosity. During the entire drying process, the porosity decreased by 0.39, while the maximum porosity variation is 0.30 in wetting process, indicating that irreversible plastic deformation exists durin2g the drying-wetting cycle process. According to the principle of Bishop’s effective stress for unsaturated soils, the effective stress tensor is given by the following expression:

where represents the total stress tensor, represents the pore air pressure, represents the degree of saturation, and represents the Kronecker delta. The term represents the net stress tensor. At the same suction level, since the degree of saturation during desiccation is greater than that during wetting, the average skeletal stress during desiccation is greater than that during wetting. Therefore, the porosity during desiccation is lower than that during wetting.

The evolution of degree of saturation with suction is presented in Fig. 3(c), which shows a similar behavior with the evolution of water content with suction. A strong hysteretic phenomenon is observed. At a specific water content or saturation, the suction of specimen along drying path is larger than that of specimen along wetting path. Especially in the middle part, the differences are significant. Along drying path, with the suction increases, the gas in the atmosphere continuously enters the soil, reaching the air intake value of the soil, making the expansive soil transition from a saturated state to an unsaturated state, the saturation (Fig. 3(c)) and void ratio (Fig. 3(b)) both decrease simultaneously, but at different rates. At the final stage, the volume shrinkage is gradually slowed, but the decrease in saturation has become more pronounced. This is due to the presence of pore gas makes the smaller reduction in the sample volume compared to the loss of water. This shows that during the drying process, the loss of water usually lags behind the stability of volume decrease. To precisely quantify the hysteresis effect, the ratio of drying suction to wetting suction at equal degrees of saturation was calculated.

Figure 3(d) shows the saturation‒porosity relationship curve, which visually demonstrates that even at the same porosity, the saturation levels are different between the desiccation and wetting processes. This finding indicates that the water content of expansive soil differs and highlights the significant coupling effect between the degrees of saturation and porosity (specific volume) as the matric suction changes.

The above analysis reveals that as matric suction increases, the water content, saturation, and porosity all decrease. Changes in the water‒mechanical properties of expansive soil lead to significant soil deformation, and expansive soil clearly exhibits a water‒mechanical coupling effect. Therefore, when constructing constitutive models for expansive soil and calculating its deformation, the water‒mechanical coupling characteristics of expansive soil should be considered.

Pore size distribution test (mercury intrusion porosimetry)

In the MIP test, the pressure and mercury intrusion volume are directly recorded. At different pressures, mercury intrudes into pores of different sizes. On the basis of the relationship between mercury intrusion pressure and pore size in the Washburn equation⁴³, the pore size can be determined from the pressure during the test. The MIP test provides the mercury intrusion pressure and corresponding pore size distribution, including the mercury intrusion pressure p at each level, the pore diameter D corresponding to a certain mercury pressure p, and the cumulative mercury intrusion volume V_M per unit mass of soil at a certain mercury pressure, which is the sum of the cumulative pore volumes with pore sizes larger than D.

During the mercury intrusion process, as the mercury intrusion pressure p increases, the measured pore size D decreases, and the cumulative mercury intrusion volume V_M increases. Since the pore size range of soil is very large (Figs. 4 and 5), the pore size data need to be logarithmically transformed to better represent the pore size distribution in the small pore size range. Figure 4 shows the MIP test results for expansive soil (considering saturated expansive soil as an example), i.e., the cumulative mercury intrusion volume graph, which includes mercury intrusion and extrusion curves. The results for the other samples (Fig. 5(a)) show the same trend, with the only differences in the mercury intrusion/extrusion volumes. An analysis of the intrusion and extrusion curves reveals that within a certain pressure range, the intrusion and extrusion curves do not overlap, indicating that some mercury remains permanently trapped in the pores of the expansive soil. The cause of this hysteresis phenomenon is the same as that in Fig. 3.

Fig. 4.

Mercury intrusion curve of saturated expansive soil.

Fig. 5.

Mercury intrusion curves of expansive soil samples at different matric suction levels.

In the initial stages of mercury intrusion, when the mercury pressure is low, the mercury intrusion volume increases slowly, and the mercury intrusion curve is relatively flat. When the mercury pressure reaches a threshold, as shown in Fig. 4, at a pore size of approximately 4000 nm (corresponding to a pressure of 45 psi), the cumulative mercury volume increases rapidly with increasing mercury pressure, and the pore size at the point of maximum slope corresponds to the dominant pore size of the sample at a specific suction level. In the later stages of mercury intrusion, the mercury pressure continues to increase to the maximum value, but the cumulative increase in mercury intrusion volume slows. When the mercury pressure reaches the maximum value, the final cumulative mercury intrusion volume V_Mmax is recorded, which represents the pore volume. The mercury intrusion curve reflects the process of mercury filling pores of different sizes in expansive soil.

Figure 5 shows the mercury intrusion curves of expansive soil samples at different matric suction levels. The cumulative mercury intrusion volume graph is shown in Fig. 5(a). Figure 5(a) shows that at lower matric suction levels, the pore volume of expansive soil is larger, with saturated expansive soil having the largest pore volume. As matric suction increases, cumulative mercury intrusion volume and the volume of expansive soil decreases, while the corresponding density increases slightly. The decreasing of total void ratio is mainly attributed to the decreasing of macro pores, especially for the pores with pore size between 0.3 μm and 10 μm (Fig. 5(f)). This graph generally illustrates the shrinkage characteristics of expansive soil during water loss. At high suction levels (3290 kPa and 38000 kPa), the cumulative mercury intrusion volume for samples with pore sizes smaller than 1 μm is significantly less than that at low suction levels. However, for samples with pore sizes larger than 2 μm, the cumulative mercury intrusion volume is greater than that at low suction levels. As the suction increases, the cumulative mercury intrusion volume decreases, and the dominant pore size at high suction levels is less pronounced than that at low suction levels. This phenomenon indicates that the pore structure of expansive soil changes continuously with variations in suction. The increase in suction causes pore deformation and total pore volume reduction because, as expansive soil loses water from a saturated state (Fig. 6), the concentration of free water in the pores first decreases⁴⁴. Then, pore water flows from the double electric layer, reducing the osmotic pressure and the ability of the double electric layer to adsorb water. As a result, osmotic swelling decreases, and the soil undergoes shrinkage deformation. When the water content decreases to a certain level, the thickness of the adsorbed water film decreases, thereby reducing the space between expansive soil particles and the degree of crystalline swelling.

Fig. 6.

Adsorption mechanism of the typical soil–water characteristic curve (SWCC) for expansive soil.

Figure 5(b) shows the relationship between the increase in mercury volume per unit mass of soil (dV_M) and pore diameter—the mercury intrusion increment graph. The graph shows significant differences in mercury volume increments for large pore sizes at different matric suction levels, whereas the differences for small pore sizes are less pronounced. Figure 5(c) shows the percentage change in the cumulative mercury intrusion volume (V_M) relative to the total pore volume (V_Mmax), reflecting the change in mercury saturation with pore size.

On the basis of the MIP test results, the cumulative pore size distribution can be obtained. The cumulative pore size distribution function is F(x), where F(D) represents the proportion of pore volume for samples with pore sizes smaller than D relative to the total pore volume. Importantly, the cumulative mercury intrusion volume (V_M) obtained from the MIP test is calculated by summing the mercury intrusion volumes from low to high pressures (i.e., from large to small pore sizes). Therefore, V_M corresponds to the cumulative pore volume with pore sizes larger than D. The pore volume with pore sizes smaller than D is denoted as V_R, and the relationship is given by the following expression:

1

By using Eq. (1), the cumulative pore size distribution curve can be plotted (Fig. 5(d)). Figure 5(d) shows that at low matric suction levels, the slope of the cumulative pore size distribution curve is small for both large and small pores. However, the slope is relatively large for intermediate pore sizes (0.5–3 μm). This finding indicates that for expansive soil samples with relatively high water contents, the dominant pore sizes are between 0.5 and 3 μm.

In the i-th stage of mercury intrusion, the volume percentage△V_i/V_Mmax represents the pore size distribution in the i-th pore size interval (D_i−1> D > D_i). Therefore, the relationship between △V_M/V_Mmax and the pore size is the pore size frequency distribution curve (Fig. 5(e)).

On the basis of the relationship between the pore size and mercury intrusion volume, the pore volume distribution function can be expressed as follows:

2

In the i-th pore size interval, the difference in the pore size range is △lgD_i=lgD_i−1−lgD_i. In a semilogarithmic coordinate system, the increase in the mercury intrusion level per unit length (△V_i/(lgD_i−1−lgD_i)) as a function of pore size can be represented as the pore volume distribution curve (Fig. 5(f)). Figure 5(f) shows that expansive soil has a distinct dual-pore structure, including interaggregate pores (macropores) and intra-aggregate pores (micropores), with dominant macro and micro pore diameters around 2 μm and 0.01 μm for saturated state and suction equal to 0, respectively. The literature provides different criteria and boundary to distinguish micropores and macropores. The commonly used approach is the valley of PSD curve can be accepted as the value that separates the inter-aggregate and the intra-aggregate pore spaces. In Fig. 5(f), the PSD does not shift with the changing of suction, and the macropore and micropore peak diameters remain unchanged. Hence, the average diameter corresponding to valley is about 0.2 μm. The identified pore boundary of 0.2 μm aligns with the microstructural classification criteria established by References^21,45. The volume distribution of the intra-aggregate pores remains consistent across the full suction range with no significant changes. Under five low-suction conditions (0–400 kPa), with the increasing of suction, total void ratio decreases significantly. This is mainly due to the reducing of macro pore volume, with decreasing in macro dominant pore size and corresponding density. Meanwhile, the dominant pore size of micro pores just decrease slightly or remains almost unchanged. The pore sizes of the samples are mainly concentrated in the micropore range of 0.2–2 μm, followed by ultramicropores smaller than 0.1 μm, with few pores in other size ranges. Under two high-suction conditions (3290 kPa and 380000 kPa), the pore size distribution density is mainly concentrated in micropores (0.2–2 μm), ultramicropores (< 0.1 μm), and medium-sized pores (2–20 μm). The proportions of ultramicropores and micropores under high-suction conditions are slightly smaller than those under low-suction conditions (0–400 kPa). A cluster of macroscopic pores is observed, with pore diameters ranging from 20 μm to 60 μm. This might due to the inhomogeneity of the intact sample. It could also be that the drying shrinkage causes the soil to crack, thereby destroy the integrity of the soil structure. During the initial wetting stages (367540 kPa→149510 kPa→38000 kPa), the pore distribution significantly changes, and the dominant pore diameter decreases, reflecting the process of macropore closure. During the wetting stage from 38,000 kPa to 3290 kPa, the pore distribution does not change significantly. During the wetting stage from 3290 kPa to 400 kPa, the pore distribution continues to change significantly, and the aggregates in the soil undergo significant expansion deformation. The expansion of aggregates partially fills the macropores, and significant expansion deformation of expansive soil occurs during this stage. Compared with the pore size distribution curve of the initial soil sample, the peak height at a pore size of 20,000 nm in the wetting curve decreases, indicating a reduction in the number of macropores of this size. However, a new peak forms at 1000 nm. As the matric suction continues to decrease, wetting progresses (400 kPa→200 kPa→100 kPa→50 kPa→0 kPa), and the number of macropores with a dominant pore size of 1000 nm continues to increase. Throughout the wetting process, the distribution of micropores (50 nm) does not change significantly. Overall, the higher the matric suction is, the lower the porosity of the expansive soil, and the smaller the proportion of large pores. Within a certain range of matric suction levels, the pore size distribution is relatively uniform. This behavior is consistent with the evolution of microstructure of compacted clay, e.g. Maryland clay⁴⁶, under different boundary conditions. These phenomena mainly arise due to differences in the properties of the liquid phase in macropores and micropores. In macropores, the interaction between the liquid phase and the solid matrix is dominated by capillary and physicochemical effects, with capillary effects being dominant and gradually decreasing with increasing water content. In micropores, the interaction between the liquid phase and the solid matrix is dominated by physicochemical effects. The adsorbed liquid phase in micropores has strong physicochemical interactions with the solid matrix, increasing the stability of the microstructure.

Scanning electron microscopy (SEM) test

SEM images of expansive soil under different drying suction conditions (5000× magnification) are shown in Fig. 7.

Fig. 7.

SEM images of expansive soil samples under different drying suction conditions (5000× magnification).

Figure 7 shows that after applying different matric suction levels, the particle morphologies and pore distributions in the soil samples change significantly. Figure 7(a) shows the sample under a matric suction level of 50 kPa. At this suction level, the interparticle suction in expansive soil is relatively small, and the soil is close to saturation. The soil surface is smooth, and the particles are mainly in face-to-face contact. The expansive soil sample exhibits a laminar flow structure, with small surface particles gradually forming a whole structure with large particles due to the influence of water. There are large pores and many well-connected cracks in the soil, and the water in the soil is primarily held in place by capillary forces. The connections between expansive soil particles are relatively weak. As matric suction increases, pore water is lost, and the amount of pore air increases, enhancing the interactions between clay mineral particles and promoting particle aggregation and rearrangement. When the matric suction level reaches 400 kPa (Fig. 7(b)), the small pores between the edges of the saturated flaky particles and edge-to-face structure particles gradually become visible, and the pores are mostly small with enhanced connectivity. The particle morphology is large and flat, with more curled flaky edges, and the contact mode between particles changes to face-to-face and face-to-edge contact. This phenomenon occurs due to particle reorganization at high matric suction levels, where large flaky structures breakdown into small granular particles that interweave and stack without fixed orientations. At this stage, the microstructure of the soil sample is mainly flocculent and detrital with good pore connectivity. Capillary pores exist between soil particles and contain a certain amount of bound water and capillary water. As matric suction continues to increase, the pore connectivity becomes more pronounced. When the matric suction increases to a relatively high level, as shown in Fig. 7(c), the amount of pore water further decreases, and the soil particle surfaces are strongly affected by adsorbed water films. The suction between particles increases, and the open medium-sized and large pores in the soil gradually close. The number of newly formed small cracks increases, the number of granular particles increases, and the arrangement tightens. The connectivity between soil particles is enhanced. Figure 7 shows that the microstructures of expansive soil samples vary significantly under different matric suction conditions. As matric suction increases, the soil structure changes from a laminar flow structure to a detrital structure, and the connected large pores are compressed and gradually replaced by small pores or microcracks. This phenomenon is the main reason for the significant volume reduction in unsaturated expansive soil during water loss. Therefore, the macroscopic changes in expansive soil samples under different matric suction conditions arise due to microscopic changes in the sample, including the closure and dissipation of large pores, the adjustment and reorganization of particle morphology, and the formation and expansion of small pores and cracks.

Conclusions

A series of experimental investigations were conducted on the retention property and microstructure behavior of expansive soil upon drying–wetting cycles under the full suction range. The soil-water characteristic curve (SWCC) of samples were obtained by combining the pressure plate method, saturated salt solution–vapour equilibrium method, and WP4C instrument method. Samples at different suction levels along the wetting path were selected for pore size distribution (PSD) tests and SEM tests to study the microstructural changes in expansive soil samples from canal embankments under the full suction range. The main conclusions are as follows:

1. The SWCC of expansive soil under the full suction range exhibits significant hysteresis due to factors such as nonuniform pore size and shape distribution, contact angle hysteresis, and trapped air. At the same suction level, the void ratio during the drying process is lower than that during the wetting process.

2. Expansive soil exhibits significant hydromechanical coupling. The pore volume of saturated expansive soil is the largest, and as matric suction increases, the volume of expansive soil decreases significantly. The water contents and microscopic pores inside the sample undergo significant changes.

3. Both MIP and SEM tests reveal that expansive soil has a distinct dual-pore structure, including interaggregate pores (macropores) and intra-aggregate pores (micropores). The boundary between the interaggregate and intra-aggregate pores is 0.2 μm. The volume distribution of the intra-aggregate pores is consistent across the full suction range with no significant changes. During the initial wetting stage, the dominant pore diameter decreases, and the macropores close. Then, during the intermediate wetting stage, the pore distribution does not change significantly. Finally, during the final wetting stage, the pore distribution continues to change significantly, and the aggregates in the soil undergo significant expansion deformation. Significant expansion of expansive soil occurs during this wetting stage.

4. Due to the differences in the liquid phase properties of macropores and micropores, the pore structure and distribution of expansive soil change significantly with variations in hydromechanical properties, including the closure and dissipation of large pores, the adjustment and reorganization of particle morphology, and the formation and expansion of small pores and cracks.

These findings have deepened the theoretical understanding of the relationship between microstructure and deformation and have also laid the foundation for establishing a full suction range hydraulic coupling constitutive model.

Author contributions

D. W.: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. M. L.: Writing – review & editing, Validation, Supervision, Project administration, Methodology, Funding acquisition. Z. W.: Writing – review & editing, Validation.

Funding

This work was supported by Natural Science Foundation of Henan Province (Grant No. 242300420225).

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.