Accelerated CO2 curing of fly ash–GGBS self-compacting geopolymer concrete: experimental, environmental and explainable machine learning assessment

The construction industry is under increasing pressure to develop sustainable cement-free materials with lower carbon emissions and enhanced durability. In this study, a high-performance self-compacting geopolymer concrete (SCGC) was developed using Fly Ash (FA) and Ground Granulated Blast Furnace Slag (GGBS) as binder precursors, combined with accelerated CO 2 curing to improve engineering performance and environmental sustainability. Five binder compositions with varying FA-to-GGBS ratios were prepared under ambient and carbon curing conditions, resulting in ten experimental mixtures. The fresh properties of the developed mixtures were evaluated according to EFNARC guidelines, while mechanical performance was assessed through compressive, split tensile, and flexural strength tests at 7, 28, and 56 days. Durability performance was investigated using water absorption, sorptivity, acid resistance, and sulfate resistance tests. Carbon sequestration behavior was quantified through mass gain analysis, while microstructural evolution was examined using Scanning Electron Microscopy (SEM), Energy Dispersive Spectroscopy (EDS), X-Ray Diffraction (XRD), and Fourier Transform Infrared Spectroscopy (FTIR). The environmental performance of the developed SCGC mixtures was evaluated through cradle-to-gate life cycle assessment (LCA) using openLCA. In addition, machine learning (ML) models including Artificial Neural Network (ANN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) were developed to predict compressive strength. The results demonstrated that accelerated CO₂ curing significantly enhanced the fresh, mechanical, and durability performance of SCGC. Among all investigated mixtures, the M3-C mixture exhibited the optimum performance, achieving a compressive strength of 68.4 MPa at 56 days along with significantly reduced water absorption and sorptivity. Microstructural investigations confirmed enhanced geopolymerization, pore refinement, and carbonate formation in carbon-cured mixtures. LCA revealed that M3-C exhibited the lowest global warming potential (148.8 kg CO 2 -eq/m 3 ), low embodied energy (2384 MJ/m 3 ), and the lowest carbon intensity (2.38 kg CO 2 /MPa·m 3 ) among all investigated mixtures. ML analysis demonstrated excellent predictive capability, with XGBoost showing the highest prediction accuracy (R 2 = 0.983). The integrated experimental, microstructural, environmental, and data-driven findings confirm that carbon-cured slag-modified SCGC has strong potential as a sustainable next-generation construction material for low-carbon infrastructure applications.