The in-situ residual soil samples were taken from a slope in Fuzhou area of Fujian Province, about 50 cm thickness of the surface mixed soil was removed before the test, and then 1 m³ soil samples were cut out and sealed with plastic film to prevent the loss of water in the soil during transportation. The test soil samples were all taken from the cut soil body to ensure the consistency and homogeneity of the residual soil structure during the test. The selected soil samples were mainly formed by residual deposition of coarse-grained granite, with brownish-red and grayish-white colors. X-ray diffraction showed that the main components of the soil rich in secondary clay minerals weathered by feldspars and quartz particles, including orthoclase, kaolinite, montmorillonite, quartz sand grains, as well as traces of iron oxides, black mica, and other minerals. Table 1 shows the basic physical property indexes of the in-situ soil, the liquid limit and plastic limit of the soil measured by disc liquid limit apparatus and rubbing strip method are 46% and 33% respectively, according to which the plasticity index of the soil can be calculated as 13. The particle-size distribution (PSD) curve of the residual soil given in Figure 1 obtained by sieving and gravimetric methods show that the soil body consists of 1.5% gravel, 42.2% sand, 43.8% silt and 12.5% clay. As illustrated in Figure 1, the PSD curve has a limited particle size d ₆₀ = 0.07 mm, effective particle size d ₁₀ = 0.004 mm, and median particle size d ₃₀ = 0.02 mm. Therefore, the uniformity coefficient C _u and the coefficient of gradation C _c for the soil are 17.5 and 1.429, respectively, which indicates that the studied soil is well graded. Meanwhile, this soil can be identified as well graded intermediate plasticity gravelly clay according to the site investigation code given by British Standards Institution (BSI) 1990.

The initial matric suction of the in-situ residual soil was 45 kPa. Before the test, the cut-out in-situ soil samples were vacuumized and saturated, and repeated hygroscopic cycle samples were prepared by the automatic unsaturated soil triaxial apparatus (see Figure 2), which could lay the groundwork for the subsequent triaxial drainage test of unsaturated residual soil. The evaporation phenomenon which may appear in the test is eliminated by the axis translation technology. Figure 3 shows the schematic diagram of the residual soil suction path under the repeated hygroscopic cycle. The maximum change of suction during the hygroscopic cycle is realized by controlling the upper and lower limits of suction, in which the upper limit value of suction force (ψ _max) is 400 kPa and the lower limit (ψ _min) is 0 kPa. Subsequently, we further set the target suction force ψ _t as the suction level of the soil sample that undergoes repeated hygroscopic cycles during the shearing process. Taking the 2^nd hygroscopic cycle as an example, the suction path in the cycle is: 0→400→ψ _t →0→400→ψ _t , and the above steps can be repeated to achieve the preparation of soil samples with different hygroscopic cycles. Meanwhile, the strength characteristics of residual soil and its internal microscopic evolution mechanism can be analyzed by conducting NMR tests on soil samples after different hygroscopic cycles. In fact, the process of performing several wetting-drying cycles was designed to evaluate the water retention curves (de Oliveira et al., 2021; de Moraes et al., 2016), strength and deformation characteristics (Xu et al., 2021; Abbas et al., 2023; Tumwiine et al., 2023), crack evolution characteristics (Amenuvor et al., 2024; Zhang et al., 2024) and infiltration characteristics of soils (Wen et al., 2019; Louati et al., 2018) as reported in different studies. The number of wetting-drying cycles chosen here was similar to those utilized by some scholars for determining the soil-water characteristic curves (Wen et al., 2020) and shear strength (Cheng et al., 2021; Xu et al., 2021) evolution patterns of structural soils under specific stress levels (i.e., confining pressure and matric suction). The deformation and pore structure of the in-situ soil can basically reach a stable state after about 3–5 cycles of wetting-drying cycles (Ng and Pang, 2000; Wen et al., 2019; Cheng et al., 2021; Xu et al., 2021). Considering the test efficiency and rationality, the maximum number of hygroscopic cycles set in this research is 7 cycles. In summary, the amount of hygroscopic cycle selected aimed at revealing the intrinsic connection between the microstructural damage characteristics of the residual soil and the deterioration behavior of the macroscopic mechanical properties.

The unsaturated triaxial test included consolidation phase, suction balance stage and shear phase. Confining pressure was controlled at 50 kPa, 100 kPa, and 200 kPa, and the target matric suction ψ _t was set at 50 kPa, 100 kPa, and 200 kPa, and the influence of 0, 1, 2, 3, 5, and 7 hygroscopic cycles was also considered. Among them, 0 cycle was set up (i.e., no change in the hygroscopic paths), and used this as a reference for analyzing the characteristics of the other number of cycles. The test scheme and the test flow chart are shown in Table 2 and Figure 4, respectively. During the suction equilibrium process, the repeated water absorption cycle test was realized by keeping the pore pressure at zero and increasing the pore air pressure and the perimeter pressure value, in which the maximum suction force was controlled to be 400 kPa. To protect the equipment, ensure that axial pressure > confining pressure > air pressure > water pressure (axial pressure is 5 kPa greater than confining pressure) throughout the whole test. When the consolidation time is not less than 24 h, the axial strain is not more than 0.01 mm/h and the back pressure volume does not exceed 0.05 mm³ within 2 h, the consolidation stage can be considered to reach the equilibrium standard. The suction equilibrium process criterion should be determined by the amount of water absorbed or lost in 24 h is less than 0.05% of the specimen volume, the shear rate of the shear process is 0.0038 mm/min, and the maximum axial strain is controlled within 25%.

MesoMR23-060H-I system was used for nuclear magnetic resonance analysis (see Figure 5), the equipment magnet was a permanent magnet with a magnetic field strength of 0.5 ± 0.08 T, and the magnetic field temperature was controlled at 32.00°C ± 0.02°C. In the nuclear magnetic resonance test, PVC pipes without iron were used to cut out the soil samples after different hygrometric cycles, the saturated soil samples together with PVC pipes were put into the instrument for testing to avoid the magnetic field generated by iron-containing substances affecting the test results. Then, the time-dependent curve of the NMR signal was then obtained based on the CPMG (carr-purcell-meiboom-gill), which was further quantified as the T ₂ (transverse relaxation time) distribution curve, and the pore size distribution was also calculated accordingly.

Figure 6 shows the stress-strain curve of unsaturated residual soil under different test conditions. Referring to the test number in Table 2, the 50-50-1w in Figure 6 can be expressed as the stress-strain curve of soil after the first water absorption path under the condition that the confining pressure and target suction ψ _t are both 50 kPa. For the same number of water absorption cycles and target suction, the shear characteristics of residual soil under low confining pressure show a more obvious strain softening phenomenon and lower peak strength (see Figure 6A), while the stress-strain curve shape of soil under high confining pressure is significantly affected by the number of hygroscopic cycles and has a higher peak strength. For example, the peak strength of soil under 50-50-1w, 100-50-1w and 200-50-1w conditions is 144 kPa, 215 kPa and 318 kPa, respectively.

The strength of the restraining action around the soil directly determines the friction strength between the soil particles. When the restraining action is weak, the friction effect caused by the dislocation and disengagement between the soil particles is not obvious. In this case, the structural strength of the soil skeleton plays a major role, showing lower failure stress and softening phenomenon. With the strengthening of the confining effect around the soil, soil particles dislocation and compaction occur under the action of partial stress, and the friction and detachment occlusion effect between soil particles is prominent. At this time, the friction strength of soil plays a major role, and the soil has a higher resistance to external loads, showing strain hardening and shrinkage characteristics. With the enhancement of constraint effect around the soil, the soil particles dislocation and compaction occur under the action of deviatoric stress, and the friction and detachment effect between the soil particles are prominent. At this time, the friction strength of the soil is significantly exerted, which means that the soil has a stronger resistance to external loads and exhibits the characteristics of strain hardening and contraction.

Under low confining pressure and the same number of water absorption cycles, the greater the high target suction, the higher the shear failure stress of the soil mass, such as the soil peak strength under the conditions of 50-50-1w, 50-100-1w and 50-200-1w is about 144 kPa, 188 kPa, and 272 kPa respectively. When the test conditions are consistent, the pore water content in the soil with low target suction is high, and the high saturation reflects that there is less soil-water bending liquid surfaces surface in the soil. With the decrease of the effective area of intergranular contact and stress level, the overall structural stability of soil decreases, which ultimately contributes to the soil exhibiting a lower yield stress during shear. At a confining pressure of 200 kPa, the shear characteristics of the soil under different suction conditions differ significantly. The peak strengths of the soil at 200-50-0w/1w, 200-100-0w/1w, and 200-200-0w/1w conditions are 326/318 kPa, 372/366 kPa, and 458/445 kPa, which indicates that the stress-strain curves of the soil showed similar strain softening and shear contraction characteristics as normal consolidated soil at 200-ψ _t -0w or 1w conditions. While at 200-ψ _t -2w∼7w, the shear characteristics of the soil have brittle and shear expansion characteristics similar to those of super-consolidated soil, for example, the peak/residual strength of the soil under the conditions of 200-50-3w, 200-100-3w, and 200-200-3w are 304/258 kPa, 350/284 kPa, and 430/350 kPa.

During the hygroscopic cycle test of residual soil, it must repeatedly go through the process of saturation→dehumidification→hygroscopicity. Although the soil samples release the suction formed during the normal consolidation process in the saturation stage, the soil sample needs to undergo the net normal stress and the upper limit matric suction (ψ _max = 400 kPa) applied during the dehumidification path, which is similar to the pre-consolidation stress required before the formation of over-consolidated soil. At the same time, as the increase of the target suction experienced by the sample, the average skeleton effective stress also increases correspondingly, resulting in the change rule of the void ratio as shown in Figure 7, so that the unsaturated soil sample after high target suction has the properties of over-consolidated soil. The above analyses show that high target suction or after multiple hygroscopic cycles, the shear properties of residual clayey soil show obvious strain softening characteristics.

It can be seen from Table 3 that both the peak and residual strength of residual soil decrease with the increase of the number of hygroscopic cycles, showing a characteristics of strain softening. With the continuous increase of the confining pressure and target suction level, the strength of soil decreases more obviously and the softening phenomenon becomes more obvious after multiple hygroscopic cycles.

The residual soil contains montmorillonite minerals. During the process of soil dehumidification to water absorption, the inter-layer expansion of the crystal cells in montmorillonite changes the soil structure (Xu et al., 2022; Xu et al., 2017; Zhan et al., 2014). At this time, the deformation caused by the structural change cannot be recovered in the subsequent water absorption cycle path, that is, irreversible expansion deformation occurs. As a result, the aggregate of the soil internal structure is obvious and has a larger void ratio, which eventually leads to the decline of the soil resistance to external load. Figure 7 shows the variation trend of soil void ratio under different times of hygroscopic cycles. It can be seen from the Figure 7 that the void ratio of residual soil increases with the advancement of hygroscopic cycle, but the ability of the soil to undergo volumetric deformation is weakened under high peripheral pressure and the target suction, that is to say, the expansibility of the residual clayey soil has a correlation between the stress and hygroscopic cycle effects. Owing to the significant water-force hysteria effect of soil-water characteristic curve of residual soil, soil under the same suction condition has lower saturation after multiple water absorption, and there are more pores affected by soil-water bending surface in low-saturation soil, which is conducive to soil stability, but the irreversible deformation generated is the main reason for the decline of soil strength (Pham and Sutman, 2023; Gao et al., 2023). The strength of the confining effect around the soil has an influence on the saturation of the soil after hygroscopic cycle, that is to say, high confining pressure conditions will make soil have higher water storage performance. Although the soil with higher saturation has fewer internal pores affected by soil-water curved liquid surface, it has a larger internal contact area between soil particles and water, which is favorable for enhancing the soil strength (Yan et al., 2023; Xu et al., 2022). As a result, the void ratio changes slightly with the increase of hygroscopic cycles number for σ = 200 kPa and ψ _t = 100 kPa, whereas the increase in the soil void ratio with σ = 50 kPa and ψ _t = 100 kPa is more pronounced.

Figure 6 shows the strength curves in the p-q plane of unsaturated soil for different numbers of hygroscopic cycles. According to the unsaturated soil shear strength formula proposed by Fredlund and Xing, the critical state line in the p-q plane of unsaturated soil in isotropic triaxial shear test can be represented by the following equation q = R ψ p + C ψ (1) Where p represents the net average stress and the magnitude is equal to (σ ₁ +σ ₂ +σ ₃ )/3-u _a ; q is the deviating stress, and its magnitude is determined by σ ₁-σ ₃. R(ψ) and C(ψ) are two critical state parameters related to the matric suction ψ, which are the slope and intercept of the p-q line, respectively. As can be seen from Figure 8, when the matric suction is 50 kPa, 100 kPa and 200 kPa, the critical state parameter R(ψ) of soil under 0, 1, 2, 3, 5, 7 hygrometric cycles changes in the order of 0.8359→0.8286→0.8158→0.8093→0.8009→0.7948, 0.8434→0.8356→0.8248→0.8255→0.8136→0.8084, 0 0.8537→0.8373→0.8343→0.8316→0.8225→0.8183, which indicates that the strength parameter R(ψ) gradually decreases with the increase of the number of hygroscopic cycles. According to the shear strength formula based on the two-stress state variable proposed by Fredlund and Xing (1994), the total cohesion c and the effective internal friction angle φ′ of the soil can be calculated by the critical state parameters R(ψ) and C(ψ) (Equation 1) of the following form: sin ⁡ φ ′ = 3 R ψ / 6 + R ψ c = 3 ‐ ⁡ sin ⁡ φ ′ · C ψ / 6 ⁡ cos ⁡ φ ′

Residual soils undergoing the same hygroscopic path cycle has greater total cohesion at higher target suction (see Table 4), but the overall increase in effective internal friction angle φ′ is not significant, with its maximum increase not exceeding 3%. The average effective internal friction angle φ′ under 50 kPa, 100 kPa and 200 kPa target suction is 21°, 21.3°, and 21.5°, respectively. At the same suction level, the total cohesion c and effective internal friction angle φ′ of residual soil decrease with the advancement of hygroscopic cycle, but the total cohesion c is more sensitive to the change of the number of hygroscopic cycles, especially the decrease of the total cohesive force at the suction level of 100 kPa reach 18.9%. In general, the different responses of soil cohesion and effective internal friction angle to the number of repeated hygroscopic cycles derived from the above analyses are similar to those of Nie et al. (2023), Xu et al. (2020a) and Feng et al. (2023). However, there is still a lack of further analysis of the variation rules of each component under the framework of unsaturated soil shear strength theory.

The experimental results (see Table 4) also show that the total cohesion c appears to be linear with the target suction ψ _t , this means that it can be expressed as follows (Fredlund and Xing, 1994) c = μ ψ t + i Where μ and i represent the slope and intercept of the ψ _t -c line, i describes the shear strength of the soil under saturation state when the suction is 0, also known as the saturated cohesion; μ is the contribution of suction to the shear strength of soil, which is equal to the tangent value of the suction friction angle φ ^b (i.e., μ = tanφ ^b ). The effect of the same target suction on the soil shear strength is different under different numbers of hygroscopic cycles (Figures 9, 10), and the relative decrease of suction friction angle φ ^b throughout the hygroscopic cycle is 1.94% → 3.95% → 1.37% → 0.83% → 0.28%, which reflects that the suction friction angle decreases significantly in the early part of the hygroscopic cycle but decreases less in the later part of the cycle. Therefore, when the target suction level is at 0–200 kPa, the suction friction angle of residual soil considering the effect of hygroscopic cycle can be regarded as a constant, with an average value of 14.7°.

Although Figure 10 also shows that the saturated shear strength of soil decreases with the increase of the water absorption cycle, the relative decline of the saturated shear strength of soil during the whole hygroscopic cycle is greater, with the decreasing rate of 7.61%→3.53%→4.88%→10.26%→6.43% in order, and shows the trend of decreasing with continuous fluctuation. Obviously, this is inconsistent with the rule that the suction friction angle tends to be stable in the later stage. The above analysis can be concluded that the suction friction angle and saturated shear strength changes in the microscopic damage mechanism is not exactly the same. The stage evolution of the residual soil shear strength in the whole life cycle may be affected by the suction friction angle in the early stage, while the stability of the soil in the later stage may be more restricted by the saturated shear strength.