Aggregates make up around 70% of concrete by volume, leading to a high river sand and gravel extraction rate. This depletion of natural resources will soon become unsustainable as the demand will inevitably overcome the availability of natural aggregates. Moreover, about 200 million tons of basic oxygen furnace (BOF) steel slag (LD slag) are produced yearly worldwide and are predominantly disposed of in landfills. The need to find a sustainable alternative to substitute natural aggregates next to the great availability of steel slag opens the opportunity for its usage as aggregate in concrete. However, when used in concrete, LD slag undergoes an expansive reaction, resulting in cracks and concrete failure. This expansive behavior is widely attributed to the expansion of free lime. The instability has been widely investigated on the macro scale, but only a limited number of studies on the microscale investigated such a hypothesis.

The present study provides a thorough microscopic investigation of LD slag and its instability phenomenon. Firstly, X-ray diffraction (XRD) and Scanning Electron Microscopy (SEM) are used to characterize the steel slag’s chemical composition and phase distribution. In this study, a phase is defined as a portion of a system with similar chemical characteristics. The main phases of LD slag were dicalcium silicate (belite), calcium ferrite with potential alumina content, and wustite phase. During characterization, it was found 2 distinct phase distributions in LD slag particles, the first with small round and isolated belite phases and another with larger belite phases that are interconnected. This difference in phase distribution is attributed to the different cooling rates of the molten steel slag.

Then, possible reasons for the volumetric instability in LD slag were microscopically investigated, namely, the hydration of free lime, belite polymorphic transformation, hydration of magnesium oxide, and release of residual stress. Free lime and belite polymorphic transformation were not observed in the LD slag analyzed in this study and, therefore, were not considered as a mechanism of the volumetric instability. Moreover, Mg-rich phases have been observed in many isolated and interconnected LD slag particles, which are likely mechanisms of damage onset. Lastly, residual stress was qualitatively confirmed, and inhomogeneous hydration was also observed, which supports the hypothesis of redistribution of residual stress as a damage mechanism of concrete with LD slag as aggregate.

All findings and instability generation in concrete have been performed in concrete samples from accelerated curing conditions, such as high-temperature underwater curing. The test was validated by comparing the damage of the laboratory samples to a 15-year-old concrete left outdoors in the field, exposed to actual environmental conditions. It was observed that the damage features were similar, indicating that the accelerated curing conditions could well replicate the damage in real conditions and could predict the behavior of the concrete in practice.

Lastly, mitigation measures have been proposed, focusing on the concrete binder and particle treatment. It was found that the instability is delayed by the increasing development and compactness of the concrete matrix, likely by delaying water ingress. Moreover, by crushing LD slag particles, and thus -partially- releasing the residual stress present, the damage in concrete was reduced, delaying cracking initiation and concrete failure. However, it was found that despite delaying the damage onset, modifying the binder and crushing the aggregates could not ultimately prevent damage and instability.