Multi-Scale Mechanisms for Permeability Evolution in Remolded Fault Gouge: From Mineral-Particle Migration to Pore Structure

Abstract

Permeability evolution in remolded fault gouge creates critical uncertainties in geotechnical parameterization for dam foundations. However, the underlying multi-scale mechanisms, including mineral migration and pore structure changes, remain insufficiently understood. This study investigates these mechanisms using remolded plastic-thrust fault gouge from the Yulong Kashi hydropower project in China. We developed an innovative sample preparation method that combines in situ mineral self-cementation and directional compaction. The study integrated multidisciplinary tests including field in situ permeability tests; seepage–stress coupling tests; and micro-scale NMR/XRD/SEM-EDS analyses. Results demonstrate that remolded samples exhibit 1–2 orders of magnitude lower permeability (10⁻⁷ cm/s) than in situ samples (10⁻⁵ cm/s). This significant reduction is primarily caused by the loss of cementing agents and the uniform compaction of remolded samples, which leads to degraded pore connectivity. SEM-EDS analysis highlighted the leaching of cementing materials (such as K⁺, Ca²⁺ ions), while XRD revealed changes in mineral composition, with chlorite dissolution being the primary mineral alteration associated with permeability decay. Additionally, artificially enhanced cohesion distorted the mechanical behavior of the samples. These findings provide an explanation for why conventional laboratory tests tend to underestimate in situ permeability and overestimate shear strength in fault zones. This study establishes microstructure-informed correction frameworks for hydraulic and mechanical parameters in fault-crossing hydraulic engineering applications

1. Introduction

Compressional fault zones are typical products of tectonic deformation. Rock masses within these zones undergo prolonged intense compression, grinding, and cementation, forming geological bodies characterized by high fragmentation, weak cementation, and structural heterogeneity [1]. The mechanical behavior and stability of these rock masses are critical to the safety of engineering projects, including dam foundations, slopes, and tunnels [2,3]. Therefore, accurately determining the in situ physical and mechanical parameters of these materials is essential for geotechnical investigation [4]. However, obtaining in situ samples from fault zones presents significant technical challenges. This difficulty arises from the fragmented and foliated nature of fault zone materials and the presence of pervasive structural discontinuities [5,6]. Conventional scoring techniques inevitably induce mechanical disturbance, stress release, and structural disruption, leading to partial loss of in situ structural characteristics even in nominally in situ samples [7,8,9]. Therefore, accurately identifying differences in mechanical properties and permeability between in situ and remolded samples from compressional fault zones constitutes a critical research priority. Such investigations are essential for enhancing the engineering representativeness and predictive accuracy of laboratory test results.

Research on permeability differences between in situ and remolded samples has yielded significant findings [10]. Zhai et al. [11] measured permeability coefficients of 0.8–8 × 10⁻⁵ cm/s in remolded fault zone samples. They inferred that the fault zone permeability coefficient is below 4.5 × 10⁻⁵ cm/s. Although this value exceeds borehole pressure test results, it provides a useful reference for the permeability in fault zone structures. Xue et al. [12] found that field-measured permeability in the Wenchuan earthquake fault zone (1.372 × 10⁻⁶ cm/s) exceeded laboratory measurements of active fault samples (1.0 × 10⁻¹⁰–10⁻⁹ cm/s) by several orders of magnitude. This field value also exceeded the average from adjacent borehole pressure tests (1.862 × 10⁻⁷ cm/s). The authors attributed this discrepancy to scale effects in fault zones and the influence of rock mass damage on field-scale permeability behavior. Matsumoto et al. [13] confirmed this phenomenon through field pumping and pressurization tests, obtaining permeability coefficients (0.52–4.8 × 10⁻⁷ cm/s) two to three orders of magnitude higher than laboratory measurements (2.6–6.8 × 10⁻¹⁰ cm/s). Moreover, Zeng et al. and Sun et al. [14,15] demonstrated through comparative experiments that in situ samples consistently exhibit higher permeability than remolded samples, attributing this difference to preserved natural structures and weak cementation. However, the differences between laboratory and in situ permeability is attributed not only to testing methods and sample size but also to factors such as sample disturbance, anisotropy, and pre-existing structural conditions [16].

To simulate the weak cementation commonly observed in fault zones, researchers have incorporated artificial cementing agents into remolded samples [17]. Yu et al. [18] prepared remolded samples using crushed host rock as aggregate with cement and gypsum binders to simulate bonding strength. Subsequently, Ma et al. [19] employed a similar methodology to prepare remolded samples, verifying the dominant influence of cementitious strength on rock creep-erosion behavior. Yu et al. [20] further demonstrated that the addition of appropriate salt concentrations can effectively reduce the cementitious strength. Yin et al. [21] attempted to reveal the importance of weakly cemented clay in permeation behavior. However, these artificial binders (e.g., cement and gypsum) exhibit inherently strong binding properties through hydration and crystallization processes. These properties fundamentally differ from those of natural weak binders (e.g., silica or calcite precipitates) formed through prolonged geochemical processes. Table 1 compares these methods and emphasizes the strengths of our technique in terms of structural fidelity, chemical authenticity, and representativeness for hydraulic and mechanical behavior.

To eliminate the influence of artificial consolidation and accurately reflect the mechanical and permeability characteristics of fault fracture zones, researchers [22] employed loose-grain compaction methods to simulate in situ structures. This method involves drying, crushing, and sieving field-collected fault breccia. The material is then proportioned to achieve target dry density [23,24] following specific grain size distributions (e.g., using the Talbot gradation index n [25] or replicating the natural gradation [26,27]). Samples are prepared through layered compaction. Although this method enables sample homogeneity, it disrupts the natural fabric, grain interlocking, and cementation developed during the geological evolution of in situ samples. Consequently, remolded samples exhibit mechanical properties that differ significantly from those of in situ samples. Zhao et al. [28] systematically compared in situ and remolded samples, examining structural characteristics, strength degradation, and particle breakage. They demonstrated that cementation of in situ samples controls initial strength and post-failure strength degradation mechanisms. Beyond mechanical properties, these sample types exhibit distinct permeability differences. Zeng et al. [14] showed that in situ samples exhibit higher permeability than remolded samples, reflecting the sensitivity of structural features to stress history and loading paths. Sun et al. [15] further found that even at identical void ratios, in situ samples maintain significantly higher permeability than remolded samples. They attributed this to fundamental differences in fabric and cementation between sample types. These studies demonstrate that remolded samples prepared by compaction without artificial binders still exhibit significant permeability differences from in situ samples [29]. This discrepancy arises because laboratory compaction produces more uniform structures than natural geological processes, resulting in different microstructural characteristics and mechanical properties.

Table 1. List of remold sample preparation methods in fault zone.

Methods	Traditional Compaction Method	Binder Simulation Method	Our Method
Definition	External mechanical energy to simulate in situ compaction	External binders to simulate the strength of the rock mass	Mineral self-cementation combined with directional compaction
Advantages	Easy operation	High bonding strength	Restoration of stress and bonding state
Shortcomings	a. Anisotropic irreversibility b. Absence of geological processes c. Mechanical crushing alters particle size distribution	a. Differences in cementation b. Bonding distribution non-uniformity	a. Complex sample preparation process
References	Ma [22,24]; Gui [23]; Huang [25]	Yu [18,20]; Ma [19]	/

Table 1. List of remold sample preparation methods in fault zone.

Methods	Traditional Compaction Method	Binder Simulation Method	Our Method
Definition	External mechanical energy to simulate in situ compaction	External binders to simulate the strength of the rock mass	Mineral self-cementation combined with directional compaction
Advantages	Easy operation	High bonding strength	Restoration of stress and bonding state
Shortcomings	a. Anisotropic irreversibility b. Absence of geological processes c. Mechanical crushing alters particle size distribution	a. Differences in cementation b. Bonding distribution non-uniformity	a. Complex sample preparation process
References	Ma [22,24]; Gui [23]; Huang [25]	Yu [18,20]; Ma [19]	/

Therefore, this study proposes a sample preparation method based on mineral self-cementation combined with directional compaction to quantify differences in pore structure and permeability between in situ and remolded samples. First, remolded samples are prepared and characterized through systematic measurements of physical parameters and mechanical properties. Second, multi-scale permeability tests are conducted: field hydraulic pressure tests, in situ permeability tests, and laboratory-scale permeability tests. Finally, microscopic analyses using nuclear magnetic resonance (NMR) and scanning electron microscopy (SEM) elucidate the mechanisms underlying permeability differences. These analyses examine pore structure, fabric morphology, and cementation characteristics to clarify how cementation type and particle arrangement control mechanical and permeation behavior. This systematic investigation of in situ versus remolded fault zone materials provides a scientific basis for assessing the engineering applicability of laboratory test results.

2. Materials and Methods

2.1. Sample Geology

Figure 1 shows the largest secondary fault (F2) in the water diversion and power generation tunnel. The study site (Figure 1a) is located in the upper section of the right-bank water-diversion tunnel at the Yulongkashi Water Conservancy Hub, Hotan Prefecture, Xinjiang, China. This facility is a major water conservancy project in the Tarim River Basin, providing integrated hydropower generation, irrigation, and flood control. During tunnel excavation (Figure 1b), multiple secondary faults were encountered, among which an F2 compressional fault is the most prominent. The fault intersects the tunnel axis at steep dip angles of 80–90°. The fault zone averages 20–50 cm in width and comprises densely compacted breccia and mylonitic rock, exhibiting typical compressional deformation features. This fault zone significantly affects surrounding rock stability and tunnel support design. Figure 1c shows an in situ sample from an F2 fault zone collected at the tunnel face. To preserve in situ structure and mechanical properties, samples were immediately sealed after collection to minimize disturbance from transportation and environmental variations. Following drying, sieving, and grain size analysis, samples were categorized and stored for subsequent physical, mechanical, and microstructural testing. Figure 1d shows field testing conducted in the F2 fault zone to directly measure the mechanical and hydraulic properties of the fault rock mass under in situ conditions. These results are critical for validating laboratory data, evaluating fault zone engineering properties, and guiding tunnel support design.

2.2. Test Equipment

To systematically investigate differences in physical, mechanical, and structural properties between remolded and in situ samples, comprehensive tests combining macroscopic mechanical characterization with microstructural analysis were conducted on both sample types (

Table 2. List of test methods and equipment.

Sample Type	Test Methods	Measuring Indicators	Model
In situ sample	Water pressure test	Permeability coefficient	/
	In situ permeability test	Permeability parameters	Self-developed instruments
	XRD diffractogram analysis	Mineral composition, Content	TDM-20 desktop
	SEM-EDS	Spatial distribution of elements	Mira3 LMH
	NMR	Pore size distribution, Porosity	MesoMR12-040H-I
Remolded sample	NMR	Pore size distribution, Porosity	MesoMR12-040H-I
	Shear test	Cohesion, Internal friction angle	YSD-10a
	Uniaxial compression test	Uniaxial compressive strength	YSD-10a
	Triaxial test	Modulus of elasticity, Poisson	WDW-100
	Indoor seepage test	Permeability parameters	/
	Seepage–stress coupling test	Permeability parameters	YSYH-3

). First, mechanical property tests were performed on remolded samples, including direct shear, uniaxial compression, triaxial compression, and permeability tests. Direct shear and uniaxial compression tests were conducted using a direct shear apparatus (YSD-10a, Mete Test Instrument Co., Ltd., Tianjin, China), while triaxial compression tests were performed using a universal testing machine (WDW-100, Beijing Zhongjian Road Industry Instrument Equipment Co., Ltd., Beijing, China). Permeability tests were conducted using both a modified permeameter and a coupled seepage–stress testing system (YSYH-3, Jiangsu Yongchang Educational Instrument Manufacturing Co., Ltd., Changzhou, China). Second, to characterize differences in pore structure, nuclear magnetic resonance (NMR, MesoMR12-040H-I, Suzhou Nuomai Analytical Instruments Co., Ltd., Suzhou, China) analysis was performed on remolded and in situ samples. Key parameters, including pore size distribution, connectivity, and porosity, were determined to enable quantitative analysis of pore-scale structural controls on physical properties. Finally, microscopic analyses of in situ samples included (1) X-ray diffraction (XRD, TDM-20, Dandong Tongda Technology Co., Ltd., Dandong, China) to identify mineral composition and relative abundances and (2) scanning electron microscopy with energy-dispersive spectroscopy (SEM-EDS, Mira3 LMH, Guangzhou Gaocai Instrument Co., Ltd., Guangzhou, China) to characterize microstructural morphology and mineral spatial distribution. This integrated multi-scale experimental approach systematically reveals fundamental differences between in situ and remolded samples. These data provide a robust foundation for understanding fault zone mechanical behavior and developing constitutive models.

The experimental program comprised five components (Figure 2):

(1)

Remolded sample preparation: Based on existing literature and site geological data, and considering the weak cementation of fault materials, a mineral self-cementation combined with directional compaction method was employed. This approach simulates diagenetic conditions to reproduce natural structural and cementation characteristics, ensuring geological representativeness of the remolded samples.

(2)

Permeability testing: Comparative permeability tests between in situ and remolded samples were conducted to systematically analyze differences in permeability coefficient and critical hydraulic gradient. These tests revealed how structural differences control hydraulic conductivity through macroscopic flow behavior.

(3)

Mechanical property testing: Uniaxial compression and direct shear tests were conducted on both remolded and in situ samples. These tests yielded strength parameters (cohesion and internal friction angle) and deformation parameters (elastic modulus) for comparative analysis of mechanical behavior.

(4)

Microstructural characterization: X-ray diffraction (XRD) and scanning electron microscopy with energy-dispersive spectroscopy (SEM-EDS) were employed to identify mineral composition, cementation characteristics, and structural features through elemental distribution and micromorphology analysis. These analyses established correlations between microstructure and macroscopic mechanical response, providing mechanistic insights into fault zone behavior.

(5)

Pore structure characterization: Nuclear magnetic resonance (NMR) was employed to quantitatively characterize pore distribution, pore size distribution, and connectivity in both in situ and remolded samples. This analysis revealed pore-scale structural mechanisms controlling differences in hydraulic and mechanical properties.

2.3. Remolded Samples

To address challenges including particle breakage (>15% mass loss in size fractions) during manual compaction and deviation from in situ grain size distribution during vibratory compaction, a two-stage method combining manual pre-compaction with mechanical precision compaction was employed. This method ensures sample structure and geometry closely replicate those of in situ samples (Figure 3). The preparation procedure is as follows:

(1)

Sample preparation: To meet testing apparatus requirements, remolded samples with dimensions of ø100 mm × h200 mm were prepared. Following standards GB/T 50123-2019 and ASTM D7012-14 [30,31], the maximum particle size should not exceed one-fifth of the sample diameter. Therefore, the maximum particle size used was 20 mm. The total mass for each grain size fraction was calculated based on the in situ grain size distribution and target dry density. The material was divided into five layers, and the required mass m_i for each grain size fraction within each layer was determined.

(2)

Water addition, mixing, and manual pre-compaction: The inner mold walls were lined with 0.2 mm PVC film to reduce boundary friction. Material was filled in layers to controlled heights, lightly compacted manually with roughened interfaces to enhance interlocking, ensuring uniform distribution in each 45 mm layer.

(3)

Mechanical compaction: The pre-compacted sample (height ≈ 225 mm) was placed on the compaction platform. Displacement-controlled compaction was then applied gradually. Compaction was ceased when the mold contacted the fixed height stop block (200 mm).

(4)

Demolding and air-drying: Compacted samples were allowed to rest in situ for 5–10 h to minimize elastic rebound. Samples were then removed and placed on an automated air-drying platform. Sample mass was recorded daily until mass stabilization was achieved.

(5)

Self-cementation conditions: During the sample preparation, only distilled water was used to mix and stir the materials, and no external cementing agents were added. After mixing, the samples were compacted into shape and then subjected to a 7-day controlled self-cementation process. This process was conducted in a constant-temperature environment (25 ± 2 °C), and the relative humidity was gradually reduced from 95% to 50% through a programmed procedure, simulating natural diagenetic conditions.

3. Experimental Results

3.1. Mechanical Properties of In Situ Samples

Before conducting research on physical and mechanical properties, relevant parameter values must be determined in situ at the site. For example, the values for cohesion and friction angle of the in situ material were determined through field shear tests conducted in accordance with the GB/T 50266-2013 Standard for Rock Testing in Engineering Geology [32]. The procedure is as follows:

Sample Preparation: In situ shear test specimens were carefully prepared from fault gouge exposures in a trench or adit (50 cm × 50 cm × 30 cm). To preserve the original structure, the specimens were carefully exposed and separated to ensure that the test specimen remained within the predetermined shear plane.

Shear Testing: After the specimen was prepared, a normal stress (σ_n) was applied and stabilized. A hydraulic jack was used to apply horizontal shear force during a slow and steady loading process until complete failure of the specimen. Displacements in both the normal and shear directions were monitored using digital micrometers.

Repetition and Analysis: Multiple tests were conducted under different normal stresses (σ_n) to obtain peak shear strength (τ_f) values. The data was then plotted on a τ_f − σ_n curve (Mohr–Coulomb failure envelope), where the intercept represents cohesion (c) and the slope gives the tangent of the internal friction angle (φ). Regression analysis was performed to derive the final values for c and φ.

3.2. Mechanical Properties of Remolded Samples

To evaluate the reliability of remolded samples, we determined their physical and mechanical parameters and compared them with corresponding values from in situ samples to verify differences in weak cementation characteristics. The test program included the following: (1) dry density (ρ) measured by the wax sealing method; (2) cohesion (c) and internal friction angle (φ) measured by direct shear testing; (3) elastic modulus (E) and Poisson’s ratio (υ) obtained from triaxial compression tests; and (4) permeability coefficient (k) measured using a constant-head permeameter and seepage–stress coupling system. Remolded samples should closely approximate the physical and mechanical behavior of in situ samples. Deviations of key parameters from in situ reference values (

Table 3. Physical and mechanical testing results for remolded samples.

Number	1	2	3	4	5	Average
Cohesion c (MPa)	0.42	0.45	0.43	0.47	0.49	0.45
Internal friction angle	34.75	32.48	31.73	32.96	30.52	32.49
Uniaxial compressive strength σc (MPa)	5.46	4.67	4.78	5.53	4.91	5.07
Modulus of elasticity E (GPa)	0.036	0.031	0.027	/	/	0.031
Poisson ratio υ	0.31	0.34	0.33	/	/	0.33

) were controlled within 10% to provide reliable material for subsequent seepage–stress coupling tests.

Following DL/T5368-2007 (Specification for Rock Testing in Hydraulic and Hydroelectric Engineering) [33], five standard cylindrical rock samples (ø100 mm × h200 mm) were tested in direct shear to ensure representative results. Total applied load and normal load were measured in real-time to calculate shear stress and internal friction angle (Figure 4). Table 2 summarizes the physical and mechanical test results for the remolded samples. Direct shear tests (n = 5) showed that cohesion ranged from 0.42 to 0.49 MPa and internal friction angle ranged from 30.52° to 34.75°. Average values were 0.45 MPa (cohesion) and 32.49° (internal friction angle). Uniaxial compressive strength tests (n = 5) yielded values ranging from 4.67 to 5.53 MPa, with an average of 5.07 MPa and low variability. Triaxial compression tests (n = 3) yielded elastic moduli ranging from 0.027 to 0.036 GPa and Poisson’s ratios ranging from 0.31 to 0.34. Average values were 0.031 GPa (elastic modulus) and 0.33 (Poisson’s ratio).

Results indicate that remolded samples exhibit stable shear strength parameters and consistent uniaxial compressive strength, demonstrating uniform preparation techniques and excellent representativeness. The remolded fault breccia material falls within the range of extremely soft rock to extremely hard soil, exhibiting low strength. The low elastic modulus and moderate Poisson’s ratio reflect the material’s elastic–plastic deformation behavior under stress. Clay minerals and uniform pore filling by fine particles during reconstitution promote structural reorganization. This enhances inter-particle bonding and structural compactness, improving mechanical parameter stability. These results validate the reliability and reproducibility of the reconstitution method, providing a foundation for comparing mechanical properties with in situ samples and for constitutive modeling.

3.3. Evolution of In Situ Permeability Characteristics

3.3.1. In Situ Water Pressure Tests

To investigate permeability characteristics of F2 fault rock mass, three boreholes were drilled in the exposed F2 fault breccia zone within the right-bank diversion chamber for in situ pressure water tests. Borehole locations were selected to account for spatial distribution of the fault zone and rock mass fragmentation, obtaining representative permeability parameters across the area. Pressure water tests followed standard procedures in the Code for Pressure Water Tests in Drilled Holes for Hydraulic and Hydroelectric Engineering (SL-2003) [34] using a three-stage pressure approach. Test data are presented in

Table 4. Results of water pressure test for F2 fault.

Number	Drill 1	Drill 2	Drill 3
Length (m)	2.70	3.00	3.00
Seepage (L/min)	11.06	24.03	11.64
Pressure (MPa)	1.03	1.06	1.09
P-Q curves	A (Laminar type)	B (Turbulent type) or E (Filling-type)	A (Laminar type)
Permeability (Lu)	3.98	7.67	3.58
Permeability coefficient (cm/s)	3.79 × 10−5	7.51 × 10−5	3.50 × 10−5

. These investigations provide evidence for evaluating seepage stability of F2 fault zone and designing seepage control measures.

Test results (Figure 5, Table 4) show that F2 fault zone permeability ranges from 3.58 to 7.67 Lu, with permeability coefficients of 3.50 × 10⁻⁵ to 7.51 × 10⁻⁵ cm/s, classifying it as weakly permeable. With only three sets of pressure-rise data available, water-injection tests could only generate pressure-rise curves, not complete pressure-fall curves. Consequently, accurately determining P-Q curve type is challenging. Specifically, pressure-rise curves for Boreholes 1 and 3 exhibit near-linear behavior consistent with Type A (laminar flow), whereas Borehole 2 exhibits a concave-up profile (convex toward the flow rate axis), indicating Type B (turbulent flow) or Type E (filling-type) characteristics. The temporal stability of curve profiles across all three test sections (no looping or displacement observed) indicates that rock fracture structure remained stable during pressure water testing.

3.3.2. In Situ Permeation Test

Figure 6 illustrates the layout for the in situ permeation test. A test block incorporating the entire F2 fault was prepared with dimensions of 120 cm (length) × 70 cm (width) × 60 cm (height). The F2 fault exhibits a width of approximately 20 cm, a dip angle of 75–80°, and an exposed elevation of approximately 2070 m. This fault is located 105 m below the reservoir’s normal storage level of 2175 m. Consequently, sample width was determined based on the fault breccia zone width and fault plane dip angle to ensure that samples accurately reflected in situ geological conditions. Experimental apparatus (nitrogen cylinders, pressure control systems, and water storage tanks) were then installed and calibrated to verify operational readiness. After calibration, the nitrogen cylinder valve was opened. Water pressure was regulated through the pressure control system to saturate the sample. Once continuous, stable flow emerged from the outlet, indicating full saturation, the in situ permeability test commenced. Pressure-measuring tubes were installed at 30 cm intervals to monitor localized pore water pressure in real-time. This experiment simulated fault permeability characteristics in real geological environments, providing data for engineering design and safety assessments.

Figure 7 presents the results of the in situ permeation test. Figure 7a presents the relationship between lg v and lg i during in situ permeation deformation failure. Seepage response was divided into three stages. The initial stage exhibited linear Darcy flow with a permeability coefficient of 5.74 × 10⁻⁵ cm/s. As hydraulic gradient increased to a critical state, permeability coefficient surged to 8.94 × 10⁻⁵ cm/s (approximately 55.7% increase), corresponding to a critical hydraulic gradient (i_cr) of 15.0. This change indicated particle migration or microstructural reorganization, leading to enlarged flow pathways. As hydraulic gradient increased further, permeability coefficient rose to 1.55 × 10⁻⁴ cm/s, exceeding 170% of the initial value. A flow surge occurred, indicating permeation failure. The corresponding failure hydraulic gradient (i_f) was 40.0.

When average hydraulic gradient reached 15.0 and 40.0, local hydraulic gradient along the flow path exhibited distinct inflection points with rapid increases or decreases (Figure 7b). This reflected non-uniform seepage pathway changes and structural instability, corroborating the phased development and failure mechanism of permeation deformation.

3.4. Evolution of Remolded Sample’s Permeability Characteristics

3.4.1. Laboratory Permeation Tests

Figure 8 presents the lg v versus lg i relationship for permeation deformation failure in remolded F2 fault zone samples under stepwise hydraulic gradient loading. Based on curve morphology, the permeation deformation process was divided into three phases: (1) linear Darcy flow, where permeability coefficient remained stable and sample structure remained intact; (2) critical flow, where loss of cementing material and fine particle migration altered pore structure upon reaching critical hydraulic gradient (i_cr); and (3) failure, where continued particle loss formed stable flow channels, transitioning to turbulent flow and causing structural failure at failure hydraulic gradient (i_f). Furthermore, due to superior structural stability, compressive dense faults did not experience overall instability from fine particle loss (unlike loose soils). Permeability coefficient increased from 5.58 × 10⁻⁶ cm/s to 3.34 × 10⁻⁴ cm/s (approximately 60-fold increase), revealing channel enlargement and pore structure evolution during seepage. Measured critical gradient (i_cr = 20.82) and failure gradient (i_f = 78.88) confirmed that fault material possessed high resistance to permeation deformation.

3.4.2. Seepage–Stress Coupling Test

Figure 9 shows the seepage–stress coupling system developed for this study. The apparatus consists of a water supply system, pressure control system, seepage chamber, confining pressure loading system, seepage monitoring system, particle collection system, and computer servo system. The pressure control system includes two independent control units for simultaneous regulation of upstream and downstream pressures. The pressure chamber is positioned horizontally to allow real-time changes in seepage direction. Test samples (ø100 mm × h200 mm) are installed and secured by vertically orienting the pressure chamber.

$$k={\displaystyle \frac{v}{i}}={\displaystyle \frac{Q}{Ai}}={\displaystyle \frac{\gamma QL}{∆P}}$$

(1)

Based on experimental data, the relationship between remolded permeability coefficient (k) and pore water pressure (p) for the F2 fault zone was plotted (Figure 10). Results were consistent with patterns observed in classic experiments by Louis et al. [35]. Permeability coefficient exhibited exponential growth as effective stress decreased, as shown in Equation (2):

$$\begin{array}{c}k {= k}_0·exp\left(-\alpha ·{\sigma }_e\right) = k_0·exp\left[-\alpha ·\left({\sigma }_0 - p\right)\right]\end{array}$$

(2)

Fitted parameters were k₀ = 3.64 × 10⁻⁵ cm/s, α = 1.52, and R² = 0.997. According to the fitted equation, when effective stress σₑ = 0 (i.e., pore water pressure equals confining pressure), permeability coefficient k₀ = 3.64 × 10⁻⁵ cm/s. According to GB 50487-2008 (Code for Geological Investigation of Hydraulic and Hydroelectric Engineering) [36], this permeability coefficient falls within weakly permeable strata. To prevent boundary leakage, confining pressure was maintained at 0.2 MPa above pore water pressure throughout the test, ensuring the latex membrane remained tightly adhered to the sample. At the final loading step (σₑ = 0.47 MPa), the permeability coefficient was 1.75 × 10⁻⁵ cm/s, representing approximately 52% decrease from k₀ at σₑ = 0. Both values were within the 10⁻⁵ cm/s order of magnitude. These results validate the fitted model effectiveness and demonstrate high precision and reliability of the testing system in the low-stress range.

4. Discussions

4.1. Permeability Differences Between In Situ and Remolded Samples

Table 5. Comparison of permeation test results.

Test Type	Sample State	No.	Permeability Coefficient k (cm/s)	Critical Hydraulic Gradient icr	Failure Hydraulic Gradient if	Effective Stress Sensitivity Coefficient α
In-situ water pressure test	In situ	1	3.79 × 10−5	/	/	/
		2	7.51 × 10−5	/	/	/
		3	3.50 × 10−5	/	/	/
In situ testing		/	5.74 × 10−5	15.0	40.0	/
Indoor routine test	Remolded	/	5.58 × 10−6	20.82	78.88	/
Seepage–stress coupling test		/	1.03 × 10−7			1.52

summarizes the permeability test results for remolded and in situ samples. Tests indicated that the permeability coefficient measured in the in situ test was 5.74 × 10⁻⁵ cm/s, demonstrating strong permeability of the natural fault fracture zone. Permeability coefficients from water-injection tests (Drillholes 1–3) ranged from 3.50 to 7.51 × 10⁻⁵ cm/s, within the same order of magnitude and in good agreement with in situ test results. In contrast, permeability coefficients from remolded samples in laboratory tests showed significant reduction: conventional tests yielded 5.58 × 10⁻⁶ cm/s, while seepage–stress coupling tests yielded 1.03 × 10⁻⁷ cm/s. These values were 1–2 orders of magnitude lower than field values, reflecting significant disruption to microfractures and pore structures during sample reconstitution. Primary reasons for these discrepancies include the following: (1) Scale effect: in situ tests reflect overall seepage behavior encompassing macroscopic fracture networks, whereas laboratory samples represent only localized microfracture systems, leading to permeability underestimation; (2) Fracture closure: primary fractures are prone to closure under compression during laboratory sample preparation and stress loading. Remolded samples, subjected to intense disturbance, exhibited the lowest permeability coefficient (k ≈ 10⁻⁷ cm/s), further confirming the controlling influence of microstructural damage on seepage capacity.

Comparisons between critical hydraulic gradient (i_cr) and failure hydraulic gradient (i_f) revealed structural discrepancies: in situ tests yielded i_cr = 15.0 and i_f = 40.0, whereas conventional laboratory tests yielded 20.82 and 78.88, respectively. These elevated values indicated pronounced suppression of particle migration in small-scale samples, underscoring the influence of testing conditions on evaluating seepage behavior.

4.2. Mechanical Property Alterations

The sample preparation method for remolded samples proposed in Section 3.2 effectively restored macro- and microstructural characteristics of in situ fault zone samples. Comparison of physical and mechanical parameters between remolded and in situ samples (

Table 6. Comparison of physical and mechanical properties between in situ and remolded samples.

Parameters	Unit	F2 Fault		Experiment
		In Situ	Reshape
Dry density (ρ)	g/cm3	2.25	2.25	Wax sealing method
Relative density (Dr)	/	/	91.2%	Vibratory method
Porosity (n)	/			NMR method
Modulus of elasticity (E)	GPa	0.03	0.031	Triaxial compression method
Poisson ratio (υ)	/	0.35	0.33
Uniaxial compressive strength (σc)	MPa	5.0	5.07	Uniaxial compression method
Cohesion (c)	MPa	0.07	0.45	Straight shear method
Internal friction angle (φ)	°	20.55	31.64

) revealed that, apart from slight variations in cohesion (c) and internal friction angle (φ), all other parameters—including dry density (ρ), porosity (n) and pore size distribution, elastic modulus (E), Poisson’s ratio (υ), and unconfined compressive strength (σ_c)—highly correlated with in situ samples. This indicated that remolded samples exhibited representativeness in both physical state and mechanical behavior.

Cohesion and internal friction angle of the in situ F2 fault zone samples measured 0.07 MPa and 20.55°, respectively, whereas remolded samples measured 0.45 MPa and 31.64°, representing increases of 5.43-fold and 54.0%, respectively. This discrepancy primarily stemmed from differences in structural characteristics and stress history between in situ and remolded samples: (1) Structural fragility of in situ samples: In situ samples retained naturally formed structural planes, micro-fractures, and in situ cementation state. These defects resulted in low shear strength, with cohesion particularly susceptible to disruption by primary structural damage. (2) Particle rearrangement and increased density during reconstitution: Reconstitution disrupted the in situ weak structure, causing particles to rearrange into a more uniform and compact configuration. This enhanced mechanical interlocking and friction between particles, raising internal friction angle. (3) Loss and redistribution of cementation: Natural cementing material in in situ samples was disrupted during reconstitution. While this temporarily weakened cohesion, subsequent uniform distribution and tighter intergranular contact facilitated formation of new shear-resistant structures, substantially enhancing cohesion.

Reconstitution significantly improved the mechanical properties of fault breccia materials by altering the soil structure, increasing density, and optimizing intergranular contact. The remolding process enhances cohesion through three primary mechanisms: particle rearrangement, destruction of the natural fabric, and mineral self-cementation. The enhanced cohesion observed in remolded samples results primarily from increased particle interlocking and bonding during compaction and self-cementation, processes absent in the natural, weakly bonded fault gouge structure.

4.3. Mineralogical Evolution and Elemental Redistribution

This section explores variations in mineral components within fault zones through two distinct aspects: disintegration characteristics and mineral composition changes. The disintegration analysis primarily focuses on the physical behavior of fault zone materials under different conditions, emphasizing how the disintegration of these materials influences their properties. The mineral composition analysis, on the other hand, examines how structural and mechanical changes in fault zone samples relate to mineralogical alterations, particularly transformations in clay minerals and cementing agents.

4.3.1. Disintegration Characteristics

Figure 11 illustrates the disintegration test apparatus for fault zone mud and the corresponding mass changes. The apparatus comprises a static water balance, a support frame, a glass cup, and a wire mesh (Figure 11a). An electronic balance connected to a computer automatically records the data. The process can be divided into three stages (Figure 11b). During 0–1000 s, water rapidly penetrates and fully saturates the sample, maintaining a relatively stable mass at 810.5 g. From 1000 to 2250 s, water interacts with cementing materials in the saturated sample, causing cement softening and dissolution that breaks down inter-particle cohesion. Fine particles detach and become suspended, resulting in a sharp mass decrease of approximately 0.071 g/s. After 2250 s, the sample fully disintegrates. Larger and medium-sized particles remain, while fine particles reach equilibrium through settling, dissolving, or suspending. The overall mass stabilizes at 722.3 g.

This disintegration process demonstrates the vulnerability of fault zone mud to water infiltration. Water infiltration gradually deteriorates the cemented structure, weakens inter-particle cohesion, and causes fine particles to detach, dissolve, or suspend. This behavior indicates that the physical and chemical characteristics of fault mud contribute to its susceptibility to disintegration under hydraulic conditions and its poor engineering properties. Consequently, fault mud exhibits reduced strength and stability under hydraulic forces. These findings provide critical insights into the seepage behavior and mechanical response of fault zones under hydraulic conditions.

4.3.2. Mineral and Elemental Changes

Figure 12 presents XRD patterns of fault mud before and after the disintegration test. After disintegration, the diffraction peaks of chlorite, kaolinite, and calcite weakened significantly. These changes result from two mechanisms: dissolution and loss of cementing minerals, and migration and alteration of clay minerals. Calcite, a primary cementing agent, dissolves upon water contact and partially converts to calcium ions (Ca²⁺) in solution. This dissolution substantially reduces calcite crystallinity, weakening its diffraction intensity. Simultaneously, clay minerals undergo migration and structural alteration. During hydration, chlorite and kaolinite soften, causing particles to detach from the cemented structure and adhere to fine particle surfaces. These detached particles either remain suspended or precipitate in solution. Consequently, clay mineral content in the residual solid phase decreases, reducing their diffraction intensities. These coupled processes—cement dissolution disrupting inter-particle cohesion and clay mineral migration destabilizing the structure—ultimately cause mechanical strength degradation and mass loss.

Figure 13 presents elemental composition changes in fault mud before and after disintegration testing, determined by SEM-EDS analysis. Iron (Fe), potassium (K), calcium (Ca), and sodium (Na) decreased significantly, with concentrations declining from 8.1%, 2.3%, 1.9%, and 1.0% to 6.2%, 1.7%, 1.6%, and 0.2%, respectively. These reductions represent relative losses of 23.4%, 26.1%, 15.8%, and 80%, respectively. These changes reflect two concurrent physicochemical processes during disintegration. Water–rock interactions induce dissolution and ion exchange in clay minerals (chlorite and kaolinite), mobilizing potassium and sodium ions. Simultaneously, cementing agents (iron oxides and calcium carbonate) dissolve and structurally deteriorate, releasing iron and calcium through leaching. Element depletion and mineral transformation collectively weaken inter-particle cohesion, modify pore structure, and reduce macroscopic mechanical strength. These elemental and structural changes elucidate the hydration-induced degradation mechanisms of fault mud. The following sections compare pore structure and permeability between in situ and remolded samples. Although remolding induces structural modifications, remolded samples exhibit more uniform structure and superior mechanical properties.

4.4. Remodeling of Pore Structure and Size Distribution

Table 7. Comparison of porosity between in situ and remolded samples.

Type	Method	Porosity n (%)
		Sample 1	Sample 2	Sample 3	Average
In situ	NMR	15.92	16.71	/	16.31
	Saturation weighing method	15.98	16.87	/	16.42
Remolded	NMR	16.58	16.94	16.70	16.74
	Saturation weighing method	16.71	17.05	16.82	16.86

presents porosity measurements obtained using nuclear magnetic resonance (NMR) technology and the saturated weight method. Results indicate that in situ sample porosity ranged from 15.92% to 16.87%, while remolded sample porosity ranged from 16.58% to 17.05%. Comparison of the two methods reveals that NMR-measured porosity values are slightly lower than those obtained via saturated weighing. However, the relative error between methods remains within 1%, demonstrating good consistency. These results indicate that NMR technology provides reliable porosity characterization and effectively complements the saturated weighing method.

Figure 14 shows the pore size distribution of the in situ and remolded samples. Both sample types have similar average porosity, yet their pore structures differ significantly. In situ samples predominantly contain medium and large pores, with strong medium-pore signals and a high proportion of large pores, demonstrating pronounced structural heterogeneity and favorable pore connectivity. In contrast, remolded samples are characterized by small and medium pores, with more uniform pore size distribution and smaller average pore diameters. Pores are primarily interconnected through minute throats.

These structural differences result in significantly lower permeability in remolded samples than in in situ samples. Despite similar porosity, in situ samples retain the open pore system formed during natural deposition and geological history, facilitating fluid flow. In contrast, remolded samples undergo structural degradation due to artificial compaction, resulting in smaller pore sizes and reduced connectivity that markedly diminish permeability. This finding elucidates the intrinsic mechanism underlying permeability differences between in situ and remolded samples in the pore structure, providing crucial insights into coupled mineral–chemical, mechanical, and flow interactions.

5. Conclusions

In this study, permeability evolution in remolded fault gouge was analyzed to understand the multi-scale mechanisms from pore structure to mineral-particle migration. A series of experiments, including field in situ permeability measurements, seepage–stress coupling experiments, and micro-scale analyses (NMR, XRD, SEM-EDS), were conducted. The findings reveal significant differences in permeability between remolded and in situ samples, primarily driven by the pore compaction and degradation of pore connectivity. The main conclusions are as follows:

(1)

The reduction in permeability of remolded samples is primarily due to the degradation of pore connectivity and migration of mineral particles. SEM-EDS analysis revealed a significant leaching of cementing agents (K⁺, Ca²⁺ ions), while XRD confirmed chlorite dissolution as the dominant mechanism responsible for permeability decay.

(2)

The multi-scale pore structure plays a crucial role in controlling permeability. Remolded samples exhibit a more uniform pore distribution, with tiny pores and reduced connectivity compared to in situ samples, leading to a substantial decrease in permeability.

(3)

In terms of mechanical properties, remolded samples show enhanced cohesion and internal friction angle compared to intact samples, mainly due to uniform particle distribution and densification during remolding. However, this enhancement can distort mechanical behavior, leading to an overestimation of shear strength in laboratory tests relative to in situ conditions.

(4)

Conventional laboratory tests tend to underestimate the in situ permeability and overestimate shear strength of fault zones, primarily due to the disruption of natural structural features and cementation in remolded samples.

The findings of this study go beyond the specific fault gouge investigated, as the identified mechanisms such as cement dissolution and pore collapse are likely applicable to other fault zones with similar mineral compositions. However, these mechanisms’ relative significance may differ with lithological variations, such as rigid particle content or structural differences, which limit the generalizability of the conclusions. Future studies could investigate these effects across different geological settings.