Effects of In-Situ Drying – Wetting Cycles on the Stress-Dependent Water Retention Behavior of Intact Loess

Understanding the effects of in-situ drying – wetting pattern on the stress-dependent water retention curve of intact loess is vital for addressing geotechnical problems in loess regions. The principal objective of this study is to investigate the in ﬂ uence of in-situ drying – wetting on the stress-dependent water retention behavior of intact loess. To meet this objective, six drying – wetting tests were carried out using a suction-and stress-controlled pressure plate extractor. Intact loess was sampled from three different depths: 1.0, 3.0, and 5.0m. For specimens from each depth, two vertical net stresses (i.e., 0 and 50kPa) were applied prior to the drying – wetting cycle. Experimental results revealed that the in-situ drying – wetting pattern greatly affected various aspects of the water retention behavior, particularly the hysteresis. The hysteresis of the specimen from 5.0m is about 82 % and 77 % larger than that of the specimens from 1.0 and 3.0 m, respectively. This is because the specimen from 5.0m has some large-size pores (i.e., > 400 μ m), which were not found in specimens from 1.0 and 3.0m. These large-size pores enhance pore nonuniformity and hence the hysteresis. Furthermore, specimens from different depths consistently showed a reduction of hysteresis when the stress was increased from 0 to 50kPa. The reduction is the most signi ﬁ cant for a specimen from 5.0m due to the collapse of large-size pores under compression.


Introduction
Understanding the stress-dependent water retention curve (WRC) of intact loess is essential for addressing many geotechnical problems in loess regions, such as rainfall-induced excessive embankment settlement [1][2][3], failure of landfill cover system [4][5][6] and slope instability [7,8]. Thus, some experimental studies have been carried out in this area. Muñoz-Castelblanco et al. [9] measured the WRC of an intact loess. The WRC showed a peculiar shape, with negligible hysteresis around the initial suction and marked hysteresis at lower and higher suctions. The zone having the smallest hysteresis was corresponding to natural variations of water content in the field. Ng et al. [10] compared the WRC of intact and compacted loess. The former had a larger hysteresis than the latter, particularly at suctions below 20 kPa, mainly because of ink-bottle effects induced by extra-large pores in the intact loess. Recently, Mu et al. [11] studied the WRCs of intact, compacted, and reconstituted loess specimens subjected to two drying-wetting cycles. The air entry value (AEV) and hysteresis of all specimens consistently changed with increasing drying-wetting cycles. It should be pointed out that each of the previous studies used intact specimens from a single depth. Soils at various depths are expected to have different WRCs because they have been subjected to different in-situ drying-wetting cycles and therefore own different pore structures [12][13][14]. So far, the effects of in-situ drying-wetting on the water retention behavior of unsaturated intact loess have not been purposely studied, in spite of some previous studies of unsaturated loess in the literature.
This study aims to investigate the effects of in-situ drying-wetting cycles on the stress-dependent WRC of an intact loess. WRCs of intact loess retrieved at 1.0, 3.0, and 5.0 m depths were measured using a pressure plate extractor. For soil from each depth, vertical net stresses of 0-50 kPa were applied. In addition, mercury intrusion porosimeter (MIP) and X-ray diffraction (XRD) tests were carried out to measure the pore size distribution (PSD) and mineral composition of the tested soil, respectively.

Test Program and Test Apparatus
Six drying and wetting tests were carried out to study the water retention behavior of an intact loess (i.e., D1S0, D1S50, D3S0, D3S50, D5S0, and D5S50). Soil samples were obtained from three different depths (i.e., 1.0, 3.0, and 5.0 m), and each of them was tested at two vertical net stresses (i.e., 0-50 kPa). It should be pointed out that the vertical net stress of 50 kPa is close to the in-situ stress state and related to many geotechnical problems, such as subgrade settlement and shallow slope failure [4,8]. Comparisons of the obtained six WRCs would reveal the influence of drying-wetting and stress on the water retention behavior.
In addition, to interpret the previous water retention behavior, a series of XRD and MIP tests were carried out to investigate the mineral composition and PSD of loess samples from 1.0, 3.0, and 5.0 m depths at their in-situ states, respectively.
The tests were conducted using a suction and stresscontrolled pressure plate extractor [15]. The suction, which equals the difference between pore air pressure and pore water pressure, was controlled by the axis translation technique. The vertical net stress was imposed using a loading rod. A dial gauge was attached to the loading rod to monitor soil volume change. In addition, the outflow/inflow rate of soil water during the drying/wetting process was measured through a ballast tube. Details of the apparatus were given by Ng and Pang [15].
The XRD patterns of these three samples are shown in Figure 1(b). Similar peak positions and peak densities are observed for those XRD patterns. The mineralogical compositions of tested loess are determined based on the XRD patterns and shown in Table 1. For the intact loess from 1.0 m, the nonclay minerals, including quartz, albite, and calcite, are identified, while the clay minerals mainly consist of nimite and muscovite. Calcite is a cementitious material that is probably acquired from reprecipitation [18]. In addition, the intact loess from 3.0 to 5.0 m contains the same minerals to that from 1.0 m due to their similar XRD patterns.
Soil water contents at the sampling locations were monitored from April to November in 2018 using the time domain reflectometry [19,20] and shown in Figure 2. This period covers both the dry and wet seasons in Yan'an. The measured results are, therefore, representative of the variations of soil water content in the field. During this period, the intact loess at 1.0 and 3.0 m depths experienced two drying and wetting cycles, corresponding to changes in the degree of saturation from 32% to 96%. In contrast, soil water contents at 5.0 m depth remained almost constant throughout the monitoring period.

Specimen Preparation and Test Procedures
Specimens were extracted from the loess blocks using an oedometer ring (61.8 mm in diameter and 19 mm in height). The specimen surface was carefully smoothed with a wire saw. The dry density and gravimetric water content of each specimen were measured and summarized in Table 2. In addition, the initial suction was measured through the null-type axis translation technique [21]: 88, 34, and 32 kPa for the specimens from 1.0, 3.0, and 5.0 m, respectively.
For XRD tests, the loess sample was ground using a McCrone Micronising mill for 20 min to obtain powder specimens [18]. The random powders were prepared through a razor-tamped surface method [22]. For MIP tests, small cubes with dimensions of approximately 5.0 × 5.0 × 5.0 mm were carefully cut from loess blocks. The prepared cuboid specimens were dehydrated by freeze-drying method to protect the structure [23].
The prepared specimens were then set up in the pressure plate extractor. The predefined vertical net stress (i.e., 0 and 50 kPa) was applied to each specimen at the constant water content condition. After that, all specimens were wetted to 0.1 kPa in steps. At each step, soil suction was controlled using the axis-translation technique, as discussed earlier. About 1-2 weeks were required to reach the equilibrium condition, in which the change in water content was less than 0.04%/day [24]. The soil volumetric water content was calculated according to the recorded vertical displacement and water content. In a similar approach, all specimens were dried to 400 kPa and finally rewetted to the initial suction in steps.
For XRD tests, diffraction patterns were obtained through the PANalytical X'Pert Pro diffractometer, with Cu Kα (λ = 1.5418 Å) radiation generated at 44 kV and 40 mA. The specimens were scanned from 2°to 45°with a rate of 1°2 θ/ min. The PSDs were measured through the Micromeritics Auto Pore IV 9500 V1.04. The applied pressure ranged from 0.5 to 413,685 kPa, corresponding to intrusion pore diameters from 0.003 to 360 μm. Figure 3(a) shows the MIP results of tested specimens at the in-situ stress state. The intruded void ratios and the actual void ratios calculated from the mass-volume relationship are labeled by i and c, respectively. The intruded void ratios are 0.75, 0.72, and 0.69 for specimens from 1.0, 3.0, and 5.0 m, respectively. Compared with the actual void ratios calculated from the mass-volume relationship, the nonintruded void ratios are almost negligible (i.e., less than 0.02) for specimens from 1.0 and 3.0 m. In contrast, the specimen from 5.0 m has a nonintruded void ratio of 0.09, corresponding to about 11.5% of its total pore volume. The nonintruded void ratio is attributed to the limitation of MIP device, which is unable to determine large-size  pores larger than 400 µm and small-size pores smaller than 0.007 µm [25]. According to the study of Ng et al. [10], the nonintruded pores of intact loess are mostly large-size pores because intact loess does not have many pores smaller than 0.007 µm. The large-size pores are not found in the intact loess from 1.0 and 3.0 m, probably because they have collapsed due to the numerous drying and wetting cycles in the field (see Figure 2). They are preserved in the specimen from 5.0 m, at which soil water content remained almost constant. Based on the results in Figure 3(a), the pore size density function is calculated and shown in Figure 3(b). The specimens from 1.0, 3.0, and 5.0 m show a similar tri-modal PSD. The dominant pore of the specimen from 3.0 m (i.e., 5.8 μm) is smaller than those of the other two specimens (i.e., 7.2 μm), mainly because it is about 3% denser (see Table 2). More importantly, it should be noted that only the specimen from 5.0 m has some nonintruded large-size pores (see Figure 3(a)). Its pores are, therefore, the most nonuniform. The influence of pore nonuniformity on soil WRC is discussed later. Figure 4(a) shows WRCs of the intact loess under zero stress. For each specimen, the AEV, void ratio after drying-wetting, desorption, and adsorption rates are determined and summarized in Table 3. Along the first wetting path, the specimens D1S0 and D3S0 became almost saturated when suction was reduced to 3 kPa. The specimen D5S0 reached saturation at a much lower suction (about 0.1 kPa), mainly because it has some large-size pores (see Figure 3(a)). In addition, the adsorption rates of D1S0 (0.035 (log kPa) −1 ) and D3S0 (0.034 (log kPa) −1 ) are about two times that of D5S0 (0.016 (log kPa) −1 ). Along the subsequent drying path, similarly, the desorption rates of D1S0 (0.034 (log kPa) −1 ) and D3S0 (0.035 (log kPa) −1 ) are 11.8% and 14.3% larger than that of D5S0 (0.030 (log kPa) −1 ), respectively. The differences between these three specimens in terms of the adsorption and desorption rates are likely related to their PSDs. As revealed in Figure 3(a), the specimen from 5.0 m has more nonuniform PSD, which would result in smaller adsorption and desorption rates [26,27]. As shown in Table 3, the AEV of intact loess at 3.0 m (i.e., 8.2 kPa) is larger than that of intact loess at 1.0 m (i.e., 5.7 kPa) and 5.0 m (i.e., 6.5 kPa), mainly because of the smaller void ratio owned by the former soil. Moreover, during the drying-wetting cycle, the volume changes of specimens D1S0, D3S0, and D5S0 are essentially elastic under zero stress. Figure 4(b) shows the WRCs of the intact loess under a vertical net stress of 50 kPa. The key parameters of each WRC are also given in Table 3. It is clear that specimen D5S50 has smaller adsorption and desorption rates than D1S50 and D3S50. This is consistent with the findings from Figure 4(a), suggesting that the observed effects of sampling depth on WRC are reliable. On the other hand, when the vertical net stress was increased from 0 to 50 kPa, the AEV increased by 10.5%, 8.5%, and 18.5% for specimens from 1.0, 3.0, and 5.0 m, respectively. The increment is the highest for the specimen from 5.0 m. A similar trend can be obtained for the adsorption and desorption rates (see Table 3). More importantly, in contrast to the zero stress, the specimens D1S50, D3S50, and D5S50 contract, and hence the irreversible contraction of 3.9%, 1.4%, and 6.5% are observed during the drying-wetting cycle. These observations suggest that the specimen from 5.0 m is most sensitive to net vertical stress because its large-size pores can be readily compressed. Figure 5 shows the relationship between suction and the degree of hysteresis. At a given suction, the degree of hysteresis is calculated by the ratio of the difference in the volumetric water contents along drying and wetting paths to their average value [28]. Parabolic shapes are observed for this relationship, with peak values appearing at a suction of 20 kPa. Under a given vertical net stress, the average degree of hysteresis of D5S0 and D5S50 (i.e., 0.226 and 0.201) is 81.4%-82.6% and 76.5%-77.6% larger than that of D1S0 and D1S50 (i.e., 0.042 and 0.035) as well as  D3S0 and D3S50 (i.e., 0.053 and 0.045), respectively. The significant hysteresis of the specimens from 5.0 m was likely attributed to their large-size pores. These pores enhance the pore nonuniformity and thus the ink-bottle effects, as illustrated by Ng et al. [10].

Degree of Hysteresis in the WRC.
For specimens from all three depths, the degree of hysteresis consistently decreases with increasing vertical net stress. The reduction is the most obvious for the specimen from 5.0 m. This is consistent with the finding in Figure 4 and probably attributed to the collapse of large-size pores.

Conclusions
In situ, the intact loess from different depths has been subjected to different drying-wetting cycles, which have a significant effect on the soil structure and water retention behavior. To understand the effects of in-situ drying-wetting cycles on the stress-dependent water retention behavior of intact loess, the intact loess from different depths (i.e., 1.0, 3.0, and 5.0 m) were tested by pressure plate and microstructure tests.  Advances in Civil Engineering 5 The intact loess sampled from 1.0 and 3.0 m depths showed almost identical hysteresis, desorption, and adsorption rates. When the sampling depth was increased to 5.0 m, the desorption and adsorption rates decreased by about 14% and 54%, respectively, and the hysteresis increased by over three times. The observed differences are likely due to the fact that the specimen from 5.0 m has some large-size pores (>400 μm), which were not found in specimens from 1.0 and 3.0 m. These large-size pores enhance pore nonuniformity, leading to larger hysteresis and smaller desorption/adsorption rates.
When the vertical net stress was increased from 0 to 50 kPa, the specimens from all depths showed a reduction of hysteresis. Furthermore, the reduction is the most obvious for the specimen from 5.0 m due to the collapse of its largesize pores.

Data Availability
Data are available from the corresponding author upon reasonable request.

Conflicts of Interest
The authors declare that they have no conflicts of interest.   Advances in Civil Engineering