Effects of Freeze–Thaw and Dry–Wet Cycles on the Collapsibility of the Ili Loess with Variable Initial Moisture Contents

Abstract

Exposed to seasonal climate changes, the loess in the Ili region of Xinjiang, which has variable engineering properties, frequently undergoes freezing–thawing (F-T) and wetting–drying (W-D) cycles. In the present research, a series of uniaxial compression tests were conducted to investigate the collapsibility characteristics of the representative loess slope in the Ili region. In parallel, scanning electron microscopy (SEM) and nuclear magnetic resonance (NMR) tests were conducted. The test results obtained from the research indicated that both F-T cycles and W-D cycles exacerbate the deterioration of the loess, with the most severe effects observed after 6–10 cycles. Under the combined physical cycles, the microstructure of the loess progressively evolves from the relatively aggregated state to the dispersed one. Meanwhile, the porosity of the loess exhibited an initial increase with the number of W-D cycles, followed by an obvious decrease. Note that the pattern of the loess experiences fluctuation, which was achieved at the given point with the increased number of F-T cycles. It is suggested that the variability in loess wetting collapse is attributed to the irreversible alteration in the microstructure attributed to the combined cycles. The main reasons for the occurrence of loess collapse are the frost heaving force and the swelling–shrinking action. The impacts of W-D and F-T cycles on the loess obtained from this research can make a contribution to the in-depth understanding about loess collapse in the Ili valley.

1. Introduction

In recent years, both the periglacial processes and permafrost phenomena in the mid- and low-latitude mountain regions have attracted much attention [1]. Located in the western Tianshan Mountains of Xinjiang, the Ili River Valley is one of the representative seasonal permafrost areas in China, where the widely distributed loess is characterized by significant thickness and loose structure [2]. Due to its specific climatic conditions, the Ili loess with variable topography and geology generally undergoes prolonged wet–dry (W-D) and freeze–thaw (F-T) cycles [3]. Consequently, the loess typically exhibits pronounced degradation of mechanical properties [4,5] and strong susceptibility to wet subsidence [6,7,8]. This is generally termed as “three small (i.e., low dry density, low moisture content, and low saturation), two large (i.e., thick collapsible soil layer, high porosity), and one zero (the absence of ancient soil layers)” [9]. As one of the most critical engineering properties, the collapsibility of the loess is characterized by abrupt changes and discontinuity, as well as irreversibility. The collapsibility of the loess refers to its sudden subsidence under self-weight or external loads when wetted by water. In fact, the collapsibility of the loess can also lead to the occurrence of landslides, roadbed subsidence, and dam leakage, which may significantly threaten the safety of engineering projects [10,11].

Currently, a large number of studies have been carried out to explore the collapsibility characteristics of the loess under different climate changes, considering variable initial moisture contents. Wang et al. verified that the secondary collapsibility coefficient of the loess increases towards a specific value with the increased number of F-T cycles [12]. Song et al. discovered that F-T cycles have a dual effect of strengthening and weakening on the loess with different dry densities [13]. Qi et al. suggested that the proportion of macropores for the over-consolidated soil decreases during the whole F-T process, mainly due to the increased number of contact points between soil particles and weakened interparticle connections [14]. Gu’s research in terms of the loess deformation and collapsibility reported that the F-T cycle disrupts the particle connections of the loess, which causes enlarged volume and porosity, followed by altered subsidence characteristics [15]. Chamberlain et al. studied the changes in the permeability and structure of four fine-grained soil samples and found that the F-T cycles lead to a decrease in the pore ratio, while the increase in vertical permeability is mainly due to the decreased volume of fine particles in the pore space [16]. Apart from these aforementioned studies, lots of investigations were also carried out to explore the collapsibility of the loess under different W-D cycles. Malusis M. A. et al. observed that the hydraulic conductivity of the soil increases with the number of W-D cycles, during which pronounced vertical shrinkage was also observed [17]. Zhang et al. found that repeated W-D cycles reduce the matrix suction of the unsaturated soil, leading to irreversible changes in the mechanical properties [18]. Liu et al. observed a decrease in the shear strength for the remodeled loess subjected to W-D cycles, while the permeability coefficient exhibited a notable increase [19]. Sillanpaa M. et al. observed the enlarged average diameter of macroparticles in the loess after W-D cycles, while there was no significant change in terms of the soil composition [20]. Mao et al. discovered that both the dry density and cohesion of the compacted loess increased, whereas the porosity showed an increase [21,22]. Zhang et al. proposed the concept of humidification deformation to describe the deformation characteristics of wet-sagged loess subjected to W-D cycles [23].

Numerous theories and assumptions were previously developed to evaluate the collapsibility of the loess, such as the capillary hypothesis [24], the salt solubility hypothesis [25], colloid insufficiency theory [26,27], water film wedging theory [28], and pressure theory [25,29], as well as structural theory [27,30,31,32,33]. Among them, structural theory, which believed that the collapsibility of the loess is controlled by its own structure rather than other parameters, has been well verified and widely accepted [34]. In practice, both the collapse and structural properties of the natural loess are primarily simulated through the artificial preparation of loess samples in the laboratory [35]. Restricted by testing instruments, it was not easy to explore the structural characteristics of the loess through systematic investigation before [36,37]. With the rapid development of modern equipment, in particular, in terms of the invention of scanning electron microscopy (SEM) and nuclear magnetic resonance (NMR), it is possible to capture the microstructure of the loess, such as the anisotropy, inhomogeneity, and nonlinearity [38]. Followed by the establishment of the classification for the microstructures and engineering properties of the loess [39,40], the research in terms of the in-behind collapsibility mechanism was subsequently conducted [41,42,43,44,45]. Moreover, other advanced testing instruments and techniques, such as micro-image processing software and quantitative analysis, were also introduced to analyze the collapsibility of the loess simultaneously [46,47,48].

As discussed above, current research generally focuses on the macroscopic properties of the loess (i.e., shear strength) subjected to various physical cycles; only limited studies have been conducted to explore the collapsibility mechanism of the loess. Moreover, single-parametric analysis is not complete and insufficient, as a large number of factors will indeed affect the collapsibility of the loess. Therefore, comprehensive research on the evolutionary laws and collapsibility mechanics of the loess is required. Upon this background, the loess collected from the Ili region of China was experimentally investigated, for which the collapsibility behavior of the Ili loess subjected to F-T and W-D cycles was investigated from the macroscopic perspective and then the microstructural features of the degraded loess were further evaluated. It is expected to elucidate the collapsibility mechanisms of the Ili loess and provide theoretical guidance for engineering construction in the Ili River Valley.

2. Experimental Program

2.1. Raw Materials and Preparation of Samples

The loess used in the present research was collected from the shallow surface layer of a landslide on the western side of Haynd Saya Gorge near Almale Town in Xinyuan County, Ili Region (Figure 1a–c). As depicted in Figure 1d, the sampling depth of the Ili loess was 1–1.5 m, which experienced freeze–thaw and dry–wet cycling under natural conditions. It can be seen from Figure 1e that the vegetation roots and pest control of the undisturbed soil were relatively developed. Considering that these natural factors may affect the mechanical properties of the loess, we specifically selected the remolded loess rather than the natural one for indoor testing. As per GB/T50123-1999 [49], several critical parameters of the Ili loess were experimentally tested (Figure 2) and the averaged values of which are listed in

Table 1. Mean values of physical parameters of the Ili loess.

Parameters	Value
Specific gravity, Gs	2.69
Density, $\rho$ (g·cm−3)	1.36
Maximum dry density, ${\rho }_d$ (g·cm−3)	1.86
Moisture content, $\omega$ (%)	18.70
Optimum moisture content, $\omega$ (%)	17.60
Void ratio, e	1.01
Liquid limit, $W_L$ (%)	28.10
Plastic limit, $W_P$ (%)	19.10
Plasticity index, $I_P$	9.00
Liquidity index, $I_L$	−0.04
Grain-size distribution (%)
Sand content (greater than 0.075 mm)	43.63
Silt content (0.005–0.075 mm)	52.16
Clay content (less than 0.005 mm)	4.22
Main minerals (%)
Quartz	28.1
Calcite	21.1
Albite	19.5
Muscovite	15
Clinopyroxene	10.5

for reference. Apart from the particle size analysis with the application of the microtrac laser particle size analyzer, X-ray diffraction (XRD) tests were also conducted. As can be seen from Table 1, the main mineral composition of the loess is quartz, calcite, albite, muscovite, and clinopyroxene.

Prior to the preparation of the test samples, the lumped loess was air-dried in a cool place and then ground with a wooden mill. These pretreated loess samples were sieved by a 2 mm mesh and then sprayed to maintain the designed moisture contents (i.e., 6%, 10%, 14%, 18%, and 22%). Subsequently, these samples were moved to a sealed container for 24 h before the laboratory tests. After the settling period, the compaction method was applied to generate the cylindrical specimens with a dimension of 79.8 mm and a height of 20.0 mm for compression tests. In parallel, some other samples with a diameter of 50.0 mm and a height of 25.0 mm were prepared for microstructure analysis. Because the aim of this study is to investigate the changes in the properties of Ilie loess under different cycling modes, these remolded samples were either subjected to W-D cycles or F-T cycles, respectively.

2.2. Freeze–Thaw Cycling (F-T) Test

The constant temperature and humidity test chamber (No. JW-2000) was applied to simulate the F-T action on the Ili loess. Referring to the local meteorological data and other related information reported by academic scholars [7], the F-T cycles were alternated ranging from −20 °C to 20 °C (i.e., −5.0 °C, −10.0 °C, −15.0 °C, −20 °C, 5.0 °C, 10.0 °C, 15.0 °C, and 20 °C). It should be pointed out that the frequencies of F-T cycles were 0, 1, 3, 6, 10, and 20 times, respectively. According to the principle of controlling variables, there were 5 test samples and 5 parallel samples for each batch of the experiment, with the number of F-T cycles regarded as a quantitative parameter. As shown in Figure 3, the freezing and melting times for each cycle were 8 h.

2.3. Dry–Wet Cycling (W-D) Test

The humidification process of the loess sample for the W-D cycling test was carried out via the titration system. In practice, the titration tube was applied to generate the loess sample with the initial moisture content. These well-humidified loess samples were then placed into an electric blast drying oven with a constant temperature of 50 °C to achieve the preset moisture content before the W-D cycling tests with the given numbers (i.e., 0, 1, 3, 6, 10, and 20 times). Note that these loess samples subjected to different W-D cycles were kept for the uniaxial compression tests and microstructure analysis. In the present research, the initial moisture content of the tested samples ranged from 6% to 22% with an increment of 4%. Similar to the F-T cycle process, except for 5 identical samples with different moisture contents for each W-D cycle, there were another 5 parallel samples prepared for comparison.

2.4. Uniaxial Compression Test

The high-pressure consolidometer produced by the Nanjing Soil Instrument Factory was adopted to conduct the uniaxial compression tests (see Figure 4). With the consideration of the real loading rate in actual engineering applications, different loading rates (i.e., 12.5 kPa, 25 kPa, 50 kPa, 100 kPa, 200 kPa, 300 kPa, 400 kPa, and 600 kPa) were adopted in the present research. To ensure the reliability and reproducibility of the test results, the uniaxial compression tests were carried out for all loess samples subjected to F-T cycles and W-D cycles.

As per GB50025-2004, the collapse coefficient (δs) is defined as the additional subsidence of water-saturated specimens with a unit thickness after stabilizing under a given pressure. The collapse coefficient can be thus calculated via Equation (1), the classification of which is listed in

Table 2. Classification of loess collapsibility.

Types	Classification Criteria (${\mathit{\delta }}_{\mathit{s}}$)	Collapse Degree
Non-collapsible loess	${\delta }_s$ < 0.015	None
Collapsible loess	0.015 ≤ ${\delta }_s$ ≤ 0.03	Slight
	0.03 ≤ ${\delta }_s$ ≤ 0.07	Medium
	${\delta }_s$ > 0.07	Strong

for reference.

$${\mathsf{\delta }}_{\mathrm{s}}={\displaystyle \frac{{\mathrm{h}}_{\mathrm{p}}-{\mathrm{h}\prime }_{\mathrm{p}}}{{\mathrm{h}}_0}}$$

(1)

where ${\mathrm{h}}_{\mathrm{p}}$ represents the height of the sample after compression and stabilization at a given pressure, mm; ${\mathrm{h}\prime }_{\mathrm{p}}$ is the height of the sample after extra settlement associated with water immersion (saturation) at a certain pressure, mm; ${\mathrm{h}}_0$ is the original height of the sample without any additional pressure, mm.

2.5. Scanning Electron Microscope (SEM) Testing

To obtain samples of microscopic changes after different numbers of F-T cycles and W-D cycles, the microscopic specimens with a dimension of 50 mm and a height of 20 mm were prepared and examined by environmental SEM (FEI Quanta 250 FEG; Manufacturer: FEI; Origin: Hillsboro, OR, USA). Note that three identical samples were prepared for each set of physical cycles and the 800× magnification SEM images were obtained for in-depth analysis via Matlab software and Image Pro Plus software. More detailed information about the quantitative analysis of the microscopic particle structure parameters will be presented in Section 4.

2.6. Nuclear Magnetic Resonance (NMR) Testing

The NMR test was carried out in the nuclear magnetic resonance analyzer (MesoMR23-60H-I). Figure 5 shows the fitted curve equation of the moisture content and signal quantity data for the calibrated samples, which suggest that the porosity of each sample can be calculated upon the monitored NMR signal. Based on the equation shown in Figure 5, the average values from three identical tests can be obtained when the number of cycles was determined as the variable.

3. Compression Test Results

3.1. Influence of F-T Cycles

The repeated water–ice and ice–water phase changes are the normal phenomenon for the loess subjected to F-T cycles. It is a fact that there is movement of the inner particles during the process of free water turning to ice, which results in the expanded volume of the loess, whereas the large-size pores generated in the freeze process will not revert to small pores when the ice melts; that is, the physical properties of the loess have been affected by the change in the microstructure discussed above.

Figure 6 shows the variation process of the loess collapsibility coefficient under different numbers of F-T cycles. As recommended by GB50025-2004, if the value of the collapsibility coefficient is under 0.015 at 200 kPa, the loess does not exhibit collapsibility characteristics. It is thus obvious from Figure 6 that all the tested samples without any freeze–thaw cycles are featured with the collapsibility, except for that with a moisture content of 22%. For the loess sample with a moisture content of 22%, collapsibility was also observed after F-T cycles when the coefficient value of 0.027 is considered. It can also be seen from Figure 6 that the collapsibility coefficients exhibited an obvious decrease after three cycles. Different from the specimens with a large moisture content (e.g., 10%, 14%, and 18%), the collapsibility of the sample with a moisture content of 6% decreased with an increasing number of freeze–thaw cycles and finally stabilized. The above observations demonstrated that the freeze–thaw cycles significantly affect the collapsibility coefficient of the tested specimens, in particular, when the moisture content is below the optimum level. It also indicated that the most significant degradation occurred after three freeze–thaw cycles. Conversely, the collapsibility coefficient of these specimens with a moisture content above the natural level exhibited a gradual increase and then kept stable until the final F-T cycles.

In summary, the freeze–thaw cycles accelerated the collapsibility of the loess, the peak of which typically occurred after 6–10 cycles. The possible reason for these differences in terms of the collapsibility characteristics is mainly attributed to the disruption and reorganization of the soil microstructure caused by the F-T cycles [16,34,50,51,52]. Subjected to F-T cycles, both the formation and expansion of pores increase, which may enhance the water storage capacity of the loess and thus result in greater swelling.

3.2. Influence of W-D Cycles

Figure 7 shows the variation process of the loess collapsibility coefficient under variable W-D cycles. It is obvious that there is an initial increase, followed by a slight decrease, and then gradually stability is achieved as the number of W-D cycles increases. All samples without any W-D cycles exhibited obvious collapsibility characteristics, except for the specimen with a moisture content of 22%. After 10 W-D cycles, the collapsibility coefficient of the sample with a moisture content of 22% under a standard pressure of 200 kPa also increased to 0.023, suggesting a delayed response in terms of collapsibility, whereas the collapsibility coefficient of the other samples all reached their maximum value after 6 W-D cycles. It is evident that the number of W-D cycles also had a significant impact on the collapsibility behavior of the loess samples.

3.3. Influence of Initial Moisture Content

It has been well noted that the loess collapsibility initiates together with the immersion rather than the reach of the saturation. Figure 8 and Figure 9 depict the variations in the collapsibility coefficient of the tested samples with different initial moisture contents under different F-T and W-D cycles, respectively.

Figure 8 and Figure 9 suggest that the values of the collapsibility coefficient for these loess samples overall decrease as the initial moisture content increases. Moreover, it can also be seen from Figure 8 and Figure 9 that the chart lines are all smooth for most specimens, except that slight irregular fluctuations were observed for those samples with moisture contents of 18% and 22% after 10 W-D cycles. This observation strongly supports that the effect of the initial moisture content is much more significant than that of physical cycling. If the initial moisture content is within the range of the optimal values, those specimens with a higher initial moisture content generally show a small collapsibility. This unique observation is mainly attributed to the dissolution of soluble salts and the absorption of water by clay minerals within the loess. With the reduced cohesion and structural integrity, the collapsibility coefficient of the loess exhibits a descending trend as well.

The collapsibility coefficient of the loess experienced an initial increase and then exhibited a decrease with the increased number of F-T cycles. Different from that for the W-D cycles, the peak value of the collapsibility coefficient was reached at a given pressure of 400 kPa. When the moisture content of the samples was larger than 14%, the loess samples without cyclic action were either non-wetted or slightly wetted. If the moisture content was less than 14%, the loess collapsibility developed from the slight level to the medium level. This observation indicates that drier loess with a low moisture content is more collapsible and pressure-sensitive because the loess with a higher moisture content is easier to saturate before immersion. The values of the collapsibility coefficient also vary with the vertical load, which exhibited an initial increase and a subsequent decrease when the vertical load was below 300 kPa. Once the axial load exceeded 300 kPa, the values of the collapsibility coefficient exhibited a decrease with the increased vertical load. Because the lateral deformation of the sample is somehow confined, the increased collapsibility coefficient attributed to the vertical load is not always obvious.

Based on the analysis of the uniaxial compression tests, it was concluded that the effect of the initial moisture content is most significant on the collapsibility of Ili loess, followed by the W-D cycling and F-T cycling.

4. Scanning Electron Microscopy (SEM) Test Results

In this section, the samples with a moisture content of 18%, which is nearest to the optimal moisture content, were selected for SEM analysis. Herein, the qualitative analysis for the microstructure examination covers the morphology of skeletal particles, structural particle characteristics, contact relationships, association modes, pore types, and the degree of cementation [31,45,53,54,55]. The SEM images of representative samples were preprocessed, optimized, and binarized.

4.1. Qualitative Analysis of Microstructure

• Freeze–Thaw Cycles

Figure 10 illustrates the qualitative analysis results of the SEM images of loess samples subjected to different F-T cycles. For these samples without any F-T cycle, the loess particles predominantly featured a powdery, ellipsoidal, globular, elongated, columnar, and irregular morphology, associated with a gravelly and mosaic appearance. The primary contacts in between were mosaic, and both the face-to-face cementation and intergranular mosaic pores attributed to soil remodeling can be observed. After six F-T cycles, the scaffold contacts tended to be predominant between particles, with intergranular voids, and the microstructure was characterized by a granular, scaffold, point-contact microcementation structure. By the tenth cycle, a semi-coagulated structure emerged with the progressive development of a granular, scaffold, and point-contact semi-colloid structure. After 20 freeze–thaw cycles, the particle structure was mainly in the form of basal and adherent types. As can be seen from Figure 10, the microstructure of the loess generally transformed into granular, dispersed, point contact with partial cementation.

It can be summarized that the microstructure of the loess evolved from a granular, mosaic, surface-cemented microcementation structure to a clustered, dispersed, point-contact cohesive structure. Attributed to reduced cohesion of the loess sample, the inner particles exhibited a dispersed trend with the increased pore space. Once the water retention is enhanced, the values of the collapsibility coefficient will subsequently increase.

• Dry–Wet Cycles

The SEM images of the loess samples under different W-D cycles are presented in Figure 11, the qualitative analysis of which shows that the microstructure of the loess transitioned from a granular, mosaic, and surface-cemented microcementation to a shelf, dispersed, and point-contact cohesion structure, as the number of W-D cycles increased. Compared to the reference samples not subjected to W-D cycles, the particles within the loess sample progressively fragmented through the W-D cycles. Because large particles broke down into smaller particles, forming aggregates during the W-D cycles, the large pores within these particle aggregates were filled up. Afterwards, the large pores exhibited a gradual decrease, while the small pores showed a slight decrease during the particle aggregation process. Due to the breaking down of large particles and the agglomeration of small particles, the number of medium pores increased. This pattern suggests that the degrading and disintegration of the loess particles are much more obvious than that of the cohesion during dry–wet cycles, resulting in the increased number of large and medium pores in turn. Subjected to the combined action of salt dissolution and clay particles’ dispersion, the reduced cohesion also weakened the microstructure and facilitated the collapsibility of the loess.

4.2. Quantitative Analysis of Microstructure

It was in 1996 that the classifications of microscopic structural elements, including grain morphology and pore properties, were proposed [56,57]. Subsequently, quantitative parameter analysis was widely applied to investigate the collapsibility mechanism of the loess [58,59]. Upon previous studies and our own experimental results [44], it is believed that the loess collapsibility is closely related to changes in the pore structure. Therefore, the fractal dimension of the shape distribution, pore area ratio, mean pore diameter, and grain roundness were selected as the main parameters to evaluate the micro-mechanism of loess collapsibility, the observation of which can be seen in Figure 12.

• Freeze–Thaw Cycles

It is obvious in Figure 12 that the fractal dimension of the pore shape distribution increases with the number of F-T cycles, which indicates that the pore shapes become more complex. Although there are some fluctuations, the total pore surface area generally increased, suggesting that the formation and melting of ice crystals led to changes in the spatial position between soil particles during the F-T cycles. The mean pore diameter of the loess sample exhibited smaller variations and a narrower range of change, mainly attributed to the relatively balanced squeezing and rearrangement effects of the F-T cycles. It should be noted that the change in the pore diameter was less pronounced than that of other parameters. However, the significant impact of small variations in the average pore diameter could not be ignored. Subjected to the compressive and shear forces exerted by ice crystals during the F-T cycles, more irregular and angular particles were generated, resulting in the significant decrease in particle roundness.

• Dry–Wet Cycles

As depicted in Figure 12, the changes in selected microscopic parameters under the W-D cycles are more pronounced than those under the F-T cycles, which aligns with the results obtained from the uniaxial compression tests mentioned previously. Furthermore, both the fractal dimension of the pore shape distribution and the total pore surface area exhibited a significant upward trend as the number of W-D cycles increased. This observation indicates that the pore shapes become more complex, and the pore surfaces increase in area, mainly attributed to the repeated expansion and contraction of loess particles during the hydration and dehydration process. Concurrently, the particle roundness decreases with fluctuations and more angular or irregular particles were generated. The mean pore diameter, however, exhibits larger variations, initially increasing and then decreasing overall, reflecting the dynamic changes in the loess structure under W-D cycles.

It can be concluded that the effect of the W-D cycles seems to be more pronounced when the changes caused by the F-T cycles were compared. Because the changes in the microscopic parameters (i.e., the pore shape, pore surface area, particle roundness, and pore diameter) will affect the geotechnical properties of the loess (e.g., the collapsibility, hydraulic conductivity, and mechanical strength), detailed analysis of the microstructure is crucial for mitigating potential geotechnical hazards in the Ili River Valley.

5. Nuclear Magnetic Resonance (NMR) Test Results

5.1. Effects of F-T Cycles on Porosity Variation

It can be seen from Figure 13 that the largest value of porosity was reached at the sixth cycle, showing a similar pattern to the variation in the collapsibility coefficient of the loess under the F-T conditions presented in Section 2.1. In the initial process of the F-T cycles, the change in porosity is subtle until a noticeable increase occurs by the third cycle, which suggests that structural damage to the loess starts to intensify at this point, attributed to the expansion and subsequent disruption of loess particle connections caused by freezing. Upon thawing, the large pores formed by freezing and expansion between particles decrease as the particles settle under gravity, resulting in a reduction in original large pores and an increase in small pores. After the sixth cycle, the pore size of the loess starts to diminish and then stabilizes, indicating a gradual stabilization of the disruptive effects of the F-T cycles.

5.2. Effects of W-D Cycles on Porosity Variation

The porosities of loess samples after variable W-D cycles are plotted in Figure 13, the pattern of which aligns well with the collapsibility coefficient changes in the loess under the W-D cycles discussed in Section 3.2. Different from those samples without any W-D cycles, the porosities of these samples exposed to W-D cycles showed a significant increase in the initial cycles (N = 1 and N = 3). As the number of W-D cycles increased, the porosity change rate diminished, which is mainly attributed to the compaction of the loess and the migration of small particles. This effect on the porosity is more pronounced in the early stages of W-D cycles. As the number of W-D cycles increased, the established water transport channels reduced the movement of moisture and stabilized the internal particle structure.

6. Theoretical Discussions

6.1. Collapsibility Mechanism of Loess Under F-T Cycles

After multiple F-T cycles, the micro-textures of the loess particles underwent significant changes, which is consistent with the observation reported by previous research [60]. The transformation of loess particles under freeze–thaw action primarily occurs through the expansion and wedging forces exerted by ice crystal formation during freezing, and the scouring and dissolution effects of water during thawing. During the whole F-T cycles, these transformative processes recur, leading to microstructure changes in the loess particles. The reduced collapsibility coefficient of the loess at the initial F-T cycling process is mainly attributed to the disruption of original cementation between small particles and the reduction in the cohesive force in between [54]. On the other hand, the frost heaving and migration forces generated during the F-T process also weaken the adhesive forces between different particles, promoting the migration of moisture within the loess. These effects combine to decrease the water adsorption of the loess and thus reduce the collapsibility coefficient before subsequent F-T cycles.

However, the repetitive F-T action induces continuous movement of free water within the soil and diminishes particle occlusion as the number of F-T cycles increased. This irreversible process results in the reduction in the cohesive force (see Figure 14). With the increased number of F-T cycles, the loess porosity exhibited an initial increase, followed by a decrease and an increase at the final state. Contributing to the breakage of large particles and the accumulation of small particles, the number of medium-size pores increased while the number of large and small pores showed an obvious decrease. As suggested by the microscale experiments, the decomposition of loess particles and development of micro-cracks under F-T cycles resulted in the change in particle roundness. As depicted in Figure 14, the changed roundness weakened the contacts between the loess particles, accelerating the transition from the aggregated state to a looser one. Furthermore, the increased pore space and water storage capacity of the loess provide favorable conditions for loess collapsibility.

6.2. Collapsibility Mechanism of Loess Under W-D Cycles

Mineralogical analysis indicates that the loess in the Ili River Valley contains a high proportion of hygroscopic minerals, including calcite, dolomite, chlorite, sodium feldspar, and mica. Due to the presence of these hygroscopic minerals, under wetting–drying cycles, the minerals undergo a dynamic state of hydration. During repeated cycles of hydration and dehydration, the distribution of mineral components becomes more uniform, leading to changes in particle units and alterations in the internal structure of the soil.

When the loess undergoes repeated W-D cycles, the internal particles experience absorption and release many times along with structural expansion and contraction. The cyclic state leads to the loss of hydrophilic minerals between soil particles from their original state. This gradual reduction in cementitious material within the soil will significantly weaken the bond strength and consequently decrease the internal cohesion of the loess (see Figure 15). As revealed by the NMR test results, the loess porosity initially increases and then decreases with more W-D cycles, as the continuous wet–dry effect breaks down large particles and aggregates small particles. These combined factors make the loess structure transition from compact to loose, enlarging the pore space and enhancing conditions for wetting. As a result, the collapsibility coefficient of the loess increases after the W-D cycles, indicating progressively more severe collapsibility.

6.3. Comparison Analysis

It is believed that the formation and melting of ice crystals are the primary causes for the changes in the loess microstructure under F-T cycles. During the formation process of ice crystals, the generated expansive forces were applied on the surrounding particles. Once the ice crystals transform back into water during the melting process, the cementation between particles would be weakened along with the scavenged and dissolved ice crystals, resulting in variations in the collapsible coefficient of the loess. The ingress and egress of water are regarded as the key factors driving changes in the microstructure of the loess under W-D cycles. During the drying process, the distance between particles increased, associated with the reduced cementation attributed to the water migration out of the loess, whereas the water migrated into the loess and filled the gaps between particles during the wetting process, which altered the contact modes of loess particles. This repeated wetting and drying process led to the breakage of particles and the formation of aggregates. These changes in terms of the microstructure also impact the collapsibility of the loess.

As per the aforementioned discussions, it is believed that the collapsibility characteristics of Ili loess are fundamentally related to the irreversible changes in its microstructure [61], although the response mechanics of which are somehow different. Under freeze–thaw cycles, the frost heaving and migration forces weaken the adhesive between particles, promoting the migration of moisture within the loess. Conversely, the repeated wetting and drying processes cause the particles to continuously expand and contract, leading to the detachment of hydrophilic minerals, a reduction in cementation, and weakened adhesion. Despite these differences existing, both cycling processes share a common effect; that is, the expansion and disintegration of aggregate units exceed compression and polymerization, resulting in increased porosity and soil volume expansion. Furthermore, both cycles transform the soil microstructure from aggregated to disaggregated, reducing particle cohesion and enabling conditions for particle slippage during collapsibility.

Through in-depth exploration and analysis of the intrinsic relationship between microstructural changes and macroscopic mechanical properties, we can gain a more comprehensive understanding of the collapsible behavior mechanisms of loess either under F-T cycles or W-D cycles.

6.4. Correlation Between Landslide Geological Environment and Laboratory Results

Due to the sharp rise in temperatures from April to May each year in the Ili River Valley, the accelerated melting speed of ice and snow results in repeated diurnal dry–wet and freeze–thaw cycles. During this period, the physical and mechanical properties of loess slopes undergo damage related to dynamic and static water level changes. The deformation characteristics of hillsides are mainly manifested as surface spalling deformation, featuring creep slides, collapses, and debris flows. According to the principle of convective heat transfer between air and soil, when the surface soil temperature decreases, the warm water migrates from deep soil to the surface.

During the repeated freeze–thaw process, internal soil moisture crystallization leads to the observation of layered ice lenses from the surface (see Figure 16a,b). Subjected to the combined effects of dry–wet and freeze–thaw cycles, temperature stresses may cause slope deformations, such as reticulate spalling, as shown in Figure 16c, and the laminar spalling depicted in Figure 16d. Different from its counterparts, the blocky spalling shown in Figure 16e was also attributed to the separation of large blocks. As can be seen from Figure 16f, some loess slopes exhibiting convex, layered hard crusts may detach from the slope during the winter period. As the surface hard crusts undergo melting and collapse due to freeze–thaw and dry–wet cycles, progressive deformation also develops from the slope toe to complete instability (see Figure 16g,h).

The low temperature in winter and rapid warming in spring exacerbate the collapsibility of the loess slope within the Ili River Valley. The experimental results presented in this research indicated that both freeze–thaw and dry–wet cycles would increase the collapsibility of the loess to some extent. During the process of collapsibility, the loess undergoes significant deformations such as creep slides, debris flows, and spalling deformation, which align closely with the actual deformation observed in practice. This observation verified that collapsibility is one of the important causes of slope instability. Based on the research outcomes obtained from this research, some engineering measures, such as waterproofing and reinforcement techniques, should be well considered in future research to mitigate the influence of collapsibility and maintain the stability of the loess slopes.

7. Conclusions

The collapsibility characteristics of remolded loess subjected to freezing–thaw (F-T) cycles and wetting–drying (W-D) cycles were experimentally and theoretically investigated. Critical parameters evaluated in the present research covered the moisture contents and number of F-T and W-D cycles. Apart from the indoor uniaxial compression tests, the microstructure of the loess subjected to different climate states was also explored with the application of scanning electron microscopy (SEM) and nuclear magnetic resonance (NMR) instruments. The following conclusions are drawn:

When the moisture content of the loess is less than the optimum moisture content, the values of the collapsibility coefficient exhibited an initial decrease, followed by fluctuation until finally being stable as the number of F-T cycles increased. On the contrary, when the moisture content is greater than the optimal moisture content, the collapsibility coefficient of the soil sample gradually increases with an increasing number of F-T cycles, eventually stabilizing.

When the moisture content of the loess is less than the optimal moisture content, the maximum values of the collapsibility coefficients of the loess were reached after six W-D cycles. On the contrary, the collapsibility coefficient of the 22% moisture content soil sample only reaches its peak after 10 W-D cycles, indicating that the effects of the W-D cycles are lagging more on a large moisture content of loess samples.

The microstructure of the loess gradually transforms from a granular, embedded, and surface-cemented structure to a dispersed, point-contact cemented structure with the increased number of physical cycles.

The porosity of the loess samples experiences an increase and then decreases with the increased W-D cycles and F-T cycles at the initial stage.

The frost heaving and migration forces generated by freeze–thaw cycles weaken the adhesive between loess particles and promote the migration of moisture within the loess.

Affected by dry–wet cycles, the loess particles undergo expansion and contraction, leading to the detachment of hydrophilic minerals between particles and the reduction in cementitious material. The weakened bond strength ultimately accelerated the collapsibility of the loess.

The aforementioned research has explored the collapsibility processes of the Ili loess with the consideration of the types and number of physical cycles as well as the moisture contents of the loess samples from the macroscopic and microscopic perspectives. However, it should be noted that the properties of the loess vary across different regions and the universality of the findings obtained from this research requires further validation. Meanwhile, both meso- and microscale studies should be integrated with practical engineering applications to validate its accuracy. These considerations serve as directions for future research endeavors.