Mechanism of MgCl₂-Regulated Hydration, Pore-Structure Evolution, and Strength Development in Cement-Stabilized Loess

Abstract

To investigate the engineering performance and microstructural evolution of magnesium chloride–cement composite-stabilized loess, laboratory tests were conducted on loess treated with different cement contents and MgCl₂ dosages. Compaction tests, unconfined compressive strength (UCS) tests after 7 and 28 days of curing, X-ray diffraction (XRD), scanning electron microscopy (SEM), and nuclear magnetic resonance (NMR) analyses were performed to establish the relationship among hydration products, pore evolution, and strength development. The results showed that MgCl₂ increased the maximum dry density and reduced the optimum moisture content, thereby improving the compactability of cement-stabilized loess. The UCS exhibited a non-monotonic response to increasing MgCl₂ dosage, with an initial decrease followed by an increase and a subsequent decline. The optimal MgCl₂ content was 1.7% for 8% cement and 1.8% for both 10% and 12% cement. Integrated XRD, SEM, and NMR analyses revealed that an appropriate MgCl₂ dosage regulates the hydration and microstructural evolution of cement-stabilized loess through the formation of Mg-bearing reaction products, enhanced interparticle bonding, and refinement of the pore structure. The resulting reductions in total porosity and the proportion of medium and large pores were closely associated with the observed strength enhancement, establishing a clear hydration–pore structure–strength relationship. Although the 12% cement–1.8% MgCl₂ mixture achieved the highest strength, the 10% cement–1.8% MgCl₂ mixture provided a more favorable balance between mechanical performance and cement consumption. These findings provide mechanistic insight into MgCl₂-regulated hydration and pore-structure evolution and support the optimized use of MgCl₂ as an auxiliary modifier for cement-stabilized loess.

1. Introduction

Loess is a widely distributed aeolian deposit characterized by a loose structure, high porosity, and metastable particle arrangement. Owing to its strong sensitivity to water infiltration and external loading, loess often exhibits significant collapsibility, strength degradation, and deformation instability, which may threaten the safety and durability of infrastructure constructed in loess regions. Houston et al. [1] investigated the collapse behavior of soils caused by wetting and emphasized the importance of controlling moisture-induced deformation. Derbyshire [2] reviewed geological hazards associated with loess terrains and highlighted the engineering risks associated with the instability of loess deposits. Dijkstra et al. [3] further analyzed the engineering properties and slope stability of Chinese loess, demonstrating the necessity of improving its mechanical performance for safe infrastructure construction.

Cement stabilization is one of the most widely used methods for improving the engineering properties of loess. Cement hydration generates calcium silicate hydrate (C-S-H), calcium aluminate hydrate (C-A-H), and calcium aluminosilicate hydrate (C-A-S-H), which enhance particle bonding and fill interparticle pores. Axel et al. [4] demonstrated that the strength improvement of cement-stabilized loess was closely associated with increased particle bonding and reduced pore connectivity. Yuan et al. [5] further related the evolution of mechanical properties to changes in the microstructure during cement hydration. Studies incorporating silica micropowder and other curing agents have also shown that additional reactive components can promote hydration or secondary reactions and refine the pore structure [6,7]. Nevertheless, increasing cement content does not necessarily provide a proportionate improvement in engineering performance and is accompanied by greater cement consumption and associated environmental burdens. These limitations have motivated the use of additional stabilizing materials and chemical modifiers.

Previous modification approaches can broadly be interpreted in terms of two complementary mechanisms: enhancement of cementitious reactions and regulation of the soil–binder microstructure. Pozzolanic materials such as rice husk ash and silica fume can consume calcium-bearing hydration products and generate additional cementitious phases, thereby contributing to pore refinement and strength development [8,9,10]. In contrast, chemical and organic modifiers such as calcium lignosulfonate and lignin–lime systems can alter particle interactions, aggregation, and interparticle bonding [11,12]. Although these approaches have demonstrated improvements in strength or microstructure, their effectiveness depends strongly on additive chemistry, soil mineralogy, binder content, and curing conditions. More importantly, much of the existing work has focused on individual macroscopic or microscopic indicators; observations of strength enhancement or pore refinement alone do not fully identify the reaction products responsible for these changes or explain how hydration evolution is quantitatively associated with pore restructuring and mechanical development. Consequently, an integrated interpretation linking reaction chemistry, pore evolution, and macroscopic performance remains necessary for composite-stabilized loess.

Magnesium chloride (MgCl₂) has attracted attention as a potential chemical modifier because of its hygroscopicity, high solubility, and ability to alter ionic interactions within soils and cementitious systems. Previous studies, however, have examined MgCl₂ under substantially different material conditions. Soltani-Jigheh et al. [13] reported improvements in highly plastic swelling clay associated with microstructural modification, whereas Siddiqua and Bigdeli [14] showed that MgCl₂ treatment could reduce soil collapse potential by improving particle interactions. Yeganeh Rikhtehgar and Teymür [15] further reported improved freeze–thaw behavior of treated clay. These findings demonstrate the potential of MgCl₂ for soil modification, but they primarily concern untreated or non-Portland-cement soil systems. Conversely, studies of cementitious materials have shown that Mg²⁺ can directly interact with hydration products. Liu et al. [16] demonstrated that magnesium exposure alters the chemical composition and stability of C-S-H, while Liu et al. [17] reported that MgCl₂ affects the hydration and strength development of magnesium oxychloride cement-solidified soft clay. Thus, existing studies provide evidence for both soil-scale physicochemical modification and cement-scale hydration reactions, but the coupling of these effects in MgCl₂-modified Portland-cement-stabilized loess remains insufficiently understood.

Although MgCl₂ has shown potential for soil modification, its incorporation into Portland-cement-stabilized soil requires careful dosage control. Prolonged exposure of hardened cementitious materials to MgCl₂-containing environments can induce calcium leaching, destabilize C-S-H, and promote the formation of Mg-bearing phases such as M-S-H, ultimately resulting in microstructural deterioration and strength loss. Hara et al. [18] investigated the strength reduction mechanism of cement-treated soil under seawater exposure and attributed the deterioration primarily to the transformation of C-S-H into M-S-H, providing direct evidence of the potentially detrimental effect of prolonged magnesium exposure on cement-treated soils. However, such long-term external chemical attack differs from the controlled incorporation of a limited MgCl₂ dosage during the preparation of a soil–cement composite. At an appropriate dosage, Mg²⁺ may modify particle interactions and participate in Mg-bearing reaction products, while the associated microstructural changes may promote denser particle packing and pore refinement. In contrast, excessive Mg²⁺ and Cl⁻ can alter the stability of calcium-bearing hydration phases and interfere with the normal development of cementitious products. Han et al. [19] showed that MgCl₂ affects the early strength development of cement-treated soil, whereas studies of cementitious systems have demonstrated that prolonged magnesium exposure can destabilize C-S-H and alter its composition [13,18,20]. Therefore, MgCl₂ was selected in the present study not on the assumption that its effect is uniformly beneficial, but because its dosage-dependent dual effect provides an opportunity to identify the transition between beneficial physicochemical modification and detrimental alteration of cement hydration products. However, previous investigations have generally considered either strength behavior, cement hydration chemistry, or soil microstructure separately. It therefore remains unclear how variations in MgCl₂ dosage simultaneously affect hydration products, pore-size redistribution, and strength development in cement-stabilized loess, and whether the observed optimum dosage can be explained by the competition between beneficial microstructural densification and unfavorable alteration of hydration products.

The use of chloride-containing additives also requires consideration of potential durability and environmental limitations. Because MgCl₂ is highly soluble, repeated water infiltration or long-term exposure to moisture may promote ionic migration or leaching, potentially reducing the persistence of salt-related modification effects. Moreover, prolonged exposure of Portland cementitious phases to Mg²⁺ may promote C-S-H decalcification and the formation of Mg-bearing phases with different binding characteristics, which could adversely affect long-term mechanical stability [13,20]. From an environmental perspective, the mobility of chloride ions raises concerns regarding their transport into surrounding soils and water systems, particularly under conditions of substantial infiltration or runoff. Chloride-containing materials may also require additional attention where stabilized soils are used in proximity to reinforced concrete or metallic infrastructure because of the potential corrosion risk associated with chloride migration. Therefore, the engineering benefits of MgCl₂ should be evaluated together with dosage control, exposure conditions, long-term durability, and environmental compatibility. These issues also indicate that short-term strength improvement alone is insufficient for evaluating the broader applicability of MgCl₂-based stabilization.

The mechanical behavior of cement-stabilized soils is fundamentally governed by the combined evolution of hydration products and pore structure. Bernal et al. [21] highlighted the importance of gel structure in controlling the mechanical performance of cementitious materials, while Glasser et al. [20] emphasized the role of chemical reactions in the long-term degradation of cement-based materials. Scrivener et al. [22] systematically described Portland cement hydration and its relationship with material properties, and Taylor [23] established the fundamental characteristics of major cement hydration products. For cement-treated soils, Horpibulsuk et al. [24,25] and Muntohar further demonstrated that binder content and hydration development strongly influence strength evolution. These findings indicate that macroscopic strength cannot be interpreted independently of the evolution of cementitious phases and the resulting microstructural rearrangement.

Accordingly, complementary microstructural techniques are necessary to establish this relationship. SEM can directly characterize particle contacts, cementitious bonding, and local pore morphology, and quantitative image analysis can provide apparent pore-structure parameters [26,27]. However, two-dimensional SEM observations alone cannot represent the pore system of an entire specimen. Diamond [28] similarly emphasized that individual pore-characterization techniques may introduce method-dependent limitations. Combining phase identification by X-ray diffraction (XRD), local morphological characterization by scanning electron microscopy (SEM), and whole-specimen pore characterization by nuclear magnetic resonance (NMR) therefore provides a more comprehensive basis for relating hydration products and pore evolution to macroscopic mechanical behavior.

Against this background, the present study investigates MgCl₂–cement composite stabilization of Lanzhou loess with particular emphasis on the coupled relationship among hydration products, pore-structure evolution, and strength development. Unlike previous studies that have mainly considered these responses separately, compaction and unconfined compressive strength (UCS) tests are integrated with XRD, SEM, and NMR analyses to determine how MgCl₂ dosage affects engineering properties, Mg-bearing reaction products, particle bonding, total porosity, and pore-size distribution. By comparing different cement contents and MgCl₂ dosages, the study seeks to identify the dosage-dependent transition between beneficial modification and excessive-addition effects and to establish a hydration–pore structure–strength framework for explaining the resulting mechanical behavior. The results provide a mechanistic basis for optimizing MgCl₂ dosage in cement-stabilized loess. Because the present investigation focuses primarily on curing-stage behavior up to 28 days, long-term durability, chloride leaching, and environmental compatibility remain important subjects for further investigation.

2. Materials and Methods

2.1. Materials

2.1.1. Test Loess

The loess used in this study was collected from a site in Lanzhou New Area, China. The soil was brownish yellow in color and was classified as sandy soil. After collection, the loess was passed through a 5 mm sieve and then oven-dried at 105 °C to a constant mass. The dried soil was stored for subsequent testing. The basic physical properties of the loess are presented in

Table 1. Basic physical properties of the test loess.

Natural Density, ρ (g/cm3)	Natural Water Content, w (%)	Liquid Limit, wL (%)	Plastic Limit, wP (%)	Plasticity Index, IP (%)	Maximum Dry Density, ρd,max (g/cm3)	Optimum Moisture Content, wopt (%)
1.68	1.8	25.1	19.1	6.0	1.86	18.6

. In addition, the mineralogical composition of the untreated loess was characterized by X-ray diffraction (XRD) prior to the addition of cement and MgCl₂. The XRD results indicated that quartz and calcite were the principal crystalline phases, together with a possible minor gismondine phase. Quartz constituted the primary siliceous mineral framework of the loess, whereas calcite represented an indigenous Ca-bearing carbonate phase. These initial mineral constituents provide the mineralogical basis for interpreting the subsequent hydration reactions and the evolution of Mg-containing products in the cement–MgCl₂–loess system. The mineral phases of the raw loess are summarized in

Table 2. Mineral phases identified in the untreated Lanzhou loess by XRD.

Mineral Phase	Chemical Formula	Chemical Formula	Chemical Formula
Quartz	SiO2	46-1045	P3221
Calcite	CaCO3	05-0586	R-3c
Gismondine	CaAl2Si2O8·4H2O	20-0452	P21/c

.

2.1.2. Magnesium Chloride

The magnesium chloride used in this study was obtained from The magnesium chloride used in this study was obtained from Weifang Haizhiyuan Chemical Co., Ltd., Weifang, Shandong, China. The material had a purity of at least 98% and appeared as white granular particles. It has a density of 2.323 g/cm³ and a relative molecular mass of 95.21. Magnesium chloride is highly hygroscopic, readily soluble in water, and soluble in ethanol. Its aqueous solution is weakly acidic, with a pH ranging from approximately 5.5 to 6.5.

2.1.3. Cement

The cement used in this study was commercially available P·O 42.5 ordinary Portland cement (Saima brand) produced by Sinoma Gansu Cement Co., Ltd., Baiyin, China, and conformed to GB 175-2023. Its main physical and chemical properties are summarized in

Table 3. Physical and chemical properties of the P·O 42.5 ordinary Portland cement.

Property	Value	Property	Value
Specific surface area (m2/kg)	345.7	Magnesium oxide content (%)	2.240
Soundness	Qualified	Sulfur trioxide content (%)	1.160
Initial setting time (min)	180	Chloride ion content (%)	0.024
Final setting time (min)	240	Loss on ignition (%)	4.23
3-day compressive strength (MPa)	25.6	28-day compressive strength (MPa)	52.8

. The cement had a specific surface area of 345.7 m²/kg and qualified soundness, with initial and final setting times of 180 and 240 min, respectively. The 3-day and 28-day compressive strengths were 25.6 and 52.8 MPa, respectively, indicating satisfactory early- and later-age strength development. The MgO and SO₃ contents were 2.240% and 1.160%, respectively, while the chloride ion content and loss on ignition were 0.024% and 4.23%, respectively. These properties provide the basic material characteristics relevant to the hydration behavior and strength development of the cement-stabilized loess.

2.2. Experimental Program and Procedures

A full-factorial experimental program was designed to determine the optimum mixture proportion of magnesium chloride and cement and to clarify the influence of magnesium chloride dosage under different cement contents. The experimental procedures were performed in accordance with the Code for Soil Test of Railway Engineering (TB 10102-2023) [29]. The experimental matrix is summarized in

Table 4. Experimental matrix and characterization program.

Cement Content (% Dry Loess Mass)	MgCl2 Dosage (% Cement Mass)	Compaction	Specimens Selected for XRD/SEM/NMR	UCS Curing Age
8	0, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9	Conducted for all mixtures	MgCl2-free and optimum MgCl2 dosage	7 d, 28 d
10	0, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9	Conducted for all mixtures	MgCl2-free and optimum MgCl2 dosage	7 d, 28 d
12	0, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9	Conducted for all mixtures	MgCl2-free and optimum MgCl2 dosage	7 d, 28 d

. Three cement contents (8%, 10%, and 12% of the dry loess mass) were combined with MgCl₂ dosages ranging from 0 to 1.9% of the cement mass. Compaction and UCS tests were conducted for all mixtures, with UCS evaluated after 7 and 28 days of curing. The MgCl₂-free reference specimens and the mixtures corresponding to the optimum MgCl₂ dosage at each cement content were further selected for XRD, SEM, and NMR characterization after 28 days of curing.

The cement contents of 8%, 10%, and 12% were calculated as percentages of the dry loess mass, whereas the MgCl₂ dosage was calculated as a percentage of the cement mass. For each mixture, the amount of mixing water was determined according to the corresponding optimum moisture content obtained from the compaction test. The optimum moisture contents of the investigated mixtures ranged from 12.60% to 14.10%. The required MgCl₂ was completely dissolved in the prescribed mixing water before being mixed with the dry loess and cement; therefore, the water used to prepare the MgCl₂ solution was included in, rather than added to, the total mixing-water content.

2.2.1. Compaction Test

To determine the maximum dry density and optimum moisture content of the stabilized soils with different mixture proportions, compaction tests were conducted in accordance with the Code for Soil Test of Railway Engineering (TB 10102-2023) [29]. A 102 mm compaction mold with a volume of 947 cm³ was used. The Z1 heavy compaction method was adopted, in which each specimen was compacted in five layers, with 25 blows applied to each layer. After compaction, two representative soil samples were taken from each test point and placed in aluminum boxes. The samples were then oven-dried at 105 °C to a constant mass, and the corresponding moisture content, dry density, maximum dry density, and optimum moisture content were calculated.

2.2.2. Unconfined Compressive Strength Test

The stabilized loess specimens were prepared according to the designed mixture proportions, with the target dry density and water content determined from the compaction test results. After mixing, the materials were sealed in plastic bags and conditioned for no less than 12 h to allow a more uniform distribution of moisture within the soil. The conditioned mixtures were then statically compacted into cylindrical specimens measuring 50 mm in diameter and 50 mm in height. After molding, each specimen was sealed in a black plastic bag, labeled with its mixture proportion and curing age, and transferred to a standard curing room maintained at 20 °C and a relative humidity of at least 98%. For each mixture proportion, six replicate specimens were prepared in accordance with the specification requirements. Unconfined compressive strength (UCS) tests were conducted after 7 and 28 days of curing using a TC-200F automatic pavement material testing machine (Beijing Tianchang Tongda Instrument Co., Ltd., Beijing, China). The specimens were loaded axially under displacement control at a constant rate of 1.0 mm/min, corresponding to an axial strain rate of approximately 2%/min for the 50 mm-high specimens. Axial load and displacement were continuously recorded during loading, and the maximum axial stress obtained from the stress–strain response was taken as the UCS. The mean strength of the six replicate specimens was reported as the UCS of the corresponding mixture. For each mixture proportion, six replicate specimens were prepared in accordance with the specification requirements. Unconfined compressive strength tests were conducted after 7 and 28 days of curing. The testing procedure and determination of the representative UCS value were performed in accordance with the Specification for Soil Test of Railway Engineering (TB 10102-2023). The representative UCS value for each mixture proportion and curing age was determined from the six replicate test results according to the specification requirements.

2.2.3. SEM Sample Preparation and Image Acquisition

After the unconfined compressive strength tests, representative fragments of the stabilized loess specimens were collected for scanning electron microscopy (SEM) analysis. To preserve the original microstructure as much as possible, block samples with dimensions of approximately 5 mm × 5 mm × 5 mm were selected from the interior region near the failure surface, while areas affected by obvious macroscopic cracking, edge damage, or severe compression disturbance were avoided. The selected samples were immersed in anhydrous ethanol to terminate cement hydration and facilitate dehydration and were subsequently placed in a fume hood to remove residual ethanol and moisture. The dried samples were carefully polished using 1200-grit abrasive paper to obtain relatively flat observation surfaces and were then fixed onto sample stubs using conductive adhesive. Gold coating was applied using a Gold coating was applied using a GVC-1000 magnetron ion sputter coater (KYKY Technology Co., Ltd., Beijing, China) to improve surface conductivity.to improve surface conductivity. The coated specimens were examined using a The coated specimens were examined using a ZEISS GeminiSEM 500 field-emission scanning electron microscope (Carl Zeiss Microscopy GmbH, Jena, Germany).During SEM observation, several fields were initially examined at low magnification to identify representative internal regions and to avoid repeated imaging of the same area. Regions containing specimen edges, obvious preparation-induced damage, large mechanical failure cracks, poor focus, or significant imaging artifacts were excluded. SEM images were subsequently acquired at different magnifications and saved in TIFF format. For quantitative comparison of pore characteristics, images obtained at the same magnification (500×) were selected from representative regions of each specimen. The selected SEM images were analyzed using Image-Pro Plus 6.0 following the image-binarization procedure proposed by Feng et al. [27]. After grayscale segmentation and binarization, regions identified as pores or surface depressions were separated from the solid matrix. The number of discrete pore/depression regions and their total projected area were then quantified.

2.2.4. XRD Analysis

At curing ages of 7 and 28 days, the stabilized loess samples after the unconfined compressive strength tests were collected and ground to pass through a 0.080 mm round-hole sieve. An appropriate amount of powdered sample was weighed and packed into a grooved flat glass holder. Slight pressure was applied to ensure firm contact between the powder and the holder. To reduce the influence of preferred orientation, the powder samples were prepared without applying unidirectional compaction, so that the particles could maintain a relatively random arrangement. The mineralogical composition of the stabilized loess was determined using a The mineralogical composition of the stabilized loess was determined using a D8 ADVANCE X-ray diffractometer (Bruker AXS GmbH, Karlsruhe, Germany). The test was conducted in continuous scanning mode, with the scanning interval set at 5–80° and the scanning rate controlled at 5°/min. The obtained diffraction patterns were then imported into JADE 9.0 for phase identification. The relative peak intensities were evaluated mainly from the peak heights, and the potential mineral phases as well as hydration products in the stabilized loess were identified accordingly.

XRD phase identification was performed using JADE 9.0 by matching the measured diffraction patterns with reference PDF cards. Because some cement hydration products and Mg-bearing phases exhibit low crystallinity and overlapping diffraction peaks, unique phase assignment may involve uncertainty. Therefore, phase identification in this study was based on the combined consideration of characteristic diffraction peaks and the overall diffraction pattern rather than on a single peak. For weak or overlapping peaks, the identified phases were interpreted qualitatively and in conjunction with the known chemical composition of the cement–MgCl₂–loess system. No quantitative phase analysis was performed; therefore, variations in peak intensity were used only to discuss qualitative phase evolution rather than to determine absolute phase contents.

2.2.5. Nuclear Magnetic Resonance Test

After reaching the designed curing ages, the specimens were removed from the curing chamber and visually inspected. Only specimens with intact surfaces and without visible cracks, corner loss, or obvious artificial disturbance were selected for nuclear magnetic resonance (NMR) testing. Prior to testing, the specimens were vacuum-saturated with deionized water to improve the reliability of the NMR measurements. The specimens were placed in a vacuum saturation device, and residual air within the internal pores was removed under vacuum. While maintaining the vacuum condition, deionized water was slowly introduced until the specimens were completely submerged, and the vacuum state was maintained for a sufficient period to facilitate water penetration into the internal pore spaces. The chamber was then returned to atmospheric pressure, and the specimens were continuously immersed in deionized water for an additional 12 h to further improve the degree of saturation. After saturation, the specimens were removed, and free water on the surfaces was gently absorbed using moist filter paper to minimize interference from surface water films during NMR signal acquisition. The specimens were then immediately sealed with plastic film to reduce moisture loss before testing. NMR measurements were performed using a NMR measurements were performed using a MacroMR12-150H-I low-field nuclear magnetic resonance analyzer (Niumag Electronic Technology Co., Ltd., Shanghai, China). and imaging system. During testing, each sealed specimen was placed at the center of the radio-frequency coil to ensure stable signal acquisition. The transverse relaxation signals of the pore water were acquired and processed using the supporting analysis software to obtain the T₂ relaxation-time distribution. Under saturated conditions, the transverse relaxation behavior of pore water is primarily influenced by surface relaxation and can be approximately expressed as:

$${\displaystyle \frac{1}{T_2}}={\rho }_2\left({\displaystyle \frac{S}{V}}\right)$$

(1)

where ${\rho }_2$ is the transverse surface relaxivity and $S/V$ is the pore surface-to-volume ratio. Accordingly, shorter T₂ components are generally associated with water confined in smaller pores, whereas longer T₂ components correspond to relatively larger pores. The calibrated NMR signal was used to determine the total water-filled porosity, while the T₂ distribution was used to characterize variations in pore-size distribution among the different specimens. The processed data, including total porosity and pore-size distribution, were subsequently exported for comparative analysis.

3. Results

3.1. Compaction Behavior of Stabilized Loess

The test results for dry density and moisture content are shown in Figure 1a,b. The incorporation of magnesium chloride increased the maximum dry density of the cement-stabilized loess and reduced its optimum moisture content. For the specimens with 8% cement, the maximum dry density increased from 1.80 to 1.85 g/cm³ as the magnesium chloride content increased, whereas the optimum moisture content decreased from 14.10% to 12.80%. For the specimens with 10% cement, the maximum dry density generally ranged from 1.84 to 1.88 g/cm³, and the optimum moisture content decreased from 13.30% to 12.60%. For the specimens with 12% cement, the maximum dry density varied between 1.84 and 1.88 g/cm³, while the optimum moisture content decreased from 13.80% to 12.80%.

These results indicate that the combined incorporation of magnesium chloride and cement improved the compaction characteristics of loess to a certain extent. Compared with the 8% and 12% cement groups, the 10% cement group exhibited a relatively higher maximum dry density, a lower optimum moisture content, and more stable compaction behavior. Therefore, the 10% cement group showed better compaction performance than the 8% cement group and a more balanced compaction response than the 12% cement group.

3.2. Unconfined Compressive Strength of Stabilized Loess

Figure 2a presents the 7-day unconfined compressive strengths of the specimens prepared with different mixture proportions. For the specimens containing 8%, 10%, and 12% cement, the strength generally followed a decrease–increase–decrease pattern as the magnesium chloride content increased. At a magnesium chloride content of 1.3%, all three groups showed lower strengths than the corresponding cement-stabilized specimens without magnesium chloride, suggesting that a small MgCl₂ addition was not sufficient to improve early-age strength. With a further increase in MgCl₂ content, the strength began to recover and then reached a peak within a certain dosage range. In the 8% cement group, the strength increased to 1.750 MPa at 1.7% MgCl₂, which was 7.362% higher than that of the corresponding specimen without magnesium chloride. For the 10% and 12% cement groups, the maximum strengths were both obtained at 1.8% MgCl₂, reaching 2.297 and 2.318 MPa, respectively, with increases of 14.449% and 14.345% compared with their magnesium-chloride-free counterparts. However, when the MgCl₂ content was further raised to 1.9%, the strength of each group decreased again. These results indicate that MgCl₂ improved the early strength of cement-stabilized loess only within an appropriate dosage interval, rather than producing a continuous strengthening effect with increasing dosage. Overall, the cement content still played the leading role in controlling the 7-day strength, while a suitable amount of MgCl₂ provided an additional improvement in early-age mechanical performance.

Figure 2b shows the 28-day unconfined compressive strength results for specimens with different mixture proportions. In comparison with the 7-day values, the 28-day strengths were generally higher, suggesting that the stabilized loess continued to gain strength during later curing. For the 10% and 12% cement groups, the strength response to magnesium chloride addition showed a similar decrease–increase–decrease pattern. The highest strengths in these two groups were both obtained at 1.8% magnesium chloride, reaching 2.936 and 3.148 MPa, respectively; these values were 10.210% and 1.271% greater than those of the corresponding specimens without magnesium chloride. The 8% cement group showed a slightly different behavior. Its strength increased with increasing magnesium chloride content before reaching 2.245 MPa at 1.7% magnesium chloride, which represented an increase of 11.139% relative to the magnesium-chloride-free specimen. Further addition of magnesium chloride then caused a decline in strength. Combining the 7-day and 28-day results, the suitable magnesium chloride content can be identified as 1.7% for the 8% cement group and 1.8% for the 10% and 12% cement groups. Among all mixtures, the specimen with 12% cement and 1.8% magnesium chloride produced the highest strength, indicating the best overall mechanical performance. Nevertheless, the mixture with 10% cement and 1.8% magnesium chloride also achieved a relatively high strength while reducing cement consumption. Although the 12% cement–1.8% MgCl₂ mixture achieved the highest strength, the 10% cement–1.8% MgCl₂ mixture maintained high strength with lower cement consumption.

For the 8% cement group, different strength development characteristics were observed at curing ages of 7 and 28 days. At 7 days, the early-age enhancement effect of magnesium chloride was unstable for specimens with a low cement content. Except for the specimen with 1.7% magnesium chloride, the strengths of the other magnesium chloride-treated specimens were lower than that of the specimen without magnesium chloride. The strength of the control specimen was 1.630 MPa, whereas the maximum strength of 1.750 MPa was obtained at a magnesium chloride content of 1.7%, corresponding to an increase of 7.362%. However, when the magnesium chloride content further increased to 1.8% and 1.9%, the strengths decreased to 1.598 and 1.563 MPa, respectively. This result indicates that, at an early curing age, the strength of 8% cement-stabilized loess could be improved only within an appropriate magnesium chloride dosage range. At 28 days, the adaptability of the 8% cement group to magnesium chloride was markedly improved. When the magnesium chloride content ranged from 1.3% to 1.8%, the strengths of the specimens were higher than that of the specimen without magnesium chloride, indicating that an appropriate magnesium chloride dosage provided a more stable contribution to the later-age strength of specimens with a low cement content. The maximum strength of 2.245 MPa was achieved at a magnesium chloride content of 1.7%, which was 11.139% higher than that of the control specimen, 2.020 MPa. However, when the magnesium chloride content increased to 1.9%, the strength decreased to 1.972 MPa, which was lower than that of the control specimen. This finding suggests that excessive magnesium chloride was unfavorable for further strength development at later curing ages. Overall, the 8% cement group showed only localized strength enhancement at the early curing age, whereas the strengthening effect of an appropriate magnesium chloride dosage became more pronounced after 28 days of curing.

From the comparison among different cement contents, the unconfined compressive strength increased markedly when the cement content increased from 8% to 10%. At 7 days, the average strength of the 10% cement group was approximately 0.443 MPa higher than that of the 8% cement group. At 28 days, the average difference further increased to approximately 0.583 MPa, indicating that the 10% cement group exhibited substantially better early- and later-age strength than the 8% cement group. However, when the cement content was further increased from 10% to 12%, the strength increment became less pronounced. At 7 days, the average strength of the 12% cement group was only approximately 0.069 MPa higher than that of the 10% cement group. At 1.8% magnesium chloride, the strength gap between the two groups was only 0.021 MPa. After 28 days of curing, the 12% cement group showed a more distinct strength advantage over the 10% cement group, and the average difference increased to approximately 0.363 MPa. Even so, this increase remained smaller than the strength gain obtained when the cement content was raised from 8% to 10%. Although the 12% cement–1.8% MgCl₂ mixture achieved the highest strength, the 10% cement–1.8% MgCl₂ mixture maintained relatively high mechanical performance with a lower cement content. Therefore, it may provide a more favorable balance between strength performance and cement consumption within the range of mixtures investigated in this study.

3.3. XRD Analysis and Phase Composition

The X-ray diffraction patterns of the 28-day specimens are presented in Figure 3. As shown in Figure 3a, the cement-stabilized loess without MgCl₂ was dominated by quartz and calcite, together with diffraction features associated with cement hydration products such as C-S-H. Quartz was mainly inherited from the original loess and constituted the primary mineral framework, whereas calcite may originate from both indigenous carbonate minerals and carbonation of Ca-bearing hydration products during curing. After the incorporation of MgCl₂ (Figure 3b), additional diffraction features assigned to Mg-bearing calcite, amesite-type Mg-bearing aluminosilicate phases, M-A-H-related phases, and CaCl₂·6H₂O were observed. Compared with the corresponding MgCl₂-free specimens, the occurrence of these Mg- and Cl-bearing phases suggests that MgCl₂-derived ions participated in the phase evolution of the cement–loess system rather than remaining solely in the pore solution. In particular, the diffraction features assigned to (Mg,Ca)CO₃ are consistent with the possible incorporation of Mg²⁺ into carbonate-related phases, while the amesite-type and M-A-H-related features suggest possible interactions between Mg²⁺ and Al- and Si-bearing constituents in the loess–cement matrix. The detection of CaCl₂·6H₂O further indicates that Cl⁻ may also participate in the formation of chloride-containing crystalline products.

From the perspective of phase evolution, the XRD results indicate that MgCl₂-derived ions participated in the chemical evolution of the cement–loess system rather than remaining solely in the pore solution. Compared with the MgCl₂-free specimens, the MgCl₂-modified specimens exhibited additional diffraction features assigned to Mg-bearing calcite, amesite-type Mg-bearing aluminosilicate phases, M-A-H-related phases, and CaCl₂·6H₂O. These changes provide evidence for the chemical involvement of MgCl₂ in the stabilized system.

According to Han et al. [19], MgCl₂ can participate in reactions within cement-treated soil and promote the formation of Mg-bearing and chloride-containing products, including M-S-H, M-A-H, CaCl₂·6H₂O, Mg₂(OH)₃Cl·4H₂O, and C₃A·CaCl₂·12H₂O. These reactions were reported to produce a combined effect of alteration of existing cementitious phases and crystallization of newly formed products. In the present study, the diffraction features assigned to M-A-H-related phases and CaCl₂·6H₂O, together with Mg-bearing calcite and amesite-type Mg-bearing aluminosilicate phases, are consistent with this reaction framework. Therefore, the XRD results support the participation of MgCl₂-derived Mg²⁺ and Cl⁻ in hydration-, carbonation-, and aluminosilicate-related phase evolution. Because some cement hydration products and Mg-bearing phases exhibit low crystallinity, broad diffraction features, and overlapping characteristic peaks, distinguishing structurally or chemically similar hydration phases by XRD alone may involve uncertainty. Therefore, the assignments of M-A-H-related, amesite-type Mg-bearing aluminosilicate, and other Mg-bearing phases in this study are interpreted qualitatively based on the overall diffraction patterns and reference PDF data rather than as definitive identification of individual hydration products. Nevertheless, the differences between the MgCl₂-free and MgCl₂-modified specimens support the chemical participation of MgCl₂-derived ions in the phase evolution of the cement–loess system.

The preceding test results showed that the unconfined compressive strength increased after an appropriate amount of MgCl₂ was incorporated, and that more interparticle filling materials and cementitious products were observed in the SEM images. These findings, from both macroscopic mechanical and microscopic structural perspectives, further indicate that MgCl₂ did not simply act as an inert external salt. Instead, it participated in the formation and evolution of the internal structure of the stabilized loess.

This finding indicates that magnesium chloride may exert a dual influence on cement-stabilized loess. When the magnesium chloride dosage is within an appropriate range, magnesium ions can improve the internal soil structure through ion exchange, promotion of particle flocculation, and crystallization within the pore spaces. These effects are favorable for forming a denser particle arrangement, reducing the proportion of large pores, lowering the overall porosity, and consequently increasing the strength of the stabilized loess. However, magnesium chloride does not always produce a positive effect in cement-based stabilization systems. When the magnesium chloride dosage is excessive, the increased concentration of magnesium ions may change the chemical environment of the pore solution, interfere with the stability of cementitious products, and further affect the continued hydration process during later curing. Therefore, the influence of magnesium chloride on cement-stabilized loess can be understood as a dosage-dependent effect, namely enhancement at a suitable dosage and deterioration when the dosage becomes excessive. This interpretation is also consistent with the unconfined compressive strength results, where the strength increased at first and then decreased as the magnesium chloride content continued to rise.

3.4. Microstructural Characteristics and Pore Evolution

Figure 4 shows SEM images at 500× magnification for the specimens treated with 8% cement, 8% cement + 1.7% MgCl₂, 10% cement, 10% cement + 1.8% MgCl₂, 12% cement, and 12% cement + 1.8% MgCl₂. The SEM images show that the cement-only specimens still contained numerous irregular pores and locally unsupported structures. The particle boundaries were relatively distinct, and the overall structure remained comparatively loose. As the cement content increased from 8% to 12%, the amount of hydration products and filling materials attached to particle surfaces gradually increased, and the pore size decreased to some extent. This indicates that cement hydration products promoted interparticle bonding and pore filling. Compared with the cement-only specimens, the composite-stabilized specimens containing an appropriate amount of MgCl₂ exhibited a marked increase in fine particles and flocculent cementitious products on particle surfaces. Interparticle bridging was enhanced, large pores were subdivided and transformed into smaller pores, and the soil structure became significantly denser. Among the tested mixtures, the specimen containing 12% cement and 1.8% MgCl₂ exhibited the densest microstructure, followed by the specimen containing 10% cement and 1.8% MgCl₂. Although the specimen containing 8% cement and 1.7% MgCl₂ showed a clear improvement compared with the 8% cement-only specimen, some pores were still retained.

To describe the microstructural changes more quantitatively, the SEM images of each specimen were analyzed with Image-Pro Plus 6.0, and the pore number as well as the corresponding apparent pore-area ratio were calculated from the processed images. The micrographs were binarized according to the procedure proposed by Feng et al. [27]. As shown in Figure 5, the white regions in the binarized images denote pores, whereas the black regions correspond to soil particles. SEM images obtained at the same magnification were processed using Image-Pro Plus 6.0 following the binarization procedure proposed by Feng et al. After image segmentation, pore and surface-depression regions were represented in the binary images and their total projected area and number were quantified. The apparent pore-area ratio, $A_r$, was calculated as $A_r=A_p÷A_t\times 100\%$, where $A_p$ is the total projected area of the identified pore/depression regions and $A_t$ is the total area of the analyzed SEM image. This parameter represents a two-dimensional surface-area characteristic derived from SEM images and should not be interpreted as the three-dimensional volumetric porosity of the specimen.

The pore number, pore area, and percentage of pore area obtained from these images are summarized in

Table 5. Number, total area, and apparent area ratio of pore/depression regions identified from binarized SEM images.

Soil Sample	Total Pore Area (Pixels)	Number of Pores, N	Pore/Depression Area Ratio, Ar (%)
8% cement-stabilized loess	303,243.26	615	41.13
Loess stabilized with 8% cement and 1.7% MgCl2	283,371.98	506	38.51
10% cement-stabilized loess	270,302.52	879	36.53
Loess stabilized with 10% cement and 1.8% MgCl2	254,493.39	864	34.47
12% cement-stabilized loess	234,456.87	872	31.92
Loess stabilized with 12% cement and 1.8% MgCl2	218,323.56	809	29.53

. For the specimens stabilized only with cement, the porosities of the 8%, 10%, and 12% cement groups were 41.13%, 36.20%, and 31.92%, respectively, indicating that the pore volume decreased progressively as the cement content increased and that the internal structure became more compact. It should be noted that the 10% and 12% cement groups contained more pores than the 8% cement group, but their total porosities were obviously lower. This result suggests that the evolution of the pore system was not simply controlled by a decrease in pore number. Rather, some larger pores were gradually subdivided into a greater number of smaller pores, while the development of large pores and connected pores was effectively restricted, thereby improving the overall structural stability. After an appropriate amount of MgCl₂ was added, the porosities of the corresponding specimens further decreased to 38.51%, 34.47%, and 29.53%, respectively. This further confirms that the combined action of MgCl₂ and cement strengthened the filling, coating, and cementation among soil particles, allowing the pore structure to be further refined. Among all mixtures, the specimen containing 12% cement and 1.8% MgCl₂ had the lowest porosity and the densest microstructure, which agrees well with its highest unconfined compressive strength. The specimen containing 10% cement and 1.8% MgCl₂ showed the second-best pore refinement effect, suggesting that this mixture provides a more balanced choice in terms of microstructural densification and material economy. Overall, the incorporation of an appropriate amount of MgCl₂ not only reduced the total porosity of the stabilized loess but also promoted the transformation of large pores into smaller pores and improved the pore-size distribution. These microstructural improvements contributed to the increase in the overall strength of the stabilized loess.

3.5. Pore-Structure Characteristics Based on Nuclear Magnetic Resonance

The porosity and pore-size distribution of the entire specimens were determined by nuclear magnetic resonance tests for the cement-only specimens and the specimens treated with cement and the optimum MgCl₂ dosage. The porosity results are listed in Table 3. The cumulative pore-radius distribution curves are shown in Figure 6. In Figure 6, the MgCl₂-treated specimens with the optimum mixture proportions are denoted as stabilized soils k1–k3, whereas the cement-only specimens without MgCl₂ are denoted as cement-stabilized soils k1–k3.

As shown in

Table 6. Total porosity of stabilized loess specimens determined by nuclear magnetic resonance.

Soil Sample	Porosity, n (%)
8% cement-stabilized loess	22.65
Loess stabilized with 8% cement and 1.7% MgCl2	20.45
10% cement-stabilized loess	20.02
Loess stabilized with 10% cement and 1.8% MgCl2	18.15
12% cement-stabilized loess	19.89
Loess stabilized with 12% cement and 1.8% MgCl2	18.05

and Figure 6, the pores in all specimens were dominated by small pores, followed by medium pores, whereas large pores accounted for the smallest proportion. This indicates that the pore structure of the stabilized loess was generally controlled by small pores.

For the cement-only specimens, the total porosities of the specimens with 8%, 10%, and 12% cement were 22.65%, 20.02%, and 19.89%, respectively. The 8% cement specimen exhibited the highest porosity, indicating that its internal pores were more developed and that its structure was relatively loose. When the cement content increased to 10%, the total porosity decreased markedly, suggesting that increasing the cement content was beneficial for improving the pore structure of the soil. When the cement content was further increased to 12%, the porosity decreased slightly to 19.89%. However, the difference between the 10% and 12% cement specimens was relatively small, indicating that the pore-reducing effect became less pronounced with a further increase in cement content.

After the optimum MgCl₂ dosage was incorporated, the total porosity of all three specimen groups further decreased. The total porosities of the specimens containing 8% cement + 1.7% MgCl₂, 10% cement + 1.8% MgCl₂, and 12% cement + 1.8% MgCl₂ were 20.45%, 18.15%, and 18.05%, respectively. Compared with the corresponding cement-only specimens, the porosity decreased by 2.200, 1.870, and 1.840 percentage points, corresponding to relative reductions of 9.713%, 9.341%, and 9.251%, respectively. These results indicate that an appropriate MgCl₂ dosage further reduced the total porosity of the cement-stabilized loess and produced a denser specimen structure. Among the composite-stabilized specimens, the specimen containing 12% cement and 1.8% MgCl₂had the lowest porosity, at 18.05%, indicating the densest microstructure. The specimen containing 10% cement and 1.8% MgCl₂ had a porosity of 18.15%, which was only 0.100 percentage points higher than that of the 12% cement + 1.8% MgCl₂ specimen. This small difference suggests that the incorporation of 1.8% MgCl₂ under the 10% cement condition also achieved effective pore-structure optimization.

From the perspective of pore-size composition, the incorporation of MgCl₂ not only reduced the total porosity but also decreased pores of different size classes to varying degrees. For the specimen with 8% cement, after 1.7% MgCl₂ was added, the proportion of small pores decreased from 17.672% to 16.951%, medium pores decreased from 4.416% to 3.282%, and large pores decreased from 0.563% to 0.218%. Among these pore classes, the reduction in large pores was the most pronounced, indicating that MgCl₂ had a stronger inhibitory effect on larger pores. For the specimen with 10% cement, after 1.8% MgCl₂ was incorporated, the proportion of small pores decreased from 16.523% to 15.505%, medium pores decreased from 3.308% to 2.476%, and large pores decreased from 0.189% to 0.169%, indicating further refinement of the pore structure. For the specimen with 12% cement, after 1.8% MgCl₂ was added, the proportion of small pores decreased from 16.109% to 15.422%, medium pores decreased from 3.457% to 2.418%, and large pores decreased from 0.324% to 0.210%. This result shows that medium and large pores were both effectively suppressed after composite stabilization.

Combined with the SEM observations, the cement-only specimens still contained a certain number of interparticle pores and locally unsupported structures, and the particle boundaries remained relatively distinct. This indicates that although cement hydration products provided partial filling and cementation effects, their improvement of the pore structure was still insufficient. After an appropriate amount of MgCl₂ was incorporated, more surface attachments and flocculent cementitious products were observed on the particles, the interparticle connections became tighter, and pore connectivity was reduced. Consequently, the soil structure gradually transformed from a relatively loose state to a denser state. This result suggests that the combined action of MgCl₂ and cement further enhanced interparticle cementation and pore-filling effects, thereby improving the internal pore distribution of the stabilized loess. Although the absolute pore-related values obtained from SEM image analysis and NMR were different, both methods exhibited a consistent tendency toward pore refinement with increasing cement content and the incorporation of an appropriate MgCl₂ dosage. This difference in absolute values is mainly related to the distinct measurement scales and principles of the two techniques. SEM image analysis characterizes the projected area of pores and surface depressions within selected two-dimensional microscopic regions, whereas NMR evaluates the water-filled pore system of the entire saturated specimen based on the transverse relaxation behavior of pore water. Therefore, the SEM-derived apparent pore-area ratio should not be directly equated with the NMR-derived volumetric porosity. Rather, the two techniques provide complementary evidence of pore-structure evolution at local microscopic and whole-specimen scales.

3.6. Correlation Between Porosity and Strength

Figure 7 shows the relationship between total porosity and 28-day unconfined compressive strength. From left to right, the porosity values correspond to specimens treated with 8% cement, 10% cement, 12% cement, 8% cement + 1.7% MgCl₂, 10% cement + 1.8% MgCl₂, and 12% cement + 1.8% MgCl₂. A good correspondence was observed between total porosity and 28-day unconfined compressive strength. In general, lower porosity corresponded to higher strength, indicating that pore-structure improvement was one of the main reasons for strength enhancement. For the cement-only specimens, the specimen with 8% cement had the highest porosity and the lowest strength. As the cement content increased, the porosity decreased and the strength increased accordingly. After an appropriate amount of MgCl₂ was incorporated, the porosity of each specimen group further decreased, and the corresponding 28-day strength increased. This indicates that MgCl₂ further optimized the soil structure on the basis of cement stabilization. Among all specimens, the specimen containing 12% cement and 1.8% MgCl₂ had the lowest porosity and the highest strength, indicating that it had the densest internal structure. The porosity of the specimen containing 10% cement and 1.8% MgCl₂ was very close to that of the 12% cement + 1.8% MgCl₂ specimen, and its strength also remained at a relatively high level, suggesting that this mixture also achieved a favorable stabilization effect. In contrast, the specimens containing 8% cement and 8% cement + 1.7% MgCl₂ had relatively higher porosities and lower strengths, indicating that the degree of densification remained limited under the low cement content condition.

However, the strength development of the stabilized loess cannot be explained solely by changes in total porosity. The pore-size distribution also plays an important role in controlling the mechanical response. The NMR results showed that the incorporation of an appropriate MgCl₂ dosage reduced the proportions of medium and large pores, with the reduction in larger pores being particularly evident in several mixtures. The decrease in these relatively large pore spaces may reduce local structural discontinuities and stress concentration within the soil matrix, thereby contributing to a more uniform load-transfer framework. Meanwhile, SEM observations revealed closer particle contacts, increased interparticle bridging, and a denser cementitious matrix after the incorporation of an appropriate MgCl₂ dosage. These observations suggest that strength enhancement was associated not only with a reduction in total porosity but also with pore refinement and improved interparticle bonding. Therefore, the mechanical performance of the MgCl₂–cement-stabilized loess should be interpreted as the combined result of total porosity, pore-size distribution, and the degree of interparticle cementation rather than as a simple function of total porosity alone.

4. Discussion

The experimental results indicate a clear dosage-dependent effect of MgCl₂ on the mechanical performance and microstructural evolution of cement-stabilized loess. The UCS exhibited a non-monotonic response with increasing MgCl₂ dosage, showing an initial decrease, followed by an increase and a subsequent decline. At the optimum MgCl₂ dosage, NMR measurements showed reductions in total porosity and in the proportions of medium and large pores, while SEM observations revealed closer particle contacts, enhanced interparticle bridging, and a denser internal structure. In addition, the XRD patterns of the MgCl₂-modified specimens exhibited diffraction features assigned to Mg-bearing calcite, amesite-type Mg-bearing aluminosilicate phases, M-A-H-related phases, and chloride-containing crystalline products. These experimentally observed changes were consistent with the corresponding variation in UCS.

Based on these observations and the reaction mechanisms reported in previous studies, the strength enhancement at an appropriate MgCl₂ dosage may result from the combined effects of particle-scale physicochemical interactions and phase evolution. Mg²⁺ may promote particle aggregation through ion exchange and compression of the diffuse double layer, while Mg²⁺ and Cl⁻ may also participate in reactions involving cement-derived Ca-, Si-, and Al-bearing constituents. According to Han et al. [19], MgCl₂ can participate in reactions within cement-treated soil and promote the formation of Mg-bearing and chloride-containing reaction products. The Mg-bearing and chloride-containing diffraction features observed in the present study are consistent with such a reaction framework. These processes may contribute to enhanced interparticle bonding, pore filling, and pore refinement, thereby providing a plausible explanation for the observed strength increase.

At excessive MgCl₂ dosages, the reduction in UCS may be associated with an unfavorable change in the chemical environment of the cementitious matrix. Previous studies have suggested that high Mg²⁺ concentrations may alter or destabilize Ca-bearing hydration products such as C-S-H and promote the formation of Mg-bearing phases with different binding characteristics. Excessive Mg²⁺ and Cl⁻ may therefore disturb the normal evolution of the cementitious structure and reduce the beneficial effect of pore refinement. However, these specific reaction pathways were not directly quantified in the present study and should therefore be regarded as plausible mechanisms rather than experimentally confirmed reactions. Moreover, because some hydration products exhibit low crystallinity and overlapping diffraction peaks, the identification and evolution of individual Mg-bearing phases based on XRD remain qualitative.

The effect of MgCl₂ was also dependent on the cement content, indicating an interaction between the availability of cementitious constituents and the MgCl₂ dosage. For the 8% cement group, the optimum MgCl₂ dosage was 1.7%, whereas the optimum shifted slightly to 1.8% for the 10% and 12% cement groups. One possible explanation is that increasing the cement content provides a larger amount of Ca-, Si-, and Al-bearing hydration products and reactive constituents within the stabilized matrix. Consequently, a slightly higher MgCl₂ dosage can be accommodated because more cement-derived phases are available to interact with Mg²⁺ and Cl⁻ and to participate in the evolution of Mg-bearing and chloride-containing products. At an appropriate dosage, these interactions may accompany enhanced interparticle bonding and pore filling, thereby contributing to pore refinement and strength development.

At the lower cement content of 8%, the amount of available cement hydration products is more limited, and the stabilized matrix is less developed. Under this condition, excessive MgCl₂ may more readily disturb the balance of cement hydration and therefore the optimum dosage occurs at a slightly lower level. In contrast, the 10% and 12% cement mixtures contain a greater quantity of hydration products and a denser cementitious framework, allowing a slightly higher MgCl₂ dosage before adverse effects become dominant. However, the additional benefit of MgCl₂ becomes less pronounced at the highest cement content. For example, although the 12% cement–1.8% MgCl₂ mixture exhibited the highest UCS and lowest total porosity, the relative 28-day strength increase caused by MgCl₂ was substantially smaller than that observed for the 10% cement group. This suggests that once a relatively dense cementitious skeleton has already been established, further MgCl₂-induced pore refinement produces a diminishing mechanical benefit. Therefore, the optimum MgCl₂ dosage is controlled not only by the amount of MgCl₂ itself but also by the availability of cement-derived reactive phases and the initial degree of pore refinement in the cement-stabilized matrix. Overall, although the 12% cement–1.8% MgCl₂ mixture achieved the highest strength, the 10% cement–1.8% MgCl₂ mixture provided a more favorable balance between mechanical performance and cement consumption.

From an engineering perspective, the findings provide evidence that MgCl₂ can serve as an auxiliary modifier for cement-stabilized loess Compared with further increasing the cement content, the incorporation of an appropriate MgCl₂ dosage may provide a means of maintaining favorable mechanical performance at a relatively lower cement content. Nevertheless, the long-term durability of MgCl₂–cement-stabilized loess requires further consideration because MgCl₂ introduces soluble Mg²⁺ and Cl⁻ into the stabilized matrix. Under repeated wetting–drying conditions, these ions may migrate with pore water and undergo dissolution–redistribution–recrystallization processes, potentially leading to local salt accumulation and changes in the pore structure. Freeze–thaw exposure may further interact with salt migration by promoting moisture redistribution and repeated volumetric changes, which could gradually weaken the stabilized structure. In addition, prolonged exposure to Mg²⁺ may alter the stability of Ca-bearing hydration products, while mobile Cl⁻ may present additional durability concerns when the stabilized soil is used adjacent to reinforced concrete or metallic facilities. Therefore, although the present results demonstrate favorable short-term mechanical and microstructural performance within the investigated 28-day curing period, the long-term durability of the MgCl₂-containing system cannot be fully evaluated from the current tests. Future studies should systematically investigate strength retention, mass loss, chloride migration and leaching, pore-structure evolution, and phase changes under repeated wetting–drying and freeze–thaw cycles to establish the long-term applicability of MgCl₂–cement stabilization.

5. Conclusions

(a) The combined stabilization of loess with MgCl₂ and cement modified the compaction characteristics of the specimens. At the MgCl₂ dosages identified as optimal from the UCS results, namely 1.7% for the 8% cement group and 1.8% for the 10% and 12% cement groups, the corresponding maximum dry densities were 1.85, 1.85, and 1.86 Mg/m³, respectively, while the optimum moisture contents were 13.5%, 13.2%, and 13.6%, respectively. These results indicate that the compaction response of MgCl₂-modified cement-stabilized loess depends on both cement content and MgCl₂ dosage.

(b) MgCl₂ exhibited a distinct optimum-dosage effect on the unconfined compressive strength of cement-stabilized loess. At curing ages of 7 and 28 days, the strength of each specimen group generally decreased first, then increased, and finally decreased with increasing MgCl₂ content. The preferred MgCl₂ dosage for the 8% cement group was slightly lower than those for the 10% and 12% cement groups. In terms of strength performance, the combination of 12% cement and an appropriate MgCl₂ dosage produced the highest unconfined compressive strength, whereas the combination of 10% cement and an appropriate MgCl₂ dosage provided a more balanced performance in terms of strength improvement and material consumption. For the 8% cement group, only the specimen with 1.7% MgCl₂ showed strength improvement at 7 days, whereas specimens with 1.3–1.8% MgCl₂ exhibited enhanced strength at 28 days. This indicates that an appropriate MgCl₂ dosage contributed more effectively to later-age strength development. However, a MgCl₂ content of 1.9% caused strength reduction.

(c) The strength increased markedly when the cement content increased from 8% to 10%, whereas the strength increment became smaller when the cement content further increased from 10% to 12%. This indicates that the strengthening effect of cement exhibited a marginally decreasing tendency. Overall, the mixture containing 10% cement and an appropriate MgCl₂ dosage maintained favorable strength performance while using less cement than the 12% cement mixture. At the optimum MgCl₂ dosages, the 7-day UCS values of the 8%, 10%, and 12% cement groups increased by 7.36%, 14.45%, and 14.35%, respectively, compared with the corresponding MgCl₂-free specimens, while the corresponding increases at 28 days were 11.14%, 10.21%, and 1.27%. These results indicate that the strengthening effect of MgCl₂ depended on both cement content and curing age rather than increasing continuously with MgCl₂ dosage.

(d) At the optimum MgCl₂ dosages, the total porosity of the 8%, 10%, and 12% cement groups decreased by 9.713%, 9.341%, and 9.251%, respectively, compared with the corresponding cement-only specimens. Based on the grouped pore-size distribution, the approximate mean pore radii of the MgCl₂-modified specimens were 0.089, 0.083, and 0.084 μm, respectively. The proportions of medium and large pores were also reduced, indicating an overall refinement of the pore structure. The conclusions of this study should be interpreted within the investigated experimental conditions. The results were obtained for the specific Lanzhou loess used in this study under controlled laboratory curing conditions, with cement contents of 8–12% and MgCl₂ dosages of 0–1.9% of the cement mass. Therefore, the identified optimum MgCl₂ dosages and the associated strength and pore-structure responses may vary for loess with different mineralogical compositions, initial properties, or environmental conditions. In addition, the present study was conducted at the laboratory scale and focused primarily on short-term mechanical and microstructural behavior. Future research should therefore investigate the long-term durability of MgCl₂-modified cement-stabilized loess under representative environmental exposure conditions, including repeated wetting–drying, freeze–thaw cycles, and salt migration or leaching, with particular attention to the evolution of mechanical properties, pore structure, and hydration phases during prolonged exposure. Field-scale validation is also required to assess the effectiveness, durability, and applicability of the proposed stabilization system under actual construction and service conditions before the identified mixture proportions can be broadly applied in engineering practice.