Air-Void Stability in Self-Compacting Concrete: Linking Fresh-Air Retention with Hardened Pore Structure Through a Synthetic Dispersion Approach

Air entrainment in self-compacting concrete (SCC) is governed by coupled interactions between chemical admixtures, empirical workability behaviour, aggregate-skeleton geometry and early air-bubble stability. In highly flowable mixtures, the hardened air-void system cannot be assessed reliably from total air content alone because bubble escape, redistribution and coalescence in the fresh state may change the final pore structure. This study evaluates the link between early fresh-air retention and hardened air-void characteristics in 25 SCC mixtures arranged according to a five-level Graeco-Latin square design. The analysed factors were air-entraining admixture (AEA) dosage (0.00–0.20% by mass of cement), binder type, water-to-binder ratio (0.29–0.41) and the volumetric paste-to-aggregate filling parameter φ (1.1–1.5). The aggregate skeleton was kept constant to separate paste-composition and volumetric-filling effects from aggregate grading. Fresh concrete was characterised by slump-flow diameter, T50 flow time, density and air content after 5 and 15 min; these quantities were treated as empirical workability and early-retention indicators, not as direct rheological parameters. Hardened concrete was examined after 28 days according to EN 480-11 using total hardened air content A, spacing factor L, micropore content A300 and specific surface α. The slump-flow diameter ranged from 50 to 79 cm, fresh air content after 5 min from 1.6% to 8.6%, air loss between 5 and 15 min from 0.41 to 1.12 percentage points, hardened air content from 1.20% to 8.59%, and spacing factor from 0.13 to 0.44 mm. Strong correlations were obtained between fresh and hardened air contents (A5 vs. A: r = 0.920, R2 = 0.846, p < 0.001, 95% CI for r: 0.824–0.964; A15 vs. A: r = 0.922, R2 = 0.849, p < 0.001, 95% CI for r: 0.828–0.965), while hardened air content was strongly and inversely related to spacing factor (A vs. L: r = −0.907, R2 = 0.822, p < 0.001, 95% CI for r: −0.958 to −0.797). The recalculated ANOVA showed that statistical significance was response-dependent: w/b was significant for early air loss ΔA (F = 4.190, p = 0.040, partial η2 = 0.677) and micropore content A300 (F = 4.058, p = 0.044, partial η2 = 0.670), whereas binder type showed near-threshold tendencies for fresh and hardened air contents. No single factor was statistically significant for all air-void descriptors. The SDI-based approach is therefore presented as a bounded explanatory framework, not as an externally validated prediction model. Direct durability claims, including freeze–thaw resistance, require separate experimental verification.