Introduction

Seasonally frozen soil regions occupy a substantial part of the global landscape, with survey data showing a total area of approximately 44,682,000 km2, representing 30% of the global land area1. Freeze–thaw cycles, a common natural phenomenon in these regions, significantly influence soil pore characteristics, which critically affect the strength and deformation properties of the soil2,3,4. For engineering projects in seasonal frost regions, the effects of freeze–thaw cycles on soil pore characteristics must be carefully considered, especially in embankment and tunnel engineering5,6. Numerous studies have investigated this issue, finding that for various types of clayey soil, cohesion initially decreases during the first few freeze–thaw cycles and stabilizes after approximately 5–9 cycles. The internal friction angle increases, peaking at around 5–7 cycles, and then remains stable7,8,9,10.

In recent years, the pulsed NMR technique has gained popularity in geotechnical engineering due to its efficiency and accuracy. Applications include evaluating soil pore size distribution (PSD), detecting the local structure of rock matrices, and measuring unfrozen water content in frozen soil11,12,13. The NMR technique reliably detects protons in water molecules, with pore structures represented by proton transverse relaxation time (T2) and longitudinal relaxation time (T1) distributions.

Moreover, soil permeability is strongly influenced by its intrinsic pore characteristics14. Comprehensive analysis shows that clayey soil permeability is highly sensitive to freeze–thaw cycles, mainly due to their substantial effects on soil microstructure, including alterations in pore size and porosity15,16. These microstructural changes ultimately govern the macroscopic permeability behavior of the soil. In seasonally frozen soil regions, engineering projects using naturally compacted clay as waterproof linings—such as in landfills, tailings dam systems, and clay-core rockfill dams—rely heavily on the permeability performance of the clay layer, which directly affects the safety and stability of the entire waterproofing system. Experimental studies have shown that soil permeability is significantly influenced by factors such as temperature, dry density, and external pressure17,18,19,20. Zhu et al.21 studied the permeability and mechanical behavior of expansive soils under freeze–thaw cycles, identifying regular patterns of characteristic changes and detailing the development of internal soil fissures over time. Teng et al.22 analyzed the effects of freeze–thaw cycles on soil structure and its physical and mechanical properties, finding that these cycles increase soil permeability and reduce the plasticity index. They also observed contrasting trends in soil density and void ratio changes. These findings highlight the sensitivity of clayey soil permeability to freeze–thaw cycles, largely due to their pronounced effects on soil microstructure, such as changes in pore size and porosity. These microstructural changes ultimately dictate the macroscopic permeability behavior of the soil.

In summary, substantial progress has been made in understanding the effects of freeze–thaw cycles on the pore structure and permeability of clayey soils. Nevertheless, research on the evolution of permeability characteristics, particularly the correlation between pore structure and permeability, remains limited. This study investigates the permeability characteristics of remolded Jilin ball clay under varying dry densities and water contentsthrough comparative experiments under freeze–thaw cycles. Using nuclear magnetic resonance (NMR) technology, the study explores how freeze–thaw cycles alter the pore structure of remolded ball clay, elucidating the intrinsic relationship between pore structure evolution and permeability. The goal is to provide a theoretical foundation and data support for designing clay structures in seasonally frozen regions.

Test materials

Powdery Jilin ball clay was used as the test material in this experiment. The clay, formed through Tertiary sedimentation, consists mainly of disordered kaolinite and quartz particles, with minor contributions from illite. Its mineral composition is shown in Fig. 1. Laser particle size analysis reveals that Jilin ball clay has small particle sizes, a large specific surface area, and a concentrated particle size distribution. As presented in Table 1, the average particle diameter of Jilin ball clay is 0.67 μm, classifying it as a natural nanoscale material with high physicochemical activity. The fundamental physical properties of Jilin ball clay are summarized in Table 2.

Fig. 1
figure 1

The XRD spectrum of Jilin ball clay.

Full size image
Table 1 Particle size distribution of Jilin ball clay.
Full size table
Table 2 Basic physical property of Jilin ball clay.
Full size table

Experimental design

Specimen preparation

To examine the effects of freeze–thaw cycles on the pore structure and permeability of ball clay under different dry densities and water contents, the detailed experimental plan is provided in Table 3. The specimen preparation procedure was as follows: (1) Distilled water was evenly sprayed onto dry powdered ball clay, mixed until reaching a water content of 17.7%, sealed, and left to rest for 12 h; (2) Specimens with dry densities of 1.60 g/cm3, 1.65 g/cm3, and 1.70 g/cm3 were compacted in sample rings layer by layer in three layers; (3) For the saturated specimens (B1–B6), the procedure commenced with the evacuation of compacted samples within a sealed chamber for a duration of three hours. Subsequently, distilled water was introduced into the chamber to submerge the samples, and the evacuation process was extended for an additional 3 h. Ultimately, the chamber was exposed to atmospheric conditions, allowing the samples to remain immersed in distilled water for a further 12 h. To meet the size requirements for nuclear magnetic resonance (NMR) testing and standard variable head permeability tests (SVHPT), two specimen sizes were prepared. After specimen preparation, all samples were placed in a controllable freeze–thaw chamber, wrapped in plastic to prevent moisture loss. The chamber was set to − 30 °C for 12 h during freezing and 15 °C for 12 h during thawing, providing sufficient time for the sample to reach thermal equilibrium12,23. This completed a 24-h cycle (as shown in Fig. 2). The number of freeze–thaw cycles is listed in Table 3, with 0 cycles indicating no exposure to freeze–thaw conditions. After completing the designated number of cycles, the specimens were tested using NMR and SVHPT.

Table 3 Experiment arrangements.
Full size table
Fig. 2
figure 2

Nuclear magnetic resonance (NMR) equipment and samples.

Full size image

Nuclear magnetic resonance (NMR) testing method

To examine the pore structure characteristics of Jilin ball clay subjected to multiple freeze–thaw cycles, a series of Nuclear Magnetic Resonance (NMR) tests were performed on both saturated and unsaturated samples after varying numbers of cycles. These experiments employed a 23-MHz MiniMR system, developed collaboratively by the Institute of Rock and Soil Mechanics at the Chinese Academy of Sciences and Niumag Corporation, China. As depicted in Fig. 2, the comprehensive NMR system consists of a sample tube, a magnet unit, a radio frequency (RF) system, and a data acquisition and analysis system. Upon completion of the freeze–thaw cycles, the specimens were unwrapped and inserted into the sample tube. The T2 spectrum relaxation time serves as an indicator of the size and distribution of the internal pores within the sample24,25; the specific relationship between the T2 relaxation time and the pore size R is detailed as follows:

$$\frac{{1}}{{{\text{T}}_{{2}} }}{ = }\rho_{{2}} \frac{{3}}{{\text{R}}}$$
(1)

In the equation, R represents the pore radius, and ρ2 denotes the surface relaxation coefficient, which is closely related to the soil’s physical properties. Therefore, T2 is directly proportional to the pore radius (R), reflecting the distribution of pore water within the porous medium.

Experimental results analysis

Pore size distribution characteristics of soil samples under freeze–thaw cycling conditions

Based on previous studies26,27, the T2 relaxation time of soil pores are calculated using Eq. (1), classifying the pores into micropores (T2 ≤ 10 ms), mesopores (10 ms < T2 ≤ 100 ms), and macropores (T2 > 100 ms). Figure 3 shows the T2 distribution and pore size distribution curves for samples B5 and F1 after several freeze–thaw cycles. The results show that all samples exhibit a single-peak micropore characteristic, this is attributed to Jilin ball clay’s small and uniform particle size (mean particle diameter: 0.67μm) and large specific surface area28. As the number of freeze–thaw cycles increases, the T2 curve peak for group B5 (saturated condition) shifts slightly to the right, indicating an enlargement of pore size, consistent with the pore size distribution for group B5. In contrast, the T2 distribution peak for group F1 (unsaturated condition) shows minimal variation, indicating weaker changes in pore size. The pore size distribution of F1 samples remains stable, showing no distinct patterns.

Fig. 3
figure 3

T2 Time distribution curves and pore size distribution curves of samples under different water content conditions (NFT stands for number of freeze–thaw cycles).

Full size image

Surface characteristic changes of specimens

Figure 4 shows the evolution of surface structural characteristics of samples B5 and F1 under different water content conditions after various freeze–thaw cycles. The results show that after one freeze–thaw cycle (NFT = 1), micro-cracks appeared on the surface of the B5 samples. As the number of freeze–thaw cycles increased to six (NFT = 6), the cracks on the surface of B5 samples grew significantly, forming distinct primary cracks and multiple irregular fractures. Increasing the freeze–thaw cycles to ten (NFT = 10) caused localized surface collapse in the B5 samples. After fifteen freeze–thaw cycles (NFT = 15), the B5 samples showed significant large-scale collapse and localized disintegration on the surface, indicating a severe loss of structural integrity. In contrast, despite an increase in freeze–thaw cycles, the F1 samples retained a smooth surface without significant cracks. This phenomenon indicates that samples at optimal water content remain structurally stable, showing no significant changes after multiple freeze–thaw cycles.

Fig. 4
figure 4

Sample surface characteristics with different freeze–thaw times.

Full size image

Relationship between void ratio and freeze–thaw cycles

Figure 5 shows the variation of the specimen’s void ratio with the number of freeze–thaw cycles. Under saturated conditions, the void ratio of the specimens is strongly influenced by both the number of freeze–thaw cycles and dry density. The initial void ratio values are recorded at 0.644, 0.691 and 0.745, corresponding to the B5, B3 and B1, respectively. Upon culminating the 15th freeze–thaw cycle, the void ratio values ascended to 0.858, 0.867, and 0.891 for each respective dry density. Specimens with lower dry densities exhibit higher void ratios, yet they demonstrate the smallest magnitude of increase in void ratio compare to the specimens with higher dry densities. The growth rates of void ratio are 33.23%, 25.47%, and 19.61% corresponding to the B5, B3 and B1, respectively. The most significant increase in void ratio occurs when the number of freeze–thaw cycles reaches two. As the number of freeze–thaw cycles increases, the rate of void ratio increase slows down. An exponential correlation exists between the void ratio and the number of freeze–thaw cycles for the B1, B3, and B5 groups, expressed as:

$${\text{e = }}\alpha {\text{N}}_{{{\text{FT}}}}^{\beta }$$
(2)

where e represents the void ratio under saturated conditions, NFT denotes the number of freeze–thaw cycles, and α and β are coefficients related to the inherent properties of the soil.

Fig. 5
figure 5

The relationship between the sample’s void ratio and the number of freeze–thaw cycles.

Full size image

The results demonstrate that inter-particle junctions within clay exhibit increased flexibility in samples with relatively lower dry densities, resulting in more pronounced pore distributions. Despite this, the soil’s structural integrity remains largely unaffected by the stresses induced by frost heave. However, the soil structure shows increased variability with rising dry density. In contrast to samples with lower dry densities, those with higher dry densities may experience an increase in porosity due to freeze–thaw cycles11,28. This is likely because particles may not return to their exact original positions during these cycles, and the particle-to-particle bonding capacity may be reduced, rendering the soil structure slightly looser than before the freezing process.

Unlike the significant changes in void ratio observed under saturated conditions, the F1 specimens show a different trend. As the number of freeze–thaw cycles increases, the void ratio of the F1 specimens first decreases, then increases, and finally stabilizes, with a small overall variation. During the initial phase of freezing, the unfrozen water within the specimens migrates towards the colder end, forming ice lenses at the freezing front29. This process results in the volumetric expansion of ice, which exerts radial compressive stress on the surrounding particles, temporarily compacting the soil mass and reducing the void ratio. Between the 3rd and 5th freeze–thaw cycles, frost heave forces dislodge particles from their original positions. Furthermore, stress concentrations at the ice-soil interface are susceptible to the formation of micro-cracks, leading to an increase in the void ratio. By the 6th freeze–thaw cycle, these changes tend to stabilize, and the void ratio reaches a steady state.

The impact of freeze–thaw cycles and void ratio on soil permeability characteristics

Figures 6 and 7 illustrate the patterns of permeability coefficient variation for specimens B2, B4, and B6 under saturated moisture conditions in response to freeze–thaw cycles and their correlation with void ratio. As shown in Fig. 6, the permeability coefficients of the specimens generally increase after freeze–thaw cycles, exhibiting an exponential rise with the number of cycles. Although the void ratio tends to stabilize after multiple freeze–thaw cycles, the permeability coefficient does not stabilize and instead increases exponentially with the void ratio, as shown in Fig. 7. This phenomenon can be attributed to the freeze–thaw effect on the pore structure of saturated soil, which increases the void ratio and consequently enhances the permeability coefficient. Figure 8 provides a detailed depiction of this process. Initially, the pores within the samples are uniformly distributed as micropores. However, following 15 freeze–thaw cycles, there is an increase in the void ratio due to the progressive degradation of the soil structure. This freeze–thaw action induces a transformation of the particles from a dispersed to an aggregated state, resulting in the formation of stable seepage channels23. Despite the void ratio remaining within the micropore range and eventually stabilizing, there is a notable increase in the permeability coefficient of the specimens. In contrast, the permeability coefficient of the F2 group specimens shows only a slight increase after the initial freeze–thaw cycle and quickly stabilizes, with no apparent functional relationship between the permeability coefficient and void ratio. This suggests that under optimal water content and maximum dry density conditions, the impact of freeze–thaw cycles on soil structure and permeability is limited.

Fig. 6
figure 6

Impact of freeze–thaw cycles on the sample’s permeability coefficient.

Full size image
Fig. 7
figure 7

Impact of void ratio on the sample’s permeability coefficient.

Full size image
Fig. 8
figure 8

The evolutionary law of pore structure of specimens after freeze–thaw cycles.

Full size image

Influence of dry density and freeze–thaw cycles on permeability coefficient

Figure 9 shows the effect of varying freeze–thaw cycles on the ratio of the permeability coefficient to the initial permeability coefficient (k/k0, where k0 is the permeability coefficient of the specimen with NFT = 0) for specimens B2, B4, and B6 under saturated moisture conditions, along with its relationship to void ratio. A ratio approaching 1.0 indicates minimal influence of freeze–thaw cycles on the permeability coefficient, while a significant deviation from 1.0 suggests a notable change in the permeability coefficient. As seen in the figure, an increase in dry density leads to a linear decrease in the permeability coefficient ratio, with the slope of this relationship steepening as the number of freeze–thaw cycles increases. Specifically, after 15 freeze–thaw cycles, the permeability coefficient ratio for specimen B6 increases to 6.14, indicating a sixfold rise in the permeability coefficient compared to its original value. Specimens B4 and B2 show ratios of 9.20 and 12.81, respectively, further emphasizing the significant effect of freeze–thaw cycles on the permeability of saturated soil. The results indicate that increasing the dry density of soil can effectively reduce the impact of freeze–thaw cycles on its permeability characteristics.

Fig. 9
figure 9

The relationship between the permeability coefficient ratio of saturated samples and dry density.

Full size image

Figure 10 presents a three-dimensional surface plot of the permeability coefficient k for saturated specimens as a function of freeze–thaw cycles (NFT) and dry density. By fitting this surface, a comprehensive expression for k(NFT, ρd) can be obtained:

$$k\left( {x,y} \right) = {194}.{8} + {3}.{72}x - {299}.{1}y - 0.0{2}x^{{2}} - {1}.{9}xy + {67}.{3}y^{{2}}$$
(3)
Fig. 10
figure 10

The relationship between the permeability coefficient of saturated samples and the dry density and number of freeze–thaw cycles.

Full size image

Here, x represents the number of freeze–thaw cycles (NFT), and y denotes the dry density ρd. This three-dimensional surface model offers an intuitive visualization of the relationship between the permeability coefficient and both the number of freeze–thaw cycles and dry density.

Conclusion and discussion

The freeze–thaw process in unsaturated soils represents a complex multiphysical coupling phenomenon, encompassing water–ice phase transitions, the migration of unfrozen water, and the redistribution of pore ice. The evolution of the soil’s pore structure exerts a significant influence on its permeability characteristics, while alterations in permeability further constrain water migration and the development of frost heave pressure. Nevertheless, the intricate physical and mechanical properties of unsaturated soils have precluded this study from exploring the evolution patterns of pore structure and permeability characteristics during freeze–thaw cycles. As a result, a quantitative relationship between the microstructure and macroscopic permeability has yet to be established. This study utilizes nuclear magnetic resonance (NMR) testing and standard variable head permeability tests (SVHPT) to examine the impact of freeze–thaw cycles on the microstructure and permeability characteristics of Jilin ball clay at varying dry densities and water contents. The key findings are as follows:

  1. 1.

    As the number of freeze–thaw cycles increases, the peak of the T2 curve for samples with saturated water content conditions exhibits a slight rightward shift and becomes more pronounced. This observation suggests a gradual increase in pore size and a trend towards uniformity. After a certain number of freeze–thaw cycles, the pore size distribution of samples with varying dry densities converges, and the increase in void ratio gradually decelerates and eventually stabilizes.

  2. 2.

    In contrast to specimens subjected to saturated conditions, those exposed to unsaturated conditions exhibited a relatively stable pore size distribution throughout the freeze–thaw cycles, with no significant systematic changes observed.. The void ratio of the specimens initially decreased, subsequently increased, and eventually stabilized, with minimal overall variation.

  3. 3.

    Under saturated conditions, freeze–thaw cycles induce a transition of soil particles from a dispersed to an aggregated state, resulting in an exponential increase in the permeability coefficient with the number of freeze–thaw cycles and the void ratio. Conversely, the permeability coefficient of samples under unsaturated conditions increases slightly and quickly stabilizes, showing no clear relationship with the void ratio.Furthermore, as dry density increases, the permeability coefficient ratio of saturated samples decreases linearly, with the slope of this relationship becoming steeper as the number of freeze–thaw cycles increases.

  4. 4.

    The three-dimensional surface model of the permeability coefficient under saturated conditions demonstrates the regulatory influence of dry density on permeability and its interaction with the number of freeze–thaw cycles. This study underscores the critical importance of meticulously controlling soil dry density and moisture content in the design of clay structures in seasonally frozen regions to ensure engineering stability and safety.