The Ili loess used in the present research was collected from Xinyuan County, Xinjiang, and its basic physico-mechanical parameters are listed in Table 1 for reference. The preparation procedures of samples for both the freeze–thaw and dry–wet cycling tests were identical. In practice, the soil samples were air-dried, pulverized, and then sieved through a 2-mm mesh to generate test specimens (Figure 3). The averaged water content of the prepared specimens is 17.4% (equal to the optimal water content), and their maximum dry density was measured at 1.86 g/cm³.

The number of dry–wet cycles was set to 0, 1, 3, 5, 10, 20, and 30. During these processes, the initial water content of the loess began at 17.4%, with a minimum and a maximum (saturated) content of 10% and of 24.57%, respectively. As depicted in Figure 4, the preset water content was consistently achieved at the end of each full dry–wet cycle.

Upon completion of the designated dry–wet cycles, triaxial shear tests and microstructural analyses were conducted using a stress-controlled instrument (TFB-1). As illustrated in Figure 5, the prepared specimens were placed into the instrument, and confining pressures of 100 kPa, 200 kPa, and 300 kPa were then applied. The selected specimens were analyzed with SEM to examine surface morphology, whereas seven groups were evaluated using NMR (MesoMR23-60H-I).

The freeze–thaw cycle tests were carried out by the JW-2000 test chamber. The freezing temperature was set at −20°C, with each cycle lasting 15 h, whereas the melting temperature was set at 15°C for a constant period of 9 h. A previous research has demonstrated that the original cementation of soil samples is completely disrupted after 50 freeze–thaw cycles (Qi et al., 2014). Considering the high frequency and prolonged nature of freeze–thaw processes in the Ili Valley, the maximum number of cycles was set to 60 times. Samples subjected to 0, 5, 10, 20, 30, 45, and 60 cycles were selected for triaxial compression tests to examine the damage characteristics of Ili loess under freeze–thaw conditions. The freeze–thaw cycling path is depicted in Figure 6 for reference.

The stress–strain curves of the tested specimens with variable principal stress are depicted in Figure 7, in which different confinement conditions were also accounted for. It is evident from Figure 7 that the peak values of the principal stress for different samples are much variable. Notably, samples exposed to various numbers of dry–wet cycles exhibit a significant reduction in the difference between principal stresses compared to those that did not undergo any cycles. This indicates a weakening of the soil’s structural integrity as a result of repeated dry–wet cycling. Taking the stress–strain curves of the specimen with a confining pressure of 300 kPa, for example, the stress–strain curve of the untreated sample exhibits the highest peak, indicating that the dry–wet cycles lead to mechanical degradation of the loess, accompanied by a reduction in strength.

The maximum and minimum principal stresses were fitted using the σ1–σ3 method to determine the shear strength parameters of Ili loess under varying frequencies of dry–wet cycles. The results are summarized in Table 2, which indicates the potential degradation in mechanical properties with increased number of cycles.

Figure 8 illustrates the relationship between the friction and the cohesion values at the given number of dry–wet cycles. Compared to the untreated sample, those subjected to dry–wet cycles exhibit a reduction in the internal friction angle. This reduction varies initially but stabilizes as the number of cycles increases. It is evident in Figure 8 that the impact of dry–wet cycles is most pronounced at the initial stages, which significantly affects the cohesion of the specimens.

Figure 9 illustrates the shear strength of the tested specimens at different numbers of dry–wet cycles, in which the shear strength was calculated by the Mohr–Coulomb law. Compared to samples that were not subjected to any dry–wet cycles, there is a noticeable reduction in shear strength immediately following the initial dry–wet cycle. With the increased number of wet and dry cycles, the shear strength exhibited an initial fluctuation but then experienced a slight increase before eventually stabilizing.

Figure 10 presents the results of porosity at various dry–wet cycles, as determined by NMR testing. The data revealed a significant increase in porosity with the increased number of dry–wet cycles. Notably, significant changes in porosity occur at cycle counts of N = 1 and N = 3. As the number of dry–wet cycles continues to rise, the range of porosity variation gradually decreases. This trend is likely due to the redistribution of moisture within the specimen during the cycling process. That is, the induced motion of microparticles in the soil and the connected pores in the loess form a water conveyance channel when the specimens are suffering from the dry–wet cycles. This phenomenon is especially evident at the initial stage, where significant changes in porosity occur. As the number of dry–wet cycles increases, more water conveyance channels are formed, which likely reduces the effect of moisture on soil particle migration over time. Consequently, the internal particle structure tends to stabilize, associated with a gradual decrease in the range of porosity variation, which eventually reached a constant state.

Figure 11 illustrates the pore size distribution of the loess subjected to various dry–wet cycles, with pore diameters predominantly within the 0.01–100 μm range. These untreated Ili loess samples exhibited a bimodal pore size distribution, peaking at 1 μm and 20 μm. With the increased number of wet and dry cycles, the pore size distribution gradually transitioned to the unimodal type, and the peak value was 1 μm. As can be seen from Figure 11, the pore size distribution curve progressively shifts to the right, indicating an increase in pores, with approximately 10 μm in diameter. This observation showed that micropores gradually connected to form pores in the soil and became enlarged in the process of dry–wet cycling. Consequently, the number of micropores gradually decreases, whereas the amount of large and medium-sized pores steadily increases. The formation of enhanced water conveyance channels and larger pores will weaken the loess structure, and the created potential failure surfaces will quickly lead to sliding surfaces. This phenomenon is generally associated with reduced shear capacity and strength degradation.

Specimens subjected to various numbers of dry–wet cycles were analyzed using SEM, and the results of them are presented in Figure 12. According to the characterization by SEM, the particles became smaller, and both the abundance of particle units and pores inside the soil increased. However, the anisotropy rate decreased, and the inner structure of the soil became simple.

Figure 13 demonstrates the changes in microstructural parameters of the loess sample, suggesting significant fluctuations in the average pore size. This observation suggests that the particle units and pores become more circular in shape with improved uniformity. As the dry–wet cycle progress continues, particles and pores become more uniform and smoother, whereas the anisotropy rate of the loess decreases. This development trend implied that the internal structure of the soil gradually simplified under the influence of dry–wet cycling.

It has been well noticed that the shear strength of loess decreases gradually when it is under the dry and wet cycles (Qin et al., 2021). Initially, the effect of these cycles is pronounced, but shear strength continues to diminish as the number of cycle increases. This behavior is primarily attributed to the progressive breakdown of soil particles into smaller sizes throughout the dry–wet process. Moreover, the repeated cycles of shrinkage and expansion in loess lead to an increase in the internal pore volume and the gradual propagation of pores, which is associated with the convergence and penetration of small cracks (Wang et al., 2020; Ye et al., 2020). Note that this degradation is an accumulative process, which transited from the microscopic quantitative aspect to the macroscopic qualitative changes (see Figure 22).

The difference between the results of this experiment and the previous research results is obvious (Chou et al., 2019; Li and Xie, 2019; Liu, 2022; Qin et al., 2021; Xu et al., 2020; Yuan et al., 2017; Ma et al., 2022). As observed from this research, the internal friction angle generally decreased, except for five dry–wet cycles, whereas the cohesion fluctuated with the increasing number of cycles. The internal friction angle reflects the soil’s resistance to deformation, encompassing both sliding friction and interlocking friction between soil particles. Initially, the internal friction angle is high due to the irregular morphology of the soil grains. However, as dry–wet cycles progress, larger particles break down into smaller fragments, causing a shift in the predominant friction mode from interlocking friction to sliding friction.

Considering that the sliding friction is smaller than the occluding friction, the angle of internal friction of the soil decreases to some extent under dry–wet cycles. Meanwhile, the structure of the loess changes from an overhead structure (Liu et al., 2022b) to a mosaic structure, and then develops to a granular structure and flocculent structure. Notably, at the fifth dry–wet cycle, the soil’s shape factor reached its minimum, with particles adopting a flake-like appearance marked by rounded edges. This led to the formation of interlocking particles due to increased occluding friction, which in turn caused a moderate increase in the internal friction angle. Cohesion fluctuations were primarily linked to the contact pattern between particles and the state of fine particles (Wang et al., 2020; Liu et al., 2022c). When the contact between particles occurs in a surface-to-surface manner, the fine particles aggregate into a condensed form, resulting in increased cohesion. Conversely, the cohesion is lower and the fine particles are in a loose state when the particles’ connection mode is either the point-to-point or point-to-surface contact.

As the number of freeze–thaw cycles increased, a decrease in shear strength correlated with changes in the pore structure was observed. The Ili loess, with its distinct composition and structure, exhibited different patterns in cohesion and internal friction angle under freeze–thaw effects compared to prior studies (Liu et al., 2019; Li et al., 2020; Liu et al., 2022a; Zheng et al., 2021; Liu et al., 2021; She et al., 2019). It is apparent that the cohesion of the soil samples decreased significantly, whereas the internal friction angle increased within 10 freeze–thaw cycles. This behavior is attributed to the freeze–thaw cycles disrupting the strong cementation between grains, as the freezing and migration of water progressively weaken the inner particle bonds. As point-to-point contacts between particles become more prevalent, the reduction in cohesion becomes pronounced. Associated with the rearrangement of the soil particles, the mosaic structure developed with the increased internal friction angle. With further freeze–thaw cycling, the effects of freeze–thaw cycles on the loess gradually diminished due to the formation of new structural bonds among loess particles.

Comparative studies have previously indicated that the degradation of loess characteristics subjected to dry–wet and freeze–thaw cycles differs significantly, which is primarily attributed to differences in the material’s composition, structure, and particle characteristics, as well as differences in experimental methodologies adopted by different research workers. Chou et al. (2019), Zhang (2018), and Zhao (2019) found that the effect of dry–wet cycles is much more significant than that of the freeze–thaw cycles, whereas Li et al. (2017) reported the opposite trend. Figures 23, 24 show that the shear strength degradation for specimens subjected to freeze–thaw cycles is more significant than that of the specimens under the dry–wet cycles, when the same number of cycles is accounted for. This pronounced impact of freeze–thaw cycles is corroborated by the observation that most landslides occur in April and May during the ice and snow melting period (Cao et al., 2020).

The fundamental reason for this difference lies in the distinct deterioration mechanisms induced by dry–wet and freeze–thaw cycles on loess. The degree of strength deterioration of loess under dry–wet cycles was primarily affected by the salt content of the soil (Xiao et al., 2018). During the cycles, the water content varies from the natural state to saturation, leading to the detachment of hydrophilic minerals within the loess, which disrupts the internal cementation and alters its internal structure, promoting an increase in the content of medium-sized pores. This process is irreversible, as evidenced by the results of the NMR tests. In contrast, the water content remains unchanged during freeze–thaw cycles. When the soil freezes, the water within it generates frost heaving and migration forces, continuously reducing the interparticle bite force and the original cementation within the soil. During the thawing process, the internal ice transforms into water, prompting the migration of soil particles and altering the internal soil structure. That is, the deterioration capability of dry–wet cycles on the loess strength is jointly constrained by the water content and the salt content. Future research is thus necessary to explore the correlation between different cycling patterns.

Under natural conditions, loess is simultaneously affected by both dry–wet and freeze–thaw cycles, and the two are not independent of each other. In the present research, a series of advanced techniques such as SEM and NMR were applied to conduct a macro- and micro-comparative analysis of the deterioration characteristics of loess under different cycles. It is indicated that the deteriorating effect of freeze–thaw cycles is more pronounced than that of dry–wet cycles in the Ili Valley. This discovery not only further enriches the triggering factors of loess landslides in the Ili area but also provides meaningful theoretical support and practical guidance for the management of loess landslide disasters and local engineering construction.

In the present study, the effects of freeze–thaw and dry–wet cycles on Ili loess characteristics are independently examined. In the actual situation, however, these two actions often occur simultaneously. Given that the behavior of loess under the combined dry–wet and freeze–thaw cycles differs significantly from that under the isolated condition (Zhou et al., 2018; Li, 2017), future investigations should focus on the assessment of the characteristics of Ili loess under the combined cyclic conditions.