Effect of Freeze–Thaw Cycles and Initial Water Content on the Pore Structure and Mechanical Properties of Loess in Northern Shaanxi

Abstract

Spalling disease caused by freeze–thaw cycles often occurs on the loess cut-slopes in northern Shaanxi. The deterioration of the pore structure and mechanical strength of loess under a freeze–thaw environment is one of the critical reasons underlying the occurrence of spalling disease in the slopes. In order to explore the effects of freeze–thaw cycles on the pore structure and the mechanical strength of loess, four initial water contents (7%, 9%, 12%, and 14%) and six freeze–thaw cycles (zero, one, three, five, 10, and 20) were considered in this study. Nuclear magnetic resonance (NMR) and triaxial compression tests were carried out to analyze and reveal the mechanisms of effect causing the deterioration of the soil strength that affects the stability of loess cut-slopes. The results showed that the porosity growth increased with the initial water content and continued to increase during the freeze–thaw process until a later stage of the freeze–thaw cycle, when it gradually stabilized. The stress–strain curves were primarily influenced by the number of freeze–thaw cycles, the initial water content of the samples, and the confining pressure. Both the cohesion and the internal friction angle exhibited a decay law that showed a significant decrease, then a slow decrease, and finally stabilization during the freeze–thaw process. Small and micropores were predominant among the pore structures of the loess, while medium pores were the second most common, and large pores were the least common. With the increase in the initial water content, the pores transformed from micropore structures to medium and large pore structures. The soil strength deterioration was primarily driven by the phase changes of the pore water, as well as the water migration during the freeze–thaw process. This study will be beneficial for identifying the characteristics and types of freeze–thaw disease in cut-slope engineering in seasonally frozen loess areas and providing a theoretical reference and design basis for achieving green and sustainable development in slope engineering, management, and maintenance.

1. Introduction

“Loess” describes sediments that have formed since the Quaternary period and exhibit well-developed pore structures and vertical joints. They are primarily distributed in the middle reaches of the Yellow River in China, particularly in northern Shaanxi Loess Plateau, where they have had significant development, resulting in thick and continuous distributions [1,2]. The soil in Northern Shaanxi freezes seasonally, and the shallow surface layers of the loess are affected by the freeze–thaw environment year round. The temperature differences generated by the short-term diurnal and long-term seasonal changes have resulted in a continuous freeze–thaw cycle, deteriorating the physical and mechanical properties of the soil [3,4,5]. In addition, the impacts of human engineering activities have led to frequent spalling, landslides, and other common issues on the loess cut-slopes in the region, seriously harming the ecological environment along the loess cut-slopes and impeding the coexistence of humans and nature [6,7,8]. Therefore, understanding how the loess in northern Shaanxi degrades during freeze–thaw cycles and identifying the destructive influences on loess cut-slopes could have significant theoretical and practical implications for the engineering design and management of loess cut-slopes in areas with seasonally frozen soil.

In areas with seasonally frozen soil, geological engineering hazards caused by freeze–thaw cycles have become a significant challenge [9,10]. Many researchers have discovered that pore water freezes at low temperatures and that the process of the water–ice phase change produces a 9% volume expansion, leading to a rearrangement of the skeletal structure and a weakening of the cohesion of the soil, which may be related to the phenomenon of freeze–thaw disease in the slopes of loess road grabens [11,12,13,14]. In order to reveal the influence of a freeze–thaw environment on the mechanical properties of loess, Wang et al. [15] explained the collapse phenomenon in loess samples under freeze–thaw conditions by comparing and analyzing the volume change rate of loess during the freeze–thaw process and enhanced the freeze–thaw resistance of loess by adding amendments. Li et al. [16] conducted freeze–thaw tests on loess with different degrees of compaction, and found that water migration and redistribution occurred within the samples. The loose samples became dense, while the dense samples became loose, which was the main cause of pavement cracking and uneven settlement in compacted loess roadbeds under freeze–thaw conditions. Li et al. [17] investigated the influence of freeze–thaw cycles on the mechanical strength of saturated loess using a triaxial compression test and analyzed the strength loss evolution law of the saturated loess during the freeze–thaw process. Wang et al. [18] studied the dynamic characteristics of saturated loess in freeze–thaw environments, and they proposed the deterioration mechanism and process of saturated remodeled loess under freeze–thaw cycles by analyzing and discussing the characteristics of the soil dynamic parameters.

In fact, the root cause of these hazards is the change in the microstructure of loess under freeze–thaw cycles, which leads to a change in its engineering properties [19]. Zheng et al. [20] found that the internal pore structures of loess were relatively loose using electron microscope scanning technology, and they verified that the strength of the soil had been reduced after changes in the pore structures using a true triaxial test. They also explored the correlations between the mechanical strength and the meso-structures of the remodeled loess. She et al. [21] studied the pore structures of loess after freeze–thaw cycles using scanning electron microscopy technology, and they analyzed the changes in their structural strength to establish a relevant statistical damage model. Xu et al. [22,23,24,25] obtained the strength parameters and microfine structural parameters of loess soil samples using CT scans, SEM tests, and triaxial shear tests. After a comparative analysis, they found that the structural damage of loess during the freeze–thaw process was a main causal factor behind deformations in the slopes and foundations, and they proposed damage variables based on the porosity. Xie et al. [26] conducted microfine structural scanning and unconfined compressive strength tests on undisturbed and remodeled loess to solve engineering problems, such as freeze–thaw spalling and deformed foundation settlements, and they found a close correlation between the compressive strength and microfine structures of the loess.

The water content of the soil is one of the primary reasons for freeze–thaw diseases on loess cut-slopes in freeze–thaw conditions. In order to examine the macroscopic physical and mechanical characteristics of freeze–thaw loess, current research mostly focuses on the mechanical property deterioration and microstructural alterations. The influence of the initial water content and freeze–thaw cycles on the pore structure and mechanical strength of loess is not well studied. Therefore, after extensive field investigations, this study conducted NMR tests and triaxial compression tests on remolded loess with different initial water contents. The pore structure characteristics and mechanical strength parameters under different numbers of freeze–thaw cycles and initial water contents were obtained, and the mechanisms impacting the soil strength deterioration on the stability of loess cut-slopes were also discussed. The research results will provide scientific and effective guidance for geological hazards in seasonally frozen loess areas and effectively improve the ecological environment of vegetation along loess cut-slopes, aiming to achieve green and sustainable development for slope engineering.

2. Field Research

2.1. Overview of the Research Area

In order to comprehensively understand the freeze–thaw erosion damage of the loess cut-slopes in northern Shaanxi, this research work used the typical freeze–thaw disease slopes along the national highways G210, G242, and G341 in Yan’an and Yulin as the research object, described and recorded the development characteristics of freeze–thaw disease in the loess cut-slopes, and conducted field tests of the water content in the shallow surface layer of the slope, which was used as the reference basis for the subsequent indoor experimental research. The geographical location of the study area and the research route are shown in Figure 1.

2.2. The Storage Environment Characteristics of Seasonally Frozen Soil

Seasonally frozen soil is very sensitive to changes in the external climate environment. The freezing depth of soil depends on factors such as the winter temperature, snow depth, and number of snow days. As the most active environmental factor in winter, snow not only reflects long-term changes in the winter climate, but changes in the environmental characteristics of the soil. By sorting out the relevant information of seasonally frozen soil in northern Shaanxi, the storage environmental characteristics of the frozen soil in northern Shaanxi were obtained, as shown in Figure 2.

According to meteorological data [27,28], the frost period in northern Shaanxi occurs at the end of October or the beginning of November each year. The above figure shows that the soils in Yan’an and Yulin began to freeze in November, and the freezing depth was approx. 20 cm and 45 cm, respectively. Afterwards, the freezing depth continued to increase as the temperature decreased and the snow thickness increased. Both regions reached the maximum freezing depth in January, with a maximum freezing depth of 79 cm in Yan’an and 92 cm in Yulin. After entering the thawing period, the snow melted, the temperature rose, and the freezing depth slowly decreased until the snow completely melted and decreased to its lowest value in April. It can be seen that the shallow surface soil of the loess cut-slope in the northern Shaanxi region periodically experienced freeze–thaw cycles, and freeze–thaw damage accumulated in the process, which eventually led to frequent freeze–thaw spalling disease on the slope surface.

2.3. Developmental Characteristics of Freeze–Thaw Spalling Disease

According to the field investigation results of the four typical freeze–thaw spalling disease sites shown in

Table 1. Geographical information of the freeze–thaw spalling disease research sites.

Location	Geographic Coordinates	Environment Description	Types of Freeze–Thaw Disease
Luochuan County (National Loess Geopark)	109°25′59.90″ E 35°42′35.64″ N	Moss layer cover	Epidermal
Baota District (G341 K1257 + 200)	109°50′29.26″ E 36°43′57.04″ N	Vegetation development	Layered
Suide County (G242 K762 + 300)	110°12′44.47″ E 37°39′44.43″ N	Scattered bushes	Fragmental
Mizhi County (G242 K750 + 750)	110°11′52.89″ E 37°45′22.51″ N	Little plant cover	Granular

, the freeze–thaw spalling damage situation of loess cut-slopes in northern Shaanxi can be summarized into four types of freeze–thaw spalling patterns, from south to north according to their scale and spalling bodies, namely epidermal, layered, fragmental, and granular patterns, as shown in Figure 3.

2.3.1. Epidermal Freeze–Thaw Spalling

The epidermal freeze–thaw spalling disease of the National Loess Geopark’s cut-slope is depicted in Figure 3a. It can be seen that most of the green moss plants developed on the slope surface, which had a good integrity and a certain slope protection effect. However, due to the low temperatures in winter, the water present in the moss plant layer froze to form ice crystals during the freezing period. The surface moss layer and the bottom soil were divided by the pressure of frost heave, and as the ice crystals melted, the distance between them grew wider. After repeated freeze–thaw cycles, the surface layer will be partially peeled off and the bottom soil will be exposed. This spalling phenomenon generally arises from the local peeled surface skin that accumulates at the bottom of the slope. When it develops to a certain range, it can lead to a large surface skin spalling phenomenon, as shown in Figure 3a.

2.3.2. Layered Freeze–Thaw Spalling

As shown in Figure 3b, layered freeze–thaw spalling is one of the common freeze–thaw spalling forms in the loess cut-slopes in northern Shaanxi, which mostly occurs on the surface of the slope with short excavation times. The spalling form often begins at the bottom of the slope and develops gradually upward along the slope surface, with the sides curving toward the edge and the main portion in the middle spreading body maintaining a rectangular shape. The frost swelling effect can result in microfractures in the soil at the partition interface when the environment outside is cold. When the microfractures expand to a certain point, they separate into multiple levels with the slope body as a result of their own weight. Despite the size of the spalling scale, the layer texture was distinct, the thickness was only approximately 5 cm, and the slope surface produced a clearly striped spalling surface.

2.3.3. Fragmental Freeze–Thaw Spalling

The soil on the loess cut-slope can develop a freezing front and continuously move inward as a result of the winter’s low temperatures. Under the impact of freezing suction, the unfrozen water inside the slope will migrate to the surface layer. Since this portion of the water cannot migrate back along its original course after melting, it causes water to concentrate in the surface layer of the soil. The fractures widen as the freeze–thaw activity gets stronger. The surface layer can then be split into blocks of various sizes and shapes when the internal and external cracks are probed, which eventually presents the fragmental freeze–thaw spalling phenomenon, as shown in Figure 3c.

2.3.4. Granular Freeze–Thaw Spalling

According to the makeup of the particles, the loess region in northern Shaanxi was divided into a sand loess belt and a clay loess belt. The above-mentioned fragmental freeze–thaw spalling primarily occurred in the transition area between the clay loess belt and the sand loess belt, while the difference between granular freeze–thaw spalling and fragmental freeze–thaw spalling in the sand loess belt was primarily reflected in two aspects: the severity of soil cracking and the morphology of the spalling body. When a fissure is penetrated under the influence of freeze–thaw cycles, the soil on the slope surface will also decompose. However, due to the decrease in the natural water content, cohesion, and the increase in the sand content, the decomposed soil was cracked into finer particles, and its spalling body was primarily granular, which accumulated at the foot of the slope to form granular freeze–thaw spalling, as shown in Figure 3d.

3. Test Scheme

3.1. Sample Preparation

A portable water tester with a measuring range of 0 to 100% and a reading error of ±3% was used in the field investigation to measure the water content of the soil at the four research sites shown in Figure 3 within 1 m of the shallow surface layer. The instrument is shown in Figure 4a. The results showed that in the four test sites, the average soil water content decreased in the following ratios from south to north: 14.3% in Luochuan County, 11.8% in the Baota District, 9.5% in Suide County, and 7.3% in Mizhi County. According to the field test results, the dry density of the sample was controlled at 1.55 g/cm³, and the representative values of the water content of 7%, 9%, 12%, and 14% were used to remold the samples with dimensions of φ = 39.1 mm and h = 80 mm. A total of 96 samples were created, 24 of which were used for the NMR test and 72 for the triaxial compression test. The dried lumpy soil was crushed and put through a 2 mm geotechnical test sieve in accordance with the geotechnical test method standard (GB/T 50123-2019). A soil sample with a target water content was obtained using the spray method after determining the required water content based on the target water content. The sample was sealed with a preservative film and left for 24 h to ensure that the water content in the soil could be distributed evenly. The sample was then finished in two layers of static pressure after the mass of the soil needed to create the sample was estimated based on the controlled dry density and water content. The samples utilized in this research were created using the pressed sample method to ensure consistency in the sample manufacturing procedure.

3.2. Freeze–Thaw Cycle Tests

In the current study, the freezing temperature, melting temperature, and duration of the freeze–thaw cycle effect were not unified by regulations, so they can only be relied on based on the real site conditions and a summary of the gained experience. Referring to the minimum and maximum surface temperatures during the freeze–thaw period in the last 20 years in Yan’an and Yulin, as shown in Figure 5a, the freezing and thawing temperatures in the freeze–thaw cycle test were determined to be −20 °C and 20 °C, each lasting 12 h for one freeze–thaw cycle, as shown in Figure 5b.

According to the findings of the earlier studies, the microstructure and strength qualities of loess are mostly affected by the early phases of the freeze–thaw process, and after 20 cycles, its physical and mechanical characteristics gradually stabilize [29,30,31]. Therefore, the precise number of freeze–thaw cycles was chosen at zero, one, three, five, 10, and 20 in this investigation. To ensure the stability of the sample in the freeze–thaw process, the entire freeze–thaw process was conducted in the low-temperature freeze–thaw tester model XT5412-MTC125, which had a temperature setting range of −40 °C to 150 °C and a temperature fluctuation of 0.5 °C, as shown in Figure 4b.

3.3. Nuclear Magnetic Resonance Tests

Nuclear magnetic resonance is incredibly sensitive to H-containing fluids in porous media materials and can probe changes in the internal pore structure by capturing hydrogen atoms in soil pore water, which has been widely applied in the field of geotechnical engineering, especially in the study of changes in the internal pore structure of geotechnical materials [32,33]. Therefore, the samples were separated into six groups after undergoing the required number of freeze–thaw cycles for this study, with each group denoting four different initial water contents (7%, 9%, 12%, and 14%). As shown in Figure 4c, each group of samples was first saturated using a vacuum saturator, and then a MacroMR12-150H-I NMR analyzer provided a variable frequency magnetic field to the saturated samples. To determine the change law of the pore structure of loess with various initial water contents under the influence of freeze–thaw cycles, the relationship between the NMR signal intensity and the transverse relaxation time T₂ of each sample was measured and then converted into the distribution of the pores inside the soil.

When the pore material was completely filled with water, the relaxation process was found to exhibit three main types of relaxation mechanisms, namely free relaxation, surface relaxation, and diffusion relaxation. Therefore, the lateral relaxation time T₂ of the fluid flow within the pore can be expressed as the following [34].

$$\frac{1}{T_2}=\frac{1}{T_{2free}}+\frac{1}{T_{2surface}}+\frac{1}{T_{2diffusion}}$$

(1)

where T₂ is the lateral relaxation time (ms); T_2free is the free relaxation time, i.e., the pore fluid relaxation time measured when the fluid is in an unrestricted free space (ms); T_2surface is the surface relaxation time, i.e., the pore fluid relaxation time due to surface relaxation (ms); and T_2diffusion is the diffusion relaxation time, i.e., the pore fluid relaxation time due to the diffusion of the gradient magnetic field (ms).

When the pore structure characteristics of freeze–thaw loess were studied by obtaining the relaxation characteristics of the fluid, the free relaxation and diffusive relaxation of the fluid within the pore were relatively weak and could be neglected. Therefore, the surface relaxation time is used to approximate the lateral relaxation time T₂ and Equation (1) can be simplified as the following [34].

$$\frac{1}{T_2}={\rho }_2\frac{S}{V}$$

(2)

where ${\rho }_2$ is the lateral relaxation rate of the media surface (μm/ms) and S/V is the ratio of the surface area of the pore to the volume of the fluid (μm⁻¹).

The samples continuously developed small-scale damage as a result of the freeze–thaw cycle, which was also reflected by the NMR tests at the same time that the internal pore structure was generated and expanded. Assuming that the pore form was cylindrical, Equation (2) resulted in the following when applied to the soil’s pore structure and the water molecules within it.

$$r=2{\rho }_2{T}_2$$

(3)

where r is the pore radius (μm).

3.4. Triaxial Compression Tests

Under repeated freeze–thaw circumstances, the water in the soil experiences migration and redistribution phenomena. The water–ice phase change process irreparably damages the interior of the soil, which finally results in a major loss of mechanical qualities. When rain and snow melt and permeate into the shallow surface layer of the loess cut-slope, they cause damage to the soil by going through a cycle of freezing and thawing while still in their natural water state. In order to conduct triaxial compression tests on the samples with various initial water contents under the confining pressures of 100 kPa, 200 kPa, and 300 kPa, the samples were placed in a fully automatic temperature-controlled creep triaxial apparatus after completing the zeroth, first, third, fifth, 10th, and 20th freeze–thaw cycles, as shown in Figure 4d. In an attempt to research the effects of freeze–thaw cycles and initial water contents on the mechanical properties of loess, stress–strain curves and strength parameters of the samples during various freeze–thaw cycles were produced. Additionally, this test was consolidated undrained with a shear rate of 0.08 mm/min. It can be assumed that the sample was totally ruined when the axial strain reached 15%.

4. Results and Analysis

4.1. Effect of Freeze–Thaw Cycles on the Pore Structure of Soils

4.1.1. T₂ Distribution Curves

Figure 6 displays the T₂ spectrum distribution curves of the samples with various initial water concentrations under freeze–thaw cycles. It can be seen that the samples with lower initial water contents (7% and 9%) exhibited a bimodal pattern on their T₂ distribution curves, but those with greater initial water contents (12% and 14%) exhibited a triple peak pattern. With more freeze–thaw cycles, the T₂ distribution curve moved to the right, indicating that the expansion of the pore structure underwent a dynamic change during the freeze–thaw process.

The signal intensity of the curve in the T₂ spectrum showed the amount of hydrogen atoms in the soil pore structure, and the relaxation time position and signal intensity of the curve were connected to the size and number of pores in the soil. The total intensity of the relaxation signal for the samples with a lower initial water content increased from 2841.827 a.u. to 3203.769 a.u. for the ω = 7% sample, with a primary peak relaxation time of approximately 0.34 ms and a secondary peak relaxation time of approximately 7.37 ms for the T₂ distribution curve. The total intensity of the relaxation signal for the ω = 9% sample increased from 3049.080 a.u. to 3752.627 a.u., the main peak relaxation time was approx. 0.36 ms, and the secondary peak relaxation time was approx. 7.35 ms. It was found that the pore size under the characteristic peak of the two water content samples was essentially the same. However, the difference in the initial water content led to a significant difference in the total intensity of the relaxation signal, and this difference was mainly reflected in the number of pores in the two samples.

In the samples with a higher initial water content, the three peak relaxation times appeared around 0.49 ms, 6.85 ms, and 54.98 ms for ω = 12% and around 0.48 ms, 7.08 ms, and 55.14 ms for ω = 14%. The pattern of changes in the signal strength both before and after the freeze–thaw cycles was consistent with the pattern described above. The overall relaxation signal intensity of the samples with ω = 12% and ω = 14% after the freeze–thaw cycles was 1.20 times and 1.31 times greater than that before the freeze–thaw cycles, respectively. This indicated that the number of internal pore structures in the soil increased significantly with the progression of the freeze–thaw cycles. The distinctive pore structure inside the soil increased from two types of lower initial water content to three types of higher initial water content, according to the change in the T₂ distribution curve’s peak shape. The pore structure obviously produced a large pore structure in the process of expanding the penetration, which was primarily reflected in the third peak of the distribution curve.

4.1.2. Soil Porosity

If the sample was saturated at the time of the NMR scan and it was assumed that all of the soil’s pores were filled with water, the equation for the observed soil sample’s porosity n can be rewritten as follows [35].

$$n=\frac{V_w}{V}$$

(4)

where V_w is the volume of the water in the pore space of the soil (cm³) and V is the total volume of the soil sample (cm³).

As illustrated in Figure 7, the samples with known water contents were used to establish the correspondence between the NMR signal strength and the water volume under the identical test conditions. A linear fit to the samples produced the following relationship.

$$S_{T_2}=196.743V_w$$

(5)

where S_T2 is the total area of the T₂ spectrum with a known water content.

Equation (5) can be substituted for Equation (4) to obtain the following expression for the T₂ spectrum total area S_T2 of the water in the pores of any soil sample with porosity n.

$$n=\frac{S_{T_2}}{196.743V}$$

(6)

Therefore, by substituting the total area S_T2 of the NMR signal intensity curve from Figure 6 into Equation (6), the porosity n of the soil samples with each water content at various numbers of freeze–thaw cycles was obtained, and the calculation results are shown in Figure 8.

As shown in Figure 8, it can be observed that the porosity of the soil samples with various initial water contents changed in different ways during the freeze–thaw process. For instance, the porosity of the soil samples with initial water contents of 7%, 9%, 12%, and 14% increased by 1.915%, 3.723%, 5.775%, and 9.524%, respectively. The porosity growth value of the soil sample with ω = 14% was 4.97 times larger than that of the soil sample with ω = 7%, and it was unmistakably discovered that the porosity growth increased with the initial water content. The water in the soil sample initially froze in place, causing the initial pore structures to expand and penetrate one another. The water was also redistributed during this process, and the porosity increased. The development of the soil porosity in a freeze–thaw environment is primarily accomplished by freeze swelling during the water–ice phase change. Since the water content of the soil varied, so did the force of the freeze swelling, which resulted in variations in the rate of porosity development in the soil samples.

4.2. Effect of Freeze–Thaw Cycles on Triaxial Compression Characteristics

4.2.1. Stress–Strain Curves

In order to grasp the mechanical properties of loess with different initial water contents under a freeze–thaw environment, after the freeze–thaw cycle tests were completed for the samples with initial water contents of 7%, 9%, 12%, and 14%, the consolidated undrained triaxial compression tests under 100 kPa, 200 kPa, and 300 kPa confining pressures were conducted for the samples in saturated conditions, and the stress–strain curves were obtained, as shown in Figure 9.

As shown in Figure 9, the stress–strain curve exhibited slight hardening, illustrating the elastic deformation phase at the start of the test. The soil experienced irreversible plastic deformation after the axial strain ε reached 1%, the partial stress steadily increased, and the stress–strain curve exhibited a “straight and then curved” changing pattern. The number of freeze–thaw cycles, the initial water content of the samples, and the confining pressure all had a significant impact on the stress–strain curve. The first three freeze–thaw cycles caused the water in the soil’s internal pores to condense into ice crystals, and the volume expansion caused by the water–ice phase change destroyed the native association mode between the soil particles, significantly degrading the sample’s peak partial stress. However, from the third to the 10th freeze–thaw cycles, the internal structure gradually stabilized, the degrading effect gradually lessened, and the decay rate of the partial stress value gradually decreased and finally stabilized during the 10th to 20th freeze–thaw cycles. The effect of the initial water content on the stress–strain curves was not significant under the confining pressure conditions of 100 kPa and 200 kPa. However at 300 kPa, the peak partial stress of the stress–strain curve decreased significantly with an increase in the initial water content. The peak partial stress initially decreased from 424.79 kPa to 333.85 kPa, and the peak partial stress after freeze–thawing decreased from 342.19 kPa to 270.81 kPa. This was due to the fact that an increase in the initial water content caused the friction between the soil particles to decrease and the soil skeleton’s strength to diminish due to the lubricating effect of the water molecules. The closer occlusion friction of the soil particles, which was expressed in the macroscopic shear strength, resulted from an increase in the confining pressure. Therefore, the position of the axial strain corresponding to the peak point also shifted with an increase in the confining pressure. This was why an increase in the confining pressure resulted in a greater increase in the soil shear strength.

4.2.2. Strength Parameters

According to the Mohr–Coulomb strength theory, the Mohr stress circle corresponding to the stress–strain relationship shown in Figure 9 was plotted to obtain the cohesion c and internal friction angle φ of the samples with different initial water contents. The results are shown in Figure 10. The cohesion increased from 11.98 kPa to 23.33 kPa with an increase of 94.94% as the initial water content of the sample rose from 7% to 14%, while the internal friction angle gradually decreased from 23.58° to 19.11° with a decay rate of 18.96%. When the water content of the soil was lower than the optimal water content (the optimal water content of the soil used in this paper was 15.27%), the water in the soil mainly existed in the form of a bound water film. There were fewer free water molecules around the soil particles, and the electrostatic attraction generated by the electric double layer structure was also smaller, which reduced the binding force between the soil particles. As the water content increased, the electrostatic attraction and water film binding force between the soil particles correspondingly increased, and the water molecules in the weakly bound water film moved along the surface of the soil particles, providing lubrication for the relative motion between the particles. Its macroscopic manifestation was an increase in the cohesion and a decrease in the internal friction angle, which Wang et al. noticed during their research [36]. After the water content exceeded the optimal water content, the electric double layer structure of the soil particles reached a stable state, and the increased water mainly existed in the soil as free water. The free water molecules weakened the cementation linkage between the soil structures, which reduced the cohesion of the soil and no longer played a significant role in lubricating the relative motion between the particles [37].

According to the variation law of the cohesion and internal friction angle shown in Figure 10, the changes under the different numbers of freeze–thaw cycles were divided into three stages. Stage I was a fast decreasing stage of zero to three times, stage II was a slow decreasing stage of three to 10 times, and stage III was a stable stage of 10–20 times. In order to reflect the decay law of shear strength indicators under freeze–thaw cycles more intuitively, the decay rate of shear strength indicators was defined as follows.

$$R_{c_N}=\frac{c_0-c_N}{c_0}\times 100\%$$

(7)

$$R_{{\phi }_N}=\frac{{\phi }_0-{\phi }_N}{{\phi }_0}\times 100\%$$

(8)

where $R_{c_N}$ is the decay rate of the cohesion after N freeze–thaw cycles (%); c₀ is the initial cohesion of the sample (kPa); c_N is the cohesion of the sample after N freeze–thaw cycles (kPa); $R_{{\phi }_N}$ is the decay rate of the internal friction angle after N freeze–thaw cycles (%), φ₀ is the initial internal friction angle of the sample (°), and φ_N is the internal friction angle of the sample after N freeze–thaw cycles (°).

Substituting c and φ from Figure 10 into Equations (7) and (8), respectively, we obtained the decay law of the shear strength parameters under freeze–thaw cycles, as shown in Figure 11. Within stage I, c and φ decayed at a faster rate, especially after the third freeze–thaw cycle. R_c with an initial water content of 7%, 9%, 12%, and 14% increased from 7.6%, 4.53%, 4.73%, and 6.13% to 16.94%, 14.23%, 12.27%, and 15.52%, respectively, and R_φ increased from 2.46%, 3.44%, 2.67%, and 2.88% to 8.82%, 7.46%, 8.6%, and 7.69%, respectively, indicating that the strength parameters of the sample showed the most significant decay effect during the first three freeze–thaw cycles. This was because the volume expansion brought on by the water–ice phase change during freezing and the water migration during the freeze–thaw cycle destroyed the inter-particle association mode, the ice crystals had an extrusion effect on the soil particles, and the skeletal structure of the soil particles deteriorated, which resulted in a decline in the cementation and occlusion of the soil body. When the sample passed through stage I, the internal water did not completely return to its original path, the water migration from the soil’s interior to the cold end’s outer surface decreased, the expansion rate of the initial damage decreased during the new water–ice phase transition, and the destroyed pore structure gradually lessened, while a new stable structure was also formed. R_c and R_φ within stage II exhibited a degradation behavior with a reduced decay rate relative to stage I. Within stage III, the internal pore structure of the soil accommodated the ice crystals produced during the water–ice phase change, and the internal structure and mechanical properties of the soil were significantly reduced by the freeze–thaw effect. The average growth rates of R_c and R_φ were only 1.61% and 0.4% from the 10th to the 20th freeze–thaw processes, and the c and φ gradually achieved to a stable state. Compared to the initial values, the cohesion of the samples with an initial water content of 7%, 9%, 12%, and 14% after the freeze–thaw cycles were 8.87 kPa, 11.34 kPa, 15.33 kPa, and 18.59 kPa, respectively. The decay rates were 25.96%, 20.92%, 15.68%, and 20.32%, respectively. The internal friction angles were 20.58°, 19.49°, 17.88°, and 16.72°, and the decay rates were 12.72%, 11.85%, 13.12%, and 12.51%, respectively. The cohesion and the internal friction angle showed a similar pattern of change during the freeze–thaw process. They decreased significantly and then gradually stabilized, which was consistent with the growth trend of the porosity. This also indicated that the decay law of the strength parameters was closely related to the degree of freeze–thaw damage to the internal pore structure of the sample.

5. Discussion

5.1. Effect of Freeze–Thaw Cycles and Initial Water Content on the Pore Structure of Loess

Equation (3) was used to invert the T₂ spectral distribution curve shown in Figure 6. Based on the relaxation time T₂ corresponding to the various fugacity states of water in the pores of the soil samples and the method proposed by Lei to classify the size and type of the loess pore structure [38,39], the loess pores were classified into the following four types: r < 1 μm for micropores, 1 μm ≤ r ≤ 4 μm for small pores, 4 μm ≤ r ≤ 16 μm for medium pores, and r > 16 μm for large pores. The pore distribution curve is shown in Figure 12. The effects of the freeze–thaw cycles and the initial water content on the pore structure of loess are discussed below using only the samples with ω = 14% and N = 20 as examples.

As shown in Figure 12, the loess pore distribution was primarily composed of three peaks. The proportion of the pore structure that matched the pore size increased with an increasing peak size. Small pores comprised the majority of its pore structure, with medium pores coming in second and micropores and macropores coming in last. As shown in Figure 12a, it was evident that during the freeze–thaw cycle, the T₂ spectral area of the pore distribution curve of the soil sample with ω = 14% increased, showing that the total number of pore structures and the pore radius continued to increase, primarily demonstrating that the micro and small pores gradually decreased, and the medium and large pores continued to increase. The reason for this was that when water condensed into ice crystals at low temperatures, the forces created by the freezing and swelling of the ice crystals squeezed the surrounding soil particles. These squeezed soil particles did not completely return to their initial state after the ice crystals melted, which caused the soil to gradually expand its original pore structure while also producing new tiny pores. When the pore structures pierced one another, additional huge pore structures were generated, which ultimately changed the organization of the skeletal structure and weakened the soil connection force. Similar phenomena were seen by Zhou and Taina et al. [40,41] in their work. The pore distribution curves of the four samples with various initial water contents revealed the distribution pattern depicted in Figure 12b after the conclusion of 20 freeze–thaw cycles. With an increase in the initial water content, the curve shifted to the right and upward, showing that the percentage of pore sizes and types changed. This mostly demonstrated the transformation of a micropore structure to a medium–large pore structure. The phenomenon was particularly noticeable when the initial water content increased from 9% to 12%, and the proportion of micropores and medium–large pores changed from 85.73% and 14.27% to 60.96% and 39.04%, respectively. This was because the sample with a lower initial water content was less likely to contain ice crystals when it was frozen, which in turn resulted in a smaller amount of freezing and swelling forces in the soil and fewer pores that formed after melting. Free water began to appear in the skeleton structure as the initial water content increased. In a low-temperature environment, these water molecules quickly crystallized into ice, which destroyed the colloidal association of the soil and caused the pore structure to open up and the cohesion and internal friction angle to decrease.

5.2. Effect of Freeze–Thaw-Induced Soil Strength Deterioration on Slope Stability of Loess Cut-Slopes

Phase changes in pore water and water migration during freeze–thawing are the main causes of the decline in soil strength caused by freeze–thawing. The skeletal structure is compressed by frost swelling forces brought on by the water–ice phase transition within the pore space during the freeze–thaw process (Figure 8). The arrangement of the structural units of the soil is altered by water migration during freeze–thawing, and this water also penetrates the expanded pore structure, weakening the cementation between the particles and greatly reducing the soil’s cohesion (Figure 10). It also demonstrates how the stability of loess cut-slopes during the freeze–thaw process is significantly influenced by the water–ice phase transition and water migration. As shown in Figure 3a,b, layered body spalling occurs when the soil on a slope has a high water content, high cohesion, and fissures formed by pore penetration. This epidermal spalling also occurs when moss plants grow on the surface. It is very simple for soil on a slope to crack under the freeze–thaw cycles, and the fissures after penetration will split the soil into granular and fragmental forms, as illustrated in Figure 3c,d. This is because when the water content of the soil is low, the soil’s cohesion is also low.

The above indoor test procedure was conducted in a closed and non-replenishing environment. However, in the actual project, the freeze–thaw process of the slope soil was conducted in an open and replenishing environment, and the water migration phenomena produced by the freeze–thaw cycles were more extreme. When it rained or snowed, water infiltrated from the outside of the slope to the inside under the influence of gravity. As the water seeped into the soil pores, the clay minerals swelled when they came in contact with the water, creating an interlayer misalignment that produced migration channels for the ensuing water infiltration. The slope soil was the first to freeze when the outside temperature dropped to 0 °C. The unfrozen water in the slope body migrated to the frozen area under the influence of a temperature gradient and crystallized into ice under the influence of low temperatures. The freezing and swelling force produced by sub-condensing into ice squeezed the surrounding soil particles and further distorted the internal pore structure of the soil body. As a result, the strength of the soil gradually deteriorated with additional freeze–thaw cycles, and the loess cut-slopes suffered spalling damage. As a result, the soil strength gradually deteriorated with an increase in freeze–thaw cycles, and the loess cut-slope also exhibited various types of freeze–thaw spalling damage patterns with different water contents of the soil, such as epidermal, layered, fragmental, and granular patterns.

Within the range of seasonally frozen loess areas, freeze–thaw cycles and water content are the main reasons for inducing spalling damage on the slope surface. The research results of this paper provide a theoretical reference for studying the physical and mechanical properties of freeze–thaw loess in northern Shaanxi and also provide some reference for studying the influence of water content and freeze–thaw cycling on loess in seasonally frozen soil areas. In addition, when designing and protecting loess cut-slopes in the above-mentioned areas, the impact of the freeze–thaw loess pore structure change and mechanical property degradation on the actual engineering can be considered based on the test results of this paper. Thus, it is important to study the mechanical properties and meso-structure of loess with different initial water contents in a freeze–thaw environment. However, due to the large distribution range of loess, this paper only used the soil water content measured during the field investigation as the research variable. Subsequent research should consider a wider range of water contents, a greater number of freeze–thaw cycles, and different freeze–thaw cycle temperatures to improve the general applicability of the test results for seasonally frozen soil areas. Meanwhile, ecological protection and management should also be carried out for the types of freeze–thaw diseases discovered during the field research to achieve green and sustainable development.

6. Conclusions

This paper summarized the development characteristics and types of freeze–thaw diseases on the loess cut-slopes in the northern Shaanxi region and discussed the effects of freeze–thaw cycling and the initial water content on the pore structure and mechanical strength of loess in combination with the NMR tests and triaxial compression tests, which were of practical engineering value for the identification of freeze–thaw hazards and their stability evaluation for slope engineering in seasonally frozen loess areas. The main conclusions are as follows.

(1)

Under the influence of the freeze–thaw cycle, the T₂ distribution curves of the samples with varying initial water contents primarily displayed two forms. The samples with lower water contents (7% and 9%) displayed a double-peaked form, while the samples with higher water contents (12% and 14%) displayed a triple-peaked form. The soil sample’s porosity increased throughout the freeze–thaw process, and the higher the initial water content, the greater the increase in the porosity. Over time, when the freeze–thaw cycle was performed at a later stage, the porosity gradually stabilized.

(2)

The quantity of freeze–thaw cycles, the initial water content of the sample, and the confining pressure all had a significant impact on the stress–strain curve. When the number of freeze–thaw cycles and initial water content increased, the peak partial stress decreased significantly. As the confining pressure increased, the position of the axial strain corresponding to the peak point shifted. During the 10th–20th freeze–thaw cycles, all of the aforementioned variation processes gradually achieved a steady state.

(3)

The cohesion and internal friction angle revealed diametrically opposing magnitude relationships when the soil samples were not freeze–thawed, which was brought on by the different initial water contents. During the freeze–thaw process, the cohesion and internal friction angle exhibited a similar decay law, first declining significantly, then declining slowly, and finally tending to a steady state, which matched the three varying stages of c and φ.

(4)

The pore structure type of loess was mainly micropores and small pores, with medium pores coming in second and the largest pores coming in last. As the soil sample experienced freeze–thaw cycles, the spectral area of the pore distribution curve increased and gradually moved to the right, showing that both the overall number of pore structures and the pore radius continuously increased.

(5)

The pore space clearly varied from small to large pore structures as the initial water content increased. This phenomenon was particularly evident during the increase in the initial water content from 9% to 12%, with the proportion of micro and small pores decreasing from 85.73% to 60.96% and the proportion of medium and large pores increasing from 14.27% to 39.04%.