Figure 1 presents a schematic diagram of the sampling positions. Per the soil mechanics testing standard (GB/T50123-2019) (2019), the specific gravity, liquid limit, maximum dry density, and optimum moisture content of the soil were determined through experimentation and shown in Table 1.

The soil was placed in a drying box at 105°C to dry and then crushed. The dried soil sample was passed through a 2 mm sieve. It was then sprayed with distilled water until the water content reached 10%, stirred well, and sealed in a plastic bag for 48 h to allow the sample to be water-balanced.

Weigh according to the target dry density (1.5 g/cm³). Please place them in a mold with a diameter of 61.8 mm and a height of 20 mm, and samples are then made using the press sample method. After compaction, remove the molded samples. Eight groups of four samples each were prepared as remodeled soil samples with 10% water content and 1.5 g/cm³ density. After saturation by vacuum pumping method, one group was directly taken out and immersed in water for 1, 3, 5, 7, 10, 15, and 20 days.

Samples with different immersion times were placed in a shear box. The upper box was fixed, while the lower box could slide horizontally. Vertical pressures of 100, 200, 300, and 400 kPa were applied, respectively. Then, the horizontal shear force was gradually applied to the lower box of the shear box until the sample was sheared. The shear rate was set at v = 2 mm/min, the shear displacement was 6 mm, and the readings were recorded for every 0.2 mm shear displacement.

After drying and cooling the samples with different immersion times, they were ground into fine particles using a grinding rod. The prepared samples were stored in preservation bags. The samples were immersed in water for 0 and 20 days, respectively. The soil was tested using an X-ray fluorescence spectrometer.

The gradient of sample immersion time was 0, 1, 3, 5, 7, 10, 15, and 20 days. Appropriate quantities of samples were taken at each immersion time point for the tests. Perform particle size analysis on soil samples using a laser particle size analyzer. The impact of particle composition on soil shear strength was explored using the gray correlation entropy analysis algorithm.

The gradient of water immersion time was 0, 5, 10, and 20 days in order, and the samples were dried. A cubic rod of approximately 1 cm × 1 cm × 2 cm was cut from the middle of the ring knife sample after different immersion times and used as a scanning electron microscope sample. The specific operation steps of SEM testing were mentioned by Ni et al. (2020).

The shear stress-shear displacement curves of samples at different immersion times were obtained through direct shear tests, as shown in Figure 2. It can be observed that both vertical stress and immersion time have significant influences on the shear stress-shear displacement curves of the soil.

Based on the experimental results, the failure strength curve of the sample under different immersion days and the shear strength curve of the test under different vertical pressures were drawn. As shown in Figure 3, the shear strength of the sample decreases with the increase of immersion days, indicating that immersion has a softening effect on the shear strength of the soil. At 400 kPa, with the increase of immersion days, the cumulative attenuation value of the failure and shear strengths is 18.4 kPa. When the vertical stress is low, the attenuation of shear strength is also low. When the axial pressure is 100 Pa, the cumulative attenuation value is 10.7 kPa.

The relationship curve between cohesive force, internal friction angle, and immersion time is shown in Figure 4. The cohesive force gradually decreases with the increase of immersion time, while the internal friction angle also gradually decreases. After immersion for 20 days, the cohesive force of the clay decreased by 57.6%, while the internal friction angle decreased by 8.43%, with a smaller decrease range.

SiO₂, AL₂O₃, and Fe₂O₃ are the soil’s primary chemical binding materials (Liu zhikui et al., 2017; Mou Chunmei and Wei Yux, 2019), with a total content of 88.01% of the soil’s chemical composition.

As shown in Figure 5, after immersion for 20 days, the SiO₂ content increased by 1.69%, related to the hydrolysis equilibrium of SiO₂ in the aqueous solution. AL₂O₃ and Fe₂O₃ content was lower than those of the unsoaked samples. The AL₂O₃ and Fe₂O₃ in the soil react with the aqueous solution, destroying the original binding structure of the soil and further crushing the original soil structure. The Na₂O content decreased mainly due to the alternation adsorption of cations, which caused Na⁺ to leave the soil with the infiltrating liquid. CaO mainly exists in carbonate and sulfate minerals, and MgO and K₂O mainly exist in calcite and orthoclase, respectively. The decrease in their content reflects the dissolution of corresponding mineral components. All these chemical reactions are beneficial to weaken the structural strength of the soil.

Using a laser particle size analyzer, the experimental particle size distribution was tested, and the sand content (d > 0.075 mm), powder content (0.005 mm < d < 0.075 mm), and clay content (d < 0.005 mm) of the soil samples after different immersion times were statistically analyzed and plotted. As shown in Table 2, the particle size of the soil continuously became finer with the increase in immersion time. For example, after immersion for 20 days, the mass fraction of sand content (d > 0.075 mm) decreased by 9.47% due to its smaller base but significant decomposition, while the mass fraction of clay (d < 0.005 mm) increased by 5.83%, and the mass fraction of powder (0.005 mm < d < 0.075 mm) increased by 3.64%. Throughout the immersion process, the changes in soil particles mainly involved the transformation of medium and fine sand into powder, while powder gradually transformed into clay, and the mass fraction of sand gradually decreased. In contrast, the mass fraction of clay and powder gradually increased. Therefore, it can be concluded that once the soil is immersed in water, its structure is destroyed, the coarse particles disperse and decompose, and the content of medium and fine particles increases. With the increase of immersion days, the coarse particles further disperse and decompose, and the content of clay and powder gradually increases.

After different immersion periods, statistical analysis was conducted on the usual particle size (including the characteristic particle size with cumulative particle distribution of 10%, 30%, 60%, and 99%). The specific data and changes are shown in Table 3. It can be observed from Table 3 that with the increase in immersion time, the characteristic particle size with cumulative particle distribution of 10%, 30%, 60%, and 99% of the soil particles continuously decreased. The change of the characteristic particle size with a cumulative particle distribution of 10% and 30% slowly decreased with the increase of immersion days. The decrease of the characteristic particle size with a cumulative particle distribution of 60% was divided into the pre-immersion period (0–10 days) and the pre-and post-immersion period (10–20 days). The usual particle size with a cumulative particle distribution of 99% decreased from 0 to 1 day.

The correlation between the content of clay particles, fine particles, sand particles, and soil strength indicators was calculated using gray correlation entropy analysis. The grey correlation analysis method was used to determine various factors’ varying degrees of influence on the dependent variable in complex systems (Deng et al., 2023; Zhang et al., 2019). Zhang Qishan et al. (1996) analyzed the two significant shortcomings of the existing gray correlation methods in 1996 and, based on Deng Julong’s gray correlation theory, introduced the concept of gray entropy and proposed the gray correlation analysis method, called the gray correlation entropy analysis method. The specific introduction of the gray correlation entropy analysis method is as follows (Wang et al., 2014).

If a gray correlation factor set denoted as X contains a reference sequence (dependent variable) X 0 and comparison sequences (independent variables) X j j = 1,2 , ⋯ , k ; i = 1,2 , ⋯ , m , then X can be represented as: X = X 0 = x 0 1 , x 0 2 , ⋯ , x 0 i , ⋯ , x 0 m X 1 = x 1 1 , x 1 2 , ⋯ , x 1 i , ⋯ , x 1 m ⋯ X j = x j 1 , x j 2 , ⋯ , x j i , ⋯ , x j m X k = x k 1 , x k 2 , ⋯ , x k i , ⋯ , x k m

To normalize each sequence of X , there are two common methods. If the first data point in a sequence is not equal to zero, then each data point in the sequence is typically divided by the first data point. If the first data point is equal to zero, then the average of all data points in the sequence can be used as the divisor. Normalizing the sequences using these methods results in factor set X ′ . X ′ = X 0 ′ = x 0 ′ 1 , x 0 ′ 2 , ⋯ , x 0 ′ i , ⋯ x 0 ′ m X 1 ′ = x 1 ′ 1 , x 1 ′ 2 , ⋯ , x 1 ′ i , ⋯ x 1 ′ m ⋯ X j ′ = x j ′ 1 , x j ′ 2 , ⋯ , x j ′ i , ⋯ x j ′ m X k ′ = x k ′ 1 , x k ′ 2 , ⋯ , x k ′ i , ⋯ x k ′ m

Using the comparative sequence as a reference, the gray correlation coefficient of each comparative sequence is calculated using the following formula: γ x 0 i , x j i = min j min i x 0 ′ i − x j ′ i + ξ max j max i x 0 ′ i − x j ′ i x 0 ′ i − x j ′ i + ξ max j max i x 0 ′ i − x j ′ i

According to the mapping relationship between the distribution of gray correlation coefficients between the reference sequence and the comparison sequence, the mapping value P h h = 1,2 , ⋯ , m is obtained, where the coefficient ξ ∈[0,1] is the resolution coefficient and is usually set to 0.5. P h = γ x 0 i , x j i ∑ i = 1 m γ x 0 i , x j i P h also known as the density value of a distribution. For gray connotation sequence A = a 1 , a 2 , ⋯ , a r , ⋯ (each element is not less than 0, and the sum is equal to 1), its gray entropy is: H = − ∑ a r ⁡ ln a r

By substituting P h into Formula, we obtain the grey correlation entropy of each comparison sequence. H j = − ∑ P h ⁡ ln P h

Finally, the grey entropy correlation degree of each comparative sequence was obtained through the following formulaic calculation: E r X j = H j lnm

Sequence X a is more closely aligned with the reference sequence than sequence X b , and as a result, it makes a greater contribution to the alignment of sequence X 0 .

As shown in Figure 6, the results indicate the highest gray correlation degree between cohesion and sand content, followed by silt and clay particles. The gray correlation degree of the three is more significant than 0.95, all of which have vital statistical significance. The gray correlation degree between the internal friction angle and the silt content is the highest, followed by clay particles, and the sand content is the lowest. The gray correlation degree of the three is more significant than 0.95, all of which have vital statistical significance.

As depicted in Figure 7, SEM microstructure images of samples immersed for 0, 5, 10, and 20 days are presented, the magnification factor being ×2000. With the increase in immersion time, the soil particles were still composed of cohesive or aggregated flakes, forming a stacked structure. However, the original compactness of the soil weakened, and the orientation and arrangement could be more explicit. The particles within the field of view were severely fragmented, the area decreased, and the boundary between particles and pores blurred. The stacking of particles and pores increased, and there were more small voids at the edges of particle-to-particle or particle-to-edge contact, with some local inter-particle pores connected. Overall, the soil structure within the field of view became loose as the immersion time increased.

There was a general decreasing trend in pore area and an increasing trend in pore number. This indicates that as the number of days of immersion increased, the colloidal particles (soluble salts) in the soil dissolved rapidly, large pores decreased, and tiny pores increased. The overall soil structure within the field of view becomes loose as the immersion days increased.