The sandstone samples that were used in the tests were obtained from a quarry in Sichuan Province, China. The samples were fresh and unweathered, with a medium to fine-grained sandy texture. Mineralogy, particle size, and rock texture were studied in standard optical thin sections using an Axio Imager M2 polarizing microscope (Figure 1). Mineralogy is dominated by quartz (∼70%), followed by feldspar (∼25%) and muscovite (∼5%). The modal analysis results were converted into volume fractions using published values for mineral densities and were used to calculate the average grain density of the rock. Table 1 shows the physical parameters of 10 prepared specimens. The density of the oven-dried intact rock was measured to be 2,384.9 kg/m³±18.2 kg/m³. Together with the grain density (2,645 kg/m³), the average porosity of the rock is 9.84 ± 1.36%. Figure 1 shows the sample grains to be angular, with grain sizes predominantly within the range of 0.2∼0.3 mm. They showed no preferred alignment, were mostly sub-angular to sub-rounded, and were moderately well-sorted. The grains show evidence of compaction by pressure solution, with a small amount of siliceous cement between the grains. This rock may be termed an arkosic sandstone based on its composition and texture. With respect to a pure quartz sandstone, a feldspathic sandstone is expected to be more susceptible to damage from chemical interactions with ionic solutions.

Standard intact cylindrical specimens with sizes of Φ 50 mm × 100 mm length and Φ 50 mm × 25 mm length were prepared through drilling, sawing, and grinding. To ensure textural homogeneity between the prepared specimens, the longitudinal wave velocity and mass of each specimen were tested before the experiment. A UTA-2000A ultrasonic testing analyzer was used to obtain the longitudinal wave velocity of the rocks, with a sensor frequency of 35 kHz, a sampling frequency of 10 MHz, and a time accuracy of 0.1 μs. The experimental results for these specimens are shown in Table 1. Specimens with longitudinal wave velocities of 3952 m/s ± 118 m/s and masses of 465 ± 5 g were selected for each experiment. A total of 10 standard intact cylindrical specimens with dimensions Φ50 mm × 100 mm length, were used: five for unconfined compression tests and five for triaxial compression tests. A further five specimens of Φ50 mm × 25 mm length were prepared for Brazilian splitting tests.

For the direct shear tests, 85 cylindrical specimens of dimensions Φ50 mm × 100 mm length were prepared by making a rough tensile fracture normal to the cylinder axis near the midpoint of each specimen. Fracturing was done using a homemade specimen splitting tool (Cheng et al., 2022). For each specimen, a tensile fracture was made near the axial midpoint using a NYL-60 pressure testing machine, as shown in Figure 2. The splitting device consists of two identical steel plates bearing cylindrical grooves. A rectangular groove perpendicular to the specimen axis is set at the midpoint in the axial direction of each cylindrical groove and a steel bar with a triangular section was placed in the rectangular groove. When the splitting device wrapped the specimen, the concentrated linear load provided by the upper and lower steel bars produced a tension fracture in the specimen. The preparation process is shown in Figure 2. All of the prepared specimens were stored in an oven before shear testing.

A total of 75 shear tests were carried out on 15 groups of specimens, each group was treated with different dry-wet cycles in three kinds of pH soaking solutions. Five fractured specimens in each group were subjected to direct shear tests under five normal stresses (3, 6, 9, 12, and 15 MPa). To avoid the influence of different moisture contents on the shear test results while they were conducted, the specimens after dry-wet cycling treatment were oven-dried before the shear test. In addition, 10 fractured specimens were tested after oven drying treatment and taken as the reference group, with each test under given conditions being carried out twice.

The shear tests were carried out under constant normal stress (CNL) boundary conditions, and the specific shear test procedure was as follows: 1) the normal stress was applied up to the set value from zero in the load control mode; 2) the tangential load was applied in the displacement control mode with a shear rate of 1 mm/min; and 3) the shear process was controlled by the servo system to maintain a constant normal stress and the test was terminated when the shear displacement reached 8 mm. Four parameters were recorded (i.e., normal stress, normal displacement, shear stress, and shear displacement) in each shear test.

Water-rock interactions are usually a long-term process in practical rock engineering. However, it is difficult to obtain a substantial water-rock interaction effect simply by soaking in water for a long period of time. Therefore, the acidity and alkalinity of the soaking solution, and the ion concentration were enhanced, as proposed by Han et al. (2013) and Chen et al. (2019), to enhance the corrosion effect on the specimen in the short term during the dry-wet cycling. A set of NaSO₄ solutions with different pH values (i.e., 2, 7, and 12) were prepared as follows. A standard NaSO₄ 0.1 mol L⁻¹ solution was added into distilled water with pH = 7 to form a neutral solution. The pH = 2 acidic solution was prepared by titrating HCl solution into the neutral solution until the pH value was 2. The pH = 12 alkaline solution was prepared by titrating NaOH solution into the neutral solution until the desired pH was attained.

In this study, the specimen was soaked using the natural immersion method (e.g., Lin et al., 2005; Khanlari and Abdilor, 2015; Momeni et al., 2017). The specimen was completely immersed in chemical solution, with a rock to solution volume ratio of 1:2, with a sealed plastic wrap to prevent contact with the outside air. After soaking for 24 h, the pH value of the chemical solution in contact with the rock specimen was tested using a pH meter and then the specimen was removed. The free surface fluid on the specimen was removed by water absorption onto a filter paper and was then dried in an oven at 105°C for 24 h, which formed a dry-wet cycling event. For each of the three kinds of soaking solutions prepared, with pH values of 2, 7, and 12, five groups of specimens were subjected to 3, 6, 9, 12, and 15 dry-wet cycles, respectively. There were five specimens in each group, thus a total of 75 specimens in 15 groups were used to investigate the influence of the dry-wet cycle number and soaking solution pH value on the shear mechanical properties of the rock fracture.

When studying the influence of pore fluids on the failure of intact rocks, it is usual to vacuum-impregnate the specimen with the fluid to ensure that the fluid accesses all of the pores. However, in this study we are only interested in the effect of the various wetting fluids on the shear behavior of the exposed surface of the rock. Therefore, it was not considered necessary to submit each specimen to vacuum impregnation, which would have been inordinately time-consuming.

After dry-wet cycling treatment of the fractured sandstone specimens, the fracture surface morphology of each specimen was obtained using a Tianyuan OKIO-400 3D scanner (Cheng et al., 2022). This 3D scanner controls the scanning accuracy of fracture morphology using the global error control module. The average scanning resolution is 0.02∼0.03 mm in the vertical direction, the average sampling point spacing is 0.31∼0.15 mm, and the maximum scanning area is 400 mm× 300 mm. After scanning using the scan-line pattern shown in Figure 3, the 3D morphology parameters of each fracture were obtained by processing the scanning data, as illustrated in Figure 4.

The calculation process to obtain the fracture roughness coefficient (JRC) is as follows:

The root mean square of slope Z₂ of each profile line (Figure 3B) was calculated using: Z 2 = [ 1 ( l − 1 ) ( Δ x ) 2 ∑ i = 1 l − 1 ( y i + 1 + y i ) 2 ] 1 2 , (1) where y _i is the height coordinate of the fracture surface profile line at sampling point i, l is the number of data points, and Δx is the interval between data points.

Each 2D Z_2i of the profile line was substituted into Equation 2 (Yang et al., 2001) to calculate the roughness JRC _i of each fracture surface profile line: JR C i = 32.69 + 32.98 lg Z 2 i , (2) where Z _2i is the root mean square of the slope of the ith fracture surface profile line and JRC _i is the roughness coefficient of the ith fracture surface profile line.

Finally, the roughness coefficient JRC of the whole fracture surface was obtained with Eqn. 3: JRC = 1 m ∑ i = 1 m JR C i , (3) where m is the total number of fracture surface profile lines.

The accuracy of the JRC characterization of fracture roughness is related to the number of the profile lines. The average roughness coefficients JRC calculated from different numbers of profile lines from four specimens are shown in Figure 4A. It can be seen that JRC first increases rapidly and then tends to be stable with the increase of the number of intercepted profile lines m. When the number of profile lines is greater than 16, JRC varies only slightly. The relative errors of JRC determined by different number of profile lines calculated using Eqn. 4 are shown in Figure 4B: e = JR C m − JR C 36 JR C 36 × 100 % , (4) where e is the relative error, %; JRC _m is the JRC value obtained by intercepting m profile lines; and JRC ₃₆ is the JRC value obtained from 36 profile lines.

It can be seen from Figure 4B that when the number of profile lines exceeds 16, the calculation error is less than ±5%, and the accuracy is relatively high. Therefore, this study uses 16 profiles to calculate the fracture roughness coefficient JRC.

Uniaxial compression, conventional triaxial compression, and Brazilian splitting tests on sandstone were carried out using an RMT-150B rock mechanics test system. The axial displacement was measured by a displacement sensor with a measuring range of 5 mm, and the axial load was measured by a force sensor with a measuring range of 1000 kN. For the Brazilian splitting tests, a more sensitive force sensor with a measuring range of 100 kN was employed. The displacement control mode was used in tests with an axial loading rate of 0.002 mm/s and a confining pressure loading rate of 0.1 MPa/s. Based on five repeated uniaxial compression tests, triaxial compression tests, and Brazilian splitting tests, the basic mechanical parameters of the sandstone specimens were obtained, as shown in Table 2. These comprise the uniaxial compressive strength σ _c , tensile strength σ _t , cohesion c, internal friction angle φ ₀ , Poisson’s ratio μ, and elastic modulus E. In addition, the basic sliding friction angles φ _b of the rock fractures obtained by a tilt test of three standard cylinder specimens (Li et al., 2019) are also listed in Table 2.

An RDS-200 rock direct shear apparatus (Guo and Dong, 2019) produced by Geotechnical Consulting and Testing Systems (GCTS), as shown in Figure 5, was used for the direct shear tests. The direct shear apparatus provides normal and shear loads using an electrohydraulic servo control system. The maximum tangential and normal loads are 10 and 5 tonnes, respectively, and the load precision is 0.01 kN. The maximum tangential and normal strokes are 25 and 24 mm, respectively, and the displacement accuracy is 0.001 mm.

The Appendices 1– 4 show the shear stress-shear displacement curves obtained from direct shear tests on sandstone fractures under the oven-dry state and dry-wet cycling conditions.

Figure 6 illustrates the general form of a typical shear stress-shear displacement curve. The peak shear stress τ _p is the maximum shear stress on the curve; the peak shear displacement u _p is the shear displacement corresponding to the peak shear stress; the pre-peak shear stiffness k _s is the slope of the approximately straight-line part of the curve before the peak; and the residual strength τ _r is the shear stress value when it is roughly stable.

Table 3 presents the shear mechanical parameters and test conditions of sandstone fractures in the oven-dry state (tests D1∼D10) and under dry-wet cycling conditions (tests B1∼B75). The following parameters are tabulated: chemical solution pH value, number of dry-wet cycles N, applied normal stress σ _n , peak shear strength τ _p , pre-peak shear stiffness k _s , peak shear displacement u _p , residual shear strength τ _r , and fracture roughness coefficient JRC.

The form of the shear stress versus shear displacement curves is similar to the general schematic shown in Figure 6 in all cases, with a well-developed peak stress after an elastic loading stage, followed by a progressive decay of the shear bearing load towards a steady value of sliding friction (residual strength) as the asperities are fractured and the slip surface evolves. Pre-peak shear stiffness, peak shear stress, and residual shear stress all increase with normal stress. The data that are plotted in Figure 7 are drawn from the experimental data presented in the Appendices 1– 4 and in Table 3. This figure shows how the peak shear strength, pre-peak shear stiffness, peak shear displacement and residual shear strength of the rock fractures in the oven-dry state and under dry-wet cycling conditions vary under different normal stresses. In Figure 7, the parameter subscript p refers to a peak value, u indicates that no dry-wet cycling treatment was applied, w refers to measurements made after dry-wet cycling treatment, and ave indicates an average value.

From the shear mechanical parameters following dry-wet cycling treatment, we can observe:

1) Figure 7A shows that τ _puave is higher than τ _pwave under each same normal stress. The dry-wet cycling is inferred to change the mineral particle structure, weaken the connectivity between particles, and promote the generation and expansion of microcracks on the sandstone fracture surfaces, as proposed by Huang et al. (2020). This accumulated damage in sandstone is inferred to lead to the observed deterioration of the peak shear strength of fractures. Figure 7A also shows that the difference between the trend lines of τ _puave and τ _pwave . This implies that the greater the normal stress, the more evident is the deterioration effect of dry-wet cycling on the peak shear strength. This is inferred to indicate that higher normal stress enhances the development and growth of microcracks and meso-scale cracks in sandstone fractures, which result in a decrease in the shear bearing capacity of the fractures.

To study the influence of the number of dry-wet cycles on the shear mechanical properties of the rock fracture, the average values of peak shear strength, pre-peak shear stiffness, peak shear displacement, and residual shear strength after different numbers dry-wet cycles for each of three different pH soaking solutions and at five normal stress values are shown in Figure 8. Both the peak shear strength and the residual strength are reduced only very slightly at low normal stresses but this effect increases as the normal stress is increased. Pre-peak shear stiffness and peak shear displacement both decrease by a similar amount as the number of dry-wet cycles increases at all normal stresses.

Figure 9 shows how the pH value of the soaking solution varies with the number of imposed dry-wet cycles. The pH value of the acidic solution (initially pH = 2) increases to weaker acidity, the pH value of the alkaline solution (initially pH = 12) decreases to weakly alkaline, and the pH value of the neutral solution (initially pH = 7) changes to weakly alkaline after the second dry-wet cycle and they then remain roughly constant. The volume ratio of sandstone to soaking solution is 1 : 2. It is surprising that such large changes in the alkalinity and acidity occur with successive soaking cycles. This suggests that some chemical reactions take place between the solutions and the exposed surfaces of the rock specimens. The thin section results are dominated by quartz and feldspars, which are expected to display some reactivity, but the effect on the residual solution pH will depend on the relative ratio of the rock to solution volumes. Although there is no indication of significant amounts of carbonate minerals present, there could be some at trace levels.

In contrast to the influence of the number of dry-wet cycles imposed on specimens prior to shearing, and despite the indications that the solution pH is modified by chemical interactions with the rock (Figure 9), the experimental data show apparently no significant influence of variations in the acidity or alkalinity of the solutions in the strength parameters (Figure 10).

To further investigate the influence of normal stress, solution pH value, and number of dry-wet cycles imposed on the shear mechanical properties of sandstone fractures, orthogonal test analysis was carried out on the experimental results listed in Table 3. The orthogonal experimental design method is based on mathematical statistics and the orthogonality principle. This method involves selecting representative points from a large number of experimental data, and arranging and analyzing these in multifactor experiments using an orthogonal table (e.g., Li and Hao, 2018; Yang et al., 2020; Zhang et al., 2020). The influencing factors include the normal stress, pH value of the soaking solution, and dry-wet cycle number. The test index parameters include the peak shear strength, pre-peak shear stiffness, peak shear displacement, and residual shear strength. The standard orthogonal table, symbolized L25 (5⁶) with six factors and five levels was adopted after it was modified to L25 (5³) with three factors and five levels based on the test scheme in this paper. In this study, Design-Expert software was used for the orthogonal test analysis. The significance level of the factors was analyzed and evaluated using the P significance test (Yang et al., 2019). Only the experimental results were input to obtain the p value of the significant indigenous index representing the influencing factors.

In the variable explicitness test of a linear econometric model, we can judge by comparing the calculated t statistics with the critical value or by comparing P with the explicitness level α. Assuming that we calculate the t test value of the parameter estimation of an explanatory variable in an econometric model as t ₀:

1) When t 0 > 0 , P = 2 ∗ [ − Prob ( − ∞ < t < t 0 ) ] = 2 ∗ [ 1 − ∫ − ∞ t 0 f ( t ) d t ] , where f ( t ) is the probability density function of Student’s - distribution;

Therefore, in the judgment of the variable dominance test, P should be directly compared with α, rather than with α/2. The judgment criterion is as follows:

1) When P   > α , the original hypothesis ( H 0 : β i = 0 ) should be accepted and the alternative hypothesis ( H 1 : β i ≠ 0 ) should be rejected. Therefore, this means that the variable does not pass the explicitness test and this variable is not a significant influencing factor on the variable under consideration.

Table 4 shows the p values of the normal stress, pH value of the solution, dry-wet cycle number, and combined factors on the peak shear strength, pre-peak shear stiffness, peak shear displacement, and residual shear strength.

The variance analysis method can be used to further estimate the size of the error and to accurately estimate the contribution of each factor in the experimental results. As mentioned earlier, the size of p value reflects the significance of the influence factors on the target indicators. The greater the p value, the smaller the sensitivity of the influence factors on the target indicators. When the p value is less than 0.05, it shows that the influence factors can be determined under the condition of confidence 0.95. It can be seen from Table 4 that the influence of normal stress on peak shear strength, peak shear displacement, pre-peak shear stiffness, and residual strength is relatively significant. The influence of dry-wet cycle number on peak shear strength, pre-peak shear stiffness, and residual strength is relatively significant. However, the influence of solution pH on the four shear mechanical parameters is relatively small. In addition, the interaction between the various factors has no obvious effect on the shear mechanical parameters.

The experimental results that are presented in this paper show that wet-dry cycling degrades the shear mechanical properties of sandstone fractures. There is no single mechanism that explains the influence of wet-dry cycling on the shear mechanical behavior of rock fractures. Van Eeckhout (1976) summarized five mechanisms based on a large number of compression tests, as follows: 1) fracture energy reduction; 2) capillary tension decrease; 3) pore pressure increase (through undrained compaction); 4) friction coefficient reduction; and 5) chemical and corrosive deterioration of mineral structures. These processes, in various combinations, have been invoked to explain weakening effects observed through fluid/rock interactions by many authors (e.g., Colback and Wiid 1965; Burshtein 1969; Rutter 1972; Hadizadeh and Law 1991; Baud et al., 2000; Pellet et al., 2013; Tang et al., 2019). Most studies have been of the weakening role of the presence of water in axisymmetric compression tests on intact rocks. Some of these weakening processes are also likely to apply to shearing of fracture surfaces, which typically involves abrasion and mechanical wear because of fracturing and comminution of contacting asperities, and the progressive generation of wear products (i.e., fault gouge).

Tang et al. (2021) suggested that particular attention should be paid to the influence of water-rock interaction on rock fracture morphology, and summarized three mechanisms for the reduction of rock fracture peak shear strength, as follows: 1) water-induced degradation in rock strength, 2) water-induced changes in the frictional characteristics of the fracture surface, and 3) water-induced change in geometrical features through (for example) solution. Zhao et al. (2017) considered that the reduction of the shear strength of rock fractures is due to the reduction in both fracture energy and the sliding friction coefficient between mineral particles. The presence of water can reduce the strength and fracture toughness of rocks through chemical interactions between the rock and the pore fluid through various mechanisms of stress corrosion at crack tips. Detailed microstructural analysis has shown that the shapes of mineral particles in sandstone become smoother as the number of dry-wet cycles increases, which weakens the cemented contacts between the grains (Zhang et al. (2014). Tang et al. (2019) further indicated that the deterioration of peak shear strength was also related to the decrease of basic friction angle of fracture, the mechanism of which may arise from induced capillary forces, in addition to asperity weakening through the chemical and/or physio-chemical effects (Gutierrez et al., 2000).