Low-carbon core-shell cold-bonded lightweight aggregates using expanded polystyrene and limestone calcined clay cement: Particle-scale performance, sustainability, and physics-informed prediction

Cold-bonded core-shell lightweight aggregates (CSLWAs) using an expanded polystyrene (EPS) core with a limestone calcined clay cement (LC3) shell offer a promising route to low-carbon lightweight aggregates; however, the coupled effects of core size, shell thickness, LC3 composition, and curing regime on particle-scale performance remain insufficiently understood. This study develops EPS-LC3-based CSLWAs by combining three discrete EPS core diameters (8, 10, and 12 mm) with three LC3 shell binders of 30%, 50%, and 70% cement content (denoted LC30, LC50, and LC70). The aggregates were cured under four regimes (air, sealed, moist, and hot water) and characterized for water absorption, particle density, single-particle compressive strength, loose bulk density, and bulk compressive response; a cradle-to-gate assessment compared the embodied carbon and cost of the three binders. The shell-thickness-to-radius ratio (t/R) dominated particle performance, with a transition near t/R ≈ 0.4, below which premature shell instability was more frequent. Increasing EPS core size lowered both bulk density and crushing strength; loose bulk densities of 630–880 kg/m³ met the ASTM C330 limit for coarse lightweight aggregate, with bulk crushing strengths of 4.4–8.9 MPa. LC50 gave the best strength-density efficiency, whereas LC30 reduced embodied carbon by up to 48%. Moist and hot water curing performed best. A physics-informed neural network embedding shell-thickness, density-volume, and curing constraints predicted particle strength and density (R² = 0.955 and 0.951), reducing cross-fold R² variability by about 44% and 35%, respectively, versus a data-driven baseline. These provide a controlled framework for designing low-carbon CSLWAs with a balanced strength, density, and carbon profile and a data-efficient prediction model.