Abstract
This study investigates the microstructural changes of granite residual soil (GRS) upon compression, the aim being to understand further the deformation mechanism from a microscopic perspective and establish the relationship between the mechanical behavior and microstructural characteristics of GRS. By means of (i) experimental compression tests with different loadings and (ii) scanning electron microscopy (SEM), the structural evolution of GRS during compression tests is investigated systematically from both macroscopic and microscopic perspectives. Microstructural characteristics including particle morphology, pore size distribution, and particle preferred orientation are investigated specifically through SEM quantification methods. With an obvious turning point, the curve for natural GRS approaches the intrinsic compression line gradually when the vertical pressure exceeds the preconsolidation pressure, which indicates the significant influence of cementation bonding on the mechanical behavior of intact GRS. At the microscopic scale, the deformation of natural GRS is attributed to the compression and transformation of large pores, while the deformation of remolded GRS is also related to the transformation from mesopores to small pores. Upon compression loading, particles show higher preferred direction angle perpendicular to vertical loading, thereby facilitating the preferred orientation. With increasing vertical loading, the microfabric can no longer sustain the initial alignment pattern and tends to be rearranged and reoriented into a more stable and stronger structure. This study offers guidance for the deformation analysis of subgrade related to GRS.
1. Introduction
It is universally acknowledged that because of its unique formation process, residual soil (RS) has peculiar behavior that differs considerably from that of sedimentary soil (SS) [1, 2]. SS comes into being through transport (e.g., aeolian and fluvial) and deposition, whereas RS is the product of the weathering and decomposition of the parent rock. Therefore, the mechanical properties of RS are essentially dominated by the type of parent rock and the weathering process related to climatic conditions. RS has abnormal characteristics, having poor physical properties but satisfactory intact strength and stiffness; however, its intact strength decreases appreciably once its microstructures (e.g., cementation bonding) are destroyed upon loading and disturbance; therefore, the behavior of RS in engineering is quite complex when compared with that of SS [3]. With increasing civil engineering projects being constructed in tropical and subtropical regions (e.g., southern China and Singapore) where RS is distributed widely, its peculiar mechanical behavior has become the subject of extensive concern [4–8]. Because of its unique formation process, the deformation mechanism of RS cannot be predicted using empirical evaluation from previous studies on SS, which may bring difficulties to designing and construction in related engineering projects. Therefore, it is vitally important to understand fully the deformation mechanism of RS upon compression. Generally, the compression behavior of RS has been studied extensively in terms of the basic compressibility properties [9] and the effects of parent-rock type [10], kaolinization degree [11], particle content distribution [12], and weathering degree [13].
Studying soil microstructure is a basic prerequisite for understanding the mechanism for soil deformation, because the development of a soil’s microstructure reflects all facets of its composition, environment, and mechanical behavior [14]. However, the relationship between the compression behavior and corresponding evolution of microstructural characteristics of RS is yet to be established. This is because the conclusions of previous studies of SS are not totally applicable to RS because the latter’s properties are due mainly to the parent-rock properties and weathering process instead of the stress history. For instance, the cementation bonding of RS inherited from the parent rock plays an important role in the microstructure by interconnecting soil particles [1]. Apart from this, although several techniques including computed tomography (CT) and mercury intrusion porosimetry (MIP) have been used successfully in previous studies to investigate the microscopic characteristics of soils, most of those studies were focused on various types of SS [15–18] and have limited applicability to exploring the microstructural properties of RS. For example, MIP can only be used to obtain the microstructure of pores, whereas the pore volume of RS measured using MIP might be an overestimate because the microfissures within the RS would propagate under the high intruding pressures of MIP.
Scanning electron microscopy (SEM) is a useful method for analyzing quantitatively the gradual changes in the microscopic properties of the particles and pores of soil, through which the microstructural characteristics of SS (mainly sandy soil, soft clay, and mudrock) have been studied extensively. Pioneering studies have summarized the microstructural evolution mechanism and corresponding quantification methods [19–22], but proper and effective methods for analyzing the microstructure features of RS quantitatively from SEM images remain to be explored. For example, Hattab and Fleureau [21] drew segments manually to identify each particle to describe its dimensions and orientation, but that method is available only for clays with enriched platy kaolinite and not RS with abundant irregularly shaped particle aggregates.
Reported herein are the microstructural characteristics of granite RS (GRS) from Xiamen in southern China as established by quantifying SEM images of both intact and remolded samples under different loading stages. Of specific focus is the evolution of the pore distribution and particle orientations after one-dimensional compression, because the mechanical behavior depends largely on the changes that occur at the microfabric level [23, 24]. The present paper is organized as follows. First, the soil property indices and the test procedure are introduced; then, the methods for image processing and microstructural quantification are presented. Following this, the experimental test results are presented in detail, and the corresponding microstructural characteristics of the GRS before and after the compression test are quantified using the SEM technique. Finally, how changes in the microfabric orientation affect the void ratio and compressibility of the GRS is discussed by linking its microstructural characteristics and mechanical behavior.
2. Materials and Methods
2.1. Soil Characterization
The studied GRS samples were retrieved from a foundation pit of Subway Line 1 in Xiamen at depths of 17.5–18.5 m. Block sampling was used to reduce disturbances, and the dimensions of each GRS block were . Detailed information about the sampling conditions and other properties of the studied soil is given in Table 1. Figure 1 shows the grading of the GRS collected from the block samples, comprising 8.6% gravel, 38.1% sand, 23.3% silt, and 30.0% clay. The underground water table is relatively high (located at 1.5 m below the ground surface); therefore, the saturation degree of the presented GRS reaches 93%. X-ray diffraction analysis was also conducted to examine the mineralogy of the GRS (given in Table 1). A large amount of quartz (34.5%) contributes to most of the gravel- and sand-sized particles, which is the typical characteristic of RS. Surprisingly, the natural GRS samples with high clay content featured a high effective friction angle (), which might also be the result of the abundant unweathered quartz [2]. Kaolinite and feldspar account for 48.5% and 3.2%, respectively, and this obvious distinction of content in turn evidences the process of feldspar transforming into kaolin during the weathering of granite [25]. Also detected was a small amount of hematite (3.6%), which was shown to relate to the iron-bearing cementation within the GRS, and it may help to generate particle aggregations and hinder the rearrangement of soil particles upon loading and improve soil strength [26], and as such the unconfined compressive strength reaches 156 kPa. The plasticity chart is presented in Figure 2; repeated experimental tests were conducted, and the average value of the test results was used to obtain the Atterberg limits (given in Table 1). The plasticity index of Xiamen GRS is located above the A-line; therefore, the studied GRS could be classified as CH according to USCS [27], which is quite consistent with the high clay particle content shown in Figure 1.
2.2. Procedure for the 1D Compression Test
Standard one-dimensional compression tests were conducted to study the compression characteristics of the GRS. Each test was conducted on a natural sample and a remolded sample with identical dimensions (diameter ; height ). The natural sample was trimmed carefully following the procedures adopted by Liu et al. [2], and the remolded sample was prepared with similar water content and void ratio to those of the natural one. The remolded sample was prepared by mixing dried natural soil with deaired water, remolding in a mold, and compacting to the predetermined dimensions.
The compression tests were performed following ASTM D2435 [30]. Table 2 gives more information about each specimen and the testing program. The standard loading increment duration of 24 h suggested by ASTM was adopted. To investigate the evolution of the microstructure of the GRS samples upon compression, the natural and remolded samples were subjected to a series of parallel tests with different maximum vertical effective stress. The sample IDs listed in Table 2 summarize the soil type and maximum effective vertical stress of each sample; for example, N1600 represents the natural sample that was subjected to vertical pressure following a loading sequence until reaching 1600 kPa ultimately.
2.3. Microstructural Observations
2.3.1. SEM Sample Preparation and Observations
After one-dimensional compression testing, all the specimens were recovered from the consolidometer; then, a rectangular parallelepiped subsample with dimensions of was extracted from each compressed sample using a piercing saw coated with Vaseline (Figure 3). Note that the long side of the subsample was perpendicular to the direction of the external vertical pressure () applied in the compression test. To minimize the disturbance to the microstructure from soil shrinkage, the subsamples were dehydrated using vacuum freeze-drying, which has been shown to maintain the microfabric of soil samples successfully [14]. The subsamples were firstly frozen in liquid nitrogen and cooled to −196°C, and then, once all the water had transformed into amorphous ice, the frozen subsamples were placed in a freeze dryer at −50°C to allow all the amorphous ice to sublimate. Having obtained completely dehydrated subsamples, newly exposed planes were generated by fracturing a subsample into three samples for SEM observation, and this method has been shown to preserve the microfabric of soil samples well [21]. Exposed planes on the cross sections of dried samples, which were parallel with , were selected as the SEM observation planes to investigate the microstructural characteristics. A Quanta 250 scanning electron microscope was used to analyze randomly selected areas of each SEM observation plane; these selected areas covered different parts of the soil samples including soil particles, pores, and soil aggregates. Averaged quantitative indices of the microstructure from the three observation planes were used to evaluate the microstructural characteristics under a specific loading stage. Regarding the magnification of the SEM images, Zhao et al. [31] and Gao et al. [32] suggested the range from 1000× to 5000× as suitable for quantifying clay microstructures, and we used 2000× in the present study. Each magnified selected area from the observation planes contained approximately 500 particles on average, and their direction distribution (detailed later in Section 3.2.3) could be represented well by a rose diagram according to the conclusions from Barton [33].
2.3.2. Image Processing Method
Figure 4 shows the step-by-step method for processing the SEM images using the ImageJ software. First, the original image was transformed into a binary one by setting a threshold value; measurements with a gray level higher than this threshold were identified as particles and shown in white, while pores were conversely shown in black. In this way, soil particles and pores could be distinguished. The threshold value dominates the binarization degree, while different threshold values are applicable for processing different soil samples. For example, the widely distributed microscopic pores and flocculated structures within London clay make it difficult to separate the particles and pores accurately from SEM images through automated thresholding segmentation provided by ImageJ [19]. In the present study, the threshold was determined after repeated attempts on hundreds of SEM images of the GRS, and ultimately the average threshold value was adopted; this has also been used previously for successful threshold determination [31, 32]. Note that tiny particles of clay minerals were found to adhere to the surfaces of large particles as a result of the local weathering process, which affected the accuracy while extracting the microstructural characteristics of large particles. Therefore, the erosion and dilatation operations (shown in Figures 4(c) and 4(d)) were applied to the binary image successively to eliminate the effects of these tiny particles and narrow the bridges between two particles; this method is introduced in more detail elsewhere [34]. Because the particle aggregation produced by iron-bearing cementation bonds within the GRS might be wrongly identified as a single particle, the watershed algorithm was used to separate connected particles further as shown in Figure 4(e). Finally, the separated particles were replaced by fitted ellipses with the same areas, orientations, and centroids as those of the particles [32]. Thereafter, the geometric properties of the microfabric could be measured using the ImageJ software as discussed in Section 2.3.3. The complete procedure for identifying single particles is presented in Figure 4, and pores were extracted in the same way.
2.3.3. Microstructural Quantification
Computing the morphological parameters and orientations of the digitized particles (pores) depended on knowing the size and locations of the pixels in each SEM image. Figure 5 illustrates the quantitative methods for identifying the microfabric shape and orientation: the coverage area () was calculated by counting the pixels covering the microfabric image and then using their size (Figure 5(a)), while the microfabric orientation was characterized by the direction of the major axis of the fitted ellipse with respect to the horizontal axis, with the orientation angle () in the range of 0°–180° (Figure 5(b)). More quantitative parameters were used in this study to describe the orientation specifically as follows. (1)Fabric orientation () and index of fabric orientation (). The fabric orientation was presented as the mean direction of a set of vectors [35, 36] and was calculated aswhere is the total number of the microfabric within a SEM image, is the length of each vector (i.e., the major axis of the ellipse), and is the direction of each vector (interval 10°). The magnitude of the resultant vector was calculated as
(a)
(b)
A scalar static () reflecting the vectors’ dispersion degree was standardized by the ratio of to the sum of the lengths of all the vectors, i.e.,
The parameter characterizes the degree of orientation differently from a highly preferred orientation to a random orientation [35]. Specifically, corresponds to the maximum degree of iso-orientation, corresponds to a very oriented fabric, corresponds to a low-oriented fabric, and corresponds to a randomly oriented fabric. (2)Standard deviation of fabric orientation () and index of microfabric orientation (). Yue et al. [20] depicted the cumulative frequency curve by collecting the orientation angles of grain fabric in SEM images, and they presented the following equation to compute the values of sorting standard deviation of fabric orientation () for orientation data:where , , , and represent the angle of orientation at the 84th, 16th, 95th, and 5th percentiles on the frequency curve, respectively. This graphical method simplifies the calculation compared to those required by moment statics. The cumulative frequency curve tends to be steeper at the median value (50%), with the microfabric showing a highly preferred orientation. Furthermore, the standard deviation is normalized by the maximum value of () to calculate the index of microfabric orientation (), i.e.,
When approaches 1, the alignment of the fabric is highly oriented; corresponds to a randomly arranged fabric without preferred orientation; corresponds to a well-oriented fabric. (3)Shannon entropy (hereinafter referred to as entropy). This parameter is a measure of the spreading or dispersal of the fabric arrangement, which shows a general probability distribution aswhere is an arbitrary constant (equal to 1 in this case) that corrects the value of , is the number of bins that contain fabric orientations in the frequency distribution (in this case, 18 bins of 10° each), and is the frequency probability of fabric falling in a given bin. The dimensionless value increases with fabric dispersing into all bins and shows randomly oriented alignments.
3. Test Results and Analysis
3.1. Compression Behavior of GRS
The behavior of the natural and remolded GRS in the one-dimensional compression tests is shown in Figure 6, and the mechanical indices of these two samples are summarized in Table 3. The curve of the natural GRS has an obvious yield point and shows the tendency of remaining steady initially and then decreasing significantly (shown in Figure 6(a)). The studied natural GRS behaved as typical overconsolidated soil [2], with the apparent preconsolidation pressure of 400 kPa being determined following the Casagrande method [37]. When the vertical pressure was less than , the soil samples were barely deformed; when exceeded , the deformation developed rapidly and the compression curve tended to decrease linearly. Note that this quasi-overconsolidation is not the result of stress history (as with SS) but instead depends on the cementation bonds formed during weathering [38], which indicates the essential difference of overconsolidation behavior between RS and SS.
(a)
(b)
Unlike the natural GRS, the curve of the remolded GRS is a tilted straight line without an obvious turning point. Figure 6(a) and Table 3 show that the curve of the remolded GRS is steeper than that of the natural GRS and the remolded soil has smaller compression modulus, which indicates that the structural characteristics of the natural GRS helped to resist external loading.
However, the iron-bearing cementation tended to be damaged with gradually increasing loading, and the soil deformation was also influenced accordingly as follows: (i) for , the natural GRS deformed elastically because of the resistance from its structural strength; (ii) for , the gap between the curves for the natural and remolded GRS increased until it reached its maximum extent; and (iii) for , the curve for the natural GRS tended to approach the remolded one, demonstrating that the structural strength began to decrease, and as such the microstructural characteristics of the natural GRS gradually approached those of the remolded GRS.
To investigate further the effects of the structural characteristics (including cementation bonding) of GRS, the void index [39] was used for quantification through the location of the curve (Figure 6(b)). Also plotted in Figure 6(b) are the intrinsic compression line (ICL) and sedimentation compression line (SCL) proposed by Burland [39], both of which are especially applicable to interpreting structural effects. The preconsolidation pressure of the natural GRS is located to the right of the ICL, while the curve of the remolded GRS coincides well with the ICL, reflecting the fact that the mechanical behavior of the studied GRS was affected by not only the soil composition but also the soil fabric and interparticle bonding of the natural soil more dominantly. During the initial stage, the natural curve lies well above the ICL because of the structure of the GRS. However, after yielding, the natural curve tends to converge back toward the ICL, probably due to the breaking down of the bonding structure by excessive deformation [1]. However, the compression curve in this condition does not coincide with the ICL ultimately, indicating that under loading of 3200 kPa, the soil was partially structured and the soil fabric still functioned.
3.2. Evolution of Microstructural Characteristics
3.2.1. Microstructural Observation from SEM Images
Typical SEM images of natural and remolded GRS samples under different compression conditions are shown in Figure 7. From Figure 7(a), the natural GRS mainly presented a skeleton structure and the kaolin aggregates formed by booklet-shaped kaolinite minerals exhibited face-to-face contacts, while these aggregates interconnected with each other through face-to-face and edge-to-edge contacts. Copious interparticle pores can be seen in Figure 7(a), which also shows many clay minerals in the forms of particles and aggregates; however, quartz particles could hardly be observed in the SEM image, which could be because of the local weathering of the GRS. Once the vertical stress reached 400 kPa (equal to the preconsolidation pressure), the interparticle cement bonding began to break down and the interconnection of the aggregates tended to transform into face-to-face contacts to form larger aggregations, with the pore volume decreasing dramatically (Figure 7(b)). The higher the loading pressure, the denser the microstructural aggregates and the more evident the tendency of preferred orientation. Having been replaced by particle aggregates, single mineral particles with a flaky plate shape rarely appeared at the end of the compression (Figure 7(c)).
The remolded GRS was characterized by flocculation structures Figure 7(d)), with the soil particle size generally less than 10 μm. Unlike the soil particles of the natural soil samples with their structural anisotropy, those of the remolded samples interconnected into aggregates mainly through edge-to-edge contacts and showed disordered and loose arrangements without high preferred orientation. Although the pore volume of the remolded GRS was approximately similar to that of the natural one, the amount of large pores decreased dramatically after the soil remolding. Upon vertical loading, the remolded GRS particles showed a stronger horizontal oriented structure compared with the natural GRS, while the pores were predominantly meso- and small pores (detailed later in Section 3.2.2). Comparing the microstructural characteristics of the compressed natural and remolded GRS (Figures 6(c) and 6(f)), the natural microfabrics tended to become similar to those of the remolded GRS with increasing deformation, which is also consistent with the finding in Section 3.1 that the natural curve became parallel to the ICL at the end of the compression test.
3.2.2. Variation of Pore Size Distribution
A quantitative analysis of the corresponding pore characteristics is illustrated in Figure 8. The curves of pore size distribution before and after the compression tests of the natural and remolded GRS are shown in Figures 8(a) and 8(b), respectively. All these distribution curves are characterized by bimodal patterns: the main pore types found in the studied GRS were intragranular micropores of diameter , followed by large pores with and finally small pores () and meso pores (). However, although the volume of micropores was relatively small, the proportion of such pores in natural sample N0 (55.3%) appeared as a sharp peak in the distribution curve; the other types of pores comprised 9.8%, 5.5%, and 29.4% of the total pore-size proportion of sample N0 (Figure 8(a)). After the one-dimensional compression, the number of large pores decreased the most (by 10%); the percentages of micropores and mesopores increased by 7.1% and 2.5%, respectively, and the percentage of small pores remained nearly constant with tiny fluctuations. This variation of natural GRS pores indicates that compression of large pores contributed significantly to the deformation of the soil sample from a macroscopic perspective, which is consistent with the properties of clay according to previous studies [40]. Unlike in sedimentary clay, the pore deformation process of GRS shows a high dependency on interparticle cementation bonding, which makes it difficult for microfabrics to rearrange. During a test, large pores were compressed and became small pores preferentially, whereas because of interparticle bonding, it was difficult for mesopores to be compressed further. From a macroscopic perspective, the void ratio decreases under compression (see Table 2), but from a microscopic perspective, the amount of micropores and mesopores tends to increase.
(a)
(b)
Comparing the distribution curve of the remolded sample R0 (Figure 8(b)) with that of N0, the percentage of large pores increased from 29.4% to 34% and that of small and mesopores increased to 16.2% and 10.4%, respectively, while that of micropores decreased dramatically to 39.4%. This means that the remolding process adjusts the pore distribution characteristics partially, but they retain substantially similar trends. Comparing the distribution curves of R0 and R3200 shows that because the soil remolding destructed the iron-bearing cementation bonding, the mesopores in the remolded GRS were compressed greatly without the protection of interparticle bonding (their percentage fell by 50%), and therefore the percentage of micropores grew considerably from 39.4% to 63.2%. The changes in the pores of the remolded GRS during compression were caused by the transformation from large and mesopores to small and micro ones, which agrees well with the observation in Figure 7.
3.2.3. Analysis of Particle Orientation
Rose diagrams, which give the percentage of particles as a function of orientation, have been used widely to quantify orientation characteristics in previous studies [19, 31, 32]. Herein, global rose diagrams are plotted from the quantification of the SEM images taken from different observation areas. Each rose diagram was divided into 36 partitions, all of which are 10° intervals. Each 2000x image contains ~500 particles, which according to Barton [33] offers a good representation of the particle orientation distribution through a rose diagram.
In Figure 9, rose diagrams of the orientation measurements made on subsamples from the unloaded (0 kPa) sample and samples loaded under 200, 400, 800, 1600, and 3200 kPa are shown on the curve of the void ratio versus effective stress, along with changes in the average angles with respect to the horizontal axis. In Figure 10, the corresponding evaluation parameters for particle orientation and their development with increasing effective stress are shown. The degree of preferred particle orientation is classified as highly oriented type, lowly oriented type, and randomly oriented type, and markers in different colors (dark blue, light blue, and white) are used in Figure 10 to present the rearrangement process of the microfabrics upon compression intuitively. Note that the microstructural quantifications based on the standard deviation of fabric orientation () and index of microfabric orientation () used successfully by Yue et al. [20] show completely opposite results to those from the other parameters, and this may be due to the fact that Yue et al. segmented the particles and pores manually by applying the Photoshop software to the SEM images, which was highly subjective. Therefore, neither of these two parameters was used for the eventual orientation analysis in the present study.
Preferential orientation of particle groups appears in the unloaded natural Xiamen GRS, with particles orienting in directions between 0° and 10° and between 40° and 50°. Although the mean directional angle of natural GRS particles is calculated to be 38.8°, the rose diagram shows inhomogeneous distribution with a relatively low preferred orientation, which is also depicted in Figure 10. During the quasi-elastic deformation stage (), the particle orientations of the natural GRS remained little changed, and the rose diagrams show similar patterns with the fabric orientation slightly changed. The results indicate that when the load levels are less than the preconsolidation stress, the structural properties of natural GRS inhibit the rearrangement and reorientation of particles, and this may also be why a small account of deformation (a small change in the void ratio) occurs in Figure 9. The preconsolidation pressure is likely to be regarded as a turning point, higher than which an increasing degree of preferred orientation could be observed. With the index of fabric orientation increased from 0.20 to 0.78 and entropy decreased from 2.7 to 1.84 under loadings of 400 and 3200 kPa, respectively, the natural particles tended to rotate toward approximately the horizontal direction. After the ultimate compression stage, increasing individual particles were reoriented in the direction range of 0°–10°, with the average angle reaching 12.2°. The above discussion indicates that vertical loading induced a polarization, which means a preferred orientation particle arrangement. The polarization mechanism is completely activated once the vertical stress exceeds the preconsolidation pressure of GRS, after which the soil structure (e.g., cementation bonding) begins to degrade and particles are rearranged and reoriented into a more stable structure. They are interconnected mainly by face-to-face contacts and rotate until `perpendicular to the maximum principal loading, which is vertical in this case. Similar conclusions were also presented in previous studies [14, 41].
Unlike the phasic evolutions of the particle orientation of natural GRS, the particles of remolded soils without preconsolidation pressures and interparticle bonding show a steady growth trend toward horizontal preferred orientation. Before compression, the remolded GRS particles were classified as having random orientation (, ), which means that the remolding attenuated the particles’ orientation polarization compared with that of the natural GRS samples. With increasing vertical loading, the microfabric structure showed stronger anisotropy, with the principal orientation mode in the range of 0°–10°. The remolded GRS samples also showed higher orientation change rate; for example, under a vertical pressure of 400 kPa, the average direction angle of the remolded GRS was (smaller than for natural GRS). Particles had the highest degree of orientation at the end of the compression test, being oriented mainly with a mean preferential direction of 8.5°. Without the effect of cementation bonding, remolded soil particles tend to rearrange and rotate until reaching a more orderly and stable alignment pattern, with an even stronger orientation trend toward the horizontal plane than for natural samples.
Unlike in previous studies of sedimentary clay [1, 26], whose microstructure is characterized by structures comprising flatty clay minerals and whose particles are interconnected mainly through intermolecular forces including the van der Waals force and water bonds, GRS has more complicated particle morphology, wider pore distribution, and stronger cementation bonding due to the weathering process and thus has higher intact strength. Once the cementation bonding degrades, the particles are more easily rearranged and reoriented to the direction perpendicular to the principal loading.
4. Conclusions
The mechanisms for microstructural evolution in both natural and remolded GRS related to compression have been investigated systematically. This was done mainly through a series of one-dimensional compression tests on the two types of soil samples in progressive loading steps at the macroscopic scale, and image processing techniques were used to quantify the microfabric properties from SEM images obtained during different loading stages at the microscopic scale. The main conclusions are as follows. (1)An evident turning point was observed in the curve of the natural GRS, and the soil deformation process accelerated once the vertical loading exceeded the preconsolidation pressure of . The compression curve of the remolded GRS developed linearly, and the two curves intersected at the end of the tests. The curve of the natural GRS was located to the right of the ICL, thereby showing the microstructural characteristics of the natural GRS(2)From a microscopic perspective, the deformation process of the natural GRS upon compression was related mainly to the gradual compression of the large pores. Concurrent with the compression of the remolded soil, the large pores in the soil were transformed into micropores(3)Regarding the quantitative parameters, the fabric orientation (), the index of fabric orientation (), and the entropy () showed good applicability for the microstructural quantification on GRS. The natural GRS showed obvious preferred orientation directions between 0° and 10° and between 40° and 50°. Under increasing vertical loading, particles reoriented in the direction range of 0°–10°. The remolding process attenuated the particle orientation polarization of the GRS. However, the remolded GRS showed an increasing preferred orientation degree, with the mean preferential direction changing from 46.8° to 8.5° during the compression tests. Without the effect of cementation bonding, the remolded soil particles would be rearranged constantly toward a stable state and show a stronger orientation trend than that in the natural samples(4)The compression deformation mechanism of GRS is different from that of sedimentary clay as a result of the weathering process. During the compression tests, the microfabrics of the GRS were rearranged and reoriented constantly into a more orderly and stable structure. GRS has more complicated particle morphology, wider pore distribution, and stronger cementation bonding as a result of the weathering process and therefore shows higher intact strength
Data Availability
All data, models, or codes that support the findings of this study are available from the corresponding author upon reasonable request.
Conflicts of Interest
The authors declare that they have no commercial or associated interest that would represent a conflict of interest in connection with the reported work.
Acknowledgments
The authors are grateful for the financial support from the National Natural Science Foundation of China (Grant Nos. 41972285, 42177148, and 12102312), the Youth Innovation Promotion Association of the Chinese Academy of Sciences (Grant No. 2018363), the Science Fund for Distinguished Young Scholars of Hubei Province (Grant No. 2020CFA103), the CRSRI Open Research Program (Grant No. CKWV2021884/KY), and the Key Research and Development Program of Hubei Province (Grant No. 2021BAA186).