Weak interface strong skeleton shrinkage resistant mix design for large cement stabilized base Authors: Naitian Zhang, Zihan Yan, Hailin Wang, Dong Tang, Guozhi Zheng, Lianxia Ma, Yinbo Zhang Published: 2026 ABSTRACT SummarySemi-rigid base asphalt pavements suffer from reflective cracking caused by shrinkage of cement-stabilized materials. This study presents a “weak-interface, strong-skeleton” design for Filling-Type Large-Particle-Size Cement-Stabilized Crushed Stone (F-LSBC) using 19–26.5 mm coarse aggregates and fine aggregates below 2.36 mm. Through compaction-filling tests, mechanical strength evaluations, fatigue testing, and drying shrinkage measurements, we determined the optimal skeleton void ratio (∼43%), filler-to-coarse aggregate mass ratio (0.30), and cement dosage (5.27%). Although F-LSBC exhibits lower mechanical strength than conventional cement-stabilized macadam, it reduces cumulative drying shrinkage strain to 81% and the average shrinkage coefficient to 87% of control values. This design provides a practical, construction-friendly approach for crack-resistant pavement bases, offering a balanced trade-off between structural strength and shrinkage control for high-grade highway applications. Introduction As a representative pavement structural type, semi-rigid base asphalt pavement has been extensively applied worldwide, especially in high-grade highway networks such as those in China. According to the latest data released by the Ministry of Transport of China, by the end of 2022, the total length of expressways in the country had exceeded 177,000 km, with more than 90% of them using semi-rigid base structures.1 However, this pavement structure has long suffered from a critical engineering challenge—reflective cracking, which occurs when cracks in the underlying base layer propagate into the asphalt surface, ultimately compromising pavement durability and service performance.2,3 Semi-rigid base materials, such as cement-stabilized crushed stone, obtain their high structural strength from hydraulic cementitious bonding, yet they exhibit pronounced sensitivity to environmental variations in temperature and humidity. These fluctuations can induce internal stress accumulation, shrinkage deformation, and eventual cracking, thereby undermining the long-term integrity of the pavement structure.4 Their inherently high shrinkage coefficient results in drying shrinkage cracks in the base during construction—primarily caused by moisture evaporation—and thermal shrinkage cracks during service, arising from diurnal and seasonal temperature fluctuations. These cracks act as stress concentration zones, providing pathways for reflective cracking to propagate into the asphalt surface layer.5,6,7 Under repeated fatigue loading induced by heavy traffic, these microcracks originating in the base layer gradually propagate upward through the pavement structure, eventually reflecting onto the asphalt surface and forming visible reflective cracks.3 This progressive cracking mechanism not only accelerates surface deterioration but also compromises the pavement’s overall structural integrity and service life.8,9,10 The initiation and propagation of reflective cracks not only disrupt the continuity, integrity, and smoothness of the pavement structure—thereby diminishing driving comfort and safety—but also create preferential pathways for surface water infiltration into the underlying layers. This accelerated ingress of moisture weakens the bond between structural layers, promotes further cracking and stripping, and ultimately shortens the pavement’s service life.11,12,13 The infiltrated water generates significant dynamic water pressure under traffic loads, scouring the base material and triggering severe secondary distresses such as mud pumping, void formation, and pothole development, which collectively escalate long-term maintenance costs.5 Therefore, suppressing reflective cracking remains a critical challenge for extending pavement service life and reducing life cycle costs. To mitigate reflective cracking, various strategies have been proposed, including surface-layer reinforcement, functional interlayers, pre-set crack technology, anti-cracking base materials, and refined construction control.14,15,16,17,18,19,20,21,22,23 Among these, developing novel base materials that address shrinkage at its source is the most fundamental and cost-effective solution. Recently, optimization of aggregate gradation based on the skeleton-filling theory has become a mainstream research direction to enhance the intrinsic crack resistance of semi-rigid materials.24,25,26,27 Against this backdrop, filling-type large-particle-size cement-stabilized crushed stone (F-LSBC), or large stone base course filled with cement-stabilized macadam (CSM), has emerged as a promising solution. In F-LSBC, uniform coarse aggregates form a high-strength load-bearing skeleton through interlocking, while macroscopic voids are filled with a highly fluid, cement-stabilized fine aggregate mortar.18,28 This configuration confines drying and thermal shrinkage within discrete, isolated filler regions, dissipating shrinkage energy through dispersed microcracks and preventing continuous wide cracks.29 This concept has been successfully implemented internationally in countries like France, Germany, Brazil, and Canada.27 In China, engineering applications in regions such as Hebei Province confirm that F-LSBC effectively mitigates reflective cracking while avoiding the rutting susceptibility and low bearing capacity common to flexible bases.30 Although F-LSBC has manifested notable advantages in both theoretical and practical aspects, it still encounters formidable challenges in large-scale promotion and application. In accordance with the stipulations of the current association standard Technical Specification for F - LSBC Base (T/CECS G: K23 - 01–2019),31 the nominal maximum size of coarse aggregates can range from 26.5 to 37.5 mm, 31.5–53 mm, 37.5–63 mm, to 53–73 mm, depending on engineering conditions. In reality, such large aggregates suffer from limited material availability, complex processing, and high costs. More crucially, they present substantial challenges to the entire construction process, encompassing industrial production, transportation, paving, and compaction. In practical applications, it often becomes inevitable to rely on less advanced techniques such as manual assistance and layer - by - layer paving. This results in extremely low construction efficiency and makes it challenging to control key quality indicators such as uniformity and compaction degree, thus severely impeding the advancement of mechanized and standardized construction (as depicted in Figure S1). Furthermore, existing mix ratio design methods continue to encounter issues such as unclear mechanisms, ambiguous procedures, and high strength variability, which hinder the consistent attainment of material performance. For instance, Li et al.,30 Based on the skeleton-filling concept, an improvement research was conducted using coarse aggregates with a maximum size of 26.5 mm. Although their study verified excellent drying shrinkage performance, the 7-day unconfined compressive strength of their prepared F-LSBC was merely 1.8 MPa—far below the strength requirement of 4.0–6.0 MPa specified for heavy-traffic expressway bases in Technical Guidelines for Construction of Highway Pavement Bases (JTG/T F20-2015).32 This highlights a significant issue of insufficient strength.5,6 These research findings suggest that the existing F-LSBC technology is in urgent need of solutions to critical issues. These issues encompass an ambiguous mix ratio design, comparatively low material strength, and intricate construction processes. These challenges stem from inadequate fundamental research on the material composition of F-LSBC, which is governed by aggregate packing, filler proportions (cement and moisture contents), and the filler-to-skeleton mass ratio.33 Notably, a critical trade-off exists between skeletal stability and filler shrinkage: strong interfacial bonding enhances mechanical strength but concentrates shrinkage stress, whereas weak bonding dissipates shrinkage energy at the expense of overall strength.34,35 Therefore, the core scientific challenge lies in realizing a “weak bonding” interface within the filler to activate the anti-cracking mechanism while guaranteeing skeletal stability and structural strength. Despite previous skeleton-filling studies on F-LSBC, three critical knowledge gaps remain: (1) a quantitative design framework balancing skeleton stability and filler shrinkage via controlled interfacial bonding is lacking; (2) the construction challenges posed by ultra-large aggregates (>37.5 mm) have not been systematically addressed; and (3) a comprehensive trade-off calibration among strength, fatigue, and shrinkage performance is missing. To fill these gaps, this study proposes a quantitative “weak-interface, strong-skeleton” design framework using 19–26.5 mm coarse aggregates (construction-friendly and widely available) and fine aggregates ≤2.36 mm. The research systematically determines: (1) random packing characteristics of single-sized coarse aggregates; (2) optimal aggregate sizes and filler-to-coarse aggregate mass ratio via compaction-filling tests; (3) the critical cement dosage (c% = 30%, total cement = 5.27%) that achieves the strength-shrinkage balance; and (4) a practical, step-by-step mix design methodology for F-LSBC. Results Packing characteristics of coarse aggregates From Table 1, it can be seen that for the ordered packing of equal-sized spheres, the interparticle void fraction depends only on the packing mode and is independent of particle diameter. However, Vvoid¯ and dvoid¯ not only depend on the packing mode but also are directly related to the particle diameter: the former is proportional to the cube of the particle diameter, while the latter is proportional to the particle diameter.Table 1Five types of ordered packing characteristics of equal-sized spheresPacking modesCubic packingHexagonal packingDouble hexagonal packingPyramidal packing and tetrahedral packingPacking parameters Coordination number681012 Particle packing densityπ/6(2/3)·(π/6)(2/3)2·(π/6)(2/2)·(π/6) Interparticle void fraction1-π/61−(2/3)·(π/6)1−(2/3)2·(π/6)1−(2/2)·(π/6)Volume of the unit cubeD3D3(3/2)2D32D3/2Average interparticle void volume Vvoid¯(1-π/6)D3(3/2−π/6)D3[(3/2)2−π/6]D3(2/2−π/6)D3Interparticle void diameter dvoid¯0.732D0.528D0.265D; 0.323D0.225D; 0.414D Table 1Five types of ordered packing characteristics of equal-sized spheresPacking modesCubic packingHexagonal packingDouble hexagonal packingPyramidal packing and tetrahedral packingPacking parameters Coordination number681012 Particle packing densityπ/6(2/3)·(π/6)(2/3)2·(π/6)(2/2)·(π/6) Interparticle void fraction1-π/61−(2/3)·(π/6)1−(2/3)2·(π/6)1−(2/2)·(π/6)Volume of the unit cubeD3D3(3/2)2D32D3/2Average interparticle void volume Vvoid¯(1-π/6)D3(3/2−π/6)D3[(3/2)2−π/6]D3(2/2−π/6)D3Interparticle void diameter dvoid¯0.732D0.528D0.265D; 0.323D0.225D; 0.414D Five types of ordered packing characteristics of equal-sized spheres The aforementioned characteristics of ordered packing were derived from the ideal spherical particle model and thus differ from actual engineering conditions. The aggregate used in road engineering has irregular morphologies, and its packing falls into the random rather than the ideal ordered arrangement. Given that F-LSBC relies on the skeleton structure formed by equal-sized coarse aggregates, the subsequent section of this paper will focus on investigating the characteristics of such aggregates under random packing states. The skeleton void ratio (VCADRC) of five-graded equal-sized coarse aggregates in random packing states is presented in Figure 1. The VCADRC values of these aggregates show minimal variation, with results of 43.00%, 42.95%, 43.05%, 43.08%, and 43.11%, respectively. This indicates that the skeleton void ratio (VCADRC) of equal-sized coarse aggregates in random packing is independent of particle size—a finding consistent with the void characteristics of ordered packing for equal-sized spheres. In addition, the average VCADRC of the coarse aggregate skeleton is 43.04%, slightly higher than the theoretical void ratio of the packing with a coordination number of 8 (hexagonal packing). This suggests that the random packing of equal-sized coarse aggregates is primarily hexagonal.Figure 1Random packing characteristics of uniform coarse aggregates Random packing characteristics of uniform coarse aggregates For comparison, Han used the discrete element software PFC3D to simulate the random packing of four types of equal-sized spheres, achieving rapid compaction by applying pressure to the container’s upper and lower planes. Simulation results show that spheres with diameters of 6 mm, 12 mm, 20 mm, and 25 mm have void ratios of 39.95%, 39.93%, 40.00%, and 39.97%, respectively. Their average is 39.96%, which is basically consistent with the theoretical void ratio of hexagonal packing with a coordination number of 8 (Figure 1). This indicates that the random packing of equal-sized spheres is also dominated by hexagonal packing, and the void ratio is independent of particle size—consistent with the conclusion of this paper. However, the experimentally measured skeleton void ratio (VCADRC) of equal-sized coarse aggregates in this study is slightly higher than that of the spheres, mainly due to the irregular shape and low sphericity of the coarse aggregates. Given the above results, the average void volume Vvoid¯ and interparticle void diameter dvoid¯ for the random packing of equal-sized coarse aggregates can be characterized using the hexagonal packing model (Table 1). Based on the equations in the table, with the average particle sizes of each grade of coarse aggregates (7.125 mm, 11.35 mm, 14.6 mm, 17.5 mm, 22.75 mm) as inputs, the calculated Vvoid¯ and dvoid¯ are presented in Figure 2.Figure 2Average void volume and interparticle void diameter of equal-sized coarse aggregates under random packing (characterized by hexagonal packing) Average void volume and interparticle void diameter of equal-sized coarse aggregates under random packing (characterized by hexagonal packing) As shown in Figure 2, with the increase in coarse aggregate particle size, both the average void volume Vvoid¯ and interparticle void diameter dvoid¯ increase significantly. For instance, when the coarse aggregate particle size increases from 4.75–9.5 mm to 19–26.5 mm, the average void volume Vvoid¯ rises to approximately 33 times its original value, and the interparticle void diameter dvoid¯ becomes roughly 3 times that of the initial value. An increase in void diameter dvoid¯ signifies a weakening of the wall effect—a phenomenon where the actual packing density of particles in a confined space (here, the voids between coarse aggregates) decreases as the space size shrinks. This weakening of the wall effect enhances the packing density of fine aggregates within the voids. Therefore, when the quantity of fine aggregate remains the same, larger voids (characterized by a larger dvoid¯) can not only accommodate more fine aggregates but also improve their packing rate. Combining Figures 1 and 2, although the coarse aggregate particle size does not affect the skeleton void ratio (VCADRC), it has a significant impact on the filling characteristics, which in turn determines the skeleton-dense structure of the mineral mixture. Therefore, further research will be conducted on the influence of coarse aggregate particle size and fine aggregate grading. Based on the packing characteristics described above, the optimal combination of coarse and fine aggregate gradations was subsequently determined through the compaction-filling test. Selection of coarse- and fine-aggregate grading Based on the determined coarse aggregate particle size, preparing F-LSBC requires further clarification of the maximum particle size and the grading of the fine aggregate. The results are shown in Figure 3F and the right side of Figure 4.Figure 3Variation of skeleton void ratio (VCADRC) and void ratio in mineral aggregates (VMADRC) with fine-to-coarse aggregate mass ratio in compaction-filling test(A–E) Correspond to fine aggregates with a maximum particle size of 4.75 mm, mixed with coarse aggregates of different size ranges: (A) 4.75–9.5 mm; (B) 9.5–13.2 mm; (C) 13.2–16 mm; (D) 16–19 mm; and (E) 19–26.5 mm.(F) Shows the results for 19–26.5 mm coarse aggregates mixed with fine aggregates having a maximum particle size of 2.36 mm. The dashed vertical line in each panel marks the critical fine/coarse mass ratio corresponding to the skeleton-dense state.Figure 4Skeleton-dense structure: variation of fine-to-coarse aggregate mass ratio, VCADRC, and VMADRC with coarse aggregate particle size Variation of skeleton void ratio (VCADRC) and void ratio in mineral aggregates (VMADRC) with fine-to-coarse aggregate mass ratio in compaction-filling test (A–E) Correspond to fine aggregates with a maximum particle size of 4.75 mm, mixed with coarse aggregates of different size ranges: (A) 4.75–9.5 mm; (B) 9.5–13.2 mm; (C) 13.2–16 mm; (D) 16–19 mm; and (E) 19–26.5 mm. (F) Shows the results for 19–26.5 mm coarse aggregates mixed with fine aggregates having a maximum particle size of 2.36 mm. The dashed vertical line in each panel marks the critical fine/coarse mass ratio corresponding to the skeleton-dense state. Skeleton-dense structure: variation of fine-to-coarse aggregate mass ratio, VCADRC, and VMADRC with coarse aggregate particle size By comparing the experimental results of the two fine aggregate fractions, it is evident that when the 2.36–4.75 mm particle size fraction is removed, the rate of VMADRC decreases significantly (Figures 3E and 3F). Under the skeleton-dense state, VMADRC decreased from 35.98% to 28.86%, while the amount of fine aggregates increased from 0.125 times to 0.25 times the mass of the coarse aggregates (Figure 4). This is because the average void volume dvoid¯ of the coarse aggregates remains constant (approximately 0.528 times the coarse aggregate particle size, as shown in Table 1), allowing finer aggregates to fill the voids more thoroughly. Therefore, on the premise of maintaining the skeleton integrity of coarse aggregates, the smaller the fine aggregate particle size, the greater the filling amount of fine aggregates, the lower the VMADRC, and the better the skeleton-dense performance. Considering the practical reality that 2.36 mm is the commonly used minimum sieve aperture in engineering, this study selects fine aggregates with a maximum particle size of 2.36 mm for subsequent research. Unless otherwise specified, “fine aggregates” mentioned hereinafter refer to this specification. In summary, this study finally determines: coarse aggregate particle size: 19–26.5 mm; maximum particle size of fine aggregates: 2.36 mm. Using the compaction-filling procedure and the skeleton-preservation criterion described in the STAR Methods, we evaluated the effects of coarse-aggregate size and fine-to-coarse aggregate mass ratio on VCADRC and VMADRC. The results are presented in Figures 3 and 4. Taking the combination of 19–26.5 mm coarse aggregates and fine aggregates with a maximum size of 2.36 mm as an example (Figure 3F), the initial VCADRC of the pure coarse aggregates is 43.11%. As the fine/coarse mass ratio increases from 0 to 0.25, VMADRC decreases progressively and reaches its minimum value of 28.86% at a ratio of 0.25. Further increasing the ratio to 0.325 causes VMADRC to rise slightly (indicating over-filling) and VCADRC to increase by more than 1.0%, signaling expansion of the coarse aggregate skeleton. Therefore, according to the proposed criterion, the critical fine/coarse mass ratio is identified as 0.25. The same procedure was applied to all other coarse aggregate sizes and fine aggregate gradations, and the results are summarized in Figure 4. As shown in Figure 4, as the coarse aggregate particle size increases, the skeleton void ratio (VCADRC) remains essentially stable, with an average of approximately 43.04%. This further validates the conclusion that VCADRC is independent of particle size, consistent with the results presented in Figure 1. Meanwhile, as the coarse aggregate particle size increases, the fine-to-coarse aggregate mass ratio required to achieve the skeleton-dense structure increases gradually—indicating that more fine aggregates are needed to fill the skeleton voids, which leads to a gradual decrease in VMADRC. For example, when the coarse aggregate particle size is 4.75–9.5 mm, the mass of fine aggregates required is only 0.01 times that of the coarse aggregates (almost negligible). The corresponding VMADRC at this point is 42.43%, close to VCADRC. When the coarse aggregate particle size increases to 19–26.5 mm, however, the mass of fine aggregates required rises significantly, reaching 0.125 times that of the coarse aggregates. The corresponding VMADRC drops sharply to 35.98%, far lower than VCADRC. This is because larger particle sizes of coarse aggregates create larger voids, which require more fine aggregates to fill these voids, thereby maintaining a dense skeleton structure. This trend is consistent with Figures 3A–3E—namely, as the coarse aggregate particle size increases, the rate of VMADRC decrease accelerates. This phenomenon stems from the linear increase of average void volume dvoid¯ with coarse aggregate particle size (Figure 2), which weakens the wall effect and thereby enhances the filling efficiency of fine aggregates. Although VCADRC remains unaffected by coarse aggregate size, the increase in coarse aggregate particle size influences the distribution and filling behavior of fine aggregates, which in turn reduces VMADRC. This trend demonstrates how coarse aggregate particle size indirectly affects the overall void characteristics of the mixture. Therefore, when the fine aggregate grading is fixed, a larger coarse aggregate particle size requires more fine aggregate to fill, resulting in a smaller VMADRC and better skeleton-dense performance. It can be inferred that if ultra-large particle size aggregates (53–73 mm) are used, VMADRC will decrease further. In conclusion, to achieve a superior skeleton-dense structure, it is preferable to use large-particle-size coarse aggregates.27 In this study, five different coarse aggregate sizes were tested for their suitability in F-LSBC mixtures. The aggregates were selected based on availability and application requirements. Based on the results of the Compaction-Filling Test and the analysis of void ratios, 19–26.5 mm coarse aggregates were selected for this study. This particle size was found to be optimal for achieving a skeleton-dense structure, which is essential for minimizing shrinkage-induced cracking. Larger coarse aggregate sizes were excluded due to the increased difficulty in material procurement, mixing uniformity, and construction efficiency. Moreover, this size range of aggregates is widely available in practical engineering and ensures a more efficient construction process. Skeleton-dense combined grading and filler-to-coarse aggregate ratio As shown in Figure 5, with the increase of cement dosage, the optimum water content of the filler material increases approximately linearly, while the maximum dry density changes slightly. Water in the filler adsorbs onto the surface of fine aggregates and cement particles, forming a water film to lubricate particles and reduce friction. The larger the specific surface area of the particles, the larger the required water film area—and thus the more water needed to reach the densest state. The specific surface area of aggregates is far lower than that of cement: for example, the specific surface area of material passing through the 0.075 mm sieve is approximately 32.77 m2/kg, while that of ordinary Portland cement is usually no less than 300 m2/kg. Therefore, increasing the cement dosage raises the total specific surface area of the particles, leading to an increase in optimum water content. Thus, as the cement dosage (c%) rises, the maximum dry density remains nearly unchanged—indicating that cement dosage has little effect on the densest packing state of fine aggregate - cement mixtures.Figure 5Optimum water content and maximum dry density of filler material with different cement dosages Optimum water content and maximum dry density of filler material with different cement dosages Following the method shown in Figure S6, the compaction-filling test (Method C) was conducted on the mixture of coarse aggregates and filler material. The skeleton void ratio of coarse aggregates under wet compacted condition (VCAWCC) and the void ratio under wet compacted condition (VVWCC), with the results presented in Figures 6A–6F.Figure 6Results of compaction-filling test on coarse aggregates and filler material: variation patterns of VCAWCC and VVWCC(A–F) represent different cement dosages (c%) in the filler material: (A) 10%; (B) 15%; (C) 20%; (D) 25%; (E) 30%; and (F) 35%. The dashed vertical line in each panel indicates the critical filler-to-coarse aggregate mass ratio (0.3 for all cases) at which the skeleton-dense structure is achieved. Results of compaction-filling test on coarse aggregates and filler material: variation patterns of VCAWCC and VVWCC (A–F) represent different cement dosages (c%) in the filler material: (A) 10%; (B) 15%; (C) 20%; (D) 25%; (E) 30%; and (F) 35%. The dashed vertical line in each panel indicates the critical filler-to-coarse aggregate mass ratio (0.3 for all cases) at which the skeleton-dense structure is achieved. With increasing filler material dosage, fine aggregates and cement gradually fill the voids between coarse aggregates, causing VVWCC to decrease continuously (Figures 6A–6F). When the mass of the filler material does not exceed 0.3 times that of the coarse aggregates, the coarse aggregate skeleton remains stable, and the VCAWCC remains approximately 34%. This is consistent with the results of the previous compaction test (Figures 3 and 4). When the filler material reaches 0.3 times the mass of the coarse aggregates, the VCAWCC shows an inflection point and begins to rise, indicating that the coarse aggregate skeleton is being interfered with and expanded. Meanwhile, the VVWCC drops to the lowest value, indicating that the mixture reaches the skeleton-dense state. The parameters corresponding to the skeleton-dense state under different cement dosages are summarized in Figure 7.Figure 7Variation patterns of parameters corresponding to skeleton-dense structure with cement dosage c% in Filler Material Variation patterns of parameters corresponding to skeleton-dense structure with cement dosage c% in Filler Material In the skeleton-dense state, the mass ratio of filler material to coarse aggregates remains constant at 0.3—independent of cement dosage (Figure 7). However, as the cement dosage (c%) increases, the total mass of fine aggregates and cement decreases slightly. This is because an increase in cement dosage increases the filler material’s optimum water content (Figure 5), thereby reducing the proportion of solid materials (fine aggregates + cement) per unit mass. When the cement dosage increases from 10% to 35%, the mass of fine aggregates and cement decreases from 0.282 times that of the coarse aggregates to 0.276 times—a small decline. This reflects the limited variation range of the optimum water content and indicates that the mass of solid materials remains essentially stable. The reason is that once the particle size of coarse aggregates is determined, their average void volume Vvoid¯ remains essentially constant, and the maximum dry density of fine aggregates and cement (i.e., the bulk density in the densest state) is barely affected by cement dosage (Figure 5). Therefore, in the skeleton-dense state, the bulk volume VFA+P·Ostacking of fine aggregates and cement required to fill the voids Vvoid¯ in coarse aggregates remains essentially unchanged; consequently, their mass also stays stable. The total volume of fine aggregates and cement was calculated according to Equation 5, with results presented in Figure 7. As the cement dosage increases, this ratio fluctuates slightly around 0.224 and remains stable overall—consistent with the previous conclusions. Furthermore, since the mass of fine aggregates and cement remains essentially unchanged and the skeleton void ratio VCAWCC stays stable, it can be inferred that the void ratio VVWCC also remains essentially unchanged. To sum up, to ensure F-LSBC achieves the skeleton-dense structure, the mass ratio αCAFiller of filler material to coarse aggregates is recommended to be 0.3. Based on this ratio, the cement dosage c′%, water content w′%, and dry density ρdry′ of F-LSBC can be calculated, with specific results summarized in Table 2. Based on these findings, F-LSBC specimens conforming to the skeleton-dense structure requirements can be fabricated.Table 2Relevant parameters of F-LSBC under skeleton-dense stateFiller materialCement dosage c%10%15%20%25%30%35%Optimum water content wOP%6.47%6.88%7.48%7.96%8.27%8.61%Filler material: coarse aggregates (αCAFiller)0.3Volume of the compaction cylinder VMold/cm32177F-LSBCtotal mass MCA+Filler508650795087507150845074cement dosage c′%2.04%2.94%3.77%4.55%5.27%5.94%water content w′%1.42%1.51%1.63%1.73%1.79%1.86%dry density ρdry′/(g/cm3)2.3032.2982.2992.2902.2942.288 Table 2Relevant parameters of F-LSBC under skeleton-dense stateFiller materialCement dosage c%10%15%20%25%30%35%Optimum water content wOP%6.47%6.88%7.48%7.96%8.27%8.61%Filler material: coarse aggregates (αCAFiller)0.3Volume of the compaction cylinder VMold/cm32177F-LSBCtotal mass MCA+Filler508650795087507150845074cement dosage c′%2.04%2.94%3.77%4.55%5.27%5.94%water content w′%1.42%1.51%1.63%1.73%1.79%1.86%dry density ρdry′/(g/cm3)2.3032.2982.2992.2902.2942.288 Relevant parameters of F-LSBC under skeleton-dense state Through the Compaction-Filling Test, the optimal mass ratio αCAFiller of filler material to coarse aggregates was determined to be 0.3. Based on the selected coarse aggregate particle size and the determined fine aggregate grading, the combined grading of the cement-stabilized crushed stone F-LSBC was finally finalized. Although F-LSBC specimens with different cement dosages can be prepared accordingly, their appropriate cement dosage remains undetermined and requires further in-depth study. Mechanical properties and optimization of cement dosage The experimental data presented in Figure 8A illustrates the performance differences between F-LSBC and CSM regarding unconfined compressive strength, splitting tensile strength, and flexural strength. Figure 8B presents the ratio of F-LSBC strength to CSM strength. The graphical representation clearly reveals that F-LSBC demonstrates considerably inferior mechanical properties in all three measured parameters when contrasted with CSM. These findings suggest a marked deterioration in the material’s comprehensive performance characteristics. The results provide additional evidence that the weakening at material interfaces substantially impacts the large-scale mechanical properties of F-LSBC.Figure 8Comparison of macroscopic mechanical properties between F-LSBC and CSM(A) Evolution of unconfined compressive strength, splitting tensile strength, and flexural strength at different curing ages.(B) ratio of F-LSBC strength to CSM strength for each mechanical index. The total cement content is 5.27% for both materials, with a filler cement dosage of 30% for F-LSBC. Specimens were subjected to 7-day standard curing, 14-day standard curing, and 14-day accelerated curing.Data in (A) are presented as the mean of five replicate specimens (n = 5) for each material, curing condition, and mechanical test; The F-LSBC/CSM strength ratios in (B) were calculated from the corresponding group means in (A). Comparison of macroscopic mechanical properties between F-LSBC and CSM (A) Evolution of unconfined compressive strength, splitting tensile strength, and flexural strength at different curing ages. (B) ratio of F-LSBC strength to CSM strength for each mechanical index. The total cement content is 5.27% for both materials, with a filler cement dosage of 30% for F-LSBC. Specimens were subjected to 7-day standard curing, 14-day standard curing, and 14-day accelerated curing. Data in (A) are presented as the mean of five replicate specimens (n = 5) for each material, curing condition, and mechanical test; The F-LSBC/CSM strength ratios in (B) were calculated from the corresponding group means in (A). Fatigue performance In accordance with the experimental procedures, both F-LSBC and CSM materials underwent fatigue testing to collect lifespan data across varying stress-ratio scenarios. The initial step involved computing the mean fatigue duration for every stress level applied to each material. Following this, a statistical analysis employing linear regression was executed, incorporating the averaged fatigue duration measurements from four distinct stress ratios. Within this analytical framework, stress ratios served as predictor variables while logarithmic transformations of fatigue lifespan values functioned as response variables, enabling the equation of linear predictive models. Consequently, mathematical expressions describing fatigue behavior were developed for both F-LSBC and CSM, with the graphical representations of these findings displayed in Figure 9.Figure 9Comparison of fatigue performance between F-LSBC and CSMEach point represents the log10-transformed mean fatigue life calculated from three replicate beam specimens (n = 3) for the indicated material and stress ratio. Lines show separate linear regression fits for CSM and F-LSBC. Comparison of fatigue performance between F-LSBC and CSM Each point represents the log10-transformed mean fatigue life calculated from three replicate beam specimens (n = 3) for the indicated material and stress ratio. Lines show separate linear regression fits for CSM and F-LSBC. Drying-shrinkage performance Figure 10 presents the evolution of drying shrinkage strain with drying time for F-LSBC and CSM. As shown in the figure, the daily drying shrinkage strain εi′ of both materials decreases continuously throughout the test period, and the rate of reduction gradually diminishes, indicating a progressive stabilization of shrinkage behavior. Specifically, the daily drying shrinkage strain exceeds 20 × 10−6 during the first three days, decreases to a range of 20 × 10−6-10×10−6 between days 4 and 7, and remains generally below 10 × 10−6 after day 7 with only slight fluctuations, suggesting that drying shrinkage has largely stabilized. Correspondingly, the cumulative drying shrinkage strain εi increases rapidly in the initial stage and then exhibits a markedly reduced growth rate, demonstrating a typical shrinkage development pattern characterized by rapid early progression followed by gradual stabilization.Figure 10Variation of dry shrinkage strain with drying time(A) Cumulative drying shrinkage strain (εᵢ) as a function of drying time.(B) daily drying shrinkage strain (εᵢ′) as a function of drying time. Both panels compare the performance of F-LSBC and CSM under controlled environmental conditions (20 ± 1°C, 60 ± 5% RH) over a 14-day period.Data in both panels are presented as the mean of three replicate prismatic specimens used for dimensional-change measurements per material (n = 3). Variation of dry shrinkage strain with drying time (A) Cumulative drying shrinkage strain (εᵢ) as a function of drying time. (B) daily drying shrinkage strain (εᵢ′) as a function of drying time. Both panels compare the performance of F-LSBC and CSM under controlled environmental conditions (20 ± 1°C, 60 ± 5% RH) over a 14-day period. Data in both panels are presented as the mean of three replicate prismatic specimens used for dimensional-change measurements per material (n = 3). Consistent with the evolution of the cumulative drying shrinkage strain εi, the average drying shrinkage coefficient αd,i increases rapidly during the first three days and subsequently exhibits a progressively reduced growth rate, as shown in Figures 10A and 11. In addition, Figure 10B indicates that the daily drying shrinkage strain of F-LSBC remains slightly lower than that of CSM throughout the drying period, leading to a consistently smaller cumulative drying shrinkage strain εi for F-LSBC. This discrepancy becomes more pronounced with increasing drying time. As a result, the average drying shrinkage coefficient αd,i of F-LSBC is also markedly lower than that of CSM, as illustrated in Figure 11, and the difference between the two materials increases gradually over time. These results demonstrate that F-LSBC possesses a superior ability to restrain drying shrinkage, characterized by reduced deformation and enhanced dimensional stability.36Figure 11Variation of average dry shrinkage coefficient with drying timeValues were calculated using the mean mass-loss measurements from three reference specimens and the mean dimensional-change measurements from three separate test specimens per material (n = 3 for each measurement type). Variation of average dry shrinkage coefficient with drying time Values were calculated using the mean mass-loss measurements from three reference specimens and the mean dimensional-change measurements from three separate test specimens per material (n = 3 for each measurement type). Given that the drying shrinkage test was conducted over a 14-day period, a regression analysis of the time-dependent evolution of the cumulative drying shrinkage strain εi and the average drying shrinkage coefficient αd,i was performed to further evaluate and compare the long-term shrinkage behavior of F-LSBC and CSM. The corresponding regression results are summarized in Table 3. As discussed in the analysis of Figure 11, the shrinkage development tends to stabilize after the seventh day, indicating that the available experimental data sufficiently represent the key stages of the shrinkage process. Accordingly, regression fitting based on the present dataset is considered both appropriate and reliable.Table 3Regression equations describing the variation patterns of cumulative drying shrinkage strain and average drying shrinkage coefficient with drying time iDry shrinkage characteristicsData sourceMaterialsRegression equationR2ID no.εiFig.F-LSBCεiF-LSBC=52.582ln(i)+41.6740.9941(1)CSMεiCSM=63.651lni+54.9960.9937(2)αd,iFig.F-LSBCαd,iF-LSBC=19.64ln(i)+26.7250.9934(3)CSMαd,iCSM=21.453lni+35.70.9973(4) Table 3Regression equations describing the variation patterns of cumulative drying shrinkage strain and average drying shrinkage coefficient with drying time iDry shrinkage characteristicsData sourceMaterialsRegression equationR2ID no.εiFig.F-LSBCεiF-LSBC=52.582ln(i)+41.6740.9941(1)CSMεiCSM=63.651lni+54.9960.9937(2)αd,iFig.F-LSBCαd,iF-LSBC=19.64ln(i)+26.7250.9934(3)CSMαd,iCSM=21.453lni+35.70.9973(4) Regression equations describing the variation patterns of cumulative drying shrinkage strain and average drying shrinkage coefficient with drying time i As summarized in Table 3, all four regression equations exhibit relatively high correlation coefficients, indicating a satisfactory fitting accuracy. These models can therefore be reliably employed to predict the long-term drying shrinkage behavior of F-LSBC and CSM over extended periods (e.g., 180 days). It should be noted that although the drying shrinkage test in this study was conducted for only 14 days, the applicability of logarithmic-time-based extrapolation for predicting long-term shrinkage behavior has been validated in previous studies on cementitious materials. Bažant37 demonstrated that drying shrinkage evolution exhibits a characteristic logarithmic-time trend and that short-term experimental data can be reasonably extrapolated to estimate long-term shrinkage behavior. Their study further verified the applicability of this approach using experimental data from cement-based materials with different specimen sizes and mixture compositions. Therefore, the logarithmic regression adopted in this study is consistent with established methodologies for shrinkage prediction and provides a reasonable approximation of the long-term shrinkage behavior of cement-stabilized materials (CSMs). Based on regression equations 1 and 2, the predicted evolution of cumulative drying shrinkage strain is illustrated in Figure 12, while the corresponding variation in the average drying shrinkage coefficient, derived from equations 3 and 4, is presented in Figure 12. As shown in Figure 12, the cumulative drying shrinkage strain εi of F-LSBC at 180 days is approximately 81.36% of that of CSM. Similarly, Figure 12 indicates that the average drying shrinkage coefficient αd,i of F-LSBC is about 86.62% of that of CSM. These results confirm that F-LSBC demonstrates a stronger resistance to drying shrinkage than CSM, reflecting its superior shrinkage control performance.Figure 12The variation pattern of drying shrinkage characteristics with drying time The variation pattern of drying shrinkage characteristics with drying time It should be noted that, due to material constraints, the CSM specimens used in this study contained only 0.06% mineral powder, which is considerably lower than the 2%–5% range recommended in the Technical Guidelines for Construction of Highway Roadbases (JTG/T F20-2015). Previous studies have indicated that an increased mineral powder content leads to higher water demand and more significant shrinkage deformation. Lower mineral powder content reduces drying shrinkage. Therefore, a standard CSM (2%–5% mineral powder) would have greater drying shrinkage than our CSM control. This means that the actual drying shrinkage advantage of F-LSBC over standard CSM is even larger than reported. Very low mineral powder may slightly reduce CSM strength, but this does not affect the main conclusion that F-LSBC intentionally trades off strength for superior crack resistance. Hence, our comparison is conservative. Proposed F-LSBC mix-design procedure and numerical example Previously, the particle size of coarse aggregates and the grading of fine aggregates were determined through the Compaction - Filling Test. Subsequently, the mass ratio of coarse aggregates to filler material was finalized using the same test. Thereafter, the cement dosage was determined based on the unconfined compressive strength results. Collectively, these results facilitate the development of the mix design for cement-stabilized crushed stone F-LSBC. By integrating the aforementioned procedures, the mix design methodology for cement-stabilized crushed stone F-LSBC is formulated, as depicted in Figure 13.Figure 13Mix design methodology for F-LSBC Mix design methodology for F-LSBC As depicted in Figure 13, the mix design procedures for F-LSBC are delineated as follows:(1)Select uniformly sized coarse aggregates (with different diameters) and fine aggregates of diverse gradations. Blend them at various mass ratios and subsequently conduct the Compaction-Filling Test.(2)For each compaction - filling test, ascertain the voids in mineral aggregate (VMA) corresponding to the skeleton-dense structure under dry tamping conditions, denoted as VMADRC.(3)From the aforementioned VMADRC values, select the coarse aggregate size and fine aggregate gradation corresponding to the minimum VMADRC.(4)For the fine aggregate determined in Step (3), initially select 5–6 cement dosages to prepare the filler material. Conduct compaction tests to determine the optimum water content and maximum dry density of the filler material for each cement dosage.(5)Randomly choose one filler material (with a fixed cement dosage) and initially set 5–6 mass ratios for mixing with coarse aggregates. Perform the Compaction-Filling Test to identify the mass ratio corresponding to the skeleton-dense structure.(6)Based on the parameters determined in Step (5), including the coarse aggregate size, fine aggregate gradation, optimum water content of the filler material, and mass ratio of F-LSBC specimens with different cement dosages. (1)Select uniformly sized coarse aggregates (with different diameters) and fine aggregates of diverse gradations. Blend them at various mass ratios and subsequently conduct the Compaction-Filling Test. Select uniformly sized coarse aggregates (with different diameters) and fine aggregates of diverse gradations. Blend them at various mass ratios and subsequently conduct the Compaction-Filling Test. (2)For each compaction - filling test, ascertain the voids in mineral aggregate (VMA) corresponding to the skeleton-dense structure under dry tamping conditions, denoted as VMADRC. For each compaction - filling test, ascertain the voids in mineral aggregate (VMA) corresponding to the skeleton-dense structure under dry tamping conditions, denoted as VMADRC. (3)From the aforementioned VMADRC values, select the coarse aggregate size and fine aggregate gradation corresponding to the minimum VMADRC. From the aforementioned VMADRC values, select the coarse aggregate size and fine aggregate gradation corresponding to the minimum VMADRC. (4)For the fine aggregate determined in Step (3), initially select 5–6 cement dosages to prepare the filler material. Conduct compaction tests to determine the optimum water content and maximum dry density of the filler material for each cement dosage. For the fine aggregate determined in Step (3), initially select 5–6 cement dosages to prepare the filler material. Conduct compaction tests to determine the optimum water content and maximum dry density of the filler material for each cement dosage. (5)Randomly choose one filler material (with a fixed cement dosage) and initially set 5–6 mass ratios for mixing with coarse aggregates. Perform the Compaction-Filling Test to identify the mass ratio corresponding to the skeleton-dense structure. Randomly choose one filler material (with a fixed cement dosage) and initially set 5–6 mass ratios for mixing with coarse aggregates. Perform the Compaction-Filling Test to identify the mass ratio corresponding to the skeleton-dense structure. (6)Based on the parameters determined in Step (5), including the coarse aggregate size, fine aggregate gradation, optimum water content of the filler material, and mass ratio of F-LSBC specimens with different cement dosages. Based on the parameters determined in Step (5), including the coarse aggregate size, fine aggregate gradation, optimum water content of the filler material, and mass ratio of F-LSBC specimens with different cement dosages. For the F-LSBC specimens with varying cement dosages, conduct 7-day unconfined compressive strength tests. Obtain the curve of compressive strength versus cement dosage, and select the cement dosage at the curve’s inflection point as the final optimal value. The physical significance of this strength inflection point lies in the transition of the load-bearing mechanism of F-LSBC. At relatively low cement dosages, insufficient hydration products are generated, and the bonding between the filler and coarse aggregates remains weak. Therefore, the mechanical strength mainly depends on the cement-stabilized fine aggregate filler. With increasing cement dosage, more hydration products, especially calcium silicate hydrate (C–S–H) gels, are produced, which improves particle bonding and gradually activates the load-bearing contribution of the coarse aggregate skeleton. However, after the inflection point, the skeleton structure has already formed an effective load-transfer system, and further increasing cement dosage mainly enhances the stiffness of the cementitious matrix rather than significantly improving structural strength. Excessive cementitious bonding may increase shrinkage stress concentration and weaken the crack-resistance advantage provided by the weak-interface and strong-skeleton structure. Therefore, the cement dosage corresponding to the strength inflection point represents an optimal balance between mechanical strength development and shrinkage crack resistance. Numerical example of the proposed mix design method To further demonstrate the engineering applicability of the proposed design procedure in Figure 13, a numerical example is provided. The key design parameters obtained from the proposed procedure are summarized in Table 4. Based on the compaction-filling test results, 19–26.5 mm coarse aggregate and fine aggregate with a maximum particle size of 2.36 mm were selected. The mass ratio of filler material to coarse aggregate was determined as 0.30, and the optimum cement dosage in the filler was selected as c = 30%. The corresponding optimum water content of the filler material was 8.27%.Table 4Design parameters used in the numerical example of F-LSBC mix designDesign stepResultCoarse aggregate size19–26.5 mmFine aggregate size≤2.36 mmFiller/coarse aggregate ratio0.30Cement dosage in filler30%Optimum water content8.27%Optimum total cement dosage5.27% Table 4Design parameters used in the numerical example of F-LSBC mix designDesign stepResultCoarse aggregate size19–26.5 mmFine aggregate size≤2.36 mmFiller/coarse aggregate ratio0.30Cement dosage in filler30%Optimum water content8.27%Optimum total cement dosage5.27% Design parameters used in the numerical example of F-LSBC mix design Taking 1000 kg of dry aggregate as the calculation basis, the mass of fine aggregate is denoted as (mf), the mass of coarse aggregate as (mc), the mass of cement as (mce), and the mass of water as (mw). Since the cement dosage in the filler is defined as the mass ratio of cement to fine aggregate[mce=0.30mf] The final total cement dosage is 5.27%, namely[mcemc+mf=5.27%]and[mc+mf=1000kg] Therefore[mf=52.7kg0.30=175.7kg][mc=1000-175.7=824.3kg][mce=52.7kg] The water content of the filler material is 8.27%, calculated based on the dry mass of fine aggregate and cement. Thus[mw=(mf+mce)×0.0827]=18.9kg Accordingly, the final mix proportion of F-LSBC is:[mc:mf:mce:mw=824.3:175.7:52.7:18.9]or, normalized by the mass of coarse aggregate[mc:mf:mce:mw=1:0.213:0.064:0.023] The calculated mass ratio of filler material to coarse aggregate is:[175.7+52.7+18.9824.3=0.300]The physical properties of the selected coarse aggregates and the gradation and apparent density of the fine aggregates are reported in Tables 5 and 6, respectively. The physical and mechanical properties of the cement and the mix proportion of the control CSM are summarized in Tables 7 and 8, respectively. The compaction parameters used in the mix-design tests are listed in Table 9.which is consistent with the skeleton-dense state determined by the compaction-filling test. Meanwhile, the total cement dosage is:[52.7824.3+175.7=5.27%]indicating that the calculated mix proportion satisfies both the structural filling requirement and the optimum cement dosage requirement. Therefore, for every 1000 kg of dry aggregate, the recommended mix is 824.3 kg of 19–26.5 mm coarse aggregate, 175.7 kg of fine aggregate below 2.36 mm, 52.7 kg of cement, and 18.9 kg of water. In practical construction, the water dosage should be further corrected according to the actual moisture content and water absorption of the aggregates.Table 5Physical properties of limestone coarse aggregatesParticle size/mmApparent density/(g·cm−3)Bulk density/(g·cm−3)Water absorption/%4.75–9.52.7422.7090.449.5–13.22.7282.7010.3813.2–162.7422.7130.3916∼192.7332.7130.2819∼26.52.7542.7150.53Table 6Gradation and apparent density of limestone fine aggregatesSieve opening size/mm4.752.361.180.60.30.150.075Bulk density/(g·cm−3)Passing rate (%)4.75 mm10066.7742.9429.9824.0615.0512.022.7102.36 mm/10064.3144.9136.0322.5418.00Table 7Physical and mechanical properties of P·O 42.5 cementPropertiesDensity/(g·cm−3)Fineness/(m2·kg−1)Setting time/minCompressive strength/MPaFlexural strength/MPaInitial setting timeFinal setting time3 days28 days3 days28 daystest results3.06534515024022.654.54.89.1technical requirements≥300≥45≤600≥17.0≥42.5≥3.5≥6.5Table 8Mix proportion of the control CSM specimenParameterValueCement content (% of dry aggregate)5.27%Maximum aggregate size26.5 mmMineral powder content (<0.075 mm)0.06%Optimum moisture content3.87%Maximum dry density (g/cm3)2.391Table 9Compaction parameters (T0804-1994)Type of compaction testMethod A: compaction testMethod C: compaction - filling testCompaction objectfilling materialmixture of filling material and coarse aggregate (F-LSBC)Hammer parametersmass = 4.5 kg, diameter = 5.0 cm, drop height = 45 cm, gravitational potential energy = 19.845JCompaction cylinder dimensionsΦ 10.0 cm × H 12.7 cm = 997 cm3Φ 15.2 cm × H 12.0 cm = 2177 cm3Number of compaction blows27 blows/layer × 5 layers = 135 blows98 blows/layer × 3 layers = 394 blowsUnit volume compaction work2.687 J/cm32.677 J/cm3 Table 5Physical properties of limestone coarse aggregatesParticle size/mmApparent density/(g·cm−3)Bulk density/(g·cm−3)Water absorption/%4.75–9.52.7422.7090.449.5–13.22.7282.7010.3813.2–162.7422.7130.3916∼192.7332.7130.2819∼26.52.7542.7150.53 Physical properties of limestone coarse aggregates Table 6Gradation and apparent density of limestone fine aggregatesSieve opening size/mm4.752.361.180.60.30.150.075Bulk density/(g·cm−3)Passing rate (%)4.75 mm10066.7742.9429.9824.0615.0512.022.7102.36 mm/10064.3144.9136.0322.5418.00 Gradation and apparent density of limestone fine aggregates Table 7Physical and mechanical properties of P·O 42.5 cementPropertiesDensity/(g·cm−3)Fineness/(m2·kg−1)Setting time/minCompressive strength/MPaFlexural strength/MPaInitial setting timeFinal setting time3 days28 days3 days28 daystest results3.06534515024022.654.54.89.1technical requirements≥300≥45≤600≥17.0≥42.5≥3.5≥6.5 Physical and mechanical properties of P·O 42.5 cement Table 8Mix proportion of the control CSM specimenParameterValueCement content (% of dry aggregate)5.27%Maximum aggregate size26.5 mmMineral powder content (<0.075 mm)0.06%Optimum moisture content3.87%Maximum dry density (g/cm3)2.391 Mix proportion of the control CSM specimen Table 9Compaction parameters (T0804-1994)Type of compaction testMethod A: compaction testMethod C: compaction - filling testCompaction objectfilling materialmixture of filling material and coarse aggregate (F-LSBC)Hammer parametersmass = 4.5 kg, diameter = 5.0 cm, drop height = 45 cm, gravitational potential energy = 19.845JCompaction cylinder dimensionsΦ 10.0 cm × H 12.7 cm = 997 cm3Φ 15.2 cm × H 12.0 cm = 2177 cm3Number of compaction blows27 blows/layer × 5 layers = 135 blows98 blows/layer × 3 layers = 394 blowsUnit volume compaction work2.687 J/cm32.677 J/cm3 Compaction parameters (T0804-1994) Discussion Packing characteristics and structure-performance trade-off The results demonstrate that the packing behavior of F-LSBC is governed by both the total skeleton void fraction and the dimensions of the individual voids. Although single-sized coarse aggregates of different sizes produced a similar VCADRC, larger aggregates formed larger voids that accommodated more fine material before the skeleton expanded. Combining 19–26.5 mm coarse aggregate with fine aggregate no larger than 2.36 mm therefore reduced VMADRC while preserving the coarse-aggregate framework. The stable filler-to-coarse aggregate ratio of 0.30 further indicates that the skeleton-dense state was controlled mainly by particle packing rather than cement dosage within the range studied. This packing structure produced a clear performance trade-off. F-LSBC contained a continuous coarse-aggregate skeleton, but its compressive, splitting tensile, flexural, and fatigue performance remained below that of CSM. In F-LSBC, the cementitious filler primarily occupied the voids between coarse particles, leaving relatively weak cemented contacts at the coarse aggregate-matrix interfaces. Continued curing narrowed the relative strength difference, particularly for splitting and flexural strength, but did not eliminate it. The lower fatigue performance was also consistent with the lower flexural strength and the greater susceptibility of the interfacial transition zone (ITZ) to damage.6,38,39 Thus, “strong skeleton” describes the geometric stability of the aggregate framework rather than higher bulk strength than CSM. In F-LSBC, the initial step involves blending cement with fine aggregates to produce a filler substance, which subsequently gets incorporated with larger aggregates to yield the end product. On the other hand, CSM employs a different approach where cement is added at the same time to both fine and coarse aggregates, actively engaging in the binding process of the complete aggregate framework. This variation in mixing techniques results in a higher density of cement hydration byproducts within the fine aggregate matrix of F-LSBC, while fewer of these products are allocated to the binding surfaces of the coarse aggregate structure. Since both materials utilize identical cement quantities, the total volume of hydration products accessible for binding coarse aggregates in F-LSBC is notably less than in CSM. Consequently, the coarse aggregate framework in CSM develops a more robust and cohesive formation, demonstrating enhanced load-bearing performance and structural integrity relative to F-LSBC. Given the predominant role of the coarse aggregate skeleton, this difference significantly impacts the overall mechanical properties of the materials. Mechanical properties and the role of interfacial transition zone The distinct structural configuration significantly impacts the mixture’s ability to withstand compression, leading to markedly reduced strength characteristics in F-LSBC specimens. Experimental data presented in Figure 8B demonstrate that F-LSBC achieves less than two-fifths of the compressive capacity observed in CSM samples. This substantial performance gap primarily stems from inferior bonding conditions at particle interfaces, which critically diminishes the material’s structural integrity under loading conditions. The weakened interparticle connections within the aggregate framework consequently undermine the composite’s capacity to distribute and resist applied forces effectively. Moreover, when subjected to splitting and bending loads where tensile stresses predominate, the ITZ surrounding coarse aggregates plays a crucial role in determining composite strength. While the ITZ in F-LSBC exhibits inferior mechanical properties, it should be recognized that conventional concrete (CSM) at equivalent curing periods also demonstrates incomplete ITZ maturation. Consequently, F-LSBC demonstrates strength proportions between 40% and 55% of CSM’s values for both splitting and flexural resistance, as illustrated in Figure 8B. A particularly striking observation emerges from 7-day standard curing conditions: F-LSBC maintains a porous microstructure with severely compromised interfacial adhesion. This structural deficiency results in remarkably diminished splitting resistance. To conclude, while F-LSBC demonstrates inferior performance in unconfined conditions, nearly reaching negligible levels, merely 10.80% of CSM’s equivalent strength at identical curing duration (Figure 8). These findings underscore the paramount importance of interfacial characteristics in governing tensile behavior during early hydration stages. With prolonged curing duration, the cement hydration process advances steadily, causing a progressive buildup of hydration byproducts. While the share of these products that contribute to binding the coarse aggregate framework in F-LSBC remains comparatively limited, their total volume still experiences substantial growth as time passes. Considering that the baseline structural integrity of the coarse aggregate matrix in F-LSBC is not particularly high, the expanding quantity of hydration byproducts exerts a more noticeable reinforcing influence on its framework. Consequently, this leads to a more rapid enhancement in compressive strength compared to CSM. To illustrate, Figure 8B demonstrates that the compressive strength ratio between F-LSBC and CSM rises from 29.96% after 7 days of standard curing to 38.59% following 14 days of accelerated curing. Likewise, the total volume of hydration products allocated to the ITZ around the coarse aggregates also grows with extended curing time. As a result, the ITZ in F-LSBC exhibits a swifter strengthening trend, suggesting that the improvement in interface performance due to hydration is more pronounced in F-LSBC compared to CSM. This phenomenon manifests through faster development rates in both splitting and flexural tensile strengths, leading to progressively widening performance gaps between F-LSBC and CSM in these mechanical properties as curing time extends. Furthermore, the optimization of cement dosage in the F-LSBC mixture follows the same design rationale. Drying shrinkage was measured only at the optimal cement dosage (5.27%) in the study. However, the selection of 5.27% is based on a clear strength inflection point and a transition in failure mode from lose to integral. Higher cement contents (5.94%) would increase shrinkage and reduce the crack-dissipating capacity of the weak interface, while lower contents (e.g., 4.55%) would not fully activate the coarse aggregate skeleton, leading to lower strength. Therefore, 5.27% is considered the optimal balance between strength and shrinkage resistance. Although F-LSBC exhibits lower compression, splitting tensile, and flexural strength than CSM across all curing periods, it shows a greater capacity for mechanical property enhancement with extended curing. Nevertheless, given its comparatively modest strength values, special attention should be paid to system compatibility and load distribution characteristics when employing F-LSBC as a foundational layer in road construction projects. Although the absolute mechanical strengths of F-LSBC were measured using unconfined compression, splitting tensile, and flexural tests, they were consistently lower than those of conventional CSM. This reduction is consistent with the proposed weak-interface design, which intentionally limits interfacial bonding to promote the dissipation of shrinkage-induced strain energy while retaining the load-bearing function of the coarse-aggregate skeleton. This strength reduction, while intentional, raises a critical question: how does this weak-interface design affect the long-term performance of F-LSBC under cyclic loading? To address this, we conducted fatigue tests to evaluate the material’s resistance to repeated traffic-induced stresses. Fatigue performance The fatigue performance relationships depicted in Figure 9 demonstrate that an increased vertical axis intercept corresponds to enhanced fatigue characteristics. This implies that materials with higher intercept values can endure greater cyclic loading before failure when subjected to equivalent stress conditions. The experimental data clearly reveals that F-LSBC specimens demonstrate markedly inferior fatigue durability compared to CSM samples, suggesting their diminished capacity to withstand cyclic stresses. A fundamental determinant of fatigue behavior is the material’s bending resistance capacity. Previous studies have demonstrated that the fatigue life of semi-rigid base materials is strongly correlated with their flexural strength, and materials with lower bending strength generally exhibit poorer fatigue resistance and shorter fatigue life.6 Materials exhibiting weaker bending performance tend to display poorer fatigue characteristics, ultimately leading to premature structural failure. As evidenced in Figure 8, the bending strength measurements for F-LSBC reach only about half of those recorded for CSM specimens. This substantial mechanical property disparity aligns perfectly with the fatigue test outcomes, providing compelling evidence for the crucial relationship between bending resistance and long-term cyclic loading performance. Fatigue failure does not occur as an immediate, single-event fracture phenomenon. Rather, it begins with the formation of minute defects in areas experiencing elevated stress or strain concentrations within the material. These microscopic imperfections progressively accumulate with cyclic loading until they culminate in visible structural failure. In cement-stabilized aggregate materials, the ITZ separating coarse aggregates from the cement matrix serves as the structural vulnerability, becoming particularly prone to microscopic fracture initiation and consequently governing the progression of fatigue deterioration. When comparing F-LSBC with conventional CSMs, the ITZ surrounding coarse particles in F-LSBC demonstrates a more porous microstructure and inferior mechanical performance. Specifically, the ITZ in F-LSBC displays an elastic modulus of merely 60–80% of that found in CSM, while its hardness measures only 55% of the CSM value. Additionally, the width of the ITZ in F-LSBC varies between 55 and 90 μm, substantially exceeding the typical 40-micrometer thickness observed in CSM. These distinctive properties suggest that the boundary region in F-LSBC possesses greater susceptibility to fracture formation. The propagation of cracks becomes more pronounced, increasing the likelihood of continuous failure pathways developing during fatigue loading. As a result, the fatigue resistance of F-LSBC shows a marked decline. Drying shrinkage and crack mitigation mechanisms In F-LSBC and CSM, the primary source of drying shrinkage is the cementitious matrix. While fine aggregates remain dimensionally stable, the mortar system—an amalgamation of fine aggregates and cement paste-experiences shrinkage collectively. In contrast, coarse aggregates not only maintain dimensional stability but also impose substantial restraint on the shrinkage behavior of adjacent mortar. Consequently, regarding shrinkage principles, increased coarse aggregate content or higher mass ratios lead to reduced void spaces within the aggregate framework, thereby diminishing the quantity of shrinkage-prone material and ultimately decreasing total drying shrinkage deformation. F-LSBC demonstrates a coarse aggregate mass fraction of 82.44%, markedly exceeding CSM’s 57.58%. The void ratio within the coarse aggregate skeleton measures 33.17% for F-LSBC versus 51.56% for CSM. These findings demonstrate that F-LSBC incorporates a greater proportion of coarse. The material forms compact aggregates and develops a more densely packed skeletal framework. Consequently, it demonstrates diminished deformation due to drying shrinkage and enhanced resistance to shrinkage effects. These characteristics align well with the previously examined experimental data regarding shrinkage strain measurements and coefficient values. Beyond the reduced void ratio in the coarse aggregate skeleton, a more pivotal aspect lies in the comparatively fragile interfacial connection between coarse aggregates and cement paste in F-LSBC. Previous nanoindentation analyses revealed that the ITZ in F-LSBC measures 55–90 μm in thickness, substantially exceeding the approximately 40 μm thickness found in conventional cementitious systems. Furthermore, this zone demonstrates merely of the elastic modulus 60–75% and about 55% of the hardness compared to standard cement mortar. These compromised mechanical characteristics render the interface particularly vulnerable to microcrack development when subjected to paste shrinkage, consequently diminishing the cohesion between aggregates and paste while decreasing the probability of aggregate movement following the contracting matrix. This weakened interface phenomenon partially obstructs the transfer of shrinkage strain to the aggregate framework, thereby amplifying the coarse aggregates’ capacity to restrain paste shrinkage. Consequently, the total drying shrinkage is significantly affected. Although F-LSBC has weaker ITZ characteristics, this design feature helps reduce the transmission of shrinkage strain, thus effectively reducing drying shrinkage and enhancing crack resistance.39 From the perspective of reflective crack propagation mechanisms in semi-rigid base asphalt pavements, previous studies have demonstrated that the initiation and propagation of reflective cracks are closely related to the accumulation and transfer of shrinkage-induced tensile strain energy in the base layer.2 A higher drying shrinkage strain leads to greater elastic strain energy stored within the base, which is subsequently transferred to the asphalt surface layer under structural constraints, thereby promoting crack propagation. In this study, the significantly reduced drying shrinkage strain observed in F-LSBC indicates a corresponding reduction in shrinkage-induced strain energy within the base layer. Consequently, the driving force for energy transfer to the surface layer is effectively weakened. This result suggests that F-LSBC has the potential to mitigate shrinkage-induced cracking and reduce one of the driving forces associated with reflective cracking. The suppression of deformation also results in a “crack-without-displacement” phenomenon, in which the weakened ITZ helps limit the transfer of shrinkage strain from the filler to the coarse aggregate skeleton, thereby reducing overall shrinkage deformation. However, it should be emphasized that reflective cracking in semi-rigid base asphalt pavements is a complex distress mechanism governed by multiple interacting factors, including drying shrinkage, thermal shrinkage, fatigue loading, and interlayer bonding conditions. Although drying shrinkage is an important factor controlling the initiation of shrinkage cracks in cement-stabilized bases, it does not solely determine the reflective cracking resistance of the pavement system. Therefore, the drying shrinkage results presented in this study should be interpreted as evidence of improved shrinkage resistance and potential anti-reflective cracking behavior, rather than as a direct and complete evaluation of reflective cracking performance. A comprehensive assessment of reflective cracking resistance should further incorporate thermal shrinkage behavior, repeated traffic loading effects, crack propagation under coupled environmental-traffic conditions, and the interface performance between pavement layers. The above reduction in drying shrinkage can be further interpreted from the perspective of crack propagation energy evolution. According to the energy-based reflective cracking mechanism, shrinkage-induced tensile stress in the semi-rigid base leads to the accumulation of elastic strain energy. When this stored energy exceeds the energy storage limit of the base, part of the energy is released and transferred to the asphalt surface layer, thereby providing the driving force for reflective crack initiation and propagation. A higher damage degree in the base material can reduce the energy transferred to and accumulated in the surface layer, consequently lowering the probability of reflective cracking. In the present study, the “weak-interface, strong-skeleton” structure of F-LSBC provides a material basis for this energy regulation mechanism. The strong coarse aggregate skeleton maintains load-bearing stability, while the relatively weak ITZs between the filler and coarse aggregates allow controlled early-stage microcracking under shrinkage-induced tensile stress. These distributed microcracks dissipate part of the accumulated elastic strain energy within the base layer, rather than allowing it to be released intensively and transferred upward to the asphalt surface layer. Meanwhile, the reduced drying shrinkage strain of F-LSBC indicates that the total shrinkage-induced tensile strain energy generated in the base layer is also lower. Therefore, from an energy evolution perspective, F-LSBC mitigates reflective cracking through two coupled mechanisms: reducing the generation of shrinkage-induced strain energy and promoting internal energy dissipation through weak-interface microcracking. This provides a mechanistic explanation for the observed improvement in shrinkage resistance and its potential contribution to reflective cracking mitigation. Limitations of the study Lower drying shrinkage may reduce one source of tensile strain energy associated with reflective-crack development,2 but the present results do not constitute a direct evaluation of reflective cracking. Thermal shrinkage, repeated traffic loading, interlayer bonding, and full pavement boundary conditions were not tested. In addition, the 180-day values were extrapolated from a 14-day dataset, and the CSM control contained only 0.06% material passing 0.075 mm. Therefore, the main conclusion is that the proposed F-LSBC mixture improves laboratory drying-shrinkage resistance at the cost of mechanical and fatigue capacity. Its engineering application should depend on whether this balance satisfies the structural and durability requirements of the intended pavement base. Beyond durability, the mix proportion design and performance prediction of F-LSBC can also benefit from emerging data-driven and numerical modeling approaches. For instance, machine learning-based multi-objective optimization, as recently demonstrated for recycled brick aggregate concrete,40 could be adapted to F-LSBC to efficiently balance multiple performance objectives—such as strength, fatigue life, drying shrinkage, and cost—thereby accelerating the mix design process. Furthermore, discrete element mesoscale modeling, developed for investigating size and shape effects in recycled aggregate concrete,41 offers a promising tool to explore the mesoscale fracture mechanisms of F-LSBC, particularly the influence of ITZ properties and void distribution on its deformation and failure behavior. Incorporating these methodologies into future work would further enhance the scientific basis and engineering applicability of F-LSBC as a crack-resistant base material in modern highway pavements. Resource availability Lead contact Requests for further information should be directed to and will be fulfilled by the lead contact, Dong Tang (17799301973@163.com). Materials availability This study did not generate new unique materials. Data and code availability •All experimental data reported in this paper are available from the lead contact upon reasonable request.•This paper does not report original code.•Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. •All experimental data reported in this paper are available from the lead contact upon reasonable request. All experimental data reported in this paper are available from the lead contact upon reasonable request. •This paper does not report original code. This paper does not report original code. •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Acknowledgments This work was supported by the Research and Application of Intelligent Paving and Compaction Machine Cluster Collaborative Construction Technology for Highway Pavements in Xinjiang (grant no. 2024LQ04002), the Science and Technology Support Xinjiang Project Plan (grant no. 2024E02035), the Youth Science Foundation of Xinjiang Uygur Autonomous Region Natural Science Foundation (grant no. 202435120003), Science and Technology R&D Project of Xinjiang Transportation Investment (Group) Co., Ltd (grant no. ZKX-FWCG-202212-018), the Xinjiang Uygur Autonomous Region Tianchi Talent Young Doctor Program (grant no. 5105250182q), the 10.13039/501100004735Natural Science Foundation of Hunan Province, China (grant no. 2025JJ60342), the Open Fund of State Key Laboratory of Road Engineering in Extreme Environment (Changsha) (grant no. kfj250107), the Science and Technology Research and Development Project of Xinjiang Transportation Investment (Group) Co., Ltd. (grant no. XJJTKX-FWCG-202412-0802). General Program of the 10.13039/100009110Natural Science Foundation of Xinjiang Uygur Autonomous Region. Author contributions D.T.: visualization, validation, and methodology. Z.Y.: validation, resources, and methodology. H.W.: visualization, methodology, and investigation. L.M.: visualization, validation, and resources. Y.Z.: visualization, validation, investigation. G.Z.: review and editing. N.Z.: writing – original draft, visualization, validation, supervision, software, resources, methodology, investigation, and funding acquisition. Declaration of interests The authors declare no competing interests. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERChemicals, peptides, and recombinant proteinsP·O 42.5 ordinary Portland cementCommercial supplier, ChinaGB 175-2007Software and algorithmsOriginOriginLab Corporation, Northampton, MA, USAhttps://www.originlab.com/Microsoft ExcelMicrosoft Corporation, Redmond, WA, USAhttps://www.microsoft.com/OtherLimestone coarse aggregatesCommercial quarry, ChinaJTG E42-2005Limestone fine aggregatesCommercial quarry, ChinaJTG E42-2005 REAGENT or RESOURCESOURCEIDENTIFIERChemicals, peptides, and recombinant proteinsP·O 42.5 ordinary Portland cementCommercial supplier, ChinaGB 175-2007Software and algorithmsOriginOriginLab Corporation, Northampton, MA, USAhttps://www.originlab.com/Microsoft ExcelMicrosoft Corporation, Redmond, WA, USAhttps://www.microsoft.com/OtherLimestone coarse aggregatesCommercial quarry, ChinaJTG E42-2005Limestone fine aggregatesCommercial quarry, ChinaJTG E42-2005 Method details Materials Coarse aggregates The Technical Specification for F-LSBC(T/CECS G: K23-01-2019) requires that coarse aggregates be single-sized ultra-large particles, such as those with nominal maximum sizes of 26.5–37.5 mm, 31.5–53 mm, 37.5–63 mm, and 53–73 mm. However, in practical construction, while using larger aggregates is technically feasible and modern construction machinery can handle them, they still pose challenges in mixing uniformity, particularly in ensuring consistent filler distribution. These issues can lead to lower efficiency during the mixing and compaction stages, ultimately affecting construction quality. This is especially true when the particle size exceeds the optimal range for uniform distribution in the mixture (as shown in Figure S1). Therefore, using such large aggregates may require additional steps to ensure consistent mixing, thereby impacting overall construction efficiency. To provide a quantitative basis for the selection of 19–26.5 mm aggregates over larger sizes (e.g., 31.5–53 mm), a comprehensive literature review was conducted across four key dimensions: mixing uniformity, construction efficiency, compaction quality, and cost.(1) Based on an actual engineering analysis of cement-stabilized crushed stone bases, the coefficient of variation (CV) of key strength-related factors—including the coarse aggregate fraction, cement content, and dry density—showed significant variability and was strongly correlated with overall strength variability.When the nominal maximum aggregate size exceeds 31.5 mm, aggregate segregation becomes more pronounced, which is recognized as a critical cause of decreased compactness and strength in road bases.22,32 Specifications and engineering guidelines further require that the maximum aggregate size for cement-stabilized base courses shall not exceed 31.5 mm to ensure construction quality.32 Therefore, the use of 31.5–53 mm may introduce greater variability and potential performance instability.30(2) According to the specification T/CECS G:K23-01-2019 for F-LSBC bases, larger aggregates inherently require thicker paving layers to maintain skeleton integrity.31 A reduced layer thickness can significantly increase paving speed and reduce the number of required compaction passes. Specifications also stipulate that the minimum compacted layer thickness for cement-stabilized base courses shall not be less than 15 cm, and effective paving methods must be determined to avoid material segregation.32 The use of oversized aggregates thus reduces construction efficiency and increases project duration.30,31(3) For mixtures containing aggregates larger than 31.5 mm, achieving a compaction degree ≥98% requires stricter control of rolling passes and layer thickness.32 For skeleton-dense mixtures, the compaction degree of cement-stabilized base courses shall not be less than 98% under the heavy compaction standard.32 In contrast, for 19–26.5 mm aggregates, a compaction degree of 98–99% is achieved more reliably under standard rolling procedures. The lower compaction efficiency of large aggregates is attributed to increased interlocking resistance and greater void heterogeneity, which hinder aggregate rearrangement under compaction energy.30,42 (4) The processing cost for 31.5–53 mm aggregates is substantially higher than for 19–26.5 mm aggregates, primarily due to higher crushing energy requirements and increased wear on screening and handling equipment.31 Moreover, the greater risk of segregation during mixing and transport necessitates more frequent quality control inspections and potential rework, adding to overall production cost.22,30 Aggregate segregation can lead to decreased compactness and compromised structural integrity, ultimately increasing long-term maintenance costs.22 A comprehensive cost–performance analysis indicates that the selected 19–26.5 mm aggregates represent a balanced trade-off between structural performance, construction feasibility, and economic viability.27,31 Based on the above quantitative comparison, the selection of 19–26.5 mm coarse aggregates—rather than larger sizes such as 31.5–53 mm—is justified as a construction-friendly yet performance-effective alternative that does not compromise the intended “weak-interface, strong-skeleton” design philosophy.27,31 In view of this, this study reduces the coarse aggregate size and selects five single-sized limestone aggregates with sizes of 4.75–9.5 mm, 9.5–13.2 mm, 13.2–16 mm, 16–19 mm, and 19–26.5 mm, respectively. The physical property indices (including apparent density, bulk density, and water absorption) of these aggregates were determined in accordance with Test Methods of Aggregates for Highway Engineering (JTG E42-2005), and the specific results are listed in Table 5. Fine aggregates The filler consists of fine aggregate, cement, and water. The particle size of fine aggregate exerts a significant interference effect on the skeleton structure: the smaller the particle size, the lesser its impact on skeleton stability. Two types of limestone fine aggregates were selected in this study, with maximum nominal particle sizes of 4.75 mm and 2.36 mm, respectively. The latter was obtained by sieving the former through a 2.36 mm square mesh sieve. The detailed gradation compositions (sieve passing rates) and apparent density test results of the two fine aggregates are presented in Table 6. Cement This study uses P·O 42.5 ordinary Portland cement as the binder in the filler. Its physical and mechanical properties were tested in accordance with General Portland Cement(GB 175–2007), and all test results met the specification requirements, as detailed in Table 7. Control CSM mixture For comparison, conventional cement-stabilized macadam (CSM) specimens were prepared with the same total cement content (5.27%) as the F-LSBC mixture. The aggregate gradation followed the skeleton-dense type recommended for heavy-traffic highways in JTG/T F20-2015,32 with a nominal maximum aggregate size of 26.5 mm. The mix composition is summarized in Table 8, while the aggregate gradation is presented in Figure S2. The mineral powder content (passing the 0.075 mm sieve) was 0.06%, which is lower than the 2–5% range suggested by the specification due to limited availability of fine limestone powder. The optimum moisture content (3.87%) and maximum dry density (2.391 g/cm3) were determined by heavy compaction tests. CSM cylindrical specimens (Φ150 mm × H150 mm) were prepared by static pressing under the same compaction work as the F-LSBC specimens, and cured under identical conditions (20 ± 2°C, ≥95% RH) for 7, 14, or accelerated 14 days before testing. Calculation of the total volume of fine aggregates and cement (Equation 5)VFA+P·Ostacking=αCAFiller·(1−VCAWCC)·ρCAbulk·VMold/(1+wop%)/ρdrymax This study focuses on the trend of the bulk volume VFA+P·Ostacking of fine aggregates and cement rather than its absolute value. Thus, the ratio of VFA+P·Ostacking to the volume VMold of the compaction cylinder (2177 cm3) is used as a relative evaluation index, Theoretical packing model In F-LSBC, the coarse aggregate skeleton is formed by the packing of single-sized particles; thus, their packing characteristics need to be analyzed first. The aggregate particles are approximately spherical, allowing their packing behavior to be simplified to that of equal-sized spheres. Particle packing can be categorized into two types: ordered packing and disordered packing. Disordered packing further includes random packing and arbitrary packing. Random packing neglects particle self-weight and interparticle interactions and has only theoretical significance; arbitrary packing—defined as random packing under gravitational fields—is standard in practical engineering (e.g., aggregate grading for road infrastructure). Arbitrary packing can be viewed as a combination of multiple ordered packing units; thus, clarifying the ordered packing patterns of equal-sized spheres is a prerequisite. The ordered packing of equal-sized spheres primarily comprises five types: cubic, hexagonal (orthorhombic), double hexagonal, pyramidal, and tetrahedral packing. This study begins by investigating the simplest cubic packing configuration. The packing characteristics of particles are primarily described by three parameters: packing density, void fraction, and coordination number. The total packing volume consists of the solid volume of particles and the void volume between them. Packing density refers to the percentage of solid volume in the total packing volume, while void fraction is the percentage of void volume—together, they account for 100%. Coordination number, originally a chemical concept, is used here to characterize interparticle contacts: it defines the number of particles directly touching a given “central” particle. A particle’s role as “central” or “surrounding” depends solely on its relative spatial position. For example, cubic packing has a coordination number of 6, meaning each particle is directly in contact with six neighboring particles (see the upper schematic in Figure S3). First, the coarse aggregate particles with a particle size of D are simplified to spherical particles with a diameter of D, and their volume Vsphere is:(Equation 6)Vsphere=πD3/6 As shown in the lower part of Figure S3, eight spherical particles with a diameter of D are packed in a cubic arrangement. The lines connecting their centers form a unit cube with an edge length of D, and its volume Vcell is:(Equation 7)Vcell=D3 Within a unit cube, the eight spheres are located at its vertices, with each sphere contributing only 1/8 of its volume to the cube. Therefore, the total solid volume within the cube Vsolidcell can be expressed as:(Equation 8)Vsolidcell=Vsphere/8×8=πD3/6 From Equations 8 and 9, it can be seen that the void volume of the particles within the unit cube Vvoidcell is:(Equation 9)Vvoidcell=Vcell−Vsolidcell=(1−π/6)D3 Therefore, based on (Equation 7), (Equation 8), (Equation 9), the packing density scell and void fraction of particles within the unit cube vcell can be calculated as:(Equation 10){scell=Vsolidcell/Vcell=π/6vcell=Vvoidcell/Vcell=1−π/6 Based on the unit cube’s packing characteristics, the results can be extrapolated to the entire packing system. The overall packing density sparticle and void fraction of particles vparticle remain consistent with those in the unit cube, as shown in Equation 11.(Equation 11){sparticle=scell=π/6vparticle=vcell=1−π/6 From Equation 11, it can be seen that in the cubic packing of equal-sized spheres, vparticle is a constant independent of sphere diameter. For example, though the two stacked cubes (left and right) in Figure S3 have different diameters, vparticle remains consistent. Furthermore, in the entire packed cube, the average void volume per particle Vvoid¯ is equal to the particle void volume of the unit cube Vvoidcell, as shown in Equation 12.(Equation 12)Vvoid¯=Vvoidcell=(1−π/6)D3 Based on the geometric relationships in the lower part of Figure S3, the maximum packing diameter of secondary particles within Vvoid¯ can be derived using the properties of the cube—this diameter corresponds to the sphere diameter dvoid¯ associated with Vvoid¯—as shown in Equation 13.(Equation 13)dvoid¯=(3−1)D=0.732D From Equations 12 and 13, it can be seen that as the particle diameter D increases, Vvoid¯ increases exponentially while dvoid¯ increases linearly. This is because, under the condition that vparticle remains constant, the total volume of particle packing stays the same, and the total void volume also remains unchanged; however, the increase in particle diameter reduces the number of voids per unit volume, thereby causing both Vvoid¯ and dvoid¯ to increase accordingly. For example, in the two stacked cubes (left and right) shown in Figure S3, the left cube has larger particles and fewer voids, so both Vvoid¯ and dvoid¯ in the left cube are greater than those in the right cube. Previously, the simplest ordered packing—cubic packing of equal-sized spheres—was discussed. Additionally, four other types of ordered packing exist for equal-sized spheres: hexagonal packing, double hexagonal packing, pyramidal packing, and tetrahedral packing. By referencing the computational model of cubic packing (Figure S3), the characteristics of these other ordered packings can be derived, and the results are presented in Table 1. Dry-rodded compaction test In road engineering, aggregates are actually irregular particles, and their gradation falls into random packing rather than the ordered arrangement of ideal spheres. Given that F-LSBC relies on equal-sized coarse aggregates to form the skeleton, this paper determines the skeleton void ratio of coarse aggregates in the dry rodded condition—denoted as VCADRC via compaction tests. The test followed method T0309-2005: single-graded coarse aggregates of different particle sizes were layered into a 10 L capacity cylinder (three layers total), with each layer uniformly rodded 25 times using a tamping rod. After scraping the mouth of the cylinder flat, the total mass was weighed. The skeleton void ratio of coarse aggregates in the dry rodded state was calculated using Equation 14:(Equation 14)VCADRC=1−ρCAstacking/ρCAbulkWhere: VCADRC is the skeleton void ratio of coarse aggregates in the dry rodded state; ρCAbulk is the bulk density of coarse aggregates, in g/cm3, as listed in Table 6; ρCAstacking is the packing density of coarse aggregates, in g/cm3, measured experimentally.The compaction test diagrams for the five single-sized coarse aggregates (4.75–9.5 mm, 9.5–13.2 mm, 13.2–16 mm, 16–19 mm, and 19–26.5 mm) are shown in Figures S4A–S4E, respectively. Compaction-filling test To determine the optimal proportion of coarse and fine aggregates for forming a skeleton-dense structure, a Compaction-Filling Test was designed. The mass of coarse aggregates was kept constant, while the mass ratio of fine aggregates (K, fine aggregate/coarse aggregate mass ratio) was gradually increased. After uniformly mixing the mixture, it was filled into a capacity cylinder in three layers and compacted densely using the previously described compaction method. Once the total mass of the mixture was weighed, the skeleton void ratio (VCADRC) and the void ratio in mineral aggregates (VMADRC)—denoted as Voids in Mineral Aggregates in Dry Rodded Condition—were calculated for the dry rodded state using Equation 15, respectively.(Equation 15){VCADRC=1−mCA/ρCAbulk/VMVMADRC=1−(mCA/ρCAbulk+mFA/ρFAbulk)/VMWhere: mCA and mFA are the masses (in grams) of coarse aggregates and fine aggregates, respectively, determined experimentally; ρCAbulk and ρFAbulk are the bulk densities (in g/cm3) of coarse aggregates and fine aggregates, respectively, as listed in Tables 5 and 6. Refer to Figure S5 for the Compaction-Filling Test diagram. To ensure a reproducible determination of the critical fine-to-coarse aggregate mass ratio, the following criterion is adopted: The skeleton-dense structure is achieved at the fine/coarse mass ratio where VMADRC (voids in mineral aggregates under dry rodded condition) attains its minimum value, provided that the corresponding VCADRC (skeleton void ratio of coarse aggregates) does not increase by more than 1.0% relative to the initial VCADRC of the pure coarse aggregate skeleton. The 1.0% threshold is chosen based on the typical repeatability error range of the dry-rodded compaction test (approximately 0.5–1.0% based on laboratory experience). This ensures that any increase within this range can be attributed to normal test variability rather than genuine skeleton expansion, thereby preserving the “strong-skeleton” characteristic. This criterion is applied consistently to all compaction-filling tests in this study. Filler compaction and combined-grading procedure According to the F-LSBC material composition design, the filler material is prepared by mixing fine aggregates, cement, and water, which is then combined with coarse aggregates to form the final mixture. Therefore, in addition to the coarse aggregate particle size and fine aggregate grading, the determination of F-LSBC’s combined grading requires clarification of: (1) the mix proportion of the filler material (i.e., cement content and optimum moisture content); and (2) the mass ratio of filler material to coarse aggregates. To address this, this study first conducts a compaction test on the filler material, followed by a compaction-filling test involving both the filler material and coarse aggregates. The detailed experimental scheme is presented in Figure S6. Test methods Compaction test of filler material Determine the mix proportion of the filler material—including cement content and optimum moisture content—via compaction tests. In the experiments, cement contents of 10%, 15%, 20%, 25%, 30%, and 35% were set. For each cement content, compaction tests were conducted on the filler material to determine its corresponding optimum moisture content. Before testing, fine aggregates and cement were dried to constant weight at 105 ± 5°C, then cooled to room temperature for later use. The dried fine aggregates were uniformly mixed with water at a predetermined moisture content and sealed for impregnation for approximately 4 h. Cement was then added and mixed uniformly, followed by immediate compaction testing. To minimize the impact of cement hydration on results, cement was added before compaction, and all compaction operations were completed within 1 h. Given that the maximum particle size of the fine aggregates is 2.36 mm (falling within the fine-grained soil category), the light compaction method (Method A) was employed to improve efficiency. Specific test parameters are provided in Table 9. After compaction is complete, take two representative samples from different depths of the sample for water-content determination, and use the average of these two samples as the sample’s water content. Then, calculate the dry density ρdry of the filling material according to Equation 16.(Equation 16)ρdry=mFP/VMold/(1+w%)Where: mFP is the mass of the filling material in the compaction cylinder, in grams, determined directly after compaction; w% is the water content of the filling material, i.e., the ratio of water mass to the total mass of fine aggregates and cement; ρdry is the dry density (bulk density) of the filling material, in g/cm3; VMold is the volume of the compaction cylinder, in cm3. As shown in Table 9, the volume of the compaction cylinder for Method A is 997 cm3. Based on the aforementioned water content w% and dry density ρdry, plot the “water content - dry density curve” of the filling material, namely the compaction curve. The water content and dry density corresponding to the peak value of the curve are the optimum water content wOP% and the maximum dry density ρdrymax , respectively. Accordingly, the mix proportion parameters of the filling material can be determined: the cement dosage c% of the filling material and its corresponding optimum water content wOP%. Compaction-filling test of filler material and coarse aggregates After determining the filler proportion through the compaction test (Method A), it is mixed with coarse aggregates at preset mass ratios (0.2, 0.25, 0.275, 0.3, and 0.325). The values of these ratios are based on the tamping-filling test results. A mass ratio of 0.25 for fine aggregates to coarse aggregates results in a skeleton-dense state of the mineral aggregate mixture. The masses of each component in F-LSBC are calculated based on the compaction test and the compaction-filling test, following standard experimental procedures.” Formula-based calculations are omitted for brevity, and the corresponding parameters can be found in the experimental design section. The theoretical principles and calculations are based on established methods for determining mixture compositions in cement-stabilized materials. Specimen preparation and curing For this study, two types of cylindrical specimens were prepared based on the maximum dry density: filler material specimens and F-LSBC specimens. Given that the maximum particle size of fine aggregates is 2.36 mm (classified as fine-grained soil), the size of filler material specimens was determined to be Φ50 mm × H50 mm. In contrast, F-LSBC—containing coarse aggregates (classified as coarse-grained soil)—had a specimen size of Φ150 mm × H150 mm. In the tests, the cement content of the filler material (denoted as c%, representing the mass ratio of cement to fine aggregates) was set to six levels: 10%, 15%, 20%, 25%, 30%, and 35%. The maximum dry density and optimum water content of the filler material under each cement dosage have been determined, as shown in Figure 5. Therefore, the required mass of filler material for specimen preparation is:(Equation 17)mFiller=(1+wOP%)·ρdrymax·VMoldWhere VMold is the volume of the cylindrical specimen mold, in cm3. Since both the diameter and height of the filler material specimen are 50 mm, this volume is thus 98 cm3. Based on this, the mass of each component of the filler material can be further calculated as follows:(Equation 18){mFA=mFiller/(1+wOP%)/(1+c%)mP·O=c%·mFiller/(1+wOP%)/(1+c%)mW=wOP%·mFiller/(1+wOP%)In F-LSBC specimens, the mass ratio of filler material to coarse aggregates is fixed at 0.3. Their maximum dry density and optimum water content have been determined through tests (Figure 5). Based on these parameters, the total mass of filler material and coarse aggregates required for preparing F-LSBC specimens can be calculated, with the calculation equation as follows:(Equation 19)MCA+Filler=[1+αCAFiller·wop%/(1+αCAFiller+wop%)]·ρdry′·VMoldWhere VMold is the volume of the cylindrical specimen mold, in cm3. Since both the diameter and height of the F-LSBC specimen are 150 mm, this value is thus 2651 cm3. Subsequently, the masses of coarse aggregates and filler material can be calculated separately:(Equation 20){mCA=MCA+Filler/(1+αCAFiller)=ρdry′·VMold·[1+αCAFiller·wop%/(1+αCAFiller+wop%)]/(1+αCAFiller)mFiller=αCAFiller·MCA+Filler/(1+αCAFiller)=αCAFiller·ρdry′·VMold·[1+αCAFiller·wop%/(1+αCAFiller+wop%)]/(1+αCAFiller) After determining the mass of the filler material mFiller, the masses of its components (fine aggregates, cement, and water) are calculated using Equation 18. After determining the masses of the filler material and F-LSBC specimens, cylindrical specimens were statically pressed using a pressure machine in accordance with the Test Method for Preparation of Cylindrical Specimens of Inorganic Binder Stabilized Materials (T0843-2009), as shown in Figure S7. Specimens are formed using the static pressing method (Figure S7). First, fine aggregates, cement, and water are mixed uniformly according to the designed mix proportion, then divided into two parts: One part is directly statically pressed into filler material specimens; the other is mixed with coarse aggregates in a saturated surface-dry (SSD) state to prepare F-LSBC specimens via pressing. After molding, specimens are cured in accordance with Specification T0845-2009: they are placed in a standard curing chamber at 20 ± 2°C with relative humidity ≥95% until reaching the specified age. Upon completion of curing, specimens are immersed in water at 20 ± 2°C for saturation treatment. Macroscopic mechanical properties test The research conducted a comprehensive analysis of the mechanical characteristics of F-LSBC in comparison to traditional CSM materials, with particular emphasis on three crucial parameters: compressive strength under unconfined conditions, resistance to splitting tension, and bending capacity. For assessing compressive and splitting tensile properties, cylindrical samples were employed, whereas prismatic beams were utilized to measure flexural performance. Three distinct curing conditions were implemented for all sample types: normal curing periods of 7 and 14 days, along with an accelerated 14-day curing process. Each experimental group consisted of five identical specimens, with the mean value serving as the representative measurement. The investigation involved the preparation of 60 cylindrical samples and 30 beam specimens. All mechanical evaluations, including unconfined compression, splitting tension, and flexural tests, were performed using MTS testing equipment, strictly following the standardized testing protocols outlined in the relevant methodology guidelines. Fatigue performance test Following the standardized procedures outlined in the Highway Engineering Test Methods for Inorganic Binder-Stabilized Materials (JTG E51-2009), researchers employed an MTS testing apparatus to perform four-point bending fatigue evaluations on both F-LSBC and CSM specimens. Considering the substantial time required for fatigue testing, all samples underwent a uniform 14-day curing period under standard conditions, omitting the typical 90-day aging process. Previous studies have shown that, although the strength and fatigue life of cement-stabilized macadam at an early curing age (e.g., 14 days) have not yet reached their ultimate values, the fundamental characteristics of fatigue damage evolution have already been established and can be used for comparative analysis.38 Specifically, the study demonstrated that, during the early curing stage, the fatigue life exhibited clear and consistent relationships with loading level and loading frequency, while the evolution trends of modulus degradation, cumulative strain, and damage variables were capable of reliably reflecting the load-bearing performance of the material. Furthermore, the model fitting results indicated that early-age experimental data could be effectively used to predict fatigue damage evolution under different conditions. These findings suggest that fatigue data obtained at a curing age of 14 days are sufficiently representative for evaluating the early-age fatigue performance of semi-rigid base materials and for conducting comparative analyses among different mixtures. The experimental design focused primarily on contrasting the fatigue characteristics between F-LSBC and CSM materials, making identical curing durations for both material types adequate for achieving valid comparative results. Stress-controlled loading patterns were implemented during testing, utilizing a continuous Haversine wave pattern at a frequency of 10 Hz. Four distinct stress levels were chosen according to each material’s bending resistance capacity. The comparatively reduced flexural strength of F-LSBC makes its samples particularly vulnerable to premature failure when subjected to elevated stress conditions. Consequently, stress levels of 0.35, 0.45, 0.50, and 0.70 were selected for testing. In the case of CSM samples, different stress parameters of 0.65, 0.70, 0.75, and 0.80 were employed. The different stress ratio ranges were necessary because the flexural strength of F-LSBC is substantially lower than that of CSM (approximately 40–45% of CSM, as shown in Figure 8). Using the same stress ratio would result in absolute stress levels that are either too high for F-LSBC (causing immediate fracture before fatigue cycling) or too low for CSM (resulting in fatigue lives exceeding 107 cycles, which are impractical to measure within a reasonable testing time frame). Therefore, direct fatigue life comparison at identical absolute stress levels is not feasible with the current dataset, as extrapolation beyond the tested stress ratio ranges would introduce significant uncertainty.Each stress level was examined using three identical samples, with environmental conditions kept at around 20°C. The experimental procedure continued until clear material failure was observed. Furthermore, for test pieces demonstrating extended fatigue resistance and requiring prolonged examination periods, a protective plastic covering was applied to maintain surface moisture and prevent dehydration, ensuring consistent testing conditions and reliable results. Dry shrinkage properties test Following the standardized procedures outlined in (JTG E51-2009) for evaluating inorganic binder-stabilized road construction materials, comparative drying shrinkage analyses were performed on F-LSBC and CSM samples. The experimental protocol involved fabricating six prism-shaped test pieces per material variant, which underwent a 7-day standardized curing process. Three samples from each set were designated for moisture content determination through weight monitoring, with the other three allocated for dimensional variation assessment to quantify shrinkage deformation. All measurements were conducted in a precisely regulated environmental chamber maintaining 20 ± 1°C temperature and 60 ± 5% humidity over a two-week period. Daily observations documented both mass fluctuations and linear dimensional changes. Utilizing the prescribed calculation methods from the technical standard, four key parameters were derived: moisture evaporation rate, contraction extent, deformation strain, and mean shrinkage factor.