Investigation of the effects and mechanisms of the wetting-Induced deformation behavior of compacted loess in Yan’an, China Authors: Yexia Guo, Limei Liu, Yongpeng Nie, Zhixiong Zeng, Bowen Zheng Published: 2025 ABSTRACT Wetting-induced deformation of loess is a key factor contributing to frequent geological disasters in the Loess Plateau region. Therefore, investigating the wetting deformation behavior of loess is of great significance for disaster prevention and mitigation in this region. In this study, a series of graded wetting deformation tests were carried out to examine the evolution of loess deformation during wetting process. Scanning electron microscopy (SEM) and mercury intrusion porosimetry (MIP) techniques were also employed to analyze the microstructural changes in loess. Results indicate that pore collapse in low dry-density specimen (ρd = 1.60 g/cm³) causes large pores to gradually transition into medium and small pores during the wetting process. As both vertical pressure and water content increase, the wetting-induced deformation potential first increases and then declines. In contrast, high dry-density specimen (ρd = 1.75 g/cm³) exhibit a relatively small wetting-induced deformation potential, primarily due to their denser soil structure, and stronger cementation. Furthermore, the wetting-induced deformation of loess is essentially a combined process of large pore collapse and particle rearrangement, during which the soil structure transitions from weak to stable cementation. The obtained findings provide a theoretical foundation for understanding the mechanisms of wetting-induced geological hazards in loess regions and offer significant practical value for disaster prevention and mitigation. Introduction Loess, a Quaternary sediment formed under arid and semi-arid climatic conditions, covers about 10% of the global land area1,2. Chinese loess is characterized by its silty texture, high porosity, under-compaction, salt enrichment, and well-developed vertical joints. These attributes contribute to a distinctive soil type markedly notable structural integrity, collapsibility, and heterogeneity3–5. In engineering construction, particularly in foundation and slope engineering, the mechanical properties of loess are highly sensitive to water content6,7. Even minor changes in water content can lead to significant volumetric deformation and strength degradation8,9, resulting in engineering hazards such as foundation settlement and slope instability10,11. To mitigate the water sensitivity of loess and improve its bearing capacity and stability, compaction is commonly employed in engineering practice to improve its properties12,13. Although compaction densifies the soil structure denser and homogenizes pore distribution, significantly enhancing mechanical performance, compacted loess can still undergo sudden volumetric deformation upon wetting due to rainfall infiltration or rising groundwater levels14–16. However, loess wetting-induced deformation involves structural collapse, particle rearrangement, dissolution-cementation, and pore evolution17. The underlying deformation mechanism due to the complex interplay of multiple physical and chemical processes remains unclear. In recent years, with the rapid advancement of infrastructure construction in western China, the issue of wetting-induced deformation in compacted loess has become increasingly prominent8. Practical engineering experience demonstrates that compacted loess may experience deformation over its service life due to moisture fluctuations caused by rainfall infiltration, pipeline leakage, or rising groundwater levels, leading to roadbed settlement and structural cracking18,19. Therefore, a systematic investigation into the deformation patterns of compacted loess under varying moisture conditions and its microscopic mechanisms is of significant theoretical importance and practical engineering value. Although scholars worldwide have conducted in-depth research on the collapse mechanisms of natural loess20,21, systematic studies on the deformation characteristics, structural evolution, and controlling factors of compacted loess under moisture ingress remain relatively limited. In particular, the intrinsic relationship between the macroscopic deformation response and microscopic structural changes of compacted loess under different dry densities, pressure states, and wetting paths has not been fully elucidated. To address this gap, the present study focuses on compacted loess, systematically analyzing its wetting-induced deformation patterns by controlling dry density and vertical pressure conditions. Furthermore, microstructural evolution mechanisms are investigated using scanning electron microscopy (SEM) and mercury intrusion porosimetry (MIP). The research findings are expected to provide a scientific basis for the design, construction, and long-term performance assessment of compacted loess foundations. Materials and methods Test materials Loess soil samples were collected from a filling site in the city of Yan’an, located in the northern Loess Plateau of Shanxi Province, China. Soil samples were sealed in plastic tubes to prevent water loss and transported to the laboratory for testing. Table 1 lists the basic physical properties of the loess. The natural water content was determined by the oven-drying method. The liquid limit (LL) and plastic limit (PL) were measured using the Casagrande method and the rolling method, respectively22. Specific gravity (Gs) was determined following ASTM D85423. Particle size distribution was analyzed by a combination of sieving and the hydrometer methods24. The maximum dry density (MDD) and optimum moisture content (OMC) were determined following the ASTM D1557 standard25. The low values for LL and PI result from the low clay content of loess. According to the ASTM D2487 plasticity chart26, the loess soils are “CL,” low-plasticity clay. Table 1Physical parameters of testing Loess.Propertieswn (%)LL (%)PIGsClay (%)Silt (%)Sand (%)MDD (g/cm3)OMG (%)Values28.125.592.70157691.7814.0 Table 1Physical parameters of testing Loess.Propertieswn (%)LL (%)PIGsClay (%)Silt (%)Sand (%)MDD (g/cm3)OMG (%)Values28.125.592.70157691.7814.0 Physical parameters of testing Loess. Samples Preparation To investigate the wetting deformation mechanical characteristics of the loess, a series of graded wetting deformation tests and oedometer-collapse tests under high inundation pressure were conducted. The loess soils were first air-dried, and then sieved using a 2 mm mesh. According to the target water content, a predetermined mass of distilled water was added to the dried loess and thoroughly mixed using a paddle mixer. The water-loess mixtures were then placed in plastic bags to maintain uniform moisture distribution and cured at 20 °C (relative humidity 95%) for 48 h. After curing, the predetermined wetted mixtures were weighed and placed into an oedometer cell (20 mm in height, 61.8 mm in diameter). The inner wall of the cell was pre-lubricated with silicon grease to minimize side friction and prevent gap formation between the soil and the ring. Compacted loess specimens were prepared using a slow static compaction method, the prepared soil was placed into the mold in 3–5 layers, and each layer was gradually loaded using a hydraulic press to the targeted ρd. In accordance with the slow compression test procedure specified in GB/T 50,123 − 2019, compaction was carried out using a piston advancing at 0.1 mm/min. Upon reaching the targeted ρd, the compaction load was maintained for 1 h to mitigate effects of stress relaxation. For wetting deformation test, the loess specimens were prepared at the site water content of 10% and compacted to match in-situ compaction degrees of 98% and 90%. Wetting deformation test The wetting deformation tests were conducted adopted the pre-wetting followed by loading method. The amount of distilled water required to increase the initial water content of the prepared specimens (wi =10%) to the target levels (wi =14%, 18%, and saturated) was first calculated. Distilled water was then slowly added to both ends of the specimens until the target water content was achieved. After wetting, the specimens were sealed and allowed to equilibrate for 48 h to ensure uniform water distribution. All the specimens were divided into two groups (Table 2) according to their dry densities (1.60 and 1.75 g/cm3), and the graded wetting deformation tests were conducted using a series of WG-type triple high pressure consolidometer apparatuses in compliance with the Chinese National Standard (CNS) GB/T50123-1999 (Standardization Administration of China (SAC), Ministry of Water Resources, 1999). Prior to the oedometer-wetting deformation measurements, an effective pressure of 25 kPa was applied to the soil specimens to ensure full contact between the specimen and the apparatus. Specimens were then subjected to progressively increasing vertical pressures: 50, 100, 200, 400, 800, 1600, and 3200 kPa. Compression deformation under each pressure level was recorded. The unsaturated wetting-collapse deformation was derived from the difference between oedometer compression curves at different water content increments under a constant . The wetting-induced collapse is commonly characterized by the wetted-deformation coefficient, which can been calculated using the following equation: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\delta _{}}=\frac{{{h_1} - {h_2}}}{{{h_0}}}$$\end{document} where h0 is the initial height of a specimen, and h1 and h2 are the stabilized heights of a specimen at initial water content and increment conditions under a constant vertical pressure σv, respectively. Table 2Testing program in this Study.LabelSample IDvi (cm3)ms (g)mw (g)∆wwi (%)ρdi (g/cm3)AA-16095.939.59–101.60A-26095.9813.453.86141.60A-36096.0217.287.69181.60A-46095.9725.468.18Saturated1.60BB-160104.9410.49–101.75B-260104.8914.674.18141.75B-360105.0618.928.43181.75B-460105.1020.111.19Saturated1.75vi - volume of the prepared sample; ms - mass of dry soil required for preparation; mw - mass of water added to the sample; Δw - additional water mass required to increase the water content from a lower to a higher level; wi - initial water content of the sample (%); ρdi - initial dry density of the sample (g/cm³). Table 2Testing program in this Study.LabelSample IDvi (cm3)ms (g)mw (g)∆wwi (%)ρdi (g/cm3)AA-16095.939.59–101.60A-26095.9813.453.86141.60A-36096.0217.287.69181.60A-46095.9725.468.18Saturated1.60BB-160104.9410.49–101.75B-260104.8914.674.18141.75B-360105.0618.928.43181.75B-460105.1020.111.19Saturated1.75vi - volume of the prepared sample; ms - mass of dry soil required for preparation; mw - mass of water added to the sample; Δw - additional water mass required to increase the water content from a lower to a higher level; wi - initial water content of the sample (%); ρdi - initial dry density of the sample (g/cm³). Testing program in this Study. vi - volume of the prepared sample; ms - mass of dry soil required for preparation; mw - mass of water added to the sample; Δw - additional water mass required to increase the water content from a lower to a higher level; wi - initial water content of the sample (%); ρdi - initial dry density of the sample (g/cm³). Microstructure investigation To systematically characterize the evolution of loess microstructure during wetting-induced deformation, scanning electron microscopy (SEM) was employed to qualitative analysis, while mercury intrusion porosimetry (MIP) was used for quantitative assessment. The specimens analyzed in these tests are summarized in Table 3. To minimize microstructural alterations due to water content fluctuations during SEM sampling, the specimens were first freeze-dried. Subsequently, secondary samples were collected from the specimen centers to avoid boundary effects. The prepared samples were coated with a 200–300 Å thick gold layer to ensure electrical conductivity and facilitate observation27. For MIP analysis, the sample preparation procedure was identical to that of SEM, except that no gold coating was applied. Pore size distributions (PSD) were calculated from the measured pressure–pore diameter relationship using Washburn’s equation28: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$D= - \frac{{4\gamma \cos \theta }}{p}$$\end{document} where D is the pore diameter; γ is the surface tension of mercury (typically 0.485 N/m); θ is the contact angle between mercury and solid (typically 140°); p is the mercury intrusion pressure. Table 3Microstructure testing program in this Study.GroupTesting StageDescriptionSample IDA, BPro-test (Initial condition)Samples prepared without loading; used for baseline microstructure observationA-1, B-1A800 kPa LoadingSamples consolidated under vertical stress of 800 kPaA-1, A-2, A-3A1600 kPa LoadingSamples consolidated under vertical stress of 1600 kPaA-1 Table 3Microstructure testing program in this Study.GroupTesting StageDescriptionSample IDA, BPro-test (Initial condition)Samples prepared without loading; used for baseline microstructure observationA-1, B-1A800 kPa LoadingSamples consolidated under vertical stress of 800 kPaA-1, A-2, A-3A1600 kPa LoadingSamples consolidated under vertical stress of 1600 kPaA-1 Microstructure testing program in this Study. Results and analysis Compression characteristic of compacted loess Figure 1 shows the relationship between void ratio e and vertical pressure σv under different initial water contents. For all the compacted samples, the void ratio e decreases with the increase of vertical pressure σv. As shown in Fig. 1(a), for samples with a initial dry density of 1.60 g/cm³, the reduction in void ratio under a given pressure is more pronounced at higher initial water contents. For instance, for specimens with wi of 10% and saturated condition, the initial void ratio e is 0.687, gradually declining to 0.570 and 0.514 with increasing vertical pressure. Under low vertical pressures (50–800 kPa), the void ratio decreases markedly with rising pressure, whereas under high pressures (1600–3200 kPa), the decrease becomes less noticeable. This indicates that wetting significantly influences the compressibility of loess, inducing additional settlement as water content increases under constant vertical pressure. Moreover, wetting sensitivity is more pronounced under low pressure than under high pressure. For the higher initial dry density samples (1.75 g/cm³) in Fig. 1(b), the trend of void ratio variation is similar to that of the lower density group, with e gradually decreasing as σv increases and declining with rising wi at constant pressure. However, the reduction in void ratio is less significant with increases in both wi and σv. For example, in specimens with wi = 10% and saturated condition, the maximum and minimum void ratios are only 0.472 and 0.448, respectively. This suggests that at higher compaction degrees, initial water content has a limited influence on void ratio variation, and the wetting sensitivity of highly compacted loess is lower than that of less compacted soil. Fig. 1Compression characteristic of compacted loess. Compression characteristic of compacted loess. Wetting-induced deformation characteristics of compacted loess Fig. 2Wetting-induced deformation characteristics of compacted loess. Wetting-induced deformation characteristics of compacted loess. Figure 2 shows the relationship between the wetting deformation coefficient (δ) and vertical pressure (σv) obtained from graded wetting collapse tests. As illustrated, the compacted loess exhibits additional settlement upon wetting under constant vertical pressure. The wetting deformation coefficient δ decreases with increasing initial dry density but increases with higher water content. Similar phenomena have been reported for both compacted and natural loess in various regions8,29. As shown in Fig. 2a, for samples with a compaction degree below 90%, δ initially increases almost linearly with vertical pressure, then grows slowly until reaching a maximum at σv = 1600 kPa. Beyond this point, it stabilizes or slightly decreases. According to the Chinese National Standard GB/T50123-1999, a wetting deformation of 1.5% is adopted as the collapsibility threshold. The minimum pressure corresponding to δ = 0.015 is defined as the initial collapsibility pressure. For specimens with wi = 10%, saturation collapse occurs when σv exceeds 150 kPa, with a maximum collapsibility coefficient (δmax) of 0.0339. For water content increasing from 10% to 18%, wetting collapsibility initiates at σv >250 kPa, with δmax = 0.0279. For the increment from 10% to 14%, collapse is observed at σv >700 kPa, with δmax = 0.019. These results indicate that as water content increases, the initial collapsibility pressure decreases, and the peak wetting deformation increases. In other words, smaller increments in water content correspond to higher initial collapsibility pressures and lower peak deformation values. For soil compacted to 98% of maximum initial dry density (Fig. 2b), δ increases slowly in the low-pressure range (0-200 kPa), then rises rapidly until peaking at σv = 1600 kPa. Beyond the peak, a slight decreasing trend is observed under larger water content increments, while δ remains relatively stable under smaller increments. Notably, the post-peak trend resembles that observed under 90% compaction degree. According to the 1.5% collapsibility criterion, loess with 98% compaction and wi = 10% exhibits collapsibility only when wetted to saturation under high pressure (σv = 1600 kPa). This suggests that higher compaction degrees lead to reduced wetting deformation, and collapse occurs only under combined saturation and high pressure levels. As shown in Fig. 2c and d, for samples with initial water content of 14%, the variation of δ is similar to that at wi = 10%, increasing initially and decreasing after reaching a maximum. For wi = 14%, saturation collapsibility occurs at σv = 800 kPa with δmax = 0.0154, while no collapsibility is observed for water content increasing from 14% to 18% in samples with 98.3% compaction. Overall, at all pressure levels, δ is lower for wi = 14% than for wi = 10%. This indicates that collapsibility weakens with higher initial water content, and the increase in water content (Δw) leads to reduced collapsibility in compacted loess as vertical pressures rises. Microstructural characteristics of compacted loess SEM observation Figure 3a and b present SEM images of compacted loess specimens (A-1, B-1) with an initial water content of 10% and different dry densities before wetting deformation tests. As shown in Fig. 3a, the specimen with ρd = 1.60 g/cm³ contains skeletal particles consisting mainly of silt particles and clay aggregates, predominantly connected through point-to-point contacts and forming abundant inter-aggregate pores. Due to the limited presence of clay particles at particle contacts, the specimen exhibits only weak cementation. In contrast, the microstructure changes significantly at the higher initial dry density of ρd = 1.75 g/cm³ (Fig. 3b). Here, the skeleton is primarily composed of clay-silt aggregates, with intra-aggregate pores dominating the pore system. A substantial amount of clay particles acts as cementing material, filling inter-aggregate pores and forming a stable cemented structure. Furthermore, skeletal particles are arranged in a mosaic pattern, increasing interparticle contact area and resulting in predominantly face-to-face contacts. Figure 3c and d illustrates SEM images of compacted loess specimens with an initial water content of 10% and a initial dry density of 1.60 g/cm³ under different vertical pressures prior to wetting. Compared with the unpressureed condition in Fig. 3a, the application of 800 kPa vertical pressure (Fig. 3c) leads to partial encapsulation of silt particles by clay particles. Inter-skeleton pores are partially filled with clay, forming a relatively dense matrix with predominantly semi-cemented contacts between particles. As shown in Fig. 3d, under a higher vertical pressure of 1600 kPa, silt particles become almost entirely enveloped by clay particles. Most inter-skeletal pores are filled, resulting in a denser matrix and a more stable cementation state throughout the soil mass. SEM images of specimens wetted to 14% and 18% water content under a constant vertical pressure of 800 kPa presented in Fig. 3ee and 3 f. The microstructural changes are similar to those observed in Fig. 3d: clay particles fill pores between skeletal particles, and inter-particle cementation increases, leading to a denser matrix. When water content rises from 14% to 18%, a clear increase in clay particles acting as cementing agents can be observed, further filling inter-aggregate pores. At 18% water content, pores between skeletal particles are completely filled with clay, and the overall microstructure exhibits significantly greater compactness. Fig. 3SEM images of specimens with different dry densities under varing wetting and. Note: A1, silt grain; A2, clay-silt aggregate; A3, clay aggregate; B1, intra-aggregate pore; B2, inter-aggregate pore; C1, point to point contact; C2, edge to edge contact; C3, face to face contact. SEM images of specimens with different dry densities under varing wetting and. Note: A1, silt grain; A2, clay-silt aggregate; A3, clay aggregate; B1, intra-aggregate pore; B2, inter-aggregate pore; C1, point to point contact; C2, edge to edge contact; C3, face to face contact. Pore structure before wetting Figure 4 presents the MIP results for specimens A-1 and B-1 with dry densities of 1.60 g/cm³ and 1.75 g/cm³, respectively, before wetting. As seen in Fig. 4a, cumulative intrusion curves indicate a clear reduction in specimen porosity with increasing initial dry density. The total pore volumes measured were 0.216 mL/g for the specimen with ρd = 1.60 g/cm³ and 0.153 mL/g for that with ρd = 1.75 g/cm³. The PSD curves in Fig. 4b display a distinct three-peak pattern. Based on inflection points in the distribution curve, pore sizes in the compacted loess are classified into three categories: small pores (0.005–0.2 μm), medium pores (0.2–2 μm), and large pores (2–350 μm), using 0.2 μm and 2 μm as boundaries. This classification aligns with the pore size divisions proposed by Wang et al.30. With increasing initial dry density, both the peak diameter and intensity of the large-pore group decrease, the peak diameter decreases from 8.32 μm at ρd = 1.60 g/cm³ to 6.04 μm at ρd = 1.75 g/cm³. In contrast, for the medium-pore group, the peak intensity decreases only slightly, while the corresponding peak diameter increases modestly from 0.58 μm to 0.81 μm. Fig. 4Pore size distribution of specimens with dry densities of 1.60 and 1.75 g/cm3. Pore size distribution of specimens with dry densities of 1.60 and 1.75 g/cm3. Figure 5 illustrates the pore size distribution (large, medium, and small) of compacted loess specimens A-1 and B-1 with dry densities of 1.60 g/cm³ and 1.75 g/cm³, respectively. In the low dry-density specimen (1.60 g/cm³), large pores dominate, accounting for approximately 46% of the total pore volume. As initial dry density increases to 1.75 g/cm³, the volume of large pores decreases, while the proportions of medium and small pores increase. Under this condition, medium pores become the dominant type, comprising about 40% of the total pore volume. Fig. 5Volumetric contents of different pore types for specimens with various dry densities. Volumetric contents of different pore types for specimens with various dry densities. Pre structure evolution with vertical pressure Fig. 6Pore size distribution of specimen of 1.60 g/cm3 under different vertical pressures. Pore size distribution of specimen of 1.60 g/cm3 under different vertical pressures. Figure 6 shows the cumulative mercury intrusion curves and pore size distribution (PSD) curves for compacted loess specimen A-1 (initial dry density = 1.60 g/cm³) under vertical pressures of 0 kPa, 800 kPa, and 1600 kPa. As seen in Fig. 6a, the cumulative intrusion curves shift downward with increasing vertical pressure, indicating a reduction in total pore volume. The measured pore volumes are 0.216 mL/g at 0 kPa, 0.179 mL/g at 800 kPa, and 0.152 mL/g at 1600 kPa. In Fig. 6b, under increasing vertical pressure, both the peak intensity and peak diameter of the large-pore group decrease. Specifically, the peak pore size of large pores declines from 8.32 μm (0 kPa) to 6.67 μm at 800 kPa and 4.94 μm at 1600 kPa. In contrast, the peak intensity of the medium-pore group increases with pressure, while its corresponding peak diameter shows a slight reduction. The PSD curves for small pores remain similar in shape across all pressure levels. These results indicate that increasing vertical pressure reduces the proportion of large pores and promotes the formation of medium pores, while the small-pore structure remains largely unaffected. Figure 7 illustrates the distribution of large, medium, and small pores in compacted loess under different vertical pressures. Before loading, large pores dominate, constituting approximately 46% of the total pore volume. As vertical pressure increases, the volume of large pores decreases, while the proportions of medium and small pores gradually increase. Medium pores become the dominant pore type, accounting for about 40% of total pore volume at 800 kPa and increasing to 45% at 1600 kPa. Fig. 7Volumetric contents of different pore types for the specimens under various vertical pressures. Volumetric contents of different pore types for the specimens under various vertical pressures. Pore structure evolution with water content Figure 8 presents the cumulative mercury intrusion curves and pore size distribution (PSD) curves of compacted loess specimens under a vertical pressure of 800 kPa, comparing the un-wetted condition (wi = 10%) with specimens wetted to higher water contents (wi = 14% and 18%). As shown in Fig. 8a, although all specimens had the same initial dry density and were subjected to the same vertical pressure, their cumulative intrusion curves do not coincide, indicating differences in total intruded mercury volume. These variations reflect microstructural changes in pore characteristics induced by wetting. Figure 8b shows the PSD curves of specimens under 800 kPa vertical pressure under different wetting conditions: pre-wetted (wi = 10%), and wetted to incremental water contents (wi = 14% and wi = 18%). For large pores (2–350 μm), both the peak intensity and the corresponding pore diameter decrease with increasing water content. The peak pore size of large pores shifts from 8.31 μm (pre-wetted) to 6.27 μm at wi = 14%, and further to 3.54 μm at wi = 18%. For medium pores (0.2–2 μm), the peak intensity first increases and then decreases with rising water content, while the corresponding pore diameter decreases significantly. The peak pore sizes are 0.91 μm (wi = 10%), 0.65 μm (wi = 14%), and 0.24 μm (wi = 18%). For small pores (0.005–0.2 μm), the PSD curves show little change in morphology between the pre-wetted condition and wi = 14%, with the peak pore size remaining relatively stable. However, when water content increases to 18%, the pore size of small pores shows a noticeable increase. As shown in Fig. 9, under a vertical pressure of 800 kPa, large pores dominate the compacted loess specimen before wetting (wi = 10%), accounting for 55% of the total pore volume. As water content increases, systematic changes occur in the volumetric proportions of large, medium, and small pores. When water content rises from 10% to 14%, the volume of large pores decreases, while that of medium pores increases. This transition is reflected in a shift in the dominant pore type from large to medium pores, as also evidenced by the microstructural changes shown in Fig. 3e. With further wetting to wi = 18%, the aggregates surrounding medium pores gradually disintegrate, filling these pores and leading to the formation of more small pores. As a result, the volume of medium pores decreases while that of small pores increases (Fig. 3f). Fig. 8Pore size distribution for the specimens of 1.60 g/cm3during wetting under vertical pressure of 800 kPa. Pore size distribution for the specimens of 1.60 g/cm3during wetting under vertical pressure of 800 kPa. Fig. 9Volumetric contents of different pore types for the specimens of 1.60 g/cm3during wetting under vertical pressure of 800 kPa. Volumetric contents of different pore types for the specimens of 1.60 g/cm3during wetting under vertical pressure of 800 kPa. Discussion Microstructure evolution mechanism Effect of dry density on microstructure During specimen preparation, the aggregation of particles leads to the formation of large-sized aggregates12. Specimens with higher initial dry density require greater compaction energy. Given that aggregate strength is inversely proportional to size31, large aggregates are clearly observed in specimens with a initial dry density of 1.60 g/cm³ (Fig. 3a). Due to the lower compaction energy applied, these aggregates are loosely arranged and connected mainly through point contacts, resulting in numerous inter-aggregate pores and weak cementation (Fig. 3a). This microstructure gives rise to a high proportion of large and medium pores in low dry-density specimens (Figs. 3a and 5). Research has shown that pore strength depends on the quantity and strength of surrounding cementing materials32, implying an inverse relationship between pore strength and pore size. As initial dry density increases, large pores collapse first. Their collapse leads to aggregate disintegration, and the resulting fragments fill inter-aggregate pore spaces, forming smaller medium and small pores while enhancing overall cementation. Consequently, when initial dry density increases from 1.60 g/cm³ to 1.75 g/cm³, the volume of large pores decreases, while those of medium and small pores increase (Fig. 4b). Simultaneously, interparticle contact area expands and cementation strengthens. Effect of vertical loading and wetting on microstructure evolution During continuous loading, compaction energy increases with rising vertical pressure. The resulting evolution of pore structure resembles that induced by increasing initial dry density. Higher vertical pressure causes the collapse of inter-aggregate pores, significantly reducing pore space between aggregates and increasing inter-aggregate contact area. Specifically, as vertical pressure rises, aggregates fracture, and the fragments fill large pores, forming smaller medium and small pores. In low dry-density specimens (1.60 g/cm³), the initial structure is characterized by point contacts and weak cementation between aggregates, forming numerous unstable large pores and resulting in low structural strength. When vertical pressure increases from 0 kPa to 800 kPa, the volume of large pores decreases, while those of medium and small pores increase (Figs. 3c and 6). This transformation is reflected in the shift of dominant pore size from large to medium pores (Fig. 7). As vertical pressure continues to rise to 1600 kPa, aggregates surrounding medium pores undergo further breakage. The resulting fragments fill medium pores, forming more small pores (Figs. 3d and 7). This process transforms the microstructure from an overhead structure to a matrix structure (Fig. 3d). It is worth noting that small pores remain largely intact under increasing pressure, owing to their strong cementation and high structural stability. By comparing specimens wetted to 14% and 18% under 800 kPa, combined with SEM images (Fig. 3e and f), it is observed that wetting leads to hydration of clay particles, weakening cementation strength. Simultaneously, the leaching of calcareous cement further reduces overall bonding strength33,34. As a result, the strength of unstable large pores continuously decreases during wetting. In addition, water infiltration increases soil self-weight. Under coupled hydro-mechanical effects, loess undergoes collapse of large pores and particle rearrangement29,35. The collapse of large pores generates smaller medium and small pores, gradually increasing the volumetric fraction of finer pores. Consequently, the structure of compacted loess transitions from a skeletal to a matrix form (Fig. 3e). Moreover, the disintegration of aggregates releases clay particles, which not only increases inter-aggregate contact area but also enhances cementation degree. Ultimately, the soil structure evolves from a weakly cemented state to a more stable one, and the contact mode shifts from point-to-point to edge-to-edge (Fig. 3f). Evolution mechanism of wetting deformation potential The essential characteristic of loess collapsibility lies in the transition of its structure from an open, loose state to a denser, more stable configuration32. Natural loess typically possesses sufficient strength to withstand external loads and maintain structural stability under dry conditions3. However, upon water infiltration, the clay particles cementing the skeletal grains undergo hydration, leading to a marked reduction in strength36. This decreases the overall bearing capacity and often triggers sudden collapse. It is noteworthy that soil particles—being the hardest component of loess—are generally considered resistant to deformation or fracture8,30. Therefore, loess collapsibility is fundamentally a process of particle rearrangement toward a new stable state. In summary, collapsibility is governed by two key factors: the spatial distribution of particles and the arrangement of cementitious materials. In low dry-density specimens (ρd = 1.60 g/cm³), the soil structure contains numerous clay aggregates forming a skeletal framework, with limited clay films adsorbed on coarse particle surfaces, resulting in weak cementation (Fig. 3a). As water content increases, hydration of clay particles acts as a lubricant, facilitating particle rearrangement. This explains the relatively low initial collapse pressure observed in such specimens (Fig. 2a). The point-to-point contacts between skeletal particles form an open, metastable structure rich in large and medium pores, which provide ample space for particle movement during wetting. This microstructural configuration accounts for the pronounced wetting-induced deformation seen in low-density specimens (Fig. 2), which correlates positively with the volume of medium and large pores (Fig. 5). During wetting, weakened inter-aggregate cementation causes the collapse of large pores and the formation of smaller pores (Fig. 3e and f), leading to a gradual increase in wetting deformation with rising water content. Moreover, specimens with higher initial water content tend to exhibit more fragile structures, thereby reducing the additional deformation induced by further wetting and resulting in a higher initial collapse pressure (Fig. 2c). In contrast, high dry-density specimens (ρd = 1.75 g/cm³) undergo breakage and redistribution of clay aggregates during compaction. These particles coat the skeletal grains, forming semi-cemented or fully cemented stable structures (Fig. 3b). Upon wetting, only the surface clay particles undergo hydration, while the internal cementation remains largely unaffected, preserving much of the soil’s original strength and stiffness15. This inhibits skeletal particle rearrangement and leads to higher initial collapse pressures (Fig. 2b). Furthermore, the denser and more compact structure of high-density specimens further restrains the development of wetting-induced deformation. As a result, under high initial dry density and high vertical pressure, the wetting deformation of compacted loess is significantly reduced. These findings demonstrate that under coupled hydro-mechanical action, the wetting-induced deformation of compacted loess is governed primarily by pore structure, the properties of cementing materials, and interparticle contact relationships. Similar mechanisms have been qualitatively corroborated by scanning electron microscopy in previous studies6,8,17. In engineering practice, these findings underscore the importance of assessing the water sensitivity of low-compaction loess areas and implementing appropriate mitigation strategies. Such measures may include improving site drainage, applying soil reinforcement techniques, or employing ground modification methods to enhance overall stability. Furthermore, given the significant strength reduction of loess upon wetting, it is essential to strengthen monitoring and early warning systems—particularly during rainy seasons—and to develop comprehensive emergency response plans. These steps are crucial to ensuring the safety and long-term stability of critical engineering structures. Conclusions In this study, a series of graded wetting collapse test was performed to investigate the wetting-induced deformation characteristics of compacted loess under varying initial water contents and dry densities. Scanning electron microscopy (SEM) and mercury intrusion porosimetry (MIP) were employed to analyze microstructural evolution before and after wetting, establishing correlations between structural changes and macroscopic deformation behavior. The main conclusions are as follows: (1) Wetting significantly influences the compressibility of loess. An increase in water content induces additional settlement, with larger wetting increments leading to more pronounced deformation. Specimens with low initial dry density exhibit greater wetting-induced deformation and lower initial collapse pressure, while those with high initial dry density show reduced deformation and higher collapse resistance. (2) The structural changes in compacted loess during wetting primarily result from the hydration of cementing materials between particles, which weakens interparticle bonds and facilitates particle rearrangement, thereby causing structural instability. In contrast, increasing vertical pressure mainly disrupts the structure of clay aggregates and cementitious bonds. (3) The wetting deformation of compacted loess is fundamentally a coupled process involving the collapse of large pores and the rearrangement of soil particles, reflecting a structural transition from a weakly cemented to a more stable state. In low-density specimens, the loose structure and weak cementation allow easy disintegration of aggregates upon wetting, promoting rapid structural adjustment. In high-density specimens, however, the dense structure and strong cementation effectively suppress deformation development.