Towards Sustainable Cement Alternatives: An Interdisciplinary Study of Alkali-Activated Materials

The project aims to advance sustainable alternatives to Portland cement by developing the knowledge of alkali-activated materials (AAMs). Cement production alone accounts for nearly 8% of global CO₂ emissions, primarily due to high-temperature limestone processing. By contrast, AAMs can be synthesized from industrial by-products, natural aluminosilicates and even municipal household waste ash under milder conditions, offering significant potential for reducing CO₂ emissions and optimizing resources in the construction sector. AAMs also exhibit excellent fire resistance, making them particularly promising for applications such as underground car parks and high-rise buildings. However, their largely amorphous structure makes characterization and property prediction highly challenging. This project addresses this gap through a workflow combining synthesis, spectroscopy, diffraction, structural modelling, and machine learning potentials.
In Work Plan (WP) 1, AAMs will be synthesized and systematically characterized, with vibrational spectroscopy as a central tool. In WP2, electron diffraction and electron pair distribution function measurements performed in collaboration with Nanomegas and ULB will provide complementary structural information. WP3 applies Reverse Monte Carlo simulations to reconstruct molecular structures from these experimental data, producing structural models suitable for computational chemistry. Building on these models, WP4 focuses on training Neuroevolution Potentials for potential energy surfaces and Tensorial Neuroevolution Potentials for predicting electric dipoles and polarizability tensors, enabling efficient molecular dynamic. These potentials will be validated against ab initio molecular dynamics and further used to calculate infrared (IR) and Raman spectra. In WP5, experimental IR and Raman spectra will be compared with theoretical predictions, including a normal mode analyse, to identify correlations between spectral signatures and compositional changes (e.g., Si/Al ratio).
The integration of experimental and theoretical approaches will deliver (i) high-quality datasets on AAMs, (ii) validated ML potentials tailored to aluminosilicates, and (iii) new insights into the structural origins of spectroscopic signatures. Beyond advancing fundamental understanding, the project establishes a transferable methodology for combining diffraction, spectroscopy, atomistic modelling, and machine learning to study disordered materials.