Structural,volumetric and water retention behaviors of a compacted clay upon saline intrusion and freeze-thaw cycles

Jinguo Lin, Weilie Zou,b,*, Zhong Hn,b,**, Ziwei Zhng, Xiequn Wng

a School of Civil Engineering, Wuhan University, Wuhan, 430072, China

b Key Laboratory of Hydraulic Rock Mechanics of Ministry of Education, Wuhan University, Wuhan, 430072, China

c School of Civil Engineering and Architecture, Wuhan University of Technology, Wuhan, 430070, China

Keywords:Compacted clay Microstructure Volumetric behavior Water-retention capacity Salinization Freeze-thaw (FT) cycles

A B S T R A C T

1. Introduction

Compacted clays are widely used in barrier and liner systems for landfills and disposal repositories in geoenvironmental engineering due to their low permeability and self-healing properties against cracking(Daniel and Wu,1993;Deneele et al.,2010;Jobmann et al.,2017;Julina and Thyagaraj,2020).The intrusion of salinity from the environment (e.g. leachate from waste, fertilization, etc.) poses severe influences on the hydro-mechanical properties of compacted clays(Bolt,1955;Basma et al.,1996;Castellanos et al.,2008;Rao and Thyagaraj, 2010; Cui and Tang, 2013; Thyagaraj and Rao,2013; Mishra et al., 2019).

The as-compacted structure of clays in barriers and liners is typically homogenous and dispersed(Mitchell et al.,1965;Monroy et al., 2010; Romero et al., 2011). Upon saline intrusion, the ascompacted structure may transfer into an aggregated and densely stacked structure with clay particles becoming flocculated(Palomino and Santamarina, 2005; Lyu et al., 2020). Such alternation is more significant when the saline concentration is high(Musso et al.,2013;Mokni et al.,2014).Some researchers attribute this behavior to the suppression of adsorption water film upon the increase in the ion concentration based on the electric double layer theory (Palomino and Santamarina, 2005; Rao et al., 2006; Mokni et al., 2014; Thyagaraj et al., 2017). Such structural changes lead to a reduction in the swelling potential but an increase in the permeability of the intruded clay(Di Maio,1996;Musso et al.,2003;Rao and Thyagaraj, 2007;He et al., 2016; Zou et al., 2018a).

Barbour and Fredlund (1989) proposed two mechanisms to illustrate the effects of saline intrusion on soils’ volumetric behaviors: osmotic consolidation due to changes in the electrostatic stresses between clay particles, and osmotically-induced consolidation that is associated with the drainage of the pore fluid triggered by the osmotic pressure gradient.Some researchers regarded the osmotic pressure gradient as an equivalent stress variable to interpret the strain and settlement induced by the intrusion of salinity solution (Rao and Thyagaraj, 2007;Thyagaraj et al., 2017).

Existing researches on clays’ shrinkage behaviors upon dessication mainly focus on factors such as mineral composition, clay particle content, initial soil structure, and hydraulic and stress histories(Boivin et al.,2004;Cornelis et al.,2006;Chertkov,2007;Alaoui et al., 2011; Leong and Wijaya, 2015). Studies on the shrinkage behaviors of clays intruded by saline solutions are comparatively less reported in the literature.

For the shrinkage behavior, which is mainly indicated by the shrinkage curve (i.e. relationships between soils’ volume and moisture content during desiccation), experimental observations on salinized soils vary. For example, Zhang et al. (2017) obtained similar shrinkage curves from slurried silty clay specimens prepared with NaCl solutions of different concentrations and suggested that NaCl solutions have little impact on shrinkage curves.On the contrary,Thyagaraj et al.(2017)reported that the magnitude of osmotic suction and the type of pore fluid impose critical influences on the shrinkage behavior of compacted expansive clays.Besides, Julina and Thyagaraj (2020) evaluated the combined effects of pore fluid and cyclic wetting-drying (WD) cycles on the desiccation cracks of a compacted clay. They found that after WD cycles, clay specimens inundated with 4 mol/L NaCl solution developed more desiccation cracks than those inundated with distilled water or 0.4 mol/L NaCl solution.

The soil-water retention curve(SWRC)defines the relationships between soils’moisture content and suction and is the key property indicating soils’ water retention capacity. Capillary effects (mainly associated with soils’pore structure)dominate the SWRC in the low suction range, while adsorption effects (mainly associated with soils’ mineralogical composition, cation exchange capacity, and surficial properties) control the SWRC in the high suction range(Romero et al.,1999,2011;Birle et al.,2008;Lu and Khorshidi,2015;Lu, 2016; Han and Vanapalli, 2017; Lu and Dong, 2017; Zou et al.,2018b; Molinero-Guerra et al., 2020). Thyagaraj and Sudhakar(2010) reported that the swelling potential and water retention capacity of an expansive clay decrease upon the infiltration of salt solutions.He et al.(2016)reported for a clay that at a given suction in the high suction range, the degree of saturation of specimens increases with the increase in salt concentration. Mokni et al.(2014) obtained similar SWRC in the high suction range for the compacted illitic-kaolinitic clay samples prepared with NaNO3solutions of different concentrations. Zhang et al. (2017) found that the matric suction in the low range was not sensitive to the pore fluid.

In cold regions,the effects of cyclic freeze-thaw(FT) actions on the hydro-mechanical properties of compacted clays are of great concern.It is widely recognized that during FT cycles,the void ratio increases for dense soils while decreases for loose soils, the permeability always increases irrespective of the volumetric changes,and the shear strength declines remarkably(Chamberlain and Gow,1979;Konrad,1989;Qi et al.,2006;Gens,2010;Liu et al.,2016;Ren et al.,2017;Dalla et al.,2019).Besides,FT actions change the microstructure of soils by introducing FT-induced macropores and fissures and leads to a reduction in soils’ water retention capacity and the scale of swelling and shrinkage upon moisture fluctuation (Ding et al., 2020; Zou et al., 2020b). Saline intrusion and FT actions pose significant yet different influences on the structural,volumetric and water retention behaviors of compacted clays. However, few studies have focused on their joint effects despite their significance in the design of earthen barrier and liner systems in cold regions.

This study aims at investigating the influences of FT cycles on the microstructure, swelling and shrinkage behaviors, and water retention capacity of a compacted clay that is intruded by NaCl solutions of different concentrations. Compacted clay specimens were first subjected to different FT cycles and then submerged in pure water or NaCl solutions to record the swelling behavior.Saturated specimens were thereafter air-dried under room temperature during which the variations in the volume and suction with decreasing moisture content were measured to establish the shrinkage curve and SWRC, respectively. Mercury intrusion porosimetry(MIP)and field emission scanning electron microscopy(FESEM)tests were conducted to observe the evolution in the clay’s microstructure during FT cycles, saline intrusion and the subsequent drying processes to explain the responses of the clay from microscopic perspectives.

Fig.1. Testing sequence in this study.1D represents ‘one-dimensional’.

2. Materials and methods

2.1. Materials and specimen preparation

The clay investigated in this study was collected from Heilongjiang Province in northeastern China. Clay samples were airdried, pulverized, and then passed through a 2 mm sieve to remove larger particles and debris. The processed clay was subjected to a series of laboratory experiments to determine the basic physical properties (Table 1). X-ray analysis shows that the clay mineral content(%by mass)is 12%for the montmorillonite,3%for the illite, 2% for the kaolinite, and 1% for the chlorite.

Table 1 Physical properties of the tested clay.

Table 2 Intruded volume at different states.

The dry clay powder was thoroughly mixed with distilled water to attain the clay’s optimum water content (wopt= 22.76%). The moist soil was statically compacted to the maximum dry density(ρdmax= 1577 kg/m3) in oedometer rings of 61.8 mm in diameter and 20 mm in height. The initial void ratio of the specimens was 0.66.Compacted specimens were immediately wrapped in layers of plastic membrane to prevent evaporation.

Fig.1 shows the testing sequence in this study. As-compacted specimens were first subjected to different numbers of FT cycles(O→A). FT-impacted specimens were submerged in pure water or NaCl solutions with different concentrations to facilitate the saline intrusion and the measurement of the swelling behaviors (A→B).Saturated specimens were air-dried to the shrinkage limit (B→C)and then further dried in an oven to achieve oven-dryness. The shrinkage curves and SWRCs were measured during the desiccation.MIP and FESEM tests were performed at pointsO,A,BandCin Fig.1 to trace the structural evolution of the clay.

Fig. 2. FESEM images of (a) as-compacted specimen and (b) specimen after 10 FT cycles.

Fig. 3. MIP results before and after FT action: (a) CI curves, and (b) PSD curves.

Fig. 4. FESEM images: (a) Saturated state, NFT = 0, c = 0 mol/L; (b) Saturated state, NFT = 0, c = 3 mol/L; (c) Air-dried state,NFT = 0, c = 0 mol/L; and (d) Air-dried state, NFT = 0,c = 3 mol/L.

Fig. 5. MIP results of specimens without FT cycles: (a) CI curves at c = 0 mol/L, (b) CI curves at c = 3 mol/L, (c) PSD curves at c = 0 mol/L, and (d) PSD curves at c = 3 mol/L.

2.2. Application of FT cycles

FT cycles were applied in an environmental chamber with precise temperature control. In one FT cycle, specimens were frozen at -20°C for 12 h and thawed at 20°C for 12 h. The chosen temperature range was determined referencing the local average temperature records in Heilongjiang,which was between 23.1°C in summer and-18.6°C in winter.The 12 h is considered adequate for achieving equilibrium of temperature and moisture in small size specimens such as the disc specimens used in this study(Lu et al.,2019; Ding et al., 2020; Zou et al., 2020b).

All FT cycles were applied at free stress state.In other words,no external stress was applied. This is because the in situ soils subjected to FT actions are mainly in the shallow depth where the stress level is low. Such external stress level is much smaller than the internal stress caused by the growth of ice crystals during freezing and thus imposes a much lower impact on soils’ mechanical behaviors compared to FT cycles. In addition, during freezing,the development of cracks and swelling is most significant when there is no external stress.Thus,the nil stress state represents the worst-case scenario for the degradation in soils’ mechanical properties.

Compacted clays used in liner and barrier systems typically are very low in permeability.Upon FT processes,the migration of water in the clay layer is limited, especially during rapid temperature drop.In other words,the moisture content remains approximately constant during FTcycles.Therefore,closed-system FTcycles(i.e.no moisture exchange with the surrounding environment) are commonly used to impose FT histories to clays(Chamberlain,1973).During closed-system FT cycles, soils’ structure and hydromechanical properties experience dramatic changes during the first FT cycle. The changes level off after about 5-7 cycles (Chamberlain and Gow,1979; Qi et al., 2006). In this study, three designated FT cycles(i.e.NFT=0,1 and 10)were applied to reveal the changes in clay’s behaviors from as-compacted condition (NFT= 0) to the equilibrium condition after FT cycles (NFT= 10), highlighting the most significant changes atNFT=1.During FT processes,specimens were confined in oedometer rings and sealed in plastic membranes to facilitate the closed-system FT processes.

2.3. Saline intrusion and swelling test

After FTcycles,specimens were confined in oedometer rings and submerged in distilled water (concentrationc= 0 mol/L) or NaCl solution with differentcvalues(c=0.1 mol/L,0.5 mol/L and 3 mol/L)in an oedometer to facilitate the saline intrusion and determine the 1D swelling behavior under a token load of 1 kPa.A dial gauge and a timer were used to monitor the development of vertical deformation (i.e. the volume of specimens) with time. Specimens were soaked for at least 5 d and until the change in the dial gauge readings was less than 0.01 mm over 24 h (i.e. strain rate <0.05%per day)following ASTM D4546-14(2019).The degree of saturation of specimens after such soaking procedure was 95%-100%, confirming that the saturated condition has been achieved within the tested specimens.

2.4. Measurement of the shrinkage curve

Saturated specimens were retrieved from oedometer rings and placed on perforated plastic plates at room temperature to dry in the air at room temperature of 20°C. Their mass and dimension were periodically measured using a balance readable to 0.01 g and a digital vernier caliper readable to 0.005 mm, respectively. Such an approach measures the global volume of soils (including the volume of soil solids, voids and cracks) and is different from approaches such as the balloon method (Toker et al., 2004; Cornelis et al., 2006) which measure the volume of structureless soils or soil matrix.

It generally took 5 d for the specimens to dry from fully saturated condition to the shrinkage limit. The shrinkage limit was deemed achieved when specimens stopped shrinking and their mass stopped decreasing,per the criterion outlined in the Chinese Standards for Geotechnical Testing Method GB/T 50123-2019(2019). Specimens were thereafter transferred to an oven to dry at 105°C for 24 h to achieve oven-dryness. Mass and dimension measurements during drying were used to determine the moisture content and volumetric strain of the clay during desiccation and thus establish the shrinkage curve.

2.5. Measurement of the SWRCs

Contact and non-contact filter paper methods using Whatman 42 filter papers were employed to measure the variation in the matric suction(sm)and total suction(s)with moisture content per ASTM D5298-10 (2010), respectively. Considering that measuring suction using the filter paper method is time-consuming, only three levels of salt concentrations(i.e.0 mol/L,0.1 mol/L and 3 mol/L)were imposed on tested specimens.Specimens and filter papers were sealed in plastic containers for at least 10 d to achieve moisture equilibrium.The measuredsmandscan be used to determine the clay’s osmotic suctionso(=s-sm).Thesocan also be obtained from Van’t Hoff’s theory as follows:

whereiis the Van’t Hoff factor,andi=2 for NaCl solutions(Rao and Shivananda, 2005);Ris the universal gas constant and equals 8.314 J/(mol K); andTis the absolute temperature.

2.6. Microstructural observation

MIP and FESEM tests were carried out on specimens at ascompacted state (NFT= 0) and equilibrium state after FT cycles(NFT= 10), and specimens inundated with distilled water(c= 0 mol/L) or NaCl solution (c= 3 mol/L) at the saturated state and air-dried state. MIP tests determine the clay’s cumulative intruded volume (CI) curve (i.e. intruded volume per unit mass of soil,Vversus pore size,d, relationships) and the pore size distribution(PSD)curve(i.e.the derivatives of the CI curves,dV/d(log10d)versusdrelationships). Small soil blocks were sampled from the disc specimens and were freeze-dried in liquid nitrogen to preserve the structure before submitting to MIP and FESEM tests.

3. Results and discussions

3.1. Clay’s microstructural behaviors

3.1.1. Evolution of the microstructure during FT cycles

Fig.2 shows the FESEM images of the as-compacted specimens and specimens after 10 FT cycles.The as-compacted clay presents a homogenous structure (Fig. 2a), while after 10 FT cycles, the structure becomes segragated with visible cracks. The cracks present dimensions of approximately 20 μm. The presence of the cracks is associated with the development of ice crystals during freezing,which segregates the soil structure and causes irreversible large space and cracks(Chamberlain and Gow,1979;Qi et al.,2006).Such change is also reflected in the clay’s global volume. The void ratio(e)increased from 0.66 for the as-compacted specimen to 0.68 after 1 FT cycle and to 0.71 after 10 FT cycles.

Fig. 3 summarizes the MIP results. The PSD curve of the ascompacted specimens is generally unimodal with a peak close to 2 μm. It evolves into a bimodal fashion after 10 FT cycles which features an existing peak at 2 μm and a new peak between 10 μm and 20 μm. It is noted that 10-20 μm is the dimension of the FTinduced cracks shown in Fig. 2. Thus, this new peak indicates FTinduced cracks.The two peaks are divided atd=5 μm,which is also the turning point of the CI curves of the specimens before and after FT cycles. Burton et al. (2015) suggested using the turning point of the CI and PSD curves as the boundary between clays’macropores and micropores. It is shown in following sections that the CI and PSD curves of the clay after saturation and desiccation also turn at 5 μm. Due to this reason, 5 μm is defined as the delimiting boundary separating the macropores and micropores.

Fig. 6. FESEM images at saturated state: (a) NFT = 0, c = 0 mol/L; and (b) NFT = 0,c = 3 mol/L.

The intruded volume of macropores (VM) and micropores (Vm)can be identified from the CI curves shown in Fig. 3a and are summarized in Table 2. Micropores slightly shrink after FT cycles,which is demonstrated by a reduction in theVmfrom 0.1551 atNFT= 0 to 0.1131 atNFT= 10 (i.e. decreased by 27%) and the downshifting of the PSD curves(Fig.3b).On the contrary,due to the development of FT-induced cracks, the volume of macropores significantly increases after FT cycles. TheVMrises from 0.0143 atNFT=0 to 0.0647 atNFT=10(i.e.increased by 352%).It can be seen that FT cycles impose more significant influences on the macropore system than the micropore system.

During freezing, pore water stored in macropores freezes first while the pore water in micropores remains unfrozen (Konrad,1989; Liu et al., 2016). Due to this reason, the water potential within macropores is lower than that within micropores, which results in the migration of unfrozen water from micropores towards the macropores and facilitates the growth of ice crystals. The development of ice crystals enlarges the macropores consequently.Due to the loss of unfrozen water, the suction within micropores continues to increase, leading to the shrinkage of the micropores(Zou et al., 2020b).

3.1.2. Influences of salinization on the microstructure of specimens without FT histories

This section discusses the evolution of the microstructure of ascompacted specimens (i.e. without FT histories) during the saturation and desiccation processes. Fig. 4 shows the FESEM images of the as-compacted clay at the saturated condition and the shrinkage limit. The corresponding CI and PSD curves are summarized in Fig. 5 and theVMandVmare summarized in Table 2.

The soil structure does not change significantly upon soaking in distilled water (Figs. 2a and 4a). There is a slight increase in the volume of micropores due to the thickening of the electric double layer between soil particles (Delage et al., 2006) and macropore owing to the swelling of the soil skeleton.Correspondingly,VMandVmvalues increase and the PSD curve generally shifts upward(Table 2 and Fig. 5c).

On the contrary, as-compacted specimens soaked in NaCl solution(c=3.0 mol/L)develop visible large pore spaces(Fig.4b).It is important to note that the NaCl dissolved in the clay’s pore fluid precipitates upon freeze-drying.The solid NaCl crystals reducesVMandVm. TheVMincreases (Table 2) and the PSD curve develops a new peak between 10 μm and 20 μm(Fig.5d).Meanwhile,the soil matrix becomes denser compared to specimens soaked in distilled water(Fig. 4a and b), which is corroborated by the slight decrease inVm.In general,the global volume of the specimens swells slightly(i.e.VM+Vmincreases) after saturation.

The shrinkage of micropores and the increase in macropores are associated with saline intrusion.As shown in Fig.6a,clay particles and their structure become uniform when saturated in distilled water.However,in the 3 mol/L NaCl solution,the saline suppresses the water adsorption to the surface of clay particles and thus reduces the thickness of the electric double layer. The clay particles become densely stacked and flocculated, leaving visible space among them (Fig. 6b) that transfers to macropores (Palomino and Santamarina, 2005). This phenomenon fits the description of osmotic consolidation (Barbour and Fredlund,1989).

At air-dried state, macropores and micropores shrink significantly for specimens saturated in water and 3 mol/L NaCl solution,as can be judged from FESEM images in Figs. 4c and 4d, the decreases inVMandVmin Table 2,and the flattening of the PSD curves in Figs.5c and 5d.It is noted that the increase in macropores due to saline intrusion is reversible upon desiccation,and the shrinkage of micropores lead to an increase in the number of pores smaller than 0.05 μm (PSD curves in Fig. 5).

3.1.3. Influences of salinization on the microstructure of specimens with FT histories

This section discusses the evolution of the microstructure during saturation and desiccation for specimens that have been subjected to 10 FT cycles(i.e.with FT histories).Fig.7 shows the FESEM images of the specimens at the saturated condition and the shrinkage limit. The corresponding CI and PSD curves are summarized in Fig. 8 and theVMandVmobtained from CI curves are summarized in Table 2.

TheVmof FT-impacted specimens increases after soaking in distilled water, indicating the swelling of micropores. Meanwhile,theVmdecreases after soaking in 3 mol/L NaCl solution,indicating the shrinkage of micropores. Such behaviors are similar to that of specimens without FT histories. For macropores, the behaviors upon soaking are different. TheVMdecreases in distilled water,meaning a reduction in the macropore spaces. This is possibly associated with two reasons: (i) FT actions reduce the integrity of the soil skeleton and thus reduce the swelling or even induce the partial collapse of the soil skeleton upon soaking, and (ii) the soil matrix swells and occupies the space of macropore and FT-induced cracks.Comparing Figs.2b and 7a,it can be observed that the cracks are narrowed after soaking.

On the other hand,theVMslightly increases in the 3 mol/L NaCl solution, meaning that the volume of macropore remains stable upon saturation. This is related to the osmotic consolidation upon saline intrusion that restrains the swelling of the soil matrix.However, the PSD curve in the macropore range changes significantly after the saline intrusion and becomes similar to that of ascompacted specimens (Figs. 5d and 8d). The dimension of the FTinduced cracks is not reduced (Figs. 2b and 7b), which means that the clay’s healing(i.e.closure of cracks)is impaired during the saline intrusion.

Upon drying to the shrinkage limit, micropores shrink remarkably for specimens in both distilled water and 3 mol/L NaCl solution since their PSD curves (Fig. 8c and d) become flat and theVmdecreases. This is similar to that of specimens without FT histories.However, for the specimen saturated in distilled water, its PSD curve develops a new peak at 30 μm and theVMincreases significantly. This is due to the opening and widening of the FT-induced cracks during desiccation (Fig. 7c) that is associated with the shrinkage of the soil matrix.

The PSD curve of the specimen saturated in 3 mol/L NaCl solution also develops a similar peak close to 40 μm. However, itsVMremains almost constant during desiccation. This is the result of two reasons: (i) saline intrusion reduces the shrinkage of the soil matrix and (ii) the precipitated NaCl solid in macropores reduces theVM. Comparing PSD curves in Figs. 5d and 8c and d, it can be seen that unlike the macropore space induced by the saline intrusion that shrinks significantly upon desiccation, the macropore space developed during FT cycles (i.e. FT-induced cracks) is less sensitive to desiccation.

Fig.7. FESEM images:(a)Saturated state,NFT=10,c=0 mol/L;(b)Saturated state,NFT=10,c=3 mol/L;(c)Air-dried state,NFT=10,c=0 mol/L;and(d)Air-dried state,NFT=10,c = 3 mol/L.

3.2. Swelling behaviors

Fig. 9 shows examples of the development of swelling strain ε(the ratio of the volume of swelling to that of as-compacted specimens) versus timetfor specimens with differentNFTvalues and saturated in distilled water,and as-compacted specimens(NFT=0)saturated in NaCl solution with different concentrations. The ε-tcurves can be divided into initial swelling, primary swelling and secondary swelling phases (Rao et al., 2006; Zou et al., 2020a).Letters A and B in Fig.9a were used to identify the boundaries of the three phases that are determined based on the graphical approach introduced by Zou et al. (2020a).

The swelling strain reduces significantly after just one FT cycle and the soil’s volume is almost constant during saturation after 10 FT cycles. The ε-tcurve becomes linear after FT cycles with less determinate boundaries between the three stages (Fig. 9a). Such behavior upon FT action is related to two reasons: (i) FT-induced cracks partly accommodate the swelling strain of the soil matrix from inside of the soil,which reduces the global swelling strain and results in the closure of the FT-induced cracks, and (ii) FT action facilitates further hydration of the surface of soil particles through water flow induced by the phase transition of water.A higher level of hydration results in less swelling(Day,1994;Delage et al.,2006).

Similar to the effects of FT cycles, saline intrusion reduces the swelling of the as-compacted specimens (Fig. 9b). Such global volumetric behavior is consistent with the volumetric responses of the microstructure upon saline intrusion and is associated with the suppression of the electric double layer.Besides,the slope of the ε-tcurves reduces and the initial and primary swelling stages are prolonged in salt solutions. This is owing to the osmotic pressure gradient between the inside and outside of the specimens that reduces the velocity of flow and thus the speed of swelling.

Fig.10 shows the effects of saline intrusion and FT cycles on the final swelling strain (εf) of specimens. The εfreduces with increasingNFTandc, and the εf-NFTand εf-crelationships are nonlinear. The most significant reduction in the εftakes place during the first FT cycle and the initial increase inc. For the ascompacted specimen,the εfreduces from 7.02%to 0.22%after 10 FT cycles,constituting a reduction of 97%,to 1.28%when inundated in 3 mol/L NaCl solution,constituting a reduction of 82%,and to 0.13%when subjected to10 FT cycles and then inundated in 3 mol/L NaCl solution, constituting a reduction of 98%. It can be seen that FT action is more effective in reducing the swelling potential of the tested clay compared to saline intrusion.

Fig. 8. MIP results of specimens with FT cycles: (a) CI curves at c = 0 mol/L, (b) CI curves at c = 3 mol/L, (c) PSD curves at c = 0 mol/L, (d) PSD curves at c = 3 mol/L.

Fig. 11 shows the mass of (i) water and salt and (ii) salt that infiltrates into the specimens during inundation. The mass of intruded salt generally increases with the salt concentration,which is as expected and FT cycles appear to have little impact. NaCl solution suppresses the swelling of the soil and reduces the amount of absorbed water. Thus, the relationship of the mass of intruded water and salt withcis the interplay between two factors: reduction in the mass of water and the increase in the mass of salt.For the specimen without FT histories, the former factor governs and the mass of intruded water and salt reduces monotonically withc.When FT effects have come into an equilibrium (i.e.NFT= 10), the former factor governs at lowc(i.e. from 0 mol/L to 0.1 mol/L) and the later factor governs with a further increase inc(i.e. from 0.1 mol/L to 3 mol/L), resulting in a non-monotonic relationship.

3.3. Shrinkage behaviors

The shrinkage behavior of the tested clay during desiccation is described as the volumetric strain εv(defined by Eq. (2)) versus gravimetric water contentwrelationships (referred to as the shrinkage curve).

whereV1is the volume of the specimen during shrinkage;andV0is the volume at the as-compacted state, which is a constant for all specimens.

Fig.12 summarizes the influences ofNFTand saline intrusion on the shrinkage curves of the tested clay.A typical shrinkage curve has four stages, i.e. structural, basic, residual and zero shrinkage stages(Cornelis et al., 2006). The shrinkage curves obtained in this study present distinct basic and residual shrinkage stages.In general,their basic shrinkage stage is betweenw= 10% and 25%, as shown in Fig.12.The slope of the shrinkage curve at the basic shrinkage stage(i.e.Δεv/Δw)is used to indicate the shrinkage behavior in this study.Table 3 summarizes the εvat oven-dryness(denoted as εv0),the total shrinkage during desiccation (i.e. differences betweenεvat the saturated condition and εv0, denoted at εvtotal), and the slope of the shrinkage curve at the basic shrinkage stage.

Fig.12. Shrinkage curves of specimens with (a) NFT = 0, (b) NFT = 1, and (c) NFT = 10.

The εv0, εvtotaland the slope generally decrease with increasingNFTregardless of the concentrationcof the pore fluid. This means that FT cycles reduce the soil’s volumetric sensitivity during desiccation (i.e. shrinkage potential). On the other hand, for ascompacted and FT-impacted specimens, the εv0, εvtotaland the slope generally first increase withc(from 0 mol/L to 0.1 mol/L or 0.5 mol/L) and then decrease with further increase inc(from 0.1 mol/L or 0.5 mol/L to 3 mol/L). These behaviors were also reported by Thyagaraj et al. (2017). The initial increase in the shrinkage potential is associated with the osmotic consolidation and osmotically-induced consolidation triggered by the saline intrusion. With increasingc, although such effects still exist, the increase in the volume of undissolved salt crystals results in an increase in the final volume of solid phase in the specimens and thus the decrease in the shrinkage potential.

Fig.9. Swelling strain-time relationships:(a)Specimens with different NFT values and saturated in distilled water, and (b) As-compacted specimens saturated in NaCl solution with different c values.

Rijniersce(1983)first proposed a geometry factor,γs,to describe the relationship between the vertical and horizontal shrinkage, as written by

whereVsatis the soil bulk volume at saturated condition;ΔVis the volume change upon shrinkage;Zsatis the height of specimens at saturated condition; ΔZis the change in the height of specimens upon shrinkage; γs= 1 means 1D vertical subsidence, and γs= 3 indicates three-dimensional (3D) isotropic shrinkage.

Fig.13 summarizes the relationships between γsandw. Generally, γsincreases from 1 to around 3 with decreasingw, meaning that the shrinkage is approximately 1D when specimens are saturated and the shrinkage gradually becomes isotropic during desiccation. Such behavior is consistent with similar experimental results reported by Rijniersce (1983) and is associated with the changes in the degree of orientation of clay particles.

Clay particles after compaction demonstrate a high degree of particle orientation with mainly face-to-face contacts. Clay particles are generally perpendicular to the direction of compaction, as shown in Fig.14a.The saturation process does not significantly alter the degree of particle orientation. Upon desiccation, the shrinkage mainly takes place in the direction that is perpendicular to the orientation of clay particles.During the initial stage of desiccation,due to the high degree of particle orientation,the shrinkage mainly takes place in the vertical direction,constituting the 1D shrinkage.With further shrinkage, the orientation of clay particles gradually changes,which increases the amount of clay particles that face the lateral direction and face-to-point contacts (Fig. 14b). Due to this reason, the degree of particle orientation reduces and the lateral shrinkage develops, rendering the 1D vertical shrinkage to 3D shrinkage.

Specimens tend to demonstrate isotropic shrinkage at a higherwlevel after subjecting to FT cycles(i.e.γsincreases after FT cycles).In other words, FT cycles facilitate the horizontal shrinkage at a higherwlevel. This is owing to the effects of FT cycles on the orientation of clay particles. It appears that FT cycles reduce the degree of particle orientation and facilitate the development of lateral shrinkage during the early stage of desiccation. The NaCl solution appears to exert the opposite effect.At the samewlevel,γsof saline intruded specimens decreases with increasingc. This means that the NaCl solution increases the degree of particle orientation and thus suppresses the lateral shrinkage during the early stage of desiccation.

3.4. Water retention behaviors

Fig. 15 presents the SWRCs obtained from the filter paper method in terms of total suction,s(referred to ass-SWRCs). The concentration of the pore fluid directly influences the osmotic suction that is a part of the total suction. Consequently, thes-SWRCs are sensitive to the changes incof the NaCl solution,especially at the high concentration (c= 3 mol/L). At a certainwlevel,sincreases withc,leading to a right-upward rotation of thes-SWRCs for specimens with different FT histories(Fig.15a-c).With the increase ins, alls-SWRCs tend to converge tow= 0 ands= 106kPa, which is consistent with the assumption of Fredlund and Xing (1994) that the maximum suction at the dry condition is 106kPa.

FT cycles mainly reduce the water retention capacity of the soil in the low suction range (i.e.s< 104kPa), which is reflected by a downshifting of thes-SWRCs.This is associated with the reduction in the pore space of micropores where the majority of capillary water is stored(Monroy et al.,2010;Alaoui et al.,2011).It is noted that macropores desaturate at very low suction(typically at several kilopascals)due to their large pore radius,especially after FT cycles.In the high suction range(i.e.s> 104kPa),only adsorbed water is retained in the soil which is not sensitive to the FT-induced structural changes.Thus,thes-SWRCs remain unchanged after FT cycles(Fig.15d-f).

Fig. 16 presents the SWRCs obtained from the filter paper method in terms of matric suction,sm(referred to assm-SWRCs).Measurements were fitted using the model proposed by Fredlund and Xing (1994) as

Fig.10. Relationships of εf with (a) c and (b) NFT.

Fig.11. Mass of intruded water and salt during saturation process.

wheresresis the residual suction and takes the value of 105kPa;wsatis the moisture content at saturated condition; anda,nandmare the model parameters. Model parameters are summarized in Table 4,along with the coefficient of determinationR2achieved for each SWRC.

Table 3 Characteristics of the shrinkage curves at different concentrations.

The concentration of the pore fluid shows much less influence on thesm-SWRCs compared to thes-SWRCs. Changes in thesm-SWRCs are negligible whencincreases from 0 mol/L to 0.1 mol/L(Fig.16a-c),although the swelling and shrinkage behaviors change substantially. The slope of thesm-SWRCs slightly reduces whencfurther increases from 0.1 mol/L to 3 mol/L. This is possibly due to the changes in the soil’s macropores and micropores upon saline intrusion.Based on these observations,it can be concluded that the salinity mainly influences the clay’s volumetric and waterretention behavior through adsorption effects instead of capillary effects.

Similar to the responses ofs-SWRCs, thesm-SWRCs shift downward in the low suction range upon FT cycles, indicating a reduction in the water retention capacity (Fig. 16d-f). Thesm-SWRCs in the high suction range are less affected. It is noted that the variation in thesm-SWRCs upon FT cycles becomes less significant ascincreases.

Table 5 summarizes the measured and calculated osmotic suctions.The osmotic suction increases with salt concentration and FT cycles appear to have insignificant influence. For specimens inundated with distilled water (c= 0 mol/L), the measured osmotic suction is about 600 kPa. Such osmotic suction is associated with the intrinsically dissolved salt existing in the native clay. For specimens inundated with 0.1 mol/L and 3 mol/L salt solution,the measured osmotic suction value is close to the calculated one plus the measured osmotic suction atc=0 mol/L.This implies that the concentration of the pore fluid inside the soil equals the concentration of NaCl solution used for soaking the specimens, and the osmotic suction measured from the filter paper method is reliable.

Table 4 Model parameters of the Fredlund and Xing (1994)’s SWRC model.

Table 5 Measured and calculated osmotic suctions.

Fig.13. γs-w relationships at (a) NFT = 0, (b) NFT = 1, and (c) NFT = 10.

4. Conclusions

Effects of FT cycles and saline intrusion on the structural, volumetric and water retention behaviors of a compacted clay during swelling and shrinkage processes were investigated in this study.The following conclusions are drawn from the experimental results:

(1) Both FT cycles and salinization reduce the clay’s micropore space and increase its macropore space. FT cycles influence the pore structure through the migration of water between the micropores and macropores and the development of ice crystals while salinization influences the pore structure by altering the thickness of the electric double layer between soil particles. Salinization reduces the swelling of the soil matrix and thus suppresses the healing of the FT-induced cracks during soaking. The FT-induced macropore space develops during desiccation due to the opening and widening of the FT-induced cracks, while salinization-induced macropore space shrinks significantly during desiccation.Micropore space significantly shrinks during desiccation regardless of the FT history and saline intrusion.

(2) The swelling potential of specimens reduces significantly after FT actions because FT-induced cracks accommodate the swelling of the soil matrix and FT cycles facilitate further hydration. Salinization also reduces the swelling of specimens and prolongs their primary and secondary swelling stages.

Fig.14. Orientation of clay particles.

Fig.15. s-SWRCs at (a) NFT = 0, (b) NFT = 1, (c) NFT = 10, (d) c = 0 mol/L, (e) c = 0.1 mol/L, and (f) c = 3 mol/L.

Fig.16. sm-SWRCs at (a) NFT = 0, (b) NFT = 1, (c) NFT = 10, (d) c = 0 mol/L, (e) c = 0.1 mol/L, and (f) c = 3 mol/L.

(3) The clay’s shrinkage potential reduces remarkably after FT cycles because of the FT cracks. The shrinkage potential initially increases with the concentration of the pore-fluid due to the osmotic consolidation and osmotically-induced consolidation. The shrinkage potential tends to decrease with the further increase in the concentration because of the precipitated salt crystals. Besides, FT cycles facilitate the horizontal shrinkage at a higher water level, while the NaCl solution appears to exert opposite effects. This may be attributed to the degree of particle orientation which decreases with increasingNFTbut increases with increasing concentration of the pore-fluid.

(4) Total suction increases with the concentration of saline solution due to the contribution of osmotic suction regardless of FT cycles.However,the concentration of pore fluid shows much less influence on matric suction compared to the total suction. FT cycles reduce the water retention capacity of the clay at low suction levels but have little impact on the water retention capacity at high suction levels.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

The research was supported by the National Natural Science Foundation of China(Grant Nos.51779191 and 51809199),which is highly appreciated.