Geotechnical Properties and Stabilization Mechanism of Nano-MgO Stabilized Loess

Abstract

This study focused on the utilization of nano-MgO as an energy-saving and eco-friendly stabilizer to improve the engineering performance of loess. To this end, loess samples at various nano-MgO contents and curing times were prepared, and then standard compaction, consistency limits, and unconfined compression tests were performed. The achieved results demonstrated that adding nano-MgO increased the liquid limit, plastic limit, and optimum water content of loess, while it decreased the plastic index and maximum dry density. The unconfined compressive strength (UCS) presented an increasing trend with curing time and a “rise-fall” trend with the addition of nano-MgO. At the optimum nano-MgO content of 2%, about 72% UCS gain was to be expected with 28 days of curing. The variation of the deformation modulus was similar to that of UCS, and the strain at failure presented an opposite trend. Empirical models for these properties were formulated and validated by literature data. Finally, from NMR analyses, the improving mechanism was found to be nano-MgO induced water transformation from free water to bound water.

1. Introduction

As global infrastructure requirements grow, there always exists scope for soil stabilization, in which local soils are treated with stabilizing materials to improve their performance. In engineering practice, loess is a commonly encountered problematic soil, as it is featured with severe collapsibility and water sensitivity [1] and covers about 6% of the total landmass [2]. In these regions, to strike a balance between engineering requirements and scarcity of ideal filling materials (e.g., gravel, sand, etc.), traditional cementitious additives (cement, lime, etc.) were mainly adopted for stabilizing loess to meet the engineering requirements [3,4,5,6,7]. However, the mass production and heavy use of traditional additives were not only associated with high energy consumption, but also caused serious environmental impacts [8,9]. As public environmental consciousness awakes, greater and more urgent attention has been paid to find cleaning and efficient additives for mitigating the environmental impacts [10,11].

Nowadays, nanomaterial is an emerging additive that could be used in various aspects of civil engineering. Due to its large surface area and high reactivity, nanoparticles cause a lot of interactions between the intermixed materials and, thus, produce stronger and more applicative mixed material (nano-cement, -coating powder, -soil, etc.) [9,12,13,14]. In terms of soil stabilization, the effect of nanomaterials on geotechnical properties is a major concern. Recently, many researchers have devoted to this field and reported noticeable improvements, even with small amount of nanomaterials [15,16,17,18,19,20,21,22,23,24,25]. The more frequently utilized nanomaterials include nano-SiO₂ [16,17], nano-clay [18,19], nano-MgO [22,23], and nano-CaCO₃ [24,25]. Among them, nano-MgO (NM) is normally produced by the sintering process of magnesite, performing under a relatively low temperature (700 °C), compared to that of cement (1450 °C), which makes it more energy-saving and eco-friendly in this field [26,27,28].

To date, several studies have reported successful use of NM in improving the engineering performance of local soils. Gao et al. [29] conducted triaxial tests on silt with an NM dosage up to 6%. The results demonstrated that NM exerted positive effects on cohesion, but did not noticeably affect the internal friction angle. Majeed and Taha [30] surveyed the effect of NM on the engineering properties of soft clay. They observed that an increase in NM dosage up to 0.4% led to increases of dry density, optimum water content, and unconfined compressive strength (UCS). Wang et al. [22] studied the influence of carbonization time on clayey soil treated with NM and cement. They found a sample with 1.0% NM, and 1 day carbonization gave the best improvement effect in UCS. Yao et al. [23] reported the advantages of utilizing NM in improving the performance of a cement soft clay with an optimum dosage of 1.5%. By conducting an SEM test, they further concluded that excessive dosage of NM resulted in cracks inside the structure and, thus, decreased the improving effect. A similar negative effect of higher nanomaterial dosage on UCS was also observed in research on nano-SiO₂-treated fine soil [31] and nano-clay-treated expansive clayey soil [32]. Chen et al. [33], by experiment, reported that adding NM enhanced the dynamic performance of loess and improved freezing–thawing resistance. Gao et al. [34] concluded that NM could bring strength and stiffness increases owing to the following effects of NM: cementation effect, water-absorption effect, and pore-filling effect. Overall, existing works have shown the potential improving effects of NM in soil stabilization, even with a small dosage. However, research on applying NM to stabilize loess is still very limited, and the influences of NM content and curing time on the geotechnical properties of loess are not quite clear and require further exploration.

Furthermore, understanding the improving mechanism of stabilizers has always been important in soil stabilization, and it is the same for nanomaterials. Presently, research in this field is mainly focused on investigating the nanometric scale structure of stabilized soil through electron microscopes (e.g., SEM, TEM, and AFM) [29,35,36,37]. It has been demonstrated that nanoparticles could fill micro- to nano-pores, aggregate soil particles, and transform weak dispersed structure to harder flocculated structure. However, it is well known that nanomaterials are characterized with large surface area and high reactivity, which endows them with stronger water-absorbing capability than soil particles [14,34,38]. Therefore, the nanomaterial-induced water-state change is an important aspect for understanding the stabilization mechanism.

To the best knowledge of the authors, the utilization of NM in improving the geotechnical performance of loess has not been thoroughly studied, especially involving stabilizing mechanisms from the aspect of water-state change. In light of the above, this study conducted consistency limits, compaction, and unconfined compression tests to provide a better understanding of the effects of NM in stabilizing loess. Based on these results, the variation of UCS, strain at failure, and the deformation modulus with NM dosage and curing time were analyzed. Accordingly, empirical models were established and validated by experimental data from the literature. Furthermore, with the help of nuclear magnetic resonance (NMR) technology, the water-state change was revealed by detecting the T₂ spectra of NM-treated samples, and finally the stabilizing mechanism of NM was discussed.

2. Materials and Methods

2.1. Raw Materials

The materials utilized in the current research included loess soil, NM, and distilled water (Figure 1). The loess soil was collected from a depth of 2.0~2.5 m beneath the ground of a high-speed rail construction site in Xi’an, China (Figure 2). It was mustard yellow in color and was classified as low-plasticity clay (CL) according to the unified classification system.

Table 1. Physical indexes of the loess soil.

Properties	Index		Value
Physical index	Specific gravity	(-)	2.64
	Plastic limit	(%)	20.5
	Liquid limit	(%)	34.2
	Plastic index	(-)	13.7
Grain size distribution	2~0.075 mm	(%)	12
	0.075~0.002 mm	(%)	71
	<0.002 mm	(%)	17
Mineral components	Quartz	(%)	42
	Feldspar	(%)	32
	Muscovite	(%)	17
	Montmorillonite	(%)	5
	Illite	(%)	4

lists the physical properties determined as per ASTM D422-63, D854, and D4318 [39,40,41]. The mineral components based on the X-ray diffractometer test were also provided. The NM material was a white, nano-scale, high-purity MgO powder procured from a firm in Guangzhou, Guangdong Province, China. The physical and chemical properties are presented in

Table 2. Chemical and physical properties of the nano-MgO used.

Chemical Composition		Physical Properties
Formula	Concentration	Property	Value
MgO	99.9	Mean particle size	40 nm
CaO	0.004	Melting point	2850 °C
Fe2O3	0.002	Boiling point	3600 °C
TiO2	0.003	Bulk density	0.74 g/cm3
Other	0.001	Specific surface area	40 m2/g

.

2.2. Consistency Limits and Compaction Test

According to trial tests and previous studies, the NM content ($C_{\mathrm{NM}}$) was selected as 0%, 1%, 2%, 3%, and 4% (on dry weight basis). To prepare the NM–loess mixture, the air-dried loess was first grinded and passed through the 0.5 mm sieve, and then oven-dried with NM powder at 105 °C for 24 h. To produce a homogeneous mixture, the loess was first divided into small quantities and preliminarily mixed with a pre-specified amount of NM. Then, the small portions were combined and manually stirred until uniformity was ensured. After the mixture preparation and humectation (24 h), the standard compaction test (SCT) and consistency limits test (CLT) for all NM content were performed, following ASTM D4318 and D1557-12 [41,43]. Accordingly, the optimum water content (w_opt), maximum dry density (${\rho }_{\mathrm{dmax}}$), plastic limit (PL), and liquid limit (LL) were determined for varying NM dosages.

2.3. Unconfined Compression Test

For engineering practice, the samples for the unconfined compression test (UCT) were prepared with the w_opt and ${\rho }_{\mathrm{dmax}}$ for corresponding NM dosages.

Table 3. Mixing and curing design.

Test Item	Symbol	${\mathit{C}}_{\mathbf{NM}}$ (-)	w (-)	${\mathit{\rho }}_{\mathbf{d}}$ (g/cm3)	${\mathit{T}}_{\mathbf{c}}$ (day)
SCT	-	0%, 1%, 2%, 3%, 4%	-	-	-
CLT	-	0%, 1%, 2%, 3%, 4%	-	-	-
UCT	T01~T05	0%, 1%, 2%, 3%, 4%	wopt	${\rho }_{\mathrm{dmax}}$	1
	T06~T10				7
	T11~T15				14
	T16~T20				28
	T21~T25				42
NMR	T26~T30	0%, 1%, 2%, 3%, 4%	18.4%	1.61	28
	T31~T35	2%			1, 7, 14, 28, 42

Notes: $C_{\mathrm{NM}}$: NM content; w: water content; ${\rho }_{\mathrm{d}}$: dry density; $T_{\mathrm{c}}$: curing time; w_opt: optimum water content; ${\rho }_{\mathrm{dmax}}$: maximum dry density.

presented the mixing design and corresponding test items. To prepare these samples, the homogenous NM–loess mixture with predetermined moisture was packed in a 38 × 76 mm cylindrical mold in three layers. Before subsequent compaction, each layer’s surface was scraped to avoid the delamination effect. After being removed from the mold, the specimen was sealed with two layers of plastic film for constant moisture and then stored in a curing chamber under a constant temperature of 20 °C for 1, 7, 14, 28, and 42 days. For performing UCT, a strain-controlled YYW-2 unconfined compression apparatus (Nanjing Soil Instrument Company, Nanjing, China) was utilized (Figure 1). According to ASTM D2166 [44], the strain rate was selected as 1 mm/min and continued until a complete stress–strain response was achieved. Afterwards, some key mechanical parameters, including UCS, deformation modulus, and the strain at failure, could be determined.

2.4. Nuclear Magnetic Resonance Test

NMR is a non-destructive testing method that can be utilized to determine the state of water in unsaturated soils. In the T₂ spectrum, the relaxation time is inversely proportional to the surface tension, which is further related to the water state [45,46]. Thus, a clear idea about the bound-water and free-water could be obtained by analyzing the T₂ spectrum. In the present work, the equipment to realize NMR measurements was a 30 MHz Low-field (0.3 T) NMR spectrometer manufactured by Suzhou Niumag Analytical Instrument Corporation (Figure 1). Two groups of samples (T26~T30 and T31~T35) were prepared to reveal the effects of NM and curing, respectively. One group was cured by 28 days with varying NM contents from 0% to 4%; the other group was with 2% NM content and varying curing times. All the samples were prepared at a constant moisture of 18.4% and a dry density of 1.61 g/cm³ (the w_opt and ${\rho }_{\mathrm{dmax}}$ for the 2% NM case).

3. Results and Discussion

3.1. Consistency Limits

Figure 3 illustrates the variations of the plastic limit (PL), liquid limit (LL), and plasticity index (PI) with the addition of NM. Generally, both LL and PL increased with the addition of NM. The growth of PL was more remarkable than that of LL, therefore, resulting in a decreasing trend of PI with increasing NM. Specifically, as NM increased from 0% to 4%, PL and LL increased by 40.00% and 10.53%, respectively, while the corresponding PI decreased by 33.58%. This phenomenon suggested improved water-absorption ability, which could be attributed to the great specific surface area (SSA) and surface charge of NM particles. However, the reduction in PI with increasing fine-grained additives is interesting, but common in the design of stabilized soils [31,47]. In such situations, the decrease of PI is not due to the reduction of LL, but rather the comparatively higher growth of PL than that of LL.

3.2. Compactability

Figure 4 illustrates the standard compaction data of loess treated with 0%, 1%, 2%, 3%, and 4% NM. It can be seen that adding NM from 0% to 4% led to an increase in w_opt from 19.8% to 23.9%, yet a decrease in ${\rho }_{\mathrm{dmax}}$ from 1.64 g/cm³ to 1.57 g/cm³. Similar observations were also reported in previous works on soils treated with fine-grained stabilizer [31,48]. The increase in w_opt was probably due to the excellent water-absorption capacity of NM [49,50], and the change-down in ${\rho }_{\mathrm{dmax}}$ could be attributed to the following reasons: (1) NM particles having thicker bound-water films and lower specific gravity compared with that of soil particles; (2) the increased interparticle bonding force by adding NM, which prevented easy compaction. Therefore, the more the NM was added, the higher the w_opt was, and the lower the ${\rho }_{\mathrm{dmax}}$ was.

3.3. Mechanical Properties

3.3.1. Stress–Strain Behavior

Figure 5 presented the stress–strain response of NM-treated loess with varying NM contents and curing times. As NM content grew, the improvement in peak-strength was evident when NM content was lower than 2%. Yet, with further NM addition, a considerable drop in peak-strength was observed. For all cases, the peak-strength achieved its maximum at NM content of 2%. Moreover, the improving effect of NM was enhanced with increasing curing time; that is, the 2% NM content combined with 42 days of curing gave the best improving effect.

The shapes of stress–strain curves varied wildly with NM content. For the cases of natural loess, ductile behavior was manifested, with a gentle rise and mild decline of stress, respectively, in the pre- and post-peak phase. With the addition of NM (≤2%), both the peak-strength and growth rate rose, and the post-peak drop in stress became increasingly pronounced. In addition, the strain corresponding to peak-stress, ${\epsilon }_{\mathrm{f}}$, decreased gently with the addition of NM. As a result, the NM-treated samples exhibited much more brittle behavior than the untreated one. However, it was also found that excessive NM addition (>2.0%) would lead to declines in both strength and brittleness. As for the effects of curing, the longer the curing time was, the more brittleness the sample exhibited. Yet, the effects of curing on brittleness were less significant than that of NM.

3.3.2. Unconfined Compressive Strength

The contributions of NM content ($C_{\mathrm{NM}}$) and curing time ($T_{\mathrm{c}}$) to variations in UCS ($q_{\mathrm{u}}$) are illustrated in Figure 6. As expected, increasing the curing time considerably improved the UCS of NM-treated loess. For NM-treated loess, the gain in UCS was drastic in the early 28 days, and thereafter, it became less prominent as the average increment in $q_{\mathrm{u}}$ with $T_{\mathrm{c}}$ from 28 to 42 was just by 3.8%. So, for NM-treated loess, experiencing 28 days of curing is recommended. In this condition, the $q_{\mathrm{u}}$ of 0%, 1%, 2%, 3%, and 4% NM cases increased by 14.0%, 22.8%, 42.7%, 47.4%, and 41.1%, respectively, compared with the corresponding 1-day cured cases. It is noteworthy that, for NM-treated samples, the curing-induced growth in $q_{\mathrm{u}}$ was up to 47.4%, which was 2.386 times higher than that of the untreated one (14.0%). This indicated the strong strength growth potential of NM-treated loess after an appropriate length of curing. Generally, the improving effects of curing can be interpreted by the thixotropy effect resulting from bound-water structure change and chemical and electrical transport during the curing period [51].

The effect of NM on UCS is also observed in Figure 6. In all cases, the UCS followed a “rise-fall” path with the addition of NM, reaching its maximum at 2% NM content. To be specific, at 28 days of curing, NM-treated samples of 1%, 2%, 3%, and 4% attained UCS values of 0.4116, 0.5196, 0.4937, and 0.3934 MPa, respectively, which is 36.14%, 71.83%, 63.28%, and 30.09% higher, respectively, than that of the untreated one (0.3023 MPa). So, about 72% UCS gain was to be expected with 2% NM addition and 28 days of curing, which was suggested in engineering practice. On the other hand, it is interesting to note that adding more than 2% NM had an adverse effect on UCS. This undesirable impact of adding excessive nano-materials were also identified in previous investigations on soil stabilization [23,31,36,52].

As discussed above, the NM content $C_{\mathrm{NM}}$ and curing time $T_{\mathrm{c}}$ were two crucial factors controlling the UCS of NM-treated loess. In this part, the following function was attempted to describe the relationship between UCS and these factors:

$$q_{\mathrm{u}}=q_{\mathrm{u}\text{-}0\text{-}1}\cdot \left(a_1\cdot C_{\mathrm{NM}}^2+b_1\cdot C_{\mathrm{NM}}+1/b_2\right)\times \left(a_2\cdot \ln \left(T_{\mathrm{c}}\right)+b_2\right)$$

(1)

where $q_{\mathrm{u}}$, $C_{\mathrm{NM}}$, and $T_{\mathrm{c}}$ was expressed in MPa, %, and day, respectively; $a$ and $b$ were the regression coefficients; and the subscript 1 and 2 indicated affected by $C_{\mathrm{NM}}$ and $T_{\mathrm{c}}$, respectively. $q_{\mathrm{u}\text{-}0\text{-}1}$ was the estimated $q_{\mathrm{u}}$ value for the 0% NM and 1 day-cured condition.

In Equation (1), the $q_{\mathrm{u}}$ of NM-treated loess included three parts: the standard UCS of untreated loess represented by $q_{\mathrm{u}\text{-}0\text{-}1}$, the contribution of NM exhibited by a quadratic function, and the contribution of curing time presented by a logarithmic function. The parameters in this equation were determined from data fitting, i.e., $q_{\mathrm{u}\text{-}0\text{-}1}=0.2312$, $a_1=-1.641$, $b_1=7.244$, $a_2=8.837\times {10}^{-3}$, and $b_2=6.672\times {10}^{-2}$. Figure 7 illustrates the fitting results compared with the experimental data. It shows that the predicted results matched the experimental data well, achieving a fine determination coefficient of 0.906, which indicates that Equation (1) can be confidently used in determining the UCS of NM-treated loess.

3.3.3. Strain at Failure

Figure 8 presents the strain at failure ${\epsilon }_{\mathrm{f}}$ with various NM contents and curing times. It was observed that most ${\epsilon }_{\mathrm{f}}$-values distributed in the range of 2.4% to 6.4%. Adding NM to the natural loess led to a decline in ${\epsilon }_{\mathrm{f}}$ at low NM levels (≤2%). Thereafter, with further NM addition, an ascending trend of ${\epsilon }_{\mathrm{f}}$ was observed. Meanwhile, ${\epsilon }_{\mathrm{f}}$ generally decreased with increasing curing time. In this way, ${\epsilon }_{\mathrm{f}}$ reached its minimum (2.4%) at 3% NM content and 42 days of curing. It is noteworthy that the changing tendency of ${\epsilon }_{\mathrm{f}}$ was just the opposite of the trend of $q_{\mathrm{u}}$, indicating a probable correlation between them.

Figure 9 presents the trend of ${\epsilon }_{\mathrm{f}}$ with $q_{\mathrm{u}}$ for NM-treated loess. Generally, ${\epsilon }_{\mathrm{f}}$ presented an approximately linear correlation with $q_{\mathrm{u}}$. With a regression analysis, the ${\epsilon }_{\mathrm{f}}$-$q_{\mathrm{u}}$ relationship was well expressed by the following linear function:

$${\epsilon }_{\mathrm{f}}=\kappa \cdot q_{\mathrm{u}}+\zeta ,$$

(2)

where ${\epsilon }_{\mathrm{f}}$ and $q_{\mathrm{u}}$ were respectively expressed in % and MPa. $\kappa$ and $\zeta$ were empirical constants equal to −11.1 and 8.4585, respectively. The fair determination coefficient, $R^2=0.843$, indicated that Equation (2) could be utilized in characterizing the correlation between ${\epsilon }_{\mathrm{f}}$ and $q_{\mathrm{u}}$. Furthermore, the fair fitting results also suggested that the ${\epsilon }_{\mathrm{f}}$–$q_{\mathrm{u}}$ relation of NM-treated loess was not sensitive to NM and curing, which brought convenience in engineering application.

Figure 10 exhibits the data of nano-SiO₂ (NS)- and nano-MgO (NM)-treated clays extracted from the literature [22,31,37,53]. Twelve pairs of ${\epsilon }_{\mathrm{f}}$ and $q_{\mathrm{u}}$ were fitted according to Equation (2), and the fitting results are exhibited in this figure. It was observed that ${\epsilon }_{\mathrm{f}}$ changed with $q_{\mathrm{u}}$ as a linear function in almost every case. The higher determination coefficient, R², indicated the linear expression may be suitable to describe the ${\epsilon }_{\mathrm{f}}$–$q_{\mathrm{u}}$ relationship of nanomaterial-treated soils. Notably, the parameter $\kappa$ was negative for the cases where only NS or NM and cement were added. It was consistent with the observation of NM-treated loess in this work. However, as fibers were added, it was interesting to note that the parameter $\kappa$ turned positive. It indicated that adding fiber impacted not only on the strength, but also on the ${\epsilon }_{\mathrm{f}}$–$q_{\mathrm{u}}$ relationship. This phenomenon might be attributed to the good ductility of fiber-treated soil, which endowed them with larger failure strain and corresponding higher UCS. In terms of $\zeta$, it is positive in most cases, just the same as the results in this work. This is in accordance with the theoretical analysis, because UCS cannot be reached at 0% strain. For the cases of Ahmadi et al., the presence of negative $\zeta$ value indicated the obtained expression was not applicable in the lower UCS level.

3.3.4. Deformation Modulus

As previously discussed, the NM-treated loess exhibited elastic–plastic behavior like most geo-materials. Normally, this behavior can be described by the deformation modulus, $E_{50}$, which is expressed as the ratio of stress to strain when the strain is at half of ${\epsilon }_{\mathrm{f}}$. Figure 11 illustrated the effects of NM and curing on the deformation modulus. Generally, the deformation modulus of NM-treated loess was in the range of 7.33 to 31.9 MPa. It presented a rising tendency with increasing curing time and a “rise-fall” tendency with the addition of NM. Notably, these tendencies were similar to that of UCS in Figure 6, which indicated a probable link between $E_{50}$ and $q_{\mathrm{u}}$.

Through reviewing existing research, the $E_{50}$–$q_{\mathrm{u}}$ relationship was generally described by a simple linear function, i.e., $E_{50}=\eta \cdot q_{\mathrm{u}}$. Here, $\eta$ was a fitting parameter, and diverse ranges of $\eta$ have been provided for various cement-treated clays [54,55,56]. As for NM-treated loess, Figure 12 exhibited the variation of $E_{50}$ versus $q_{\mathrm{u}}$ of samples at various test conditions. It demonstrated that the $E_{50}$ of NM-treated loess was proportional to $q_{\mathrm{u}}$, but in a nonzero-intercept way. Therefore, their relationship could be described as,

$$E_{50}=\eta \cdot q_{\mathrm{u}}+\beta$$

(3)

where both $E_{50}$ and $q_{\mathrm{u}}$ were expressed in MPa. The regression analysis yielded $\eta =72.0$ and $\beta =-12.813$ with $R^2=0.818$. It is noteworthy that the intercept $\beta$ for NM-treated loess was negative, unlike the traditional zero-intercept model for cement-treated clays. This was probably due to the relatively weak strength and high ductility of the bonding force formed by NM, compared with that formed by cement. Consequently, the NM-treated loess exhibited high strength under relatively low stiffness, resulting in the negative intercept $\beta$.

Figure 13 presents the $E_{50}$–$q_{\mathrm{u}}$ relationships of nanomaterial-stabilized clays extracted from literature [22,31,37,53]. Based on Equation (3), information about the parameters $\eta$ and $\beta$ could be obtained together with the determination coefficient $R^2$: (1) For the NS (0~4%) stabilized clay with $E_{50}$ of 3~38 MPa and $q_{\mathrm{u}}$ of 0.4~2.1 MPa, the fitting parameter $\eta$ and $\beta$ were 54.1 and −10.83, respectively; (2) For the clay treated with NS (0~1%) and polyester fiber (PF) (0~0.5%), the $\eta$ and $\beta$ were considered to be 16.4 and −3.1473, respectively; (3) for the NM (0~1.5%) and cement (20%) treated clay, the $\eta$ was in a range of 23.1~170.1, and $\beta$ was in a range of −0.5658~−32.159; (4) for NM (0~1%) and polypropylene fiber (PP) (0~1%) treated clay, the parameters $\eta$ of 9.2~16.3 and $\beta$ of −0.6784~−5.5391 were determined. For most cases, the resulting $R^2$-values are quite high, suggesting that the proposed equation was also applicable for other nanomaterial-treated soils. Furthermore, it is noteworthy that the fitting parameter $\beta$ was negative and $\eta$ was positive for all cases. This was consistent with the results of this work, indicating it might be a common phenomenon in nanomaterial-based stabilization techniques.

3.4. Stabilization Mechanism

3.4.1. Effect of Nano-MgO

Figure 14 depicts the T₂ spectra of NM-treated loess with varying NM dosages. Since the relaxation time is inversely proportional to the surface tension, the T₂ spectrum curve could be used to reflect the states of water in soil samples. According to He et al. [57], for loess soil, the cutoff T₂ value corresponding to bound-water and free-water was at 1.65 ms (red line). That is, the left/right side area under the T₂ curve reflected the amount of bound-/free-water, respectively. As exhibited in Figure 14, the curves for all cases could be roughly divided into two peaks, P₁ and P₂, which were respectively located in the left side and right side of the red line. Notably, the left side peak P₁ was dominating, which denoted that most of the water contained in the soil belonged to bound-water. An increase in NM content led to an increase in P₁, yet a decrease in P₂. The left-shift of the T₂ curve with the addition of NM was also noticed. The above phenomena indicated the rising share of bound-water with increasing NM and curing. It could be attributed to the excellent surface effect and hydrophilicity of MgO nano-particles, which attracted more water molecules to make the proportion of bound-water enhanced [58].

Based on the above discussion and literature review, the improving mechanism of NM was primarily due to the water-absorbing and cementing effect of NM. As for the water-absorbing effect, part of the free-water transforming to bound-water would lead to a rise in the internal friction angle and a decline in the interparticle cohesion of loess soil at a given water content [59]. However, the cementing effect mainly enhanced the interparticle cohesion. In view of the unconfined compression test condition, the UCS and deformation modulus of the soil depended more on the cohesion than on the internal friction angle. Therefore, as the NM content increased, the UCS and modulus presented a rising trend when the NM dosage was small (less than 2%), because, in this condition, the improvement in cohesion induced by the cementing effects was dominating. However, as the NM content was larger than 2%, the unfavorable effects of water-absorbing became evident, resulting in the decline of cohesion and further decreasing of the UCS and deformation modulus.

3.4.2. Effects of Curing Time

The T₂ curves of samples with 2% NM at varying curing periods are plotted in Figure 15. Similarly, the samples mainly consisted of two peaks, P₁ and P₂. With increasing curing time, the T₂ curve shifted towards a higher signal in P₁ and a lower signal in P₂. Further, the left shift towards lower relaxation time was also observed. The above phenomena suggested the water transformation from free-water to bound-water with curing. It indicated a moisture redistribution procedure, which was resulting from the comparatively strong absorbability of NM than that of soil particles. This is part of the reason for the improving effects of curing on the mechanical properties of NM-treated loess. Another reason is that the cementing effect of NM could be enhanced with curing. Thus, the UCS and deformation modulus increased with curing time. Moreover, it was found that the variation of the T₂ curve during the curing period was evident at the early 14 days and became not obvious after 28 days of curing. This was in accordance with the results obtained by the unconfined compression test.

4. Conclusions

This paper investigated the geotechnical properties and water-state evolution of loess with varying NM content and curing times. The UCS, strain at failure, and deformation modulus of NM-treated loess were quantitively analyzed. In addition, the improving mechanisms of NM and curing were unfolded from the aspect of water-state changes. The main findings can be summarized as follows:

• Adding NM to loess soil led to increases in the plastic limit (from 20.5% to 28.7%), liquid limit (from 34.2% to 37.8%), and optimum water content (from 19.8% to 23.9%), but declines in the plasticity index (from 13.7% to 9.1%) and maximum dry density (from 1.64 g/cm³ to 1.57 g/cm³).
• NM and curing were two crucial factors influencing the mechanical properties of NM-treated loess. The improvement in UCS with curing was evident in the early 28 days and became insignificant thereafter. As NM content increased, the UCS followed a “rise-fall” path (from about 0.3023 to 0.5196 to 0.3934 MPa for 28d curing), indicating an optimum NM dosage at 2%. So, 2% NM-treated loess with 28 days of curing is accordingly suggested, and about 72% UCS gain is to be expected in this condition.
• The strain at failure decreased at first and then increased with the addition of NM. However, the deformation modulus presented an opposite trend. Both of them exhibited a linear relationship with UCS. Empirical models for them were established and validated by literature data.
• The stabilizing mechanism of NM-treated loess was explored from the aspect of water-state change. The enhancement of mechanical properties primarily was due to the water-absorbing and cementing effects of NM. The former caused water transformation from free-water to bound-water inside the soil and enhanced the interparticle cohesion as potently as the cementing effect did.