Effect of Sample Preparation on Saturated and Unsaturated Shear Strength of Cohesionless Soils

: The geotechnical behavior of cohesionless soils is governed by ﬁeld conditions. Such soils exist in two distinct forms, namely: disintegrated, such as fresh sediments under no overburden and/or no suction, and intact, such as old deposits with overburden and/or suction. The main contribution of this research was the successful capture of ﬁeld conditions in laboratory samples, and the determination of shear strength under saturated and dried states. Results indicated that disintegrated samples possess identical soil behavior under both saturation states. Shear stiffness and peak shear increased with increasing normal stress, and no clear failure peaks were observed, similar to loose soils. Both samples showed an initial contraction followed by dilation at low normal stresses and mostly contraction at high normal stresses. Apparent cohesion was non-existent, and the friction angle measured 44.5 ◦ in the saturated state and 48 ◦ in the dried state. The intact sample exhibited behavior similar to the disintegrated sample when saturated. Under the dried state, clear failure peaks followed by residual shear were observed, similar to dense soils. Soil response was primarily dilative at low normal stresses and largely contractive under high normal stresses. Apparent cohesion was zero, and friction angle was 42 ◦ in the saturated state and changed to 91 kPa and 36 ◦ , respectively, in the dried state. Finally, structural cohesion increased with normal stress, and the friction angle due to suction was between 0.05 ◦ and 0.02 ◦ .


Introduction
The Canadian Prairies possess cohesionless soil deposits with extensive variability derived from physical weathering due to the wind (aeolian), water (alluvial or outside river boundaries and fluvial or inside river boundaries, and ice (glacial advances and retreats) [1]. Generally, such evolving processes cause surface sediments to occur in two conditions, namely: (i) disintegrated due to the absence of soil suction (u a − u w ), such as fresh deposits of unsorted glacio-fluvial pediments [2,3] or non-stratified aeolian loess [4,5] and (ii) intact due to the presence of soil suction, such as older alluvial deposits of stratified silt [6] or stratified loess and loess plateaus [7,8]; normal stress (σ − u a ) is present in both cases. Given that the shear strength of cohesionless soils is governed by suction when unsaturated and by friction when saturated [9], it is critical to understand the engineering behavior of such soils under distinct field conditions.
The presence or absence of suction in cohesionless soils is related to post-depositional processes of cyclic saturation-desaturation [10]. The downward movement of a wetting front due to precipitation saturates the soil profile through infiltration, and the reverse process of exfiltration desaturates the same in the opposite direction [11]. During drying, water is gradually removed to increase soil suction at the air-water interface within the soil pores [12]. The thin water film around soil particles pulls the particles together to act as interparticle bonding [13]. This phenomenon of suction development forms intact cohesionless soils [2,14]. During wetting, water gradually displaces the pore air, thereby Geotechnics 2023, 3 213 reducing the negative pore pressure at the air-water interface to reduce interparticle bonding [15,16]. Clearly, interparticle bonding of suction is inversely proportional to the degree of saturation [17]. Figure 1 shows the conceptual relationship of shear stress with respect to stress state variables: normal stress is derived from overburden and governs soil density [18], whereas soil suction is derived from the atmosphere and governs soil saturation [19]. Under a given soil suction (Figure 1a), a linearly increasing relationship exists between shear stress and normal stress [20,21]. When soil suction is zero (saturated conditions), line A joins the ordinate at zero, indicating soil possesses no cohesion, and the slope pertains to the friction angle (ϕ ). When soil suction is more than zero (unsaturated condition), the intercept of line B denotes a higher value, which is known as apparent cohesion due to soil suction (c ψ ) [22,23]. Under a given normal stress (Figure 1b), a linearly increasing relationship exists between shear stress and soil suction. When normal stress is zero (surface soil), line C starts from zero because densification is non-existent [24], and the slope of line C denotes the friction angle due to suction (ϕ b ) [25]. When normal stress is more than zero (soil with overburden), line D starts from a higher intercept due to structural cohesion (c σ n ) derived from compression [26,27], while ϕ b mainly remains unchanged [9]. Geotechnics 2023, 3, FOR PEER REVIEW 2 as interparticle bonding [13]. This phenomenon of suction development forms intact cohesionless soils [2,14]. During wetting, water gradually displaces the pore air, thereby reducing the negative pore pressure at the air-water interface to reduce interparticle bonding [15,16]. Clearly, interparticle bonding of suction is inversely proportional to the degree of saturation [17]. Figure 1 shows the conceptual relationship of shear stress with respect to stress state variables: normal stress is derived from overburden and governs soil density [18], whereas soil suction is derived from the atmosphere and governs soil saturation [19]. Under a given soil suction (Figure 1a), a linearly increasing relationship exists between shear stress and normal stress [20,21]. When soil suction is zero (saturated conditions), line A joins the ordinate at zero, indicating soil possesses no cohesion, and the slope pertains to the friction angle (φ′). When soil suction is more than zero (unsaturated condition), the intercept of line B denotes a higher value, which is known as apparent cohesion due to soil suction ( ) [22,23]. Under a given normal stress (Figure 1b), a linearly increasing relationship exists between shear stress and soil suction. When normal stress is zero (surface soil), line C starts from zero because densification is non-existent [24], and the slope of line C denotes the friction angle due to suction (φ b ) [25]. When normal stress is more than zero (soil with overburden), line D starts from a higher intercept due to structural cohesion ( ) derived from compression [26,27], while φ b mainly remains unchanged [9].  Table 1 summarizes the various methods for field retrieval and laboratory preparation of cohesionless samples. Among the field methods, sample disturbance is inevitable when using tube/piston samplers (above water table) or Bishop's samplers (below water  table). Likewise, the use of soil impregnation or ground freezing techniques to minimize  Table 1 summarizes the various methods for field retrieval and laboratory preparation of cohesionless samples. Among the field methods, sample disturbance is inevitable when using tube/piston samplers (above water table) or Bishop's samplers (below water table). Likewise, the use of soil impregnation or ground freezing techniques to minimize sample disturbance is difficult to execute and may be time-consuming and uneconomical within a project. Finally, block sampling is suitable for retrieving an undisturbed sample only when there is some binding within the soil particles. Among the laboratory preparation methods, wind and water deposition can be simulated using Air Pluviation (AP), Moist Tamping (MT), Dry Deposition (DD), and Dry Tamping (DT). However, these methods have limitations in achieving high initial densities and sample uniformity. Likewise, the Water Sedimentation (WS) and Slurry Deposition (SD) are limited to saturated conditions only. Further, sample reconstitution and modified moist tamping methods mimicking intact in situ conditions are not suitable for cohesionless soils. This is because these methods do not represent cyclic saturation-desaturation of surface sediments. Although the DT is suitable for preparing disintegrated samples, a new laboratory sampling method is required to mimic the in situ intact samples in a desiccated state. Table 1. Summary of sampling methods for cohesionless soils.

Methods and References Purpose and Description Limitations
Field Retrieval Block sampling [28][29][30] To extrude undisturbed test samples from carefully trimmed and preserved in situ soil blocks.
Only applicable to soils possessing interparticle bonding through cementation and/or suction.
Sample size depends on tube diameter. Likewise, disturbance during drilling, retrieval, and extrusion.
Complex field execution and sophisticated instruments needed to flush the impregnated solution before testing.
Time-consuming, high operational cost, and special sample holders for storage and shipping.
Particle segregation. Also, non-uniformity increases with fine content due to lower falling velocities within a fixed drop height.
Non-homogeneous (±10%) density in the sample due to variable compaction of soil layers.
Dry deposition (DD) [53][54][55] To simulate surface deposition at low densities, dried soil is dropped from zero height by slowly raising the funnel.
Mostly suitable for cohesionless soils. Also, repeatability is difficult to maintain.
Dry tamping (DT) [56][57][58] To mimic surface sediments subjected to in situ loading by tamping the top of samples prepared by DD.
Susceptible to particle crushing and difficult to achieve higher initial densities.
Water sedimentation (WS) [59][60][61] To simulate the deposition of sediments in water bodies by placing the soil through a water-filled mold.
Only suitable for preparing saturated samples at low initial densities.

Slurry deposition (SD) [62-65]
To prepare uniform samples imitating in situ fabric by placing soil water mix in a container.
Applicable for saturated laboratory testing only.
Sample reconstitution [66][67][68] To replicate undisturbed in situ soil samples by consolidating slurry prepared at 1.5 times the liquid limit.
Not suitable for cohesionless soils.
Modified moist tamping [57] To mimic the unsaturated intact conditions by mixing soil with in situ water content followed by oven drying.
Does not replicate the cyclic saturation-desaturation of the cohesionless soils.
The main objective of this paper was to investigate the effect of sample preparation on the shear strength of cohesionless soils under distinct field conditions. Two types of samples were prepared [1]: disintegrated (D) and intact (I), such that each sample was tested under saturated and dried conditions. The direct shear test was employed due to its simple operation and the ability to easily test both sample types [69].

Research Methodology
Soil samples were collected from a typical fluvial deposit near Avonlea (latitude 50.0367 and longitude 105.0667), Saskatchewan, Canada. The up to 2 m thick deposit has a gentle slope of about 3 • with the horizontal and contains cohesionless soils along the valley of a flow channel. The fresh surface sediment was found to have a water content (w) of 4% and a dry unit weight (γ d ) of 11 kN/m 3 [3]. Under desiccated conditions, the underlying deposit at 1 m depth was found to have a w of 2% and a γ d of 14 kN/m 3 . The suction value under such conditions was reported to be 10 5 kPa [1]. The soil was classified as silty sand (SM) with 36% material finer than 0.075 mm and moderate consistency (liquid limit of 27% and plastic limit of 23%), as reported by [70]. A pneumatically controlled direct shear apparatus was used to conduct the testing ( Figure 2). The shear box assembly (containing the samples between two porous stones) was mounted on the apparatus and clamped. The displacements were determined by linear strain conversion transducers (HM-2310.10 for horizontal and HM-2310.04 for vertical measurement). Likewise, the shear stress was measured by an S-type transducer (HM-2300.020). The measured data were automatically recorded using a data acquisition system and transferred to a computer. The net normal stress of 25, 50, 125, and 300 kPa was applied pneumatically. Dried samples were allowed to equilibrate under each net normal stress for 24 h before shearing. In contrast, samples were inundated with water for 24 h after compression under a given net normal stress and subsequently sheared. A shearing rate of 0.05 mm/min was selected to preclude pore pressure build-up during shearing [71].
The tests were terminated after achieving 10 mm horizontal displacement (10% strain), samples were removed from the direct shear box, and the water content was measured for the saturated conditions. The shear strength at saturated and dried state was calculated by using the following two equations, respectively: Geotechnics 2023, 3, FOR PEER REVIEW 5  Figure 3 shows the different stages of Sample D during shear testing. Based on Cresswell et al. [72], a known amount of soil mass at 4% water content was poured into a direct shear ring (101.6 mm diameter and 26.8 mm height) and gently tamped with a wax rammer (50 mm diameter and 10 mm thickness) to obtain the desired density (11 kN/m 3 ) while avoiding particle crushing. The sample was prepared in the direct shear ring that provided lateral confinement ( Figure 3a) and subsequently mounted on the direct shear machine ( Figure 3b). Post-failure, the saturated sample ( Figure 3c) showed a distinct failure plane, whereas the dried sample ( Figure 3d) disintegrated due to the removal of lateral confinement. It is pertinent to mention that the investigated soil changes color from light grey (Figure 3a) to dark grey (Figure 3c) with the addition of water [70].  Figure 3 shows the different stages of Sample D during shear testing. Based on Cresswell et al. [72], a known amount of soil mass at 4% water content was poured into a direct shear ring (101.6 mm diameter and 26.8 mm height) and gently tamped with a wax rammer (50 mm diameter and 10 mm thickness) to obtain the desired density (11 kN/m 3 ) while avoiding particle crushing. The sample was prepared in the direct shear ring that provided lateral confinement (Figure 3a) and subsequently mounted on the direct shear machine (Figure 3b). Post-failure, the saturated sample ( Figure 3c) showed a distinct failure plane, whereas the dried sample (Figure 3d) disintegrated due to the removal of lateral confinement. It is pertinent to mention that the investigated soil changes color from light grey (Figure 3a) to dark grey (Figure 3c) with the addition of water [70]. Figure 3 shows the different stages of Sample D during shear testing. Based on Cresswell et al. [72], a known amount of soil mass at 4% water content was poured into a direct shear ring (101.6 mm diameter and 26.8 mm height) and gently tamped with a wax rammer (50 mm diameter and 10 mm thickness) to obtain the desired density (11 kN/m 3 ) while avoiding particle crushing. The sample was prepared in the direct shear ring that provided lateral confinement (Figure 3a) and subsequently mounted on the direct shear machine (Figure 3b). Post-failure, the saturated sample ( Figure 3c) showed a distinct failure plane, whereas the dried sample (Figure 3d) disintegrated due to the removal of lateral confinement. It is pertinent to mention that the investigated soil changes color from light grey (Figure 3a) to dark grey (Figure 3c) with the addition of water [70].  Figure 4 shows the different phases of Sample I during shear testing. As before, a known amount of soil mass at 2% water content was poured into the direct shear ring to obtain a density of 14 kN/m 3 . Next, the sample was saturated by placing it in a water bath for 24 h and desaturation in a convection oven for 24 h. Unlike Sample D, this sample was prepared in the lower half of the direct shear ring (Figure 4a) because it was capable of  Figure 4 shows the different phases of Sample I during shear testing. As before, a known amount of soil mass at 2% water content was poured into the direct shear ring to obtain a density of 14 kN/m 3 . Next, the sample was saturated by placing it in a water bath for 24 h and desaturation in a convection oven for 24 h. Unlike Sample D, this sample was prepared in the lower half of the direct shear ring (Figure 4a) because it was capable of standing without lateral support. The sample was mounted on the direct shear machine (Figure 4b). Post-failure, both the saturated (Figure 4c) and the dried (Figure 4d) samples exhibited distinct failure planes due to interparticle friction and suction bonding, respectively. The sample exhibited a color change (Figure 4c), as explained above.   Figure 5 gives the test results for Sample D. Soil response is nearly identical under both saturated and dried conditions because of the absence of soil suction to provide interparticle bonding. From the shear stress versus horizontal displacement curves, the initial slope (shear stiffness) and the peak shear stress increased with increasing normal  Figure 5 gives the test results for Sample D. Soil response is nearly identical under both saturated and dried conditions because of the absence of soil suction to provide interparticle bonding. From the shear stress versus horizontal displacement curves, the initial slope (shear stiffness) and the peak shear stress increased with increasing normal stress. This is due to densification that restricts the sliding of the soil particles along the shear plane [19,73]. Furthermore, both sample types exhibited a failure mode with no clear peaks under the given dry unit weight of 11 kN/m 3 . This is a characteristic of loose cohesionless soils [74]. The slightly lower values of peak shear stress under the saturated state are attributed to particle lubrication that reduced interparticle bonding [75]. From the vertical displacement versus horizontal displacement curves, both sample types at low normal stresses of up to 50 kPa showed an initial contraction (moving of smaller particles in the void spaces between bigger particles) followed by dilation (insufficient space for soil particles to achieve a closer configuration) [24,76]. This wavy trend was found to obtain subdued high normal stresses where both sample types primarily showed a contractive behavior because of continuous particle rearrangement due to densification. Furthermore, this contractive behavior is independent of the soil type and degree of saturation [24,77].  Figure 6 shows test results for Sample I. Soil behavior under saturated conditions was very similar to the previous sample. Due to the above-mentioned mechanisms, the shear stress versus horizontal displacement curve did not exhibit a clear peak at failure, and shear resistance increased with increasing normal stress. Likewise, the vertical displacement versus horizontal displacement curves showed contraction followed by dilation at a low normal stress of up to 50 kPa and primarily contraction at higher normal stress. Under dry conditions, where the initial data scatter is omitted, Sample I shows clear peak shear values such that the initial slopes remained nearly the same despite increasing normal stress. Thereafter, the samples showed residual shear similar to dense cohesionless soils. This trend is similar to elastoplastic softening behavior, where dense samples exhibit an initial linear portion followed by a peak at failure and then by residual values [78,79]. This behavior is attributed to the effect of soil suction that held the soil intact [80,81]. When compared with the saturated sample, soil behavior at low normal stresses of up to 50 kPa was primarily dilative because interparticle bonding due to suction precluded particle rearrangement. In contrast, soil behavior was found to be mostly contractive under high normal stresses because these stresses overcame the effect of suction and resulted in continuous particle rearrangement due to densification. Again, this contractive behavior is independent of the soil type and degree of saturation [24,77].

Results and Discussion
Under saturated states, the stress-strain behavior was identical irrespective of the sample type, although Sample I showed subdued initial dilation at a low normal stress of up to 50 kPa due to its relatively high initial γd of 14 kN/m 3 . Under dried states, Sample I  shows test results for Sample I. Soil behavior under saturated conditions was very similar to the previous sample. Due to the above-mentioned mechanisms, the shear stress versus horizontal displacement curve did not exhibit a clear peak at failure, and shear resistance increased with increasing normal stress. Likewise, the vertical displacement versus horizontal displacement curves showed contraction followed by dilation at a low normal stress of up to 50 kPa and primarily contraction at higher normal stress. Under dry conditions, where the initial data scatter is omitted, Sample I shows clear peak shear values such that the initial slopes remained nearly the same despite increasing normal stress. Thereafter, the samples showed residual shear similar to dense cohesionless soils. This trend is similar to elastoplastic softening behavior, where dense samples exhibit an initial linear portion followed by a peak at failure and then by residual values [78,79]. This behavior is attributed to the effect of soil suction that held the soil intact [80,81]. When compared with the saturated sample, soil behavior at low normal stresses of up to 50 kPa was primarily dilative because interparticle bonding due to suction precluded particle rearrangement. In contrast, soil behavior was found to be mostly contractive under high normal stresses because these stresses overcame the effect of suction and resulted in continuous particle rearrangement due to densification. Again, this contractive behavior is independent of the soil type and degree of saturation [24,77].   Table 2 show the shear strength parameters of the investigated soil under saturated and dried states. The peak shear stress versus normal stress data (Figure 7a) indicates the absence of apparent cohesion ( = 0) when the soil was either saturated and/or disintegrated. In contrast, was found to be 91 kPa for Sample I (dried), as derived from interparticle bonding of suction [82]. The friction angle for Sample D was found to be 44.5° for the saturated state because of increased lubrication. This value increased to 48° for the dried state. These values are higher than the angle of repose (measured under no stress) primarily due to enhanced interlocking between soil particles under applied normal stresses [83]. An opposite trend was observed for Sample I, in which φ′ decreased from 42° for the saturated state to 36° for the dried state because interparticle bonding due to suction reduced frictional resistance between particles [84,85]. Furthermore, the small variation in φ' (42° to 44.5°) under saturated conditions indicates that the effect of sample preparation is largely diminished due to saturation and that the discrepancy in measured values is attributed to experimental errors [61]. In contrast, the large variation inφ' (48° to 36°) under a dried state must be attributed to the effect of soil suction that is non-existent in Sample D but is dominant in Sample I. As mentioned earlier, interparticle bonding due to suction reduces frictional resistance between particles. For Sample I, which possesses suction, the lower φ′ = 36° is compensated by = 91 kPa, thereby increasing the shear strength of the soil at a given normal stress, as calculated by Equation (1).  Figure 7b shows the relationship between peak shear stress and soil suction. Both the saturated samples and the dried Sample D were plotted at an arbitrary suction of 1 kPa, Under saturated states, the stress-strain behavior was identical irrespective of the sample type, although Sample I showed subdued initial dilation at a low normal stress of up to 50 kPa due to its relatively high initial γ d of 14 kN/m 3 . Under dried states, Sample I showed higher shear stress values and clear peaks at failure, whereas Sample D gave lower shear stress values (except at 300 kPa normal stress) and vague peaks at failure. Similarly, Sample I exhibited pronounced dilation up to 50 kPa normal stress when compared with Sample D. As mentioned before, these differences are attributed to the presence of interparticle bonding and the higher initial γ d of Sample I. Figure 7 and Table 2 show the shear strength parameters of the investigated soil under saturated and dried states. The peak shear stress versus normal stress data (Figure 7a) indicates the absence of apparent cohesion (c ψ = 0) when the soil was either saturated and/or disintegrated. In contrast, c ψ was found to be 91 kPa for Sample I (dried), as derived from interparticle bonding of suction [82]. The friction angle for Sample D was found to be 44.5 • for the saturated state because of increased lubrication. This value increased to 48 • for the dried state. These values are higher than the angle of repose (measured under no stress) primarily due to enhanced interlocking between soil particles under applied normal stresses [83]. An opposite trend was observed for Sample I, in which ϕ decreased from 42 • for the saturated state to 36 • for the dried state because interparticle bonding due to suction reduced frictional resistance between particles [84,85]. Furthermore, the small variation in ϕ' (42 • to 44.5 • ) under saturated conditions indicates that the effect of sample preparation is largely diminished due to saturation and that the discrepancy in measured values is attributed to experimental errors [61]. In contrast, the large variation in ϕ (48 • to 36 • ) under a dried state must be attributed to the effect of soil suction that is non-existent in Sample D but is dominant in Sample I. As mentioned earlier, interparticle bonding due to suction reduces frictional resistance between particles. For Sample I, which possesses suction, the lower ϕ = 36 • is compensated by c ψ = 91 kPa, thereby increasing the shear strength of the soil at a given normal stress, as calculated by Equation (1). Figure 7b shows the relationship between peak shear stress and soil suction. Both the saturated samples and the dried Sample D were plotted at an arbitrary suction of 1 kPa, which is lower than the air entry value (AEV) (8 kPa) of the investigated soil [70]. The dried Sample I was plotted at 10 5 kPa suction corresponding to the desiccated state. The line joining the two states (saturated and dried) of Sample I was used to determine the values of c σ n and ϕ b . At the ordinate, the c σ n values increased by 8 kPa (from 22 kPa to 30 kPa) at 25 kPa normal stress and by up to 62 kPa (from 270 kPa to 332 kPa) at 300 kPa normal stress. This increase in c σ n is attributed to increased particle packing under increasing normal stress [86]. Furthermore, ϕ b was found to be between 0.02 • and 0.05 • based on the straight-line estimation. These values pertain to the dried state only and should not be taken as constant through the entire suction range. Generally, ϕ b reduces non-linearly with increasing soil suction [87] and diminishes the residual suction range [9,85].
Geotechnics 2023, 3, FOR PEER REVIEW 9 which is lower than the air entry value (AEV) (8 kPa) of the investigated soil [70]. The dried Sample I was plotted at 10 5 kPa suction corresponding to the desiccated state. The line joining the two states (saturated and dried) of Sample I was used to determine the values of and φ b . At the ordinate, the values increased by 8 kPa (from 22 kPa to 30 kPa) at 25 kPa normal stress and by up to 62 kPa (from 270 kPa to 332 kPa) at 300 kPa normal stress. This increase in is attributed to increased particle packing under increasing normal stress [86]. Furthermore, φ b was found to be between 0.02° and 0.05° based on the straight-line estimation. These values pertain to the dried state only and should not be taken as constant through the entire suction range. Generally, φ b reduces non-linearly with increasing soil suction [87] and diminishes the residual suction range [9,85].

Summary and Conclusions
Knowledge of field setting is critical for understanding the behavior of cohesionless soils. The main contribution of this research includes the following: (i) successful capture of in situ conditions by developing a laboratory sample preparation procedure and (ii) development of a clear understanding of the shear strength behavior under saturated and dried states. The findings of this research are summarized as follows:

•
The disintegrated samples showed identical soil responses under saturated and dried conditions. The initial slope (shear stiffness) and the peak shear stress increased with increasing normal stress, and no clear peaks were observed at the failure, similar to loose soils. Both samples showed an initial contraction followed by dilation at low normal stresses and mostly a contractive behavior at high normal stresses. Furthermore, apparent cohesion was non-existent and ϕ was found to be 44.5 • in the saturated state and 48 • in the dried state.

•
The intact sample exhibited behavior similar to the disintegrated sample in the saturated state. Under the dried state, the sample showed clear peaks at failure followed by residual shear similar to dense soils or elastoplastic softening behavior. Soil behavior was dilative at low normal stresses and largely contractive under high normal stresses. Likewise, apparent cohesion was zero, and ϕ was 42 • in the saturated state and changed to 91 kPa and 36 • , respectively, in the dried state. For the latter state, structural cohesion increased with normal stress, and ϕ b was found to be between 0.05 • and 0.02 • .

Data Availability Statement:
The authors can provide access to test data upon request.