Data-driven insights into the performance of scalable magnetic clay-based composites for pollutant removal

Magnetic clay-based composites are promising materials for pollutant remediation due to their tunable surface chemistry, ion-exchange capacity and facile magnetic recovery. However, their practical deployment remains limited by poor scalability and the lack of a predictive framework linking physicochemical properties to adsorption performance. Here, we address both challenges by combining scalable spray-drying fabrication with colloidally driven design and data-driven modelling, establishing a predictive approach to multifunctional adsorption. We developed a scalable strategy to integrate Fe₃O₄ nanoparticles into oppositely charged clay matrices via heterocoagulation or coprecipitation, followed by one-step spray-drying to produce mechanically robust micrometre-sized granules. The resulting composites retained the intrinsic magnetic behaviour of Fe₃O₄ (M-H up to 1.5 T) while exhibiting strong and selective uptake across multiple pollutant classes, including heavy-metal ions (Cu²⁺ and Fe³⁺ > 15 mg g^-1), oppositely charged dyes (rhodamine B ∼20 mg g^-1; methylene blue and methyl orange ∼7 mg g^-1) and phenols (30-40 µg g^-1). The co-aggregates were highly reusable, with < 10% efficiency loss after four reuse cycles. Notably, in all cases, chemisorption governed the overall uptake, primarily driven by ion-exchange interactions with aqueous species. A machine-learning model (R ² = 0.91) correlated adsorption capacity with material descriptors, identifying zeta potential, isoelectric point and hydrodynamic size as the dominant factors controlling performance. Together, these results moved beyond empirical material design by providing a scalable and predictive structure-property framework for magnetic clay composites, enabling their rational optimisation and practical implementation in water treatment.